Copper(II) bromide (CuBr₂, 99.999%), copper(II) nitrate trihydrate (Cu(NO₃)₂, 99%), copper(II) sulfate (CuSO₄, 99%), copper(II) chloride dihydrate (CuCl₂∙2H₂O, 99.999%), copper(I) acetate (Cu(CO₂CH₃)₂, 98%), sodium sulfide hydrate (SNa₂, 96%), sodium bromide (NaBr, 99%), hydrazine monohydrate (N₂H₄∙H₂O, 98%), sodium sulfate (Na₂SO₄, 99%), and deuterium oxide (D₂O, 99.994 atom%), were purchased from Sigma Aldrich. Isopropyl alcohol (99.5%) was purchased from Duksan and used without further purification. Anion-exchange membrane (FAA-3-75, Fumatech) and carbon paper (TGP-H-120, Toray) was purchased and used as the membrane separator and electrode support, respectively. High-purity Ar and N₂ (99.999%) were purchased from Shinyang Oxygen Industry Co., Ltd. ¹⁵N-labeled nitrogen (98 atom% ¹⁵N) for isotope measurements was obtained from Sigma-Aldrich.

For the synthesis of copper sulfide nanoparticles, 0.2 mmol of copper precursors (CuBr₂, Cu(NO₃)₂, CuSO₄, CuCl₂, or Cu(CO₂CH₃)₄) and 0.4 mmol of Na₂S were dissolved in 10 mL of deionized water. The precursor solution was magnetically stirred at room temperature for 30 min. Subsequently, chemical reduction was performed for 2 h at room temperature by adding a solution containing 1 mL hydrazine and 9 mL deionized water. Finally, the resulting solids were collected by centrifugation (9000 rpm, 10 min) and washed with deionized water.

Transmission electron microscopy (TEM), high-resolution TEM (HRTEM), and energy-dispersive X-ray spectroscopy (EDX) images were obtained using a JEM-2100F microscope operated at 200 kV. Scanning electron microscopy (SEM) images were obtained using a Stereoscan 440 (Leica Cambridge). Powder X-ray diffraction (XRD) patterns were measured using a Rigaku D-MAX/A diffractometer at 35 kV and 35 mA. X-ray photoelectron spectroscopy (XPS) results were obtained using a K-Alpha spectrometer (Thermo Electron). The specific surface area and pore size, including their distribution, were evaluated by the Brunauer-Emmett-Teller (BET) method.

Electrochemical experiments, including cyclic voltammetry (CV) and chronoamperometry (CA), were conducted using a Bio-Logic SP-300 potentiostat. CA measurements were specifically conducted to evaluate the eNRR activity of each catalyst. Each CA experiment was carried out within a potential range of 0.0 to −0.6 V_RHE for durations of either 30 min or 1 h to accumulate NH₃ produced in the electrolyte solution. A 0.5 M Na₂SO₄ aqueous solution, continuously purged with N₂ gas for at least 15 min prior to and during the experiment continuously purged with N₂ gas (250 mL/min) for at least 15 min prior to and during the experiment, was employed as the electrolyte. Electrochemical measurements were conducted at room temperature and atmospheric pressure using a three-electrode configuration comprising a saturated calomel electrode (SCE, in saturated KCl aqueous solution) as the reference electrode, a carbon rod as the counter electrode, and catalyst-coated carbon paper (2.5×2.5 cm²) as the working electrode. CA experiments were carried out in a two-compartment H-cell separated by an anion-conducting membrane (FAA-3, Fumatech), which had been pretreated with 1 M KOH aqueous solution for 24 h. During the CA measurements, N₂ gas flowed continuously through the working electrode chamber, and the outlet gas was directed into an NH₃ absorption trap containing 40 mL of aqueous H₂SO₄ solution. Additionally, electrochemical properties of catalysts aside from their eNRR activity, such as electrochemical surface area (ECSA) and HER performance, were evaluated via CV measurements performed on catalyst-coated glassy carbon (GC) electrodes. These CV measurements were conducted within a potential range from −0.3 to −0.8 V_RHE at a scan rate of 50 mV/s under an Ar atmosphere.

