Direct Synthesis of Ammonia from Nitrate on Amorphous Graphene with Near 100% Efficiency

Ammonia is an indispensable commodity in the agricultural and pharmaceutical industries. Direct nitrate‐to‐ammonia electroreduction is a decentralized route yet challenged by competing side reactions. Most catalysts are metal‐based, and metal‐free catalysts with high nitrate‐to‐ammonia conversion activity are rarely reported. Herein, it is shown that amorphous graphene synthesized by laser induction and comprising strained and disordered pentagons, hexagons, and heptagons can electrocatalyze the eight‐electron reduction of NO3− to NH3 with a Faradaic efficiency of ≈100% and an ammonia production rate of 2859 µg cm−2 h−1 at −0.93 V versus reversible hydrogen electrode. X‐ray pair‐distribution function analysis and electron microscopy reveal the unique molecular features of amorphous graphene that facilitate NO3− reduction. In situ Fourier transform infrared spectroscopy and theoretical calculations establish the critical role of these features in stabilizing the reaction intermediates via structural relaxation. The enhanced catalytic activity enables the implementation of flow electrolysis for the on‐demand synthesis and release of ammonia with >70% selectivity, resulting in significantly increased yields and survival rates when applied to plant cultivation. The results of this study show significant promise for remediating nitrate‐polluted water and completing the NOx cycle.


Introduction
Ammonia is an important chemical widely used as a fertilizer and in the manufacture of many nitrogen-containing compounds. [1] Because of its greater energy density and lower transportation cost than hydrogen, [2] ammonia is also regarded as a promising next-generation energy carrier. [3] The Haber-Bosch process, which catalyzes the reaction of nitrogen with hydrogen at temperatures of 400-600°C and pressures above 100 bar, [4,5] has been the main industrial process for ammonia production for more than a century. Because of the energy-and resource-intensive character of the Haber-Bosch method, electrochemical synthesis of ammonia via the nitrogen reduction reaction and plasmon-enabled nitrogen activation using renewable electricity are regarded as promising approaches to decarbonize the NH 3 market. [6] However, the high dissociation energy of the N≡N bond (941 kJ mol -1 ) of ox-LIG and LIG relative to those of rGO. LIG and ox-LIG exhibit an enhanced FE of nitrate-to-NH 3 conversion relative to rGO. The amorphous graphene FE exceeds 80% over a wide potential range and reaches a maximum value of nearly 100% at −0.93 V versus reversible hydrogen electrode (RHE) with an ammonia production rate of ≈3000 μg cm −2 h −1 . In situ Fouriertransform infrared (FTIR) spectroscopy and density functional theory (DFT) calculations reveal the essential role of amorphous carbon in directing the catalytic selectivity. The degree of oxidation of amorphous graphene also affects selectivity. LIG, which has an amorphous structure with lower oxygen content, shows poorer activity than ox-LIG. Our work describes the uncommonly high nitrate reduction activity of amorphous graphene and its potential for application in wastewater remediation and the agricultural industry. We also show that electroreduced nitrate electrolyte can be used to cultivate vegetables and significantly increase crop yields.

