NH₃-Induced Challenges in CO₂ Hydrogenation over the Cu/ZnO/Al₂O₃ Catalyst

Abstract

Gas sources rich in CO₂ derived from biomass/waste gasification, anaerobic digestion, or industrial carbon capture often contain impurities such as H₂S, H₂O, and NH₃, which can significantly hinder catalyst performance. Here, we show the role of NH₃ on the reverse water–gas shift (RWGS) reaction over a commercial Cu/ZnO/Al₂O₃ catalyst, examining its effects on both the catalytic activity and the catalyst structure. We found that NH₃ reversibly decreases CO₂ conversion immediately by suppressing carbonate hydrogenation and CO desorption. This effect intensifies with an increase in NH₃ concentration but decreases at higher temperatures. However, prolonged exposure (over 100 h) to RWGS conditions in the presence of 1.4% NH₃ leads to near-total and irreversible deactivation of the Cu/ZnO/Al₂O₃ catalyst. Under NH₃ exposure, the catalyst loses Cu⁺ sites on the surface, causing a spatial separation of Cu and ZnO. Finally, to address this challenge, we propose a novel strategy to mitigate NH₃ inhibition by decomposing NH₃ into N₂ and H₂.

Introduction

Since the Industrial Revolution, increasing levels of carbon dioxide in the atmosphere have driven anthropogenic climate change.¹ Consistent with the guidelines of the United Nations Framework Convention on Climate Change, many countries prioritize achieving carbon neutrality.² CO₂ hydrogenation to CO, commonly known as the reverse water–gas shift (RWGS) reaction, is among the most promising CO₂ conversion processes for CO₂ utilization.^3,4 The produced CO can be converted into various liquid fuels (gasoline, diesel, and methanol) via Fischer–Tropsch synthesis (FTS) and other syngas conversion processes.^5,6

CO₂ streams derived from biomass/waste gasification, biomass/waste anaerobic digestion, carbon capture from power plants and steel plants, and other industrial processes⁷⁻¹¹ often contain impurities such as NH₃, H₂S, and H₂O.^12,13 For example, gas from biomass steam gasification typically contains 38–56% H₂, 17–32% CO, 13–17% CO₂, 1–150 g m^–3 tar, 100–1400 ppm of NH₃, and 20–200 ppm of H₂S.¹² Additionally, trace amounts of NH₃ can occur in CO₂ streams from industrial carbon capture processes.^14,15 These impurities can act as catalyst poisons in CO₂ conversion processes.^16,17 Furthermore, the main capital costs of FTS are gasification and gas purification, and gas cleanup is particularly costly for biomass-to-liquid processes.¹⁸ Therefore, establishing benchmarks for the tolerance of catalysts to specific poisons without significant deactivation can help determine the necessary gas cleaning requirements. This insight can contribute to the development of poison-resistant catalysts and the realization of cost-effective, environmentally friendly energy utilization.¹⁹

Some studies have reported the effects of H₂O and H₂S on CO_x hydrogenation processes.²⁰⁻²³ For example, one study reported that 500 ppm of H₂S deactivated Cu/ZnO catalysts by forming ZnS, CuS, Cu₂S, and CuSO₄ phases.²³ In addition, a suitable amount of water (2.24 mol %) promoted methanol formation over Cu/ZnO/ZrO₂ catalysts during CO₂ hydrogenation by facilitating the hydrolysis of methoxy.²¹ However, the role of NH₃ in the CO₂ hydrogenation processes has not been reported. Additionally, studies have shown that the effects of NH₃ on the FTS process may be contradictory.²⁴⁻²⁶ NH₃ can adsorb onto and block cobalt sites, inhibiting the adsorption and activation of H₂, making ppm levels of NH₃ harmful to the FTS process using cobalt catalysts.²⁶ Furthermore, increasing the ammonia level (in the form of NH₄NO₃) to 400 ppm rapidly deactivated the Fe catalyst and simultaneously changed the product selectivities by oxidizing χ-Fe₅C₂ to Fe₃O₄.²⁷ Conversely, recent studies have demonstrated the positive effects of NH₃ on the FTS process. NH₃ can reduce the secondary reactions of olefins and enhance the production of primary products, particularly olefins.²⁵ NH₃ has also been shown to expand the types of FTS products. For example, the formation of long-chained aliphatic amines, nitriles, and amides was promoted via exposure to 10% NH₃ over Fe-based catalysts.²⁸

Cu-based catalysts, particularly Cu/ZnO/Al₂O₃, are widely used in CO₂ hydrogenation reactions owing to their excellent activity, selectivity, and low cost.²⁹⁻³¹ In this study, we investigated the role of NH₃ on the RWGS reaction activity and catalyst evolution over the Cu/ZnO/Al₂O₃ catalyst. We report that short-term tests revealed a significantly reversible inhibition of the CO₂ hydrogenation activity by NH₃. Operando diffuse reflectance infrared Fourier-transform spectroscopy (DRIFTS) and solid-state nuclear magnetic resonance (ssNMR) experiments revealed that NH₃ impeded carbonate hydrogenation and CO desorption. Moreover, NH₃ competed with CO₂ and H₂ for adsorption on the catalyst surface. Exposure to a H₂/CO₂ system containing 1.4% NH₃ for 100 h resulted in complete deactivation of the Cu/ZnO/Al₂O₃ catalyst. To elucidate the deactivation mechanism, changes in active metal states, catalyst structure, and physicochemical properties were investigated. Our findings indicate that the presence of NH₃ in the gas flow led to the eventual disappearance of Cu⁺ sites on the catalyst surface, causing the separation of Cu and ZnO and a consequent loss of redox properties. Finally, we designed a series of ammonia-resistant catalysts to counteract the inhibitory effect of NH₃ in the RWGS reaction.