NH₃ concentration in electrolyte samples were quantified using indophenol blue method [47]. Initially, a 1 mL aliquot of the electrolyte solution was extracted from either the working electrode chamber of the H-cell or the acid NH₃ trap. Subsequently, 1 mL of phenol solution (0.64 M C₆H₅OH, 0.38 M NaOH, and 1.3 mM C₅FeN₆Na₂O) and 1 mL of hypochlorite solution (55 mM NaOCl and 0.75 M NaOH) were added to the aliquot. After mixing for 2 h at room temperature, UV-vis absorption spectra were recorded between 350 and 900 nm (Cary UV-visible 100 spectrophotometer, Agilent). The indophenol absorbance peak at 633 nm was corrected using background absorbance measured at 875 nm, representing 0 ppm NH₃, then calibrated by subtracting the background absorbance measured at 875 nm. Furthermore, ¹H nuclear magnetic resonance (¹H NMR) spectra were measured using a Bruker Avance III HD 400 MHz spectrometer to confirm NH₃ production from isotopically labeled ¹⁴N₂ or ¹⁵N₂ gases. For ¹H NMR detection of NH₃, 0.1 mL of 1 M H₂SO₄ was added to 1 mL of electrolyte samples sealed carefully to minimize the effects of ambient NH₃ interference form the atmosphere (pH 2). Prior to measurements, 0.055 mL of D₂O (99%, Sigma) was mixed with 0.495 mL sample solution, and 64 scans were performed for each ¹H NMR measurement.

The rate of NH₃ production was calculated using eq (1), where C_NH3 is the concentration of NH₃ determined by using the calibration curve obtained from the UV-vis spectrum, V is the volume of the electrolyte, t is the reaction time, and A is the geometrical area of electrode.

Considering three electrons are required to produce one NH₃ molecule, the F.E. could be calculated using eq (2),

where F is the Faraday constant (96,485 C mol–1) and Q is the total charge passed through the electrochemical reaction in CA.

Copper sulfide catalysts with varying crystallinities and bromide content were synthesized to evaluate the facet-dependent activity and selectivity and to investigate the effects of facet-selective halide adsorption. To modulate crystallinity, various copper salts were used as precursors. Copper sulfide nanoparticles were synthesized by reacting hydrazine with copper and sulfur precursors in aqueous solution without additional stabilizing agents. When CuBr₂, Cu(NO₃)₂, and CuSO₄ were used, the resulting nanoparticles exhibited a plate-like morphology, approximately 150 nm in diameter and 5 nm thick (Fig. 1(a), (b) and (d)). Previous reports suggest that the formation of such nanoscale structures arises from the absence of organic stabilizers, with halide anions instead adsorbing onto the particle surfaces [42].

In contrast, Cu(CO₂CH₃)₄ (Cu acetate) produced aggregated, spherical nanoparticles, and CuCl₂ resulted in a mixture of plate-like and aggregated forms. XRD patterns showed distinct peaks at 28.65, 31.83, and 47.36^o, corresponding respectively to the (0 0 22), (1 0 13), and (1 1 1) planes of Cu₉S₈ (Fig. 2(a), JCPDS file no. 36-0379). These diffraction peaks were distinct in plate-shaped copper sulfide nanoparticles but absent or ambiguous when synthesized from copper acetate and CuCl₂, which instead showed amorphous peaks around 46^o associated with aggregated spherical nanoparticles. In the HRTEM image shown in Fig. 1(f), the lattice spacing of 0.198 nm corresponds to the low-index (1 1 1) plane of Cu9S8. Halide ions have a strong affinity to adsorb onto catalyst surfaces and interact intensively with surface metal atoms due to their high electronegativity [48]. Among halide ions, bromide ions preferentially adsorb onto higher-index facets due to greater adsorption site density, stabilizing these facets and reducing surface energy [43,44]. The ambiguity or absence of characteristic peaks at 28.65 and 31.83^o in the XRD patterns of nanoparticles synthesized from CuCl₂ and Cu acetate indicates that chloride and acetate ions provide weaker stabilization, particularly on higher-index crystal facets, compared to bromide ions.

Bromide ions were selectively removed from CuBr₂-derived copper sulfide or intentionally introduced into copper sulfides synthesized with alternative precursors, where necessary, to investigate their impact on eNRR activity and selectivity. When hydrazine was omitted from the synthesis using CuBr₂, the bromide ions that would normally reside on the surface of the copper sulfide were no longer present. This was confirmed by XPS analysis of the Br 3d core level: a peak at 67.5 eV, indicative of surface bromide ions, was not observed in samples synthesized without hydrazine (Fig. 2(d)). In addition to enabling bromide incorporation, hydrazine also acted as a reducing agent during synthesis, significantly affecting both the morphology and crystallinity of the resulting copper sulfide nanoparticles [49].