Preparation and Characterization of Graphene Materials
Amorphous graphene and conventional rGO were synthesized as shown in Figure 1a. ox-LIG and LIG were formed directly from polyimide by CO 2 -laser irradiation in ambient air and under a nitrogen atmosphere, respectively. The local temperature due to the photothermal effect during laser irradiation can be as high as 2500°C. [24] The scanning electron microscopy (SEM) images in Figure S1 (Supporting Information) reveal the extensive porosity of ox-LIG and LIG, which is caused by the rapid release of gaseous products. [25] The porous structure exhibits a large specific surface area, [24] which is crucial for catalytic applications. The transmission electron microscopy (TEM) images in Figure S2 (Supporting Information) show that the ox-LIG and LIG structures comprise only a few layers. Graphene oxide (GO) was prepared by a modified Hummers' method. [26][27][28] rGO was prepared by thermal reduction of GO. The TEM images of rGO in Figure S2 (Supporting Information) show single-layered and wrinkled structures. X-ray photoelectron spectroscopy (XPS, Figure S3a, Supporting Information) shows that the oxygen content of ox-LIG is greater than that of LIG and rGO with more C-O and O-C=O bonds in ox-LIG revealed by deconvolution analysis. The 291.02 eV binding energy peak in LIG is assigned to -* shakeup of the aromatic sp 2 carbons. [29] Figure S3b (Supporting Information) shows ≈148°a nd ≈20°contact angles for hydrophobic LIG and hydrophilic ox-LIG, respectively, which correlates with their surface properties. Raman spectra of ox-LIG, LIG, and rGO are depicted in Figure S4 (Supporting Information). Three characteristic signals at ≈1352 (D peak), ≈1589 (G peak), and ≈2700 cm −1 (2D peak) are observed. [30,31] The G-peak to 2D-peak intensity ratios (I G /I 2D ) of ox-LIG and LIG are 2.7 and 2.03, respectively, which indicates that the materials comprise but a few layers [32] in accord with the TEM images in Figure S2 (Supporting Information). The intensity of the 2D peak of rGO is relatively low, which may result from wrinkles and layer lamination in the very thin material ( Figure S2, Supporting Information). [33][34][35][36] A larger D-peak to Gpeak intensity ratio (I D /I G ) usually signifies a greater degree of disorder in carbon materials. [33,37] The larger I D /I G of ox-LIG relative to LIG may arise from its greater degree of oxidation. and rGO (c). Scale bars equal 5 and 0.5 nm in the normal and enlarged areas, respectively. Four-, five-, six-, seven-, and eight-membered rings are highlighted in green, blue, yellow, red, and light brown, respectively. d) PDF analyses and e) EPR signals of ox-LIG, LIG, and rGO.
The X-ray diffraction (XRD) pattern in Figure S5 (Supporting Information) contains peaks corresponding to the (002) and (100) facets. The (002) peaks of ox-LIG and LIG are located at 2 = 26°, whereas 2 = 23°for rGO. This difference suggests a larger dspacing in rGO than in ox-LIG and LIG. [38] The FTIR spectra in Figure S6 (Supporting Information) show the presence of C-O (1024-1043 cm −1 ), C-OH (1144-1163 cm −1 ), and C=C (≈1640 cm −1 ) bonds in all samples. The vibrational mode at ≈3438 cm −1 is from the O-H bond of water. [39] The HRTEM images in Figure 1b,c reveal distinctly different carbon lattices in ox-LIG and rGO. Figure 1b shows that the rGO structure comprises predominantly strained hexagons with sporadic pentagons, whereas significantly different lattices are observed in ox-LIG and LIG. The ox-LIG ( Figure 1c) and LIG ( Figure S7, Supporting Information) structures contain primarily five-, six-, and seven-membered rings embedded with a few fourand eight-membered rings. The hexagons in ox-LIG and LIG are severely distorted. The irregular atomic structure with abundant disorders and grain boundaries establishes the amorphous nature of ox-LIG and LIG. PDF analysis provides statistical insight into the interatomic distances in rGO, ox-LIG, and LIG. All samples show three characteristic graphene bands with the nearest planar pair distances (r) of ≈1.42, 2.46, and 2.84 Å ( Figure 1d) and a minor peak at r = 1.97 Å from the interlayer sp 3 bonds. [40][41][42] The nearest-neighbor bond distance in rGO is 1.46 Å, whereas it is 1.44 Å in both ox-LIG and LIG, which reflects the different degrees of strain in the three materials. The shorter carbon-carbon distance in ox-LIG and LIG may arise from the formation of multiple Stone-Wales defects due to rapid heating and cooling during laser irradiation. [42] The room-temperature EPR signal (Figure 1e) demonstrates the presence of a small population of unpaired electrons in the rGO structure, in accordance with recent literature. [42] However, a distinct EPR signal is observed in ox-LIG and LIG, which indicates the existence of abundant radicals. [29,6] ox-LIG exhibits the most intense EPR signal, which may result from greater sparking during laser irradiation. The defects and structural strain in ox-LIG and LIG induced by laser irradiation may benefit electrocatalysis.