Results

Alteration of the RWGS Reaction Pathway by NH₃ Addition

The CO₂ hydrogenation process was conducted in a tubular fixed-bed reactor (Figure S1). In the absence of a catalyst, the conversion of CO₂ was low (Figure S2). In this study, the Cu and ZnO loadings in the industrial Cu/ZnO/Al₂O₃ catalyst were 58.2 and 29.6 wt %, respectively, as determined via inductively coupled plasma–optical emission spectroscopy (ICP-OES) (Table S1). The catalyst was initially reduced in situ under a 5% H₂/N₂ flow at 500 °C for 2 h (with a heating rate of 10 °C min^–1) before the RWGS reaction. Cu/ZnO/Al₂O₃ exhibited high performance, approaching thermodynamic equilibrium, when NH₃ was absent in the reaction gas under temperatures of 300–700 °C and a H₂/CO₂ ratio of 1:1 (Figures S3 and S4). The introduction of 0.1% NH₃ had a minimal impact on the reaction (Figure 1A). However, increasing the NH₃ concentration to 1.4% inhibited CO₂ conversion from 4.4 to 0.9% at 300 °C and reduced the CO selectivity from 90.1 to 66.9%. Additionally, solid byproducts were observed at the reactor outlet and identified as ammonium carbonate ((NH₄)₂CO₃) and ammonium bicarbonate (NH₄HCO₃) through X-ray diffraction (XRD) analysis (Figures S5 and S6). Increasing the NH₃ concentration to 5% resulted in nearly complete catalyst deactivation, with the CO₂ conversion dropping to 0.06% and a substantial decrease in the CO selectivity and the CO formation rate at 300 °C (Figure 1A,C). This indicates that the catalytic activity decreases with increasing NH₃ concentrations, while the CH₄ formation rate remains largely unaffected, suggesting distinct formation mechanisms for CO and CH₄ (Figure 1C and Figure S7). Additionally, the impact of NH₃ was mitigated at higher temperatures (Figure 1A–C and Figure S3). Increasing the temperature from 300 to 400 °C led to a 5.5% reduction in CO₂ conversion caused by the presence of 5% NH₃ (from 15.6 to 10.1%) (Figure 1B). Notably, the RWGS was not significantly inhibited by NH₃ at temperatures above 500 °C (Figures S3 and S8, vide infra). The inhibitory effect of NH₃ was reversible in short-term tests (less than 60 min), as demonstrated by adding NH₃ to the reactor for a specific period and then stopping its introduction (Figure 1D and Figure S9).

Figure 1.

Role of NH₃ on the catalyst performance. (A) CO₂ conversion and CO selectivity over Cu/ZnO/Al₂O₃ at 300 °C and (B) at 400 °C (reaction conditions: 1 bar, H₂:CO₂:N₂ = 1:1:1, contact time: 0.05 s g/mL^–1). Inset: CO selectivity at different NH₃ concentrations. (C) CO formation rate during RWGS over Cu/ZnO/Al₂O₃ with 5% NH₃ addition (reaction conditions: 1 bar, H₂:CO₂:N₂ = 1:1:1, contact time: 0.05 s g/mL^–1). (D) Changes in CO₂ conversion when switching the gas composition (reaction conditions: 1 bar, H₂:CO₂:N₂ = 1:1:1, contact time: 0.05 s g/mL^–1). (E) CO₂ conversion and (F) methanol formation rate over Cu/ZnO/Al₂O₃ with 1.4% NH₃ addition for CO₂ hydrogenation to methanol (reaction conditions: 250 °C, 30 bar, H₂:CO₂:N₂ = 3:1:1, contact time: 0.03 s g/mL^–1).

We also investigated the effects of contact time and the H₂/CO₂ ratio on NH₃ inhibition (Figures S10 and S11). Notably, increasing the contact time reduced the inhibitory effect of NH₃, while NH₃ inhibition was observed across all H₂/CO₂ ratios from 1:1 to 4:1. Furthermore, increasing the H₂/CO₂ ratio enhanced the inhibitory effect of NH₃ on the RWGS reaction. Additionally, we tested the role of NH₃ in CO₂ hydrogenation to methanol at 30 bar and 250 °C with a H₂:CO₂:N₂ ratio of 3:1:1 (Figure 1E,F and Figure S12) and found that NH₃ similarly acted as a reversible poison in the process. Specifically, the presence of 1.4% NH₃ resulted in a significant decrease in the level of CO₂ conversion from 3.9 to 1.4%, a reduction in methanol selectivity from 38.3 to 16.1%, and a decrease in the methanol formation rate from 16.0 to 2.4 mmol h^–1 g_cat^–1. This indicates the broad inhibitory effect of NH₃ on the CO₂ hydrogenation reactions (vide infra).

To elucidate the instantaneous deactivation mechanism caused by NH₃ addition during the RWGS reaction, we conducted in situ DRIFTS experiments to investigate the evolution of the reaction intermediates. The in situ-reduced Cu/ZnO/Al₂O₃ catalyst was first subjected to CO₂ (30 mL min^–1) preadsorption for 30 min, followed by a 60 min N₂ purge before the DRIFT spectra were collected (Figure S13). Bands at 1538, 1476, 1395, and 1362 cm^–1 were observed, indicating the formation of carbonate species on the Cu–ZnO_x interface.³²⁻³⁴ After the gas was switched to H₂, the carbonates were rapidly consumed, leading to the formation of CO and H₂O within 5 min. Notably, when the gas was switched to 1.4% NH₃/N₂ instead of H₂ (Figure S14), carbonates were also consumed, indicating a reaction between NH₃ and CO₂.

To monitor real-time intermediates under the reaction conditions, we further employed operando DRIFTS to investigate surface intermediates on the catalyst. Under an H₂/CO₂/N₂ flow (5/5/90) at 400 °C, no distinct bands were observed (Figure 2A), indicating the rapid consumption of intermediates. Upon switching from a 5% H₂/5% CO₂ flow (lasting for 60 min) to a 5% H₂/5% CO₂/0.2% NH₃ flow, carbonate bands began to appear within 1 min, reached their maximum intensity after 5 min, and then gradually faded, disappearing after 17 min (Figure 2A). It indicates that the introduction of NH₃ initially hindered carbonate hydrogenation, attributed to the initial decrease in catalyst activity (Figure 2B). Studies have reported that adsorbed NH₃ negatively affects dissociative H₂ adsorption,^26,27 which may explain why NH₃ hindered carbonate hydrogenation in this study. After 5 min, an equilibrium between NH₃ and H₂ on the catalyst surface was achieved, leading to the gradual consumption of carbonates and stabilization of the reaction activity. This trend was also observed in experiments with the effect of NH₃ on CO₂ hydrogenation to methanol. The results further elucidate the impact of NH₃ on carbonate hydrogenation processes (Figure 1E,F).

Figure 2.

Evolution of reactive intermediates as detected via operando DRIFTS and ssNMR. (A) Evolution of the surface functional groups over Cu/ZnO/Al₂O₃ (reaction conditions: 400 °C, 5% H₂/5% CO₂/0.2% NH₃/N₂) in 17 min. (B) Illustration of NH₃ hindering carbonate hydrogenation. (C) Evolution of carbonyl groups over Cu/ZnO/Al₂O₃ (reaction conditions: 400 °C, H₂:CO₂:N₂ = 1:1:1 with 1.4% NH₃) during 10 h of TOS. (D) Ex situ MAS NMR ¹H–¹³C HETCOR spectra of Cu/ZnO/Al₂O₃ exposed in H₂/¹³CO₂ or H₂/¹³CO₂/5% NH₃ (reaction conditions: 400 °C, 4 h, H₂:¹³CO₂ = 1:1). (E) Ex situ DD-MAS NMR ¹³C spectra of Cu/ZnO/Al₂O₃ exposed to H₂/¹³CO₂ or H₂/¹³CO₂/5%NH₃ (reaction conditions: 400 °C, 4 h, H₂:¹³CO₂ = 1:1).