In the absence of hydrazine, TEM analysis revealed irregularly shaped nanoparticles (Fig. S5), and XRD patterns exhibited significantly lower crystallinity compared to samples synthesized with hydrazine (Fig. 2(a)). This comparison underscores the dual role of hydrazine—not only as a reducing agent facilitating crystal growth, but also as a condition necessary for bromide ion adsorption during synthesis. To isolate the effect of bromide ions, hydrazine was intentionally excluded in a control synthesis using CuBr₂. XPS analysis of the resulting sample showed that Br 3d signals were no longer detectable (Fig. 2(d)), indicating that bromide ions were not retained on the surface without hydrazine. These results suggest that hydrazine is essential to form well-defined crystalline structures that expose high-index facets, which serve as preferred sites for bromide adsorption. Bromide ions are known to selectively adsorb onto high-index planes due to the higher density of low-coordination surface atoms, enabling strong halide–metal interactions [43,44].

TEM and EDX analyses were also conducted to evaluate morphological changes induced by bromide addition. However, no distinct morphological differences were observed between CuBr₂, Cu(NO₃)₂, and NaBr-added Cu(NO₃)₂ samples; all retained their plate-like structures (Fig. 3(a)–(c)). Additionally, BET surface area analysis (Table S1) showed only slight increases for bromide-containing samples, suggesting that any potential changes in eNRR performance are unlikely to be attributed to differences in morphology or surface area. Instead, EDX elemental mapping confirmed that bromide ions were retained on the surfaces of NaBr-added Cu(NO₃)₂ and CuBr₂ samples (Fig. 3(d), (e)), supporting the conclusion that bromide ions modify the chemical nature of the catalyst surface rather than its physical structure.

eNRR activities of copper sulfide catalysts synthesized from various copper precursors, including Cu(NO₃)₂, CuSO₄, CuCl₂ and Cu acetate, were studied to determine the influence of different surface adsorbates or crystal structures on the eNRR performance. Hereafter, each copper sulfide catalyst is referred to by the name of its respective copper precursor. eNRR activity measurements were performed using a two-compartment H-cell at ambient temperature and pressure, with nitrogen gas (¹⁴N₂ and ¹⁵N₂)-saturated 0.5 M Na₂SO₄ aqueous electrolyte. CVs were carried out in N₂ and Ar-saturated electrolytes within a potential range from −0.3 to −0.8 V_RHE. CVs indicated different onset potentials for HER, observed at −0.6 V_RHE in Ar-purged electrolyte and −0.7 V_RHE in N₂-purged electrolyte. However, relying solely on CV to estimate eNRR activity is insufficient due to surface instability of the copper sulfide catalysts, which undergo sulfide decomposition during eNRR, as evidenced by distinct current responses near −0.6 V_RHE in N₂-saturated conditions (Fig. S3(a)) [27]. Therefore, accurate evaluation of NRR activity was achieved by measuring NH₃ accumulation after CA at potentials ranging from −0.4 to −0.8 V_RHE for 30 min or 1 h.

Among the tested copper sulfides, Cu(NO₃)₂ and CuSO₄ showed high NH₃ yield rates of 669 (j_NH3 = 0.108 mA/cm²) and 657 (j_NH3 = 0.106 mA/cm²) nmol h^–1 cm^–2, respectively, while CuCl₂ showed the highest F.E. of 19.5%. Background absorbance in NH₃ detection was corrected using Ar-saturated electrolytes, in which eNRR is inactive (Fig. 4(b) and S3(b)). Additionally, UV-Vis spectroscopy results from the NH₃ trap solution are presented in Fig. S3(c). To avoid false-positive NH₃ detection due to impurity NOₓ reduction [50], ¹H NMR analysis of electrolytes after eNRR experiments with ¹⁵N₂ gas was conducted following benchmarking protocol suggested by Andersen et al. [51]. The ¹H NMR analysis of electrolyte obtained from eNRR in ¹⁵N₂ purged electrolyte was performed. In the ¹H NMR measurements shown in Fig. 4(c), doublet and triplet peaks were observed which are evidence for existence of ¹⁵NH₃ and ¹⁴NH₃, respectively. Existence of ¹⁵NH₃ doublet supports that NH₃ in electrolyte was electrochemically produced from purged isotopically labeled nitrogen gas, while presence of ¹⁴NH₃ triplet attributed to trace atmospheric contamination or residual ¹⁴N₂ adsorbed in the reactor.