Electrochemical Measurements and Product Analysis
Nitrate reduction was first characterized using an H-cell with Ag/AgCl and platinum foil reference and counter electrodes, respectively. The working electrode was prepared by coating ox-LIG, LIG, or rGO with 1 mg cm −2 catalyst ink on cellulose paper (CP) sprayed with a layer of Au (Au/CP), as described in the Experimental Section. The supporting electrolyte was 1 m NaNO 3 , and all potentials were corrected to the RHE. Optical and SEM images of the surface and cross-section of the catalyst coating on Au/CP are presented in Figures S8 and S9 (Supporting Information). The thickness of the catalyst layer was 70-80 μm. We use Au-coated paper as the current collector because it shows high conductivity and negligible ammonia yield, so that the interference from carbonaceous or metallic electrodes [13][14][15][16][17][18] could be minimized. In addition, the large resistivity of rGO or LIGs will cause a significant voltage drop, leading to an underestimated intrinsic performance ( Figure S10, Supporting Information). The sheet resistances of catalyst-loaded Au/CP electrodes are collected in Table S2 (Supporting Information). The sheet resistance increases from 12 Ω □ −1 for bare Au to 18.8-27 Ω □ −1 for the catalyst-loaded electrodes. The electronic conductivity decreases in the sequence LIG > rGO > ox-LIG, which is attributed to the lower oxygen content of LIG and to the fewer defects in rGO relative to ox-LIG. The electrochemical impedance spectroscopy (EIS) measurement in Figure S11 (Supporting Information) presents that the charge transfer resistance at −0.4V versus RHE of rGO (22.95 ± 0.95 Ω) is larger than ox-LIG (9.95 ± 0.15 Ω) and LIG (13.19 ± 0.21 Ω). At higher overpotential (<−0.6 V), passivated film impedance due to surface polarization becomes significant. Both passivated film impedance and charge transfer resistance of rGO are the largest at the same overpotential, suggesting its poorest activity toward nitrate reduction.
We have confirmed that NH 3 is produced by the reduction of NO 3 − . Figure 2a and Figure S12 (Supporting Information) show linear sweep voltammetry (LSV) traces in Na 2 SO 4 and NaNO 3 electrolytes. The current density increases with increasing overpotential in the presence of NO 3 − , which indicates the occurrence of nitrate reduction rather than hydrogen evolution. [16] The products of nitrate reaction were quantified by colorimetry using the indophenol blue method for NH 3 and the Griess test for NO 2 − . Ultraviolet-visible (UV-vis) spectroscopy was used to calculate the FE and yields of NH 3 and NO 2 − . Calibration curves showing the relationship between concentration and UV-vis absorbance are presented in Figures S13 and S14 (Supporting Information). The quantity of NH 3 produced was also determined by 1 H NMR spectroscopy ( Figure S16, Supporting Information). The FE of NH 3 production of LIG at −0.73 V detected by the indophenol blue method is 82.4% and 84.8%, respectively. The yield rate established by colorimetry (709.5 μg cm −2 h −1 ) is close to that determined by NMR (727 μg cm −2 h −1 ), which confirms that nitrate is the source of NH 3 . The NH 3 partial current at ox-LIG is much greater than at a bare Au electrode ( Figure S17, Supporting Information), which demonstrates the negligible effect of Au/CP on the nitrate reduction reaction.
The FE and production rate of NH 3 are summarized in Figure 2b,c, respectively. NH 3 selectivity at rGO is less than that at LIG and ox-LIG. ox-LIG attains a FE of ≥85% at −0.73 to −0.93 V versus RHE and reaches nearly 100% at −0.93 V. The yield rates of NH 3 at ox-LIG and LIG are greater than at rGO. Furthermore, the ox-LIG attains a yield rate of 150% greater than rGO at high overpotentials. The yield rates of ox-LIG and LIG at −0.93 V are 2859 ± 112.7 and 2572.5 ± 75.3 μg cm −2 h −1 , respectively. The electrochemical active surface area (ECSA) method is used to normalize the activity of the catalysts. [3, [43][44][45] As shown in Figures S18 and S19 and Table S3 (Supporting Information), the ECSA of rGO is nearly twice larger than ox-LIG and LIG, while the NH 3 partial current of rGO is twice lower than ox-LIG and LIG after normalization to the ECSA. N 2 adsorption isotherm measurement also suggests a higher specific surface area of rGO ( Figure S20, Supporting Information). This indicates the higher performance of LIG and ox-LIG, which is not due to the surface area, but from intrinsically higher catalytic activity.