Furthermore, we tested NH₃ adsorption on the catalyst with and without the reaction stream. An in situ DRIFTS experiment conducted solely under 1.4% NH₃ conditions revealed weak adsorption of NH₃ on the catalyst surface. This observation is consistent with the results from NH₃ temperature-programmed desorption (NH₃ TPD) and NH₃ DRIFTS (Figures S15–S17). In contrast, pronounced NH₃ adsorption occurred when a similar operando DRIFTS experiment was conducted under H₂/CO₂/N₂ (1:1:1) with 1.4% NH₃ for 10 h. The products were simultaneously quantified via gas chromatography (GC) (Figure S18). During the 10 h time on stream (TOS), a rapid decline in catalytic activity was observed at the initial stage. Bands corresponding to NH₃ species on Cu/Zn sites (1623 cm^–1), NH₄⁺ species on Zn–OH sites (1423 and 1340 cm^–1), and NH₃ species on Zn sites (1260 and 1234 cm^–1) emerged (Figures S18–S20).^35,36 These findings highlight a pronounced affinity of NH₃ for the ZnO substrate under a CO₂-containing flow (for detailed discussions, see Figures S19–S21). This enhanced NH₃ adsorption on the catalyst surface further hindered the hydrogenation of the carbonates.

We also investigated the effect of NH₃ on carbonate hydrogenation using density functional theory (DFT) calculations. These calculations first established a stable configuration of the Cu–ZnO_x interface with adsorbed CO₂. The configuration was characterized by the stable bonding of oxygen atoms (from ZnO) to both Cu and adsorbed CO₂. The analysis highlighted that the strong adsorption of NH₃ on the Cu surface occupied the Cu atoms (with an adsorption energy of −0.72 eV, Table S2 and Figure S22), making it difficult for H₂ to adsorb near *CO₃ and provide *H (with an adsorption energy of −0.06 eV, Table S2 and Figure S22). Consequently, subsequent reactions occurred between *CO₃ and NH₃ or the decomposition products of NH₃. Additionally, the adsorption of NH₃ by Zn ions (blue atoms) at the Cu–ZnO_x interface affected the reaction intermediates (with an adsorption energy of −1.11 eV; see Figure 2B, Table S2, and Figure S22). Once NH₃ was adsorbed on Zn^δ+, CO production became extremely difficult. Moreover, recent investigations have elucidated that HCOO* on the Cu surface serves as a critical intermediate in the formation of CH₄ rather than CO.^37,38 This delineates a potential pathway for methane synthesis, where the notably low rate of CH₄ formation implies that intermediates are present in minimal quantities, rendering them elusive to detection. The adsorption site of the HCOO* species (on the Cu surface) differed from that where the CO₃* species formed (at the Cu–ZnO_x interface). This distinction may explain the unaffected methane production. Additionally, the stability of HCOO* species, unaffected by a hydrogen-rich environment, coupled with NH₃ capacity to provide hydrogen, likely underpins the sustained production of CH₄ observed.³⁸

In the CO adsorption DRIFTS experiment, the band at 2103 cm^–1, observed after N₂ purging, disappeared with the increase in temperature to 100 °C, indicating weak CO adsorption on the catalyst (Figure S23). During the CO₂ hydrogenation process, with or without NH₃, in situ DRIFTS revealed two bands at 2120 and 2167 cm^–1, corresponding to gas-phase CO (Figure 2C).³³ In the 10 h DRIFTS experiment with 1.4% NH₃, a new band emerged at 2203 cm^–1, which was absent under H₂/CO₂ conditions on the Cu/ZnO/Al₂O₃ catalyst (Figure 2C). Additionally, the NH₃ adsorption process in isolation did not manifest this band, indicating that the band is associated with species generated by the interaction between NH₃ and CO₂ (Figure S24). Thus, we ascribed the band to the *N=C=O (*NCO) species.^39,40 Subsequent in situ DRIFTS analysis on ZnO samples under a NH₃/CO₂ flow at 400 °C revealed a related band at 2210 cm^–1, while a similar examination of the Cu surface showed this band at 2200 cm^–1 (Figure S25). This suggests that the *NCO species (2203 cm^–1) preferentially adsorbed at the Cu–ZnO_x interface rather than on isolated Cu or ZnO surfaces. Additionally, the Zn^δ+ sites at the Cu–ZnO_x interface (blue atoms in Figure 2B) were the primary sites for NH₃ adsorption. The presence of *NCO species indicates that the formed CO* reacted with adsorbed NH_x species to produce *NCO species rather than desorbing as gas-phase CO. Consequently, the RWGS reaction was suppressed.

To gain deeper insights into changes in reaction intermediates during the RWGS reaction, we conducted isotopic labeling experiments. First, a mixture of H₂/¹³CO₂ was subjected to the Cu/Zn/Al₂O₃ catalyst for 4 h at 400 °C. A separate catalyst was treated with a mixture of H₂/¹³CO₂/5% NH₃ under similar conditions. Afterward, the system was cooled to room temperature and thoroughly purged with nitrogen. The resulting samples were carefully transferred into a specialized rotor within a glovebox to avoid exposure to air, and their magic angle spinning nuclear magnetic resonance (MAS NMR) spectra were recorded (Figure 2D,E and Figures S26 and S27). The ¹H–¹³C cross-polarization (CP)-MAS NMR spectrum of the catalyst under NH₃-free condition showed a peak at 170 ppm (Figure 2D), which is attributed to bicarbonate or formate species.^41,42 However, in situ DRIFTS experiments revealed the absence of formate species (Figure S28), leading us to attribute it to bicarbonate species. This peak was absent in the ¹H–¹³C CP-MAS NMR spectra of the catalyst exposed to the RWGS stream with 5% NH₃ (Figure 2D). To confirm the absence of carbonaceous species on the surface of this catalyst, a ¹³C double quantum dipolar (DD)-MAS NMR spectrum was collected (Figure 2E). While the ¹³C CP-MAS NMR signals are restricted to carbon nuclei near protons, the ¹³C DD-MAS NMR technique does not have this limitation. Consequently, both CP-MAS and DD-MAS methods are often used to assess protonated and nonprotonated carbons in carbonaceous materials.⁴³ The ¹³C DD-MAS NMR analysis revealed a distinct signal at 166 ppm for the catalyst under 5% NH₃, attributed to carbonate species.⁴⁴ The results indicate that NH₃ inhibited carbonate hydrogenation, consistent with the findings from in situ DRIFTS.

NH₃-Induced Long-Term Deactivation of Cu/ZnO/Al₂O₃

To assess the long-term effect of NH₃ on the performance of Cu/ZnO/Al₂O₃, we subjected the catalyst to 100 h of TOS of RWGS at 400 °C under two conditions: in the absence of NH₃ and in the presence of 1.4% NH₃ (Figure 3). In the absence of NH₃, CO₂ conversion decreased from 15.7 to 13.7% over the 100 h period. In contrast, in the presence of 1.4% NH₃, the CO₂ conversion rapidly dropped from 16.0 to 11.8% within the first 15 min and then gradually declined further to 0.6% over 100 h. Notably, after NH₃ addition was halted, the CO₂ conversion recovered to 3.3%, indicating that the deactivation caused by 1.4% NH₃ was largely irreversible after the long-term test. Additionally, CO selectivity gradually decreased, reaching 90.3% over the first 80 h and then rapidly dropping to 54.5% in the subsequent 20 h. This rapid decline in selectivity is attributable to the low formation rate of CO compared to the relatively stable formation rate of CH₄ (Figure S29).