In Fig. 2(a), the copper sulfides with high yield rates, such as Cu(NO₃)₂ and CuSO₄, showed a higher crystalline (1 1 1) plane than CuCl₂ and Cu acetate which had relatively low yield rate exhibited higher F.E. Copper sulfides with higher crystallinity showed clear plate-like shape and the plane direction of this plateau was (1 1 1) confirmed by HRTEM image in Fig. 1(f). The yield rate of these copper sulfides similar but the F.E. of CuSO₄, was 40% higher than that of Cu(NO₃)₂. As shown in Fig. 2(a), CuSO₄ had weak signals for high-index planes, but Cu(NO₃)₂ had stronger signals. Since active HER results in the low selectivity in eNRR, F.E. comparison showed that high-index planes were the active sites for HER. It also showed that higher yield rate of crystalline copper sulfides resulted from the existence of low-index plane. Therefore, the low-index plane dominant plate structure performed as the active sites for eNRR in copper sulfides.

By contrast, CuCl₂ and Cu acetate formed agglomerated structures rather than plate structures. The agglomerated morphologies confirmed by the broad diffraction peak corresponding to the low-index (1 1 1) plane, indicative of their amorphous nature, as shown in Fig. 1(c) and (e). These two sulfides also exhibited noticeably weak XRD peaks in high-index planes, such as (0 0 22) and (1 0 13). Copper sulfides with prominent peaks in high-index planes showed lower F.E., whereas Cu acetate and CuCl₂, which showed weak or no signals in those planes, exhibited higher F.E. despite having lower NH₃ yield. It is well established that low F.E. in eNRR is primarily due to the parasitic HER.

According to previous studies on the eNRR mechanism of copper sulfides, nitrogen molecules are adsorbed onto copper atoms and sequentially hydrogenated by protons released from adjacent sulfur atoms [30,52]. Based on the XRD results in Fig. 2(a) and the yield and F.E. data in Fig. 4(a), it is inferred that nitrogen molecules were primarily adsorbed on copper atoms located on the (1 1 1) planes and hydrogenated by nearby protons. In contrast, protons adsorbed on high-index planes were either reduced to hydrogen gas via the Tafel step or transferred to nitrogen intermediates involved in eNRR.

To summarize, the catalytic performance of copper sulfides in eNRR is governed by a trade-off between activity and selectivity: increased (1 1 1) crystallinity enhances nitrogen adsorption and NH₃ formation, whereas exposure of high-index facets facilitates HER and reduces selectivity. Screening experiments showed that it is difficult to achieve both high activity and high selectivity simultaneously using a single precursor. Enhancing one often leads to a compromise in the other, underscoring the need for targeted surface engineering strategies to overcome this intrinsic limitation.

To simultaneously enhance both activity and selectivity, a strategy was implemented to passivate the high-index planes of highly crystalline copper sulfide, which inherently exposes both low- and high-index facets. In this context, Cu(NO₃)₂ was selected as the precursor, as it enables the synthesis of copper sulfide with high crystallinity in the (1 1 1) plane, thereby ensuring enhanced eNRR activity. The issue of low F.E., attributed to the presence of HER-prone high-index planes, was addressed by introducing bromide ions, which are known to preferentially adsorb onto these high-index surfaces [43,44].

To determine the optimal amount of bromide ions, NaBr in concentrations ranging from 40 to 200 μmol was added to the mixture of copper precursor solution and sulfur source. Hydrazine was then added to carried out chemical reduction. As shown in Fig. 3(c), the addition of NaBr to Cu(NO₃)₂ did not result in any notable morphological change, and residual bromide ions were detected by TEM-EDX mapping, as shown in Fig. 3(e). The amount of adsorbed bromide ion was quantified using TEM-EDX spectra, as shown in Fig. S7. The corresponding yield rate and F.E. for Cu(NO₃)₂ with NaBr addition are presented in Fig. 5(a). The best performance in this study—a yield rate of 1230.9 nmol h^–1 cm^–2 (j_NH3 = 0.198 mA/cm²) and a F.E. of 23.34%— was achieved with the addition of 80 μmol NaBr, which resulted in surface passivation corresponding to 0.07 at% Br. When the adsorbed bromide content was below 0.07 at%, both activity and selectivity improved compared to pristine Cu(NO₃)₂, whereas excessive amounts led to a significant decline in both metrics. CuBr₂, which contains surface-residing bromide ions, exhibited a yield rate of 1130.0 nmol h–¹ cm–² (j_NH3 = 0.182 mA/cm²) and a F.E. of 9.65%, representing 68.9% and 72.3% improvements, respectively, compared to pristine Cu(NO₃)₂. These results confirm the critical role of bromide ions in enhancing both eNRR activity and selectivity. Compared with other catalysts for eNRR, bromide-added CuS_x exhibited comparable activity and higher selectivity than those reported in previous studies, as summarized in Table S4.