The electrolysis time also affects selectivity. The NH 3 FEs and yield rates of the three catalysts gradually increase upon prolonging electrolysis from 15 to 60 min (Figure 2d). The NH 3 FE of ox-LIG is surprisingly high at -0.73 V. This parameter reaches 78% after 15 min electrolysis and improves to 86% after 60 min, whereas that of rGO is only 55% after 15 min. The NH 3 FE of ox-LIG is less dependent on reduction time than that of rGO, which indicates more favorable NO 3 − to NH 3 conversion by ox-LIG. The higher hydrophobicity of LIG leads to slightly less nitrate reduction than hydrophilic ox-LIG, but LIG is still superior to rGO in this respect. The nitrate reduction reaction of the commercial amorphous carbon materials-carbon black (CB) is studied for comparison. [46,47] The absence of 2D peak in the Raman spectrum of CB ( Figure S21a, Supporting Information) reveals that CB adopts a different structure from LIG. [30,47] Figure S21c,d (Supporting Information) shows that the NH 3 FE and yield rate of CB are significantly lower than those of graphene materials, suggesting that both amorphous atomic structure and graphene feature contribute to the catalytic activity. The nitrate-to-ammonia conversion efficiency of ox-LIG at −0.83 V in the electrolytes with concentrations of 0.2, 0.4, 0.6, 0.8, and 1 m are investigated and presented in Figure S22 (Supporting Information). The NH 3 partial current density correlates with the nitrate concentration (Figure S23, Supporting Information).
Isotope labeling experiments by electrolyzing Na 14 NO 3 and Na 15 NO 3 solutions ( Figure 2e) were conducted to verify the source of NH 3 formation. 1 H NMR coupled with UV-vis absorption was used to determine the NH 3 produced. Three peaks for 14 NH 4 + at = 7.12, 6.94, and 6.77 ppm are observed when Na 14 NO 3 is used as the electrolyte. However, only two peaks corresponding to 15 NH 4 + at = 7.01 and 6.89 ppm are found upon the electroreduction of Na 15 NO 3 . Quantification of ammonia using UV-vis method shows NH 3 FE of 46% and 40% for electroreduced 0.2 m Na 14 NO 3 and Na 15 NO 3 , respectively. This result confirms that ammonia is produced solely via nitrate reduction. Figure 2f shows the results of cycling tests using ox-LIG for NO 3 − at −0.73 V. The FE and yield rate of NH 3 remain unchanged after 12 consecutive cycles, indicating the superior stability of the catalyst. Figure 2g compares the nitrate reduction performance of ox-LIG with that of other electrocatalysts, and the detailed comparison of nitrate reduction performance is summarized in Table S4 (Supporting Information). It should be noted that the test conditions in the literature are different. Thus, the comparison might not be fully conclusive. Nonetheless, ox-LIG exhibits a high FE for NH 3 production at high partial current densities that is superior or comparable to the performance of non-noble metal or alloy catalysts (Fe, Ti, CuNi) [12,17,[48][49][50] and noble metal (Au in this work and Rh [51] ) catalysts.
Nitrite ion is a minor product of NO 3 − reduction. The FE of NO 2 − production is summarized in Figure S24 (Supporting Information). The FE of NO 2 − formation decreases with increasing overpotential, which is contrary to the behavior of NH 3 . Figure S25 (Supporting Information) shows that the NO 2 − FE at −0.73 V remains nearly constant from 15 to 60 min for all three catalysts, which differs from the FE behavior of NH 3 (Figure 2d). Prolonging the electrolysis time increases the concentration of NO 2 − , which can serve as an additional nitrogen source for ammonia electrosynthesis.