Figure 3.

Role of NH₃in the stability test of RWGS. (A) CO₂ conversion and (B) CO selectivity and formation rate over Cu/ZnO/Al₂O₃ with or without 1.4% NH₃ (reaction conditions: 400 °C, 1 bar, H₂:CO₂:N₂ = 1:1:1, contact time: 0.05 s g mL^–1).

We employed XRD analysis to investigate changes in the bulk phase of the catalyst (Figure 4A and Figure S30). Notably, after 100 h of TOS in the presence of 1.4% NH₃, CuO was not detected in the XRD pattern, indicating the copper oxidation did not contribute to the deactivation mechanism. A slight shift in the diffraction peaks of copper in the fresh catalyst suggested the formation of a CuZn alloy.⁴⁵ Rietveld refinement of the XRD data revealed that the lattice parameter of Cu in the fresh catalyst was 3.633 Å, while the lattice parameter in the used catalysts was 3.618 Å, which is close to the standard lattice parameter of metallic Cu (3.615 Å, PDF 01-085-1326) (Figure S31 and Table S3). This suggests that Cu in the catalysts occurred in the metallic form. The expansion of the Cu lattice in the fresh catalyst was attributed to Zn incorporation.^46,47 These results confirm the formation of a CuZn alloy in the fresh catalyst and its disappearance after the reaction. The high Cu/Zn ratio (3.0) determined from XRD refinement of the fresh catalyst, which exceeds the ratio derived from ICP analysis (2.5), further supports the formation of a CuZn alloy (Tables S1 and S3).^46,47 Alloy formation due to Zn doping into the Cu lattice enhanced the diffraction intensity of Cu peaks in the XRD pattern while reducing the intensity of ZnO peaks, resulting in a higher Cu/Zn ratio. Although CuZn alloys are known to form upon hydrogen reduction, their inherent instability during catalytic processes makes them undetectable in postreaction catalysts.^32,45 XRD refinements of the used catalysts revealed a consistent Cu/Zn ratio of 1.8, indicating that NH₃ did not alter the phase or composition of the catalyst. Additionally, the crystallite sizes for Cu and ZnO were measured as 26.6–28.7 nm and 1.5–2.3 nm, respectively, demonstrating resistance to sintering in both the absence and presence of NH₃ (Table S3). This suggests that sintering, which is often reported as a deactivation mechanism during CO₂ hydrogenation, was not a significant issue here.^48,49 H₂ TPD results further confirmed the minimal variation in metal dispersion (Figure S32 and Table S4). Additionally, the addition of 1.4% NH₃ caused only a minor decrease in surface area, from 59 m² g^–1 (without NH₃) to 54 m² g^–1 (with 1.4% NH₃) (Table S5).

Figure 4.

Characterization of fresh and used Cu/ZnO/Al₂O₃ catalysts. (A) XRD. (B) Corresponding first derivatives of normalized spectra of Cu K-edge XANES. (C) Corresponding first derivatives of normalized spectra Zn K-edge XANES. (D) Cu LMM Auger spectra. (E) Zn LMM Auger spectra. CuZnAl_NH3-free:Cu/ZnO/Al₂O₃ exposed to H₂/CO₂/N₂ (1:1:1) for 100 h at 400 °C. CuZnAl_1.4%NH3:Cu/ZnO/Al₂O₃ exposed to H₂/CO₂/N₂ (1:1:1) with 1.4% NH₃ for 100 h at 400 °C.

We subsequently conducted X-ray absorption spectroscopy (XAS) on the used catalysts. The Cu K-edge X-ray absorption near-edge structure (XANES) results after 100 h of TOS with and without 1.4% NH₃ showed negligible differences, with both conditions prominently displaying characteristics of metallic Cu (Figure S33). A detailed analysis of the first derivatives of the Cu K-edge XANES spectra revealed a subtle shoulder at 8985 eV (Figure 4B), indicating minor oxidation of the Cu particles.⁵⁰ The Zn K-edge XANES spectra were consistent with those of ZnO, showing a prominent peak at 9662 eV in the first derivative of Zn K-edge XANES (Figure 4C and Figure S34). This peak indicates that the zinc sites retained their oxidation state.⁵¹

To quantify the valence states of the cations, linear combination fitting (LCF) of the XANES spectra was performed. For Cu species, with XRD data ruling out the presence of the CuO phase and XAS serving as a bulk phase analysis technique, we used standards of Cu foil and Cu₂O. After 100 h of TOS, the proportion of Cu⁺ (calculated as Cu⁺/(Cu⁰+Cu⁺)) in the catalyst was 33% under NH₃-free conditions and 38% in the presence of 1.4% NH₃, indicating a similar extent of slight oxidation of Cu particles (Figures S35 and S36 and Table S6). Moreover, the proportions of Zn⁰ (calculated as Zn⁰/(Zn⁰+Zn²⁺)) were 18 and 19% for the catalysts under NH₃-free and 1.4% NH₃ conditions, respectively. This indicates the formation of ZnO_x species and the presence of Zn^δ+ (0 < δ < 2) (Figures S37 and S38 and Table S6). The consistency in the oxidation states of zinc and copper across the used catalysts suggests minimal variability, indicating that NH₃ did not significantly affect the chemical states of these metals.⁴⁵

Extended X-ray absorption fine structure (EXAFS) analysis at the Cu K-edge was conducted to elucidate the local structure and electronic states of copper in the catalysts (Figures S39 and S40). Considering that EXAFS cannot differentiate between Cu–Cu and Cu–Zn scattering paths, the observed path is attributed to Cu–Cu, given the instability of CuZn alloys during the reaction process (vide supra). Both catalysts exhibited a distinct peak at 2.54 Å, corresponding to Cu–Cu scattering, and a weaker peak around 1.90 Å, corresponding to Cu–O scattering. The inverse transforms of these peaks were fitted using parameters for Cu–Cu and Cu–O scattering, respectively (Figures S41 and S44 and Table S7). Both catalysts exhibited the same Cu–Cu distance of 2.54 Å and similar coordination numbers (CNs) of 6.68 (NH₃-free) and 6.29 (1.4% NH₃). Under NH₃-free conditions, the CN for Cu–O was 0.34 at a distance of 1.89 Å, which slightly increased to 0.50 at a similar distance of 1.88 Å in the presence of 1.4% NH₃. These findings indicate mild oxidation of the used catalysts with no pronounced changes in the copper state or structure. Considering that CN can serve as an indicator of crystal size,⁵² the reduction in CN from 6.68 to 6.29 for the Cu–Cu path suggests minimal changes in Cu particle size in the presence of NH₃, consistent with the XRD results (Table S3).