In contrast, when bromide ions were absent—such as in the synthesis without hydrazine—the yield rate and F.E. dropped by 32.2% and 50.7%, respectively (Fig. 2(d) and 5(c)). To further validate the role of bromide ions for other precursors, 80 μmol of NaBr was added to both CuBr₂ and Cu acetate, as shown in Fig. 5(b) and (c). When NaBr was added to CuBr₂, the resulting NH₃ yield was similar to that observed for Cu(NO₃)₂ with 160 μmol NaBr, corresponding to only 23.9% of the yield originally obtained with CuBr₂. This is likely because the excessive bromide ions blocked not only the HER-active high-index planes but also the low-index (1 1 1) planes responsible for nitrogen adsorption. Conversely, adding NaBr to Cu acetate did not significantly affect the yield, likely due to the intrinsic lack of nitrogen adsorption sites in this material. Nevertheless, the F.E. increased by 416% and 217% for Cu(NO₃)₂ and Cu acetate, respectively, because the bromide ions selectively adsorbed on high-index planes and acted as HER inhibitors.

The stability of adsorbed bromide ions during the reaction was investigated to confirm bromide ions consistently influenced the reaction. To this end, structures and atomic ratios of the catalyst before and after eNRR were compared by TEM and TEM–EDX, respectively (Fig. S8.). After the reaction, most of the copper sulfide had converted to copper oxide, consistent with previous reports [29,31]. No Br signal was detected in the post-reaction EDX spectra. Importantly, Lv et al. have shown that Br^⁻ can remain stably adsorbed on copper oxide surfaces after similar electrochemical reactions [53]. Taken together, these results suggest that the disappearance of bromide ion in this study is not due to intrinsically weak bromide-surface interactions but is more plausibly attributed to detachment during the copper sulfide to copper oxide phase transformation. Consequently, the intrinsic stability of surface-adsorbed Br⁻ should be reassessed using a more robust sulfide catalyst that preserves its sulfide phase throughout eNRR.

To further confirm the effect of bromide ions on HER activity, HER performance of Cu(NO₃)₂ was evaluated in Ar-saturated Na₂SO₄ electrolytes containing 0, 1, 5, and 10 mM NaBr (Fig. 4(d) and (e)). Cu(NO₃)₂ was selected to examine the influence of bromide ions in the electrolyte on HER activity. As shown in Fig. 3(d), the HER onset potential shi7ed positively and the current density at −1.4 V_RHE decreased with increasing NaBr concentration, indicating that bromide ions suppress HER [54]. For comparison, the HER activity of Pt/C, a benchmark HER catalyst, was also evaluated under the same conditions (Fig. 4(e)). Unlike Cu(NO₃)₂, the HER performance of Pt/C was not significantly affected by NaBr, since HER on Pt predominantly occurs on the (1 1 1) plane, which is less sensitive to halide adsorption. These results further confirm that bromide ions selectively passivate HER-active high-index planes, thereby enhancing eNRR performance through facet-selective suppression of HER.

In conclusion, the crystallinity of the (1 1 1) plane plays a crucial role in determining eNRR activity, while the selective adsorption of bromide ions on high-index planes effectively suppresses HER and improves overall selectivity. This facet-selective adsorption strategy may be extendable to other catalyst systems. For instance, in the study by Sun et al. [55], mercaptan molecules were introduced to MoS₂ to suppress HER by increasing surface hydrophobicity, which enhanced both activity and selectivity. Given that MoS₂ is a well-known HER-active material, bromide ions—being less harmful and easier to handle—could also serve as effective HER inhibitors for such systems. Therefore, facet-selective halide adsorption presents a promising approach for balancing activity and selectivity in catalysts that are inherently active toward both eNRR and HER.

In future studies, facet-selective adsorption approaches can be extended to catalysts that are highly active in both eNRR and HER — that is, systems that deliver a high NH₃ yield but suffer from low F.E. [56,57]. This strategy offers a promising route to balance between activity and selectivity.