To rule out the possible interference of metal impurities, we first measured the metal contents in electrolyte and catalysts by inductively coupled plasma-optical emission spectroscopy (ICP-OES). As shown in Table S5 (Supporting Information), there are no active metal impurities in electrolyte and ox-LIG and LIG. However, 0.73 mg g −1 of Fe is found in rGO due to the massive chemical use in the sophisticated preparation process. [53] To further rule out the possible contaminations from the reference electrode and the platinum counter electrode, a control experiment at a constant current density of 15 mA cm −2 with carbon paper as the counter electrode and without reference electrode was performed ( Figure S26, Supporting Information). The performance is consistent with the yield rate and selectivity of ox-LIG using the three-electrode configuration (Figure 2b,c and Figure S15 (Supporting Information)).
In situ FTIR spectroscopy [54] was used to provide insight into the mechanism of nitrate reduction and the detection of adsorbed reaction intermediates. Potential-dependent transmittance spectra of ox-LIG and rGO are presented in Figure 3a. The peak at ≈1633 cm −1 is attributed to the O-H bending of water. [54,55] This signal is more significant for ox-LIG at higher overpotentials, possibly due to the enhanced adsorption of water [55] and the formation of hydroxide from electroreduction. The new band at 1242 cm −1 indicates the formation of NO 2 − . [54] The peaks at 1444, 1294, and 1116 cm −1 at −1 V of ox-LIG are assigned to molecularly adsorbed NH 3 . [55][56][57] These peaks shift to 1442, 1292, and 1109 cm −1 at −0.8 V, respectively, possibly due to the change of adsorption configuration under varied potentials. [55,58] The peak at 1350-1380 cm −1 merges at high overpotential in both ox-LIG and rGO, assigned to NO 3 − possibly due to the change of local concentration due to highly polarized surface. [52,59] For control sample of Au-coated prism, only O-H bending at ≈1633 cm −1 and nitrate signal at 1350-1380 cm −1 can be found ( Figure S27, Supporting Information), which is consistent with the poor nitrate reduction of Au ( Figure S17, Supporting Information). The in situ FTIR spectra tested in 1 m NaCl are shown in Figure 3b. The reducing potential induces negligible fluctuation of the peak intensities. It confirms that the increase of peaks in 1116-1444 cm −1 in FTIR spectra in NaNO 3 (Figure 3a) results from the nitrogen species rather than the oxygen species in the graphenic materials. The NH 3 band on ox-LIG, which emerges at 0 V, is enhanced at greater overpotentials. By contrast, rGO generates a very weak NH 3 band only at high overpotentials. This suggests that ox-LIG has a smaller onset overpotential and greater NH 3 Figure 3. In situ FTIR spectra and theoretical calculations. a,b) Potential-dependent transmittance spectra of ox-LIG and rGO in 1 m NaNO 3 (a) and 1 m NaCl (b). The reference spectrum was recorded at +0.2 V vs RHE. c) DFT-calculated models of LIG, ox-LIG, and rGO. The red, brown, and pink spheres represent oxygen, carbon, and hydrogen, respectively. d) Free energy diagrams of complete NO 3 − to NH 3 conversion on LIG, ox-LIG, and rGO.
yield than rGO. The nitrite band on rGO is more intense than that on ox-LIG, which agrees with the greater NO 2 − FE of rGO ( Figure S24, Supporting Information). The in situ FTIR spectra confirm the greater catalytic activity of ox-LIG for the reduction of nitrate.