In addition, Zn K-edge EXAFS analysis revealed Zn–O and Zn–Zn scattering paths, similar to those observed in standard ZnO (Figures S45 and S46). The inverse transforms of the peaks were fitted using Zn–O and Zn–Zn scattering parameters at distances of 1.97 and 3.21 Å, respectively (Figures S47–S50). The CN for the Zn–O path showed a minor increase from 0.61 at 1.95 Å under NH₃-free conditions to 0.74 at 1.96 Å with a 1.4% NH₃ exposure. Conversely, the CN for the Zn–Zn path slightly decreased from 1.81 at 3.21 Å (NH₃-free) to 1.23 at 3.23 Å (1.4% NH₃) (Table S7). The Zn K-edge EXAFS analysis results, under both NH₃-free and 1.4% NH₃ conditions, suggest consistent Zn–O CNs, indicating that zinc remained in its oxidized state throughout the experiment. Additionally, the size of the ZnO structures, likely nanoparticles, showed no significant changes, consistent with the XRD results (Table S3). Consequently, these findings imply that exposure to NH₃ did not significantly affect the electronic or structural states of zinc or copper in the catalysts. Further analysis is required to fully understand the impact of NH₃ on the catalysts (vide infra).

We then employed X-ray photoelectron spectroscopy (XPS) to elucidate the elemental states of the surface of the catalysts. The samples were carefully transferred in an airtight holder to prevent exposure to air. The XPS spectra of the O 1s revealed distinct peaks at 531.9 and 533.3 eV, corresponding to lattice oxygen (O_L) and oxygen vacancies or defects (O_V), respectively (Figure S51 and Table S8). The O_V/(O_V+O_L) ratio indicated that the quantity of surface defect due to the formation of ZnO_x⁵³ remained unchanged. This suggests a consistent ZnO_x content, consistent with the XANES findings. Defects have been reported to promote the formation of ZnO_x, which stabilizes a ZnO_x overlayer on the Cu surface and acts as a cocatalyst in CO₂ hydrogenation.^54,55 During the CO₂ hydrogenation to methanol process, the oxygen-vacant ZnO_1–x/Cu interface, rather than Cu alone, CuZn alloy, and stoichiometric oxygen fully terminated ZnO/Cu interfaces, exhibits much higher intrinsic activity for methanol production.^56,57 Furthermore, studies have proposed that Cu–ZnO_x, rather than the CuZn alloy, mainly drives the catalytic activity,⁵⁸⁻⁶⁰ a notion supported by the unstable nature of the CuZn alloy under CO₂ hydrogenation conditions.⁴⁵

Cu 2p XPS, supported by XRD and XANES, confirmed the absence of Cu²⁺species in all samples (Figure S52). To further evaluate the state of Cu after CO₂ hydrogenation, Auger electron spectroscopy, which is highly sensitive to distinguishing between Cu⁰ and Cu⁺ states, was employed (Figure 4D).⁶¹ Analysis of the Cu LMM Auger region revealed that the fresh catalyst contained 10% Cu⁺ sites, while the used catalyst under NH₃-free conditions had 23% Cu⁺ sites (Table S9). Notably, in the Cu LMM Auger spectrum of the catalyst used under 1.4% NH₃ conditions, no Cu⁺ sites were detected, indicating that only metallic Cu was present on the catalyst surface (vide infra). Furthermore, the Zn 2p core-level and LMM Auger electron spectra of the used catalysts (Figure 4E, Figure S53, and Table S10) were consistent with those of the fresh catalysts, indicating the predominant presence of oxidized states.⁶² All catalysts exhibited similar proportions of Zn^δ+, corroborating the presence of ZnO_x and suggesting the absence of significant changes in the ZnO phase, consistent with the XANES results.

High-angle annular dark-field–scanning transmission electron microscopy (HAADF-STEM) combined with energy-dispersive X-ray (EDX) spectroscopy was employed to elucidate the structural attributes of the catalysts. First, the fresh Cu/ZnO/Al₂O₃ catalyst exhibited a uniform dispersion of Cu and Zn (Figure 5A). EDX analysis after 10 h of TOS, both with and without NH₃, revealed no substantial differences compared with the fresh catalyst (Figure S54). However, at a prolonged TOS of 100 h, the catalyst used under NH₃-free conditions continued to show a uniform distribution of Cu and Zn (Figure 5B). In contrast, the catalyst exposed to 1.4% NH₃ displayed a noticeable separation between Cu and ZnO (Figure 5C and Figure S55). When the catalyst was exposed solely to NH₃ (without CO₂ and H₂) for 100 h, this separation was not observed (Figure 5D). This separation is crucial for understanding the catalytic mechanisms in the RWGS reaction, as the Cu–ZnO_x interface plays a pivotal role.^59,63,64 Hydrogen is weakly chemisorbed on metallic Cu, as Cu is known to quickly dissociate and recombine with hydrogen.⁶⁵ The dissociated H is transferred to ZnO_x, where it reacts with carbonates. Therefore, the separation of Cu and ZnO complicates the reaction between the dissociated hydrogen and carbonates. Cu–ZnO_x interfaces accelerate reaction rates compared with surfaces with only Cu.^{54,60,66−68} Reaction rates on Cu surfaces with deposited ZnO nanoparticles can be 5 to 18 times higher than those on the most reactive Cu surfaces.⁶⁰ Additionally, intermediates involved in methanol formation undergo faster hydrogenation owing to the interaction between Cu and Zn, compared with reactions occurring solely on Cu surfaces.⁴⁵

Figure 5.

HAADF–STEM analysis and EDX mapping of Cu/ZnO/Al₂O₃. (A) Fresh Cu/ZnO/Al₂O₃. (B) Catalyst used for 100 h of TOS under NH₃-free conditions, (C) catalyst used for 100 h of TOS in the presence of 1.4% NH₃. (D) Catalyst exposed to 1.4% NH₃/N₂ (in the absence of CO₂ and H₂) for 100 h. Scale bar: 10 nm.

According to our earlier observations, NH₃ exhibited a pronounced affinity for adsorption at the Zn^δ+ sites at the Cu–ZnO_x interface. This interaction weakened the bonding between Cu and O from ZnO_x, leading to an elongation of the Cu–O bond. Consequently, the binding energy at the interface between ZnO_x and Cu decreased, shifting the valence state of Cu toward a more neutral (zero) state, as confirmed via XPS analysis. This reduction in the binding energy between Cu and ZnO_x suggests a fundamental mechanistic pathway for the observed detachment phenomena. Moreover, H₂ temperature-programmed reduction (H₂ TPR) results indicated a significant decrease in H₂ consumption under the 1.4% NH₃ condition (Figure S56 and Table S11), suggesting a reduction in Cu⁺ sites, which aligns with the observed separation of Cu and ZnO. Additional evidence from thermodynamic calculations showed that the reaction between Cu₂O and NH₃ had a significantly lower Gibbs free energy compared with the reaction between Cu₂O and H₂ at temperatures above 200 °C (Figure S57). This finding suggests that NH₃ enhanced the tendency to prolong Cu–O bonds, leading to the formation of Cu⁰ sites. This was particularly evident in cases where NH₃ had a stronger adsorption affinity to the catalyst surface compared with H₂ (−1.11 eV of NH₃ at the Cu–ZnO_x interface and −0.72 eV of NH₃ on the Cu surface, compared with −0.06 eV of H₂ on the Cu surface, Figure S22 and Table S2). It should be noted that the conversion from Cu⁺ to Cu⁰ sites was observed only on the surface and not in the bulk phase. XANES and XRD analyses, which were employed to investigate changes in the bulk phases of Cu and Zn, showed no substantial transformations.

We conducted an operando DRIFTS experiment for 100 h of TOS to elucidate the evolution of reaction intermediates (Figures S58 and S59). Spectroscopic analysis revealed a significant shift in the band from 2203 to 2210 cm^–1 during the reaction, which was attributed to the *NCO species. This shift was correlated with a rapid decline in activity at the initial stage and gradual deactivation over the long-term test, as quantified via GC (Figure S58). Notably, the *NCO species adsorbed on the ZnO surface was identified at 2210 cm^–1 (Figure S25). This shift suggests a gradual transformation of the adsorption sites from the Cu–ZnO_x interface to the ZnO surface. Additionally, the initially subtle band at 1448 cm^–1 became more pronounced as the reaction progressed, indicating increased NH₃ adsorption. This change highlights a rise in the NH₃ adsorption intensity on both ZnO and Cu surfaces (Figures S20 and S59).

Subsequent CO₂ adsorption experiments were conducted on the used catalyst (Figure S60), which had undergone a 100 h test in a CO₂/H₂/N₂ (1:1:1) stream with 1.4% NH₃. A notable shift in the adsorption band from 1538 cm^–1 (carbonate on the Cu–ZnO_x interface) to 1528 cm^–1 (carbonate on the Cu surface) occurred, indicating the separation of ZnO from Cu.⁶⁹ Additionally, the increased intensity of bands at 1335 cm^–1 (carbonate on the ZnO surface) confirmed the increased presence of these species on the ZnO surface,^21,70 as supported by direct CO₂ adsorption on ZnO samples (Figure S21). Moreover, a semiquantitative comparison of the intensity of carbonate bands revealed a marked decrease in signal intensity at the Cu–ZnO_x interface (Figure S60), indicating a reduction in the exposed Cu–ZnO_x interface owing to the separation of Cu and ZnO. Prior to the reaction, weak signals from carbonate on ZnO were present but were overshadowed by the stronger signals from the Cu–ZnO_x interface. After the reaction, as the carbonate signals at the Cu–ZnO_x interface decreased, the signals from the ZnO surface became more pronounced.

DFT-Computed Changes in Reaction Pathways due to NH₃ Addition

The effect of NH₃ on the reaction was calculated by using DFT. The computational model featured a Cu(111) surface loaded with ZnO particles. First, the adsorption of NH₃ on the active sites was evaluated. The results revealed lower adsorption energy for NH₃ (−1.11 eV at the Cu–ZnO_x interface) compared with that of CO₂ (−1.00 eV at the Cu–ZnO_x interface), CO (−0.95 eV on the Cu surface), and H₂ (−0.06 eV on the Cu surface) (Figure S22 and Table S2). This favorable adsorption of NH₃ facilitated its occupation of the reaction sites (Figure 6A I–III). The temperature-dependent adsorption free energy diagram obtained from calculations revealed that as the temperature increased, the adsorption free energies of NH₃, CO₂, and CO on the Cu–ZnO_x interface also increased. This suggests a decreased adsorption capacity at the Cu–ZnO_x interface, indicating a weakened inhibitory effect (Figure 6B and Figures S3 and S61). For reactions on the Cu(111) surface, the formation of HCOO* (with an energy change of −0.75 eV) was found to be more favorable than the formation of *COOH (with an energy change of 0.04 eV). Conversely, at the Cu–ZnO_x interface, HCO₃* emerged as the favorable intermediate with a lower reaction energy of −0.91 eV (Figure S62). Additionally, HCOO* is more likely to be an intermediate in CH₄ production,^38,71 which further supports the observation that CH₄ generation from the Cu(111) surface remained unaffected, while CO formation was suppressed at the Cu–ZnO_x interface.

Figure 6.

DFT calculation of NH₃ effects on the reaction. (A) Adsorption configuration of the reaction gas (I–III) and charge density difference diagrams for CO₂ adsorption and NH₃–HO structure (IV and V; blue represents electron dissipation, yellow represents electron accumulation; isosurface value: 0.005 e bohr^–3). (B) Adsorption free energy of the reaction gases at the Cu–ZnO_x interface. (C) Projected density of states with and without NH₃ adsorption and with H₂O adsorption. (D) Free energy of the reaction at the Cu–ZnO_x interface. (E) Summary of the effect of NH₃ on the reaction.

The CO₂ configuration with the lowest adsorption energy at the interface was further examined, and the results revealed the formation of a CO₃* intermediate. The adsorption of NH₃ will affect the interaction between the CO₃* intermediate and the interface, which is reflected in the weakening of the Zn–O bond. The charge density difference diagram of CO₂ at the interface after NH₃ adsorption indicates that CO₂ acquires electrons from both ZnO and Cu, leading to elongation in the Zn–O bond (Figure 6A IV and Figure S63). NH₃, through its lone pair of electrons, influenced the distribution of Zn d orbitals, leading to increased Zn–N overlap peaks below the Fermi level and causing Zn–O bond stretching. This interaction facilitated the separation of O from Zn, a process that cannot be achieved with bulk O atoms or O in H₂O, which also contains lone-pair electrons (Figure 6C and Figure S64). Moreover, during the hydrogenation process, hydrogenation of the O atom at the edge of the CO₃* intermediate was identified as the most favorable pathway (Figure S65). The hydrogenation led to detachment of the O atom from the edge and the Cu surface. Additionally, NH₃ adsorption on the Cu surface did not affect the transition-state energy of the hydrogenation process, which remained approximately 0.42 eV in both scenarios (Figure S66).