Theoretical Calculations
We performed DFT calculations of reaction pathways [7,60] to understand the origin of the high NH 3 selectivity. Our experimental data suggest that both ox-LIG and LIG with amorphous lattice show higher activity than higher-crystallinity rGO, and ox-LIG is more active than LIG. Thus, here we mainly focus on the activity of amorphous carbon and the effect of oxygen; it does not rule out the possible contribution from the nitrogen of LIGs, whose content is only 1-2% (Table S1, Supporting Information). The three models of graphene are shown in Figure 3c. One model comprises five-, six-, and seven-membered rings to represent the amorphous nature of LIG. A hydroxyl group is fixed at the surface of LIG to represent the local structure of ox-LIG. The other model comprises only six-membered rings to represent rGO. Fig-ure 3d shows the change in adsorption energy of all possible intermediates during nitrate reduction. We first compare the activity of rGO and LIG. There are two energies unfavorable steps, which are the adsorption of NO 2 OH and the transition of *NO → *NOH. Both energy barriers are smaller for LIG. The reduction of *NO to *NOH has been reported as a key step for some catalysts in producing NH 3 . [7,61,62] *NO to *NOH conversion is an uphill process for rGO (+0.84 eV), whereas this energy is reduced to +0.66 eV for ox-LIG. *NOH is ultimately reduced to *N followed by hydrogenation to yield *NH, *NH 2 , and NH 3 . The conversion of *NOH to *N remains unfavorable for rGO, but it becomes a downhill process for LIG. We further compare the energies of ox-LIG and LIG (Figure 3d) to understand the effect of the hydroxide group in catalyzing nitrate-to-ammonia conversion. In general, their energy barriers are more favorable than rGO. The adsorption of NO 2 OH on ox-LIG is lower than LIG by 0.24 eV. Early studies suggested that the hydroxide group could stabilize the adsorption of *NO x species by hydrogen bonds. [63] We observed a hydrogen bond distance of 2.97 and 2.55 Å for *NO 2 OH and *NO 2 on ox-LIG, respectively, which agrees with the literature. [63] To conclude, we attribute the enhanced activities of LIG and ox-LIG to the amorphous structure that can undergo structural deformation to stabilize the adsorbed species, whereas rGO has lower activity due to its aromaticity. The existence of hydrogen bonds in ox-LIG further improves the reaction pathway.

Agricultural Applications
Herein, we demonstrate that electrochemically reduced nitrate can be used to promote the growth of plants in recognition of the fact that ammonia is a potent chemical for increasing crop yields. [64,65] Figure 4a contains a schematic for the remediation of nitrate-polluted water to close the NO x cycle. Nitrate from polluted groundwater and industrial wastewater is electroreduced to NH 3 after purification and concentration and then reused as a fertilizer for vegetable growth or as a fuel. The influence of NaNO 3 on vegetable growth before and after electrolysis is also explored.
Because ox-LIG can produce large quantities of NH 3 at high current densities, we first demonstrate the use of the flow electrolyzer for the on-demand synthesis and controlled release of ammonia. [68] The flow cell ( Figure S29, Supporting Information), which uses peristaltic pumps to improve mass transport, [67,68] can produce massive amounts of ammonia at large and stable currents (Figure 4b). Figure 4c summarizes the nitrate reduction performance of ox-LIG using the flow cell. The NH 3 FE reaches 72% at a current density of 200 mA cm −2 and flow rate of 0.5 mL min −1 . Movie S1 (Supporting Information) shows the feasibility of instant ammonia acquisition from nitrate reduction. Cabbages and radishes are then cultivated using the electroreduced NaNO 3 as fertilizer. The pH of the electrolysis product is neutralized with atmospheric CO 2 before irrigation. Figure 4d,f and Figures S30-S32 (Supporting Information) show that the two vegetables flourish more productively when cultivated with reduced electrolyte than with NaNO 3 . The statistical results in Figure 4e,g show that the electroreduced material significantly promotes vegetable growth and stimulates crop yield by more than 200% relative to NaNO 3 . The survival rate also improves by 30% for cabbage and 86% for radish. The incorporation of trace elements and optimization of the NH 3 concentration from the flow electrolyzer may further improve crop yields. These practical applications show that ammonia from electroreduced nitrate is conducive to vegetable growth and offers an effective means of closing the NO x cycle.

Conclusions
Ammonia plays an essential role in the agricultural and pharmaceutical industries. [71] This study describes the first example of NO 3 − -to-NH 3 reduction by a nonmetallic catalyst with an activity comparable to that of state-of-the-art metallic catalysts. HRTEM imaging, PDF analysis, in situ FTIR spectroscopy, and DFT calculations reveal the unique atomic features of amorphous graphene that facilitate the adsorption of intermediates and formation of NH 3 during NO 3 − reduction. The elevated NH 3 selectivity of amorphous graphene also enables the construction of a flow electrolyzer coupled with an in situ sensing platform [70] to create a smart agricultural system based on electrolytically produced ammonia. [66,71] Controlled release of ammonia can be achieved by using direct laser-writing techniques to fabricate graphene-based sensors and in situ detection modules. Our results will stimulate the design and development of nonmetal electrocatalysts for nitrate reduction with important implications for wastewater remediation.

Experimental Section
Fabrication of LIG: A 120 μm-thick polyimide (PI) film was irradiated with a 10.6 μm CO 2 laser marking machine (Minsheng Laser #MSDB-FM60 CO 2 Laser Marker, 60 W) in ambient and N 2 atmospheres to fabricate ox-LIG and LIG, respectively. The laser power, speed, pulses per dot, and line spacing were set at 1.8 W, 1000 mm s −1 , 5, and 0.03 mm, respectively. The laser was operated in the vector mode. LIG powder was scraped from the PI film and collected for catalyst fabrication.
Fabrication of rGO: GO was prepared by a modified Hummers' method. [26][27][28] rGO was prepared by annealing GO under Ar at 1000°C in a tube furnace. The GO color changed from brown to black upon reduction.
Catalyst Synthesis and Working Electrode Preparation: Gold was sputtered onto tailored CP (area = 1 cm 2 ) with a sputter coater (SCD 050). The current and spraying time were set at 60 mA and 120 s, respectively. The resulting Au/CP film, which was used as catalyst support, exhibited enhanced tenacity and conductivity. CB (Vulcan XC-72) powders were purchased from Fuel Cell Store, USA. 5 mg of catalyst (ox-LIG, LIG, rGO, or CB) and 500 μL Nafion solution (Sigma-Aldrich, 5 wt% Nafion 117) were mixed and sonicated for 30 min. The homogenous catalyst ink was gently dropped onto the Au/CP at 1 mg cm −2 loading and dried at room temperature.
General Characterization: SEM and TEM images were collected with QUATTRO S and Themis Z microscopes, respectively, from Thermo Fisher. Contact angles were measured with a Ramé-Hart Model 190 goniometer. Raman spectra were obtained at 514 nm with a LabRAM HR800 confocal laser micro-Raman spectrometer. UV-visible spectra were obtained by use of a Shimadzu 1700 spectrophotometer. XRD patterns were obtained with a SmartLab diffractometer. XPS spectra were collected with a Thermo ESCALAB 250Xi spectrometer. NH 3 quantification and isotope tracking were conducted by 1 H NMR spectroscopy at 600 MHz with a Bruker spectrometer ASCEND AVANCE III HD. EPR spectra were obtained by an ADANI SPINSCA X spectrometer at room temperature. In situ FTIR spectra were obtained with a Perkin-Elmer Spectrum 100 instrument. The purities of electrolyte and catalysts were measured via an ICP-OES Optima 8000 spectrometer. PDF analysis was extracted from high-energy synchrotron X-ray total scattering by direct Fourier transform of the reduced structure function [F(Q), up to Q ≈ 24.7 Å] using the 11-ID-C beamline atAdvanced Photon Source (APS) of Argonne National Lab (ANL) (X-ray wavelength = 0.1173 Å). The G(r) functions, G(r) = 4 r[ (r) − o ], where (r) and o are the local and average atomic number densities, were computed with PDFgetX3 software.
Electrochemical Measurements: Electrochemical nitrate reduction was carried out in a three-compartment H-cell. The cathode and anode compartments were separated by a Nafion-117 membrane (Fuel Cell Store). Platinum foil and a LF-2 Ag/AgCl electrode (calibrated before use) were used as the counter and reference electrodes, respectively. LSV was conducted from 0 to −1.3 V versus RHE in 1 m NaNO 3 at a scan rate of 10 mV s −1 . All potentials were referenced to the RHE by the following equation (1) The cell was purged with high-purity Ar at a flow rate of 10 cm 3 min −1 during electrochemical measurements. The exiting Ar gas was collected in dilute HCl solution. Electrochemical experiments were carried out with a CHI 760E potentiostat employing 95% resistance compensation. Electrolytes in the flow-cell electrolyzer were purged with Ar gas for 0.5 h before experiments and were circulated through the working and counter electrode compartments using peristaltic pumps (Longer, BT100-2J) operating at 0.5 mL min −1 . and radishes (f) growth. Scale bars: 10 cm. The blue pots were cultivated with the as-prepared NaNO 3 ; the yellow pots were cultivated with electroreduced NaNO 3 . e,g) Statistics of cabbage (e) and radish (g) growth in terms of leaf weight, fruit weight, and survival rate.