The reaction free energy profile at the Cu–ZnO_x interface is depicted in Figure 6D. For the initial hydrogenation step, the energy change at the interface was −0.48 eV without NH₃ adsorption, which is more negative than the value in the presence of NH₃ (−0.37 eV). This indicates that NH₃ decreases the tendency for the hydrogenation reaction. In the subsequent step involving H₂O generation, hydrogenation on the pre-existing −OH structure showed a higher energy barrier (>1.05 eV). Consequently, it is more probable that hydrogenation would proceed on another oxygen atom of the CO₂ molecule, accompanied by hydrogen transfer to regenerate H₂O with an energy change of <0.60 eV. However, the presence of NH₃ hindered the process of hydrogen transfer due to structural variations induced by hydrogenation on the −OH structure compared to when NH₃ was absent (Figure S67). NH₃ induced the separation of the O atom with a free energy change of −0.25 eV, thereby hindering CO formation and increasing the formation energy barrier of HOHOCO* from 0.54 to 0.60 eV. Additionally, NH₃ adsorption enhanced the interaction between N and Zn, which increased the distance between Zn and Cu (Figure 6C and Figure S63). This increased distance made it more difficult for the COO* intermediate to be stabilized without being anchored by adjacent Zn atoms (Figure S68). Thus, the reaction tended to proceed with the detachment of bulk O, leading to CO₂ production. In the absence of NH₃, the formation of a CO–O–Zn intermediate (COO*) resulted in CO production.

Hydrogenation can occur in different directions on the O atom, leading to the adsorption of generated H₂O on the Zn surface. However, detachment of the O atom would still occur. Additionally, CO₂ adsorbed at the edge Zn site can be hydrogenated in two steps to produce CO. Unfortunately, NH₃ often showed a higher affinity for occupying this Zn site (Figure S69). Analysis of charge density difference diagrams and electron transfer numbers indicated that the degree of anchoring between Cu and ZnO decreased in the presence of NH₃, leading to an increased dispersion potential (Figure S70). This observation is consistent with the experimentally observed separation between Cu and ZnO. Once Cu and ZnO are separated, it becomes difficult to re-establish the Cu-ZnO_x interface. After the separation, the adsorption free energy of CO₂ at the Zn–Cu interface increases, suggesting that this site is unsuitable for CO₂ to undergo further reaction, that is, oxygen remains at the Zn site, allowing the Zn–O–Cu structure to reform and be reduced to CO (Figure S71). In this case, CO₂ is more likely to react at the Zn site on the ZnO edge, producing CO.

Note that when the catalyst was exposed solely to NH₃ (without CO₂ and H₂), the separation of Cu and ZnO did not occur (Figures S22 and S72). When NH₃ adsorbs at the Zn^δ+ site, the Zn–O–Cu structure remains intact, suggesting that the interface is not disturbed. The distance between Zn and the nearest Cu atom is 3.41 Å. Upon CO₂ adsorption at the interface, the O–Cu bond in the Zn–O–Cu structure is significantly weakened or even broken. Oxygen bonds with the carbon atom of CO₂, forming a Zn–O–C structure, and the distance between Zn and Cu increases to 4.18 Å. Moreover, during this process, NH₃ adsorption elongates the Zn–O bond, weakening the orbital overlap of Zn–O (Figure 6C), which creates favorable conditions for the subsequent cleavage of the Zn–O bond. As a result, the increased distances between Zn–O and Zn–Cu facilitate catalyst separation.

Furthermore, NH₃ and the −OH structures within HCO₃* species generated during the reaction exhibited significant electron transfer effects (Figure 6A V), indicating the possibility of a potential chemical reaction between NH₃ and the −OH groups. In comparison to the hydrogenation reaction with a higher energy barrier, the NH₃ adsorption process only showed a significantly lower energy change of −0.01 eV, making it more favorable for occurrence. This favorable energy change resulted in the formation of NH₄ structures and accelerated the conversion of intermediate HCO₃ into ammonium bicarbonate or ammonium carbonate, as supported by experimental results (Figure 6D,E, and Figure S73). Moreover, the catalyst interface lowered the energy barrier for decomposition of NH₃ into NH_x and H, allowing decomposition at temperatures below 700 °C (Figure S74). NH_x structures adsorbed on the Cu surface can potentially attack the C⁺ site, possibly leading to the formation of *NCO structures (Figure 6E and Figure S75). Intergrated crystal orbital Hamilton population (ICOHP) analysis indicated that the N–C bonding in *NCO was excessively strong, which inhibited the CO desorption (Figure S76).

Design of the NH₃-Tolerant Catalyst in RWGS

According to the thermodynamic equilibrium calculations for CO₂ hydrogenation in the presence of NH₃ (Figure 7A and Figure S77), CO₂ conversion considerably increased with higher NH₃ concentrations, suggesting that NH₃ participated in the RWGS reaction to produce CO, H₂O, and N₂. Thermodynamic equilibrium calculations indicated that NH₃ underwent nearly complete decomposition into N₂ and H₂, leaving minimal residual NH₃ (Figure 7A). A stream with either 5 or 100% NH₃ alone resulted in nearly complete NH₃ decomposition starting at 300 °C (Figure S78). In our RWGS experiments, extending the contact time at 400 °C, which promoted the reaction toward equilibrium, effectively suppressed the detrimental effects of NH₃ (Figure S10). Conversely, increasing the H₂/CO₂ ratio suppressed NH₃ decomposition because the higher H₂ concentration inhibited NH₃ decomposition, in accordance with Le Chatelier’s principle. This, in turn, potentially enhanced the role of NH₃ in inhibiting the RWGS reaction (Figure S11).

Figure 7.

Design of the NH₃-tolerant catalyst in RWGS. (A) NH₃ conversion calculated from the thermodynamic equilibrium of CO₂ hydrogenation under different NH₃ concentrations. (B) NH₃ conversion calculated from the thermodynamic equilibrium of the reaction between CO₂ and NH₃ (CO₂:NH₃ = 3:2, 1 bar, Gibbs reactor). (C) Experimental results of the reaction between CO₂ and NH₃ (1 bar, CO₂:N₂ = 1:2, 5% NH₃, contact time: 0.05 s g mL^–1). Role of 5% NH₃ on NH₃-tolerant catalysts in the hydrogenation of CO₂ over (D) Ru–Cu/ZnO/Al₂O₃, (E) Ni–Cu/ZnO/Al₂O₃, (F) β-Mo₂C, and (G) Pt/TiO₂ (reaction conditions: H₂:CO₂:N₂ = 1:1:1, contact time: 0.05 s g mL^–1).

Furthermore, thermodynamic equilibrium calculations for the NH₃–CO₂ reaction, conducted at various theoretical NH₃/CO₂ ratios, showed an enhancement in CO₂ conversion to CO with increasing temperatures (Figure 7B and Figure S79). NH₃ was completely decomposed above 350 °C, resulting in the formation of N₂ and H₂. Additionally, we experimentally investigated the direct interaction between CO₂ and NH₃. The results demonstrated that CO₂ could react with NH₃ at 700 °C even without a catalyst (Figure 6C and Figures S80 and S81). This phenomenon was further confirmed by DRIFTS experiments (Figure S14). When employing the Cu/ZnO/Al₂O₃ catalyst, the reaction between CO₂ and NH₃ commenced at 600 °C, elucidating why higher temperatures mitigate the inhibitory effect of NH₃ (Figure 1C,D and Figure S3).