Theoretical Calculations: Spin-polarized DFT calculations were performed using the plane-wave basis Vienna ab initio simulation package code. [28,72] The generalized gradient approximation in the Perdew-Burke-Ernzerhof formulation was used with a cutoff energy of 600 eV. An ≈18 Å vacuum space was added as interlamination to eliminate interaction between layers. For reactions involving the transfer of a H + /e − pair, the free energy of the pair was set to one-half the free energy of gaseous H 2 (H + + e − ↔ ½H 2 ). [75] The binding energies of nitrate were corrected with a gas-phase reference based on HNO 3 . [7,74] The Gibbs free energy changes (ΔΔG) were calculated using the computational hydrogen electrode model and defined as [75,76] where ΔE b is the reaction energy, and ΔZPE and TΔS are changes in the zero-point energy and entropy of a species, respectively. The reaction energy was calculated as where E tot , E molecular, , and E adsorbate are the total energy of molecular with adsorbate, the energy of clean molecular, and the energy of adsorbate in the gas phase, respectively. Product Analysis: The quantities of NH 3 and NO 2 − in the electrolyte were determined by colorimetry using the indophenol blue method and the Griess test, respectively. [1,77] Two reagents were prepared for NH 3 calibration. Reagent A comprised 2 mL of 1 m NaOH solution containing 5 wt% salicylic acid and 5 wt% sodium citrate. Reagent B was prepared by mixing 1 mL 0.05 m NaClO and 0.2 mL 1 wt% sodium nitroferricyanide. Reagents A and B were added to 2 mL NH 4 NO 3 solution with concentrations of 0.1, 0.2, 0.3, 0.4, and 0.5 mm. The absorbance at 650 nm was recorded by UV-vis spectrophotometry after a 2 h incubation period. The quantity of NH 3 produced was determined from the amount of indophenol blue formed from 2 mL electrolyte, 2 mL reagent A, and 1 mL reagent B. NO 2 − calibration was based on the reaction of the Griess reagent and NaNO 2 at concentrations of 0.05, 0.1, 0.2, 0.3, and 0.4 mm. The Griess reagent was prepared by dissolving 0.1 g N-(1-naphthyl)ethylenediamine dihydrochloride, 1 g sulfonamide, and 2.94 mL H 3 PO 4 in 50 mL deionized (DI) water. The Griess reagent, electrolyte, and DI water were mixed in a 1:1:2 volume ratio and allowed to stand for 15 min. The absorbance at 540 nm was used to determine the concentration of NO 2 − . Dilution of the electrolyte prior to UV-vis measurement was needed when NH 3 and NO 2 − concentrations in the electrolyte were high. Quantification of ammonia was confirmed by 1 H NMR spectroscopy using benzoic acid as an internal standard. The Faradaic efficiencies of NH 3 and NO 2 − were calculated as follows where F is the Faraday constant, C(NH 3 ) and C(NO 2 − ) are the molar concentrations of NH 3 and NO 2 − , V is the electrolyte volume, and Q is the charge (in Coulombs) passing through the catalyst.
Isotope Labeling Experiments: 0.2 m Na 15 NO 3 was used as the electrolyte in isotope labeling experiments. 15 NH 4 + was measured by NMR spectroscopy at 600 MHz after 1 h electroreduction to establish the nitrogen source of NH 3 produced by electrolysis.
Plant Growth: Equal quantities of cabbage and radish seeds were sowed in rows in culturing pots. The control groups were watered with 5 mm NaNO 3 . The test groups were watered with electrolyte produced by electroreduction of 1 m NaNO 3 in a flow cell operating at 0.5 mL min −1 . The electrolyte was diluted 200-fold after reduction, and CO 2 gas was purged through the solution to adjust the pH to ≈7. The crop yield was weighed, and the number of viable plants was recorded after 40 days.
Statistical Analysis: Data for FE calculation were expressed as mean ± standard error (SE), n = 3. 16 zones with seeds from the same source were selected for plant growth. Dead plants were not included in the crop yield calculation.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.