To mitigate the adverse effects of NH₃, we modified the catalysts to facilitate NH₃ decomposition. Considering that Ru and Ni have been reported as efficient catalysts for NH₃ decomposition,^72,73 we synthesized catalysts with Ru (5%) or Ni (5%) loaded on Cu/ZnO/Al₂O₃ through the impregnation method. The modified catalysts showed high resistance to a high concentration (5%) of NH₃ (Figure 7D,E). Notably, both Ru and Ni demonstrated catalytic activity in CO₂ methanation, even at low temperatures (300–400 °C).^74,75 Therefore, the Ni/Ru-modified catalysts exhibited lower CO selectivity than the unmodified Cu/ZnO/Al₂O₃, which exhibited high CO selectivity (≥90% at 300–700 °C).

Studies have reported that β-Mo₂C and Ni/CeO₂ are effective catalysts for NH₃ decomposition.^76,77 Another study suggested that the NH₃ decomposition rates for transition metals followed the order Ru > Ni > Pt > Pd.⁷³ Consequently, we synthesized and evaluated the performance of β-Mo₂C, Ni/CeO₂, Pt/TiO₂, and Pd/Al₂O₃ (Figure 7F,G and Figure S82), all of which are recognized for their effectiveness in CO₂ hydrogenation.³ Among the catalysts, β-Mo₂C, Ni/CeO₂, and Pt/TiO₂ exhibited notable resistance to NH₃ during the CO₂ hydrogenation. Specifically, β-Mo₂C maintained a stable performance even at 5% NH₃ concentration (Figure 7F). At 300 °C, only a slight activity loss occurred, with conversion decreasing from 7.1 to 6.1%, owing to the low NH₃ decomposition activity at this temperature. From 400 to 700 °C, CO₂ conversion slightly increased under 5% NH₃ condition, attributed to the reaction between CO₂ and NH₃ (or NH₃-derived H₂), while CO selectivity remained at 100% across this temperature range of 300–700 °C. Similarly, Pt/TiO₂ and Ni/CeO₂ demonstrated excellent NH₃ resistance, showing only minimal activity loss at 300 °C (Figure 7G and Figure S82). At temperatures above 500 °C, a slight increase in the level of CO₂ conversion was observed. In contrast, Pd/Al₂O₃ showed relatively low activity in the presence of 5% NH₃, with a noticeable reduction in activity across the 300–600 °C temperature range. This decrease was due to the lower NH₃ decomposition rate of Pd compared to those of the other metals such as Ru, Ni, and Pt.

Discussion

This work elucidates the impact of NH₃ on industrial Cu/ZnO/Al₂O₃ catalysts during the RWGS reaction. In short-term tests, NH₃ acts as a reversible poison by competing for adsorption on the catalyst surface and hindering the hydrogenation of carbonate, thereby reducing the catalyst activity. In contrast, during prolonged CO₂ hydrogenation reactions over 100 h, NH₃ works as an irreversible poison. As a result, the interaction of NH₃ with the catalyst surface triggers the separation of Cu and ZnO attributed to the loss of Cu⁺ sites on the surface, resulting in a gradual decline in the overall catalytic performance and eventual complete deactivation. Notably, we alleviate the adverse effects of NH₃ on Cu/ZnO/Al₂O₃ by catalytically decomposing NH₃ into N₂ and H₂. This finding lays the foundation for the rational design of a group of inherently NH₃-resistant catalysts.

Methods

Catalyst Synthesis

The industrial Cu/ZnO/Al₂O₃ catalyst was purchased from Tianjin Kaite New Material Technology Co., Ltd. The concentrations of Cu, Zn, and Al in the catalysts were determined via ICP-OES (iCAP 7400, Thermo Fisher; Table S1).

The 5% Ni–Cu/ZnO/Al₂O₃ catalyst was synthesized through a wet impregnation method. First, 25 mg of nickel nitrate hexahydrate (Ni(NO₃)₂·6H₂O, 99.99%, Aladdin Co., Ltd.) was dissolved in 10 mL of deionized water. This solution was then added dropwise to a suspension of Cu/ZnO/Al₂O₃ in 100 mL of deionized water. The resulting mixture was continuously stirred and dried at 80 °C overnight. The obtained catalyst was then reduced at 500 °C for 2 h (heating rate of 10 °C min^–1) under a 5% H₂/N₂ atmosphere prior to use. The same synthesis protocol was employed to prepare the 5% Ru–Cu/ZnO/Al₂O₃ catalyst, with 3 mL of ruthenium nitrosyl nitrate (Ru(NO)(NO₃)_x(OH)_y, x+y = 3, 0.15 g L^–1 Ru, supplied by Aladdin Co., Ltd.) replacing the nickel nitrate hexahydrate.

The 5% Pd/Al₂O₃ catalyst was synthesized through the incipient wetness impregnation method.⁷⁸ γ-Al₂O₃ powder (surface area of 200 m² g^–1) and Pd(NH₃)₄(NO₃)₂ (purity of 98%) were purchased from Aladdin Co., Ltd. After impregnation, the catalysts were dried at 100 °C overnight. The dried catalyst was then calcined at 500 °C for 2 h (heating rate of 10 °C min^–1) in a 6.7% O₂/N₂ atmosphere and subsequently reduced at 500 °C for 30 min (heating rate of 10 °C min^–1) under a 2% H₂/N₂ atmosphere.

The CeO₂ nanorod was synthesized through a modified reported method.⁷⁹ First, Ce(NO₃)·6H₂O (4.8 g, Sigma-Aldrich Co., Ltd., 99.999%) was dissolved in 10 mL of deionized water, and then the resulting solution was added to a NaOH solution (Greagent, ≥96%). The mixture was stirred for 1 h and then diluted with 140 mL of water. The resulting solution was transferred to a stainless-steel autoclave (300 mL) and maintained at 100 °C for 24 h. The obtained solid was centrifuged, rinsed, and then calcined at 700 °C to yield CeO₂ nanorods. 5% Ni/CeO₂ was prepared through a wet impregnation method using Ni(NO₃)₂·6H₂O (analytical reagent) as the precursor. After impregnation, the solid was calcined at 700 °C for 2 h (heating rate of 10 °C min^–1) and then reduced at 500 °C for 2 h (heating rate of 10 °C min^–1).

Similarly, the 1% Pt/TiO₂ catalyst was prepared through a wet impregnation method with PtCl₄ and TiO₂ (both from Aladdin Co., Ltd.). The resulting solid was first reduced in 30% H₂/N₂ at 300 °C for 3 h (heating rate of 10 °C min^–1) to remove chloride. The solid was then calcined in air at 400 °C for 4 h (heating rate of 10 °C min^–1). Prior to testing, the catalyst was pretreated in a H₂ flow at 500 °C for 2 h (heating rate of 10 °C min^–1).

Mo₂C was synthesized via a temperature-programmed carburization procedure.⁸⁰ First, (NH₄)₆Mo₇O₂₄·4H₂O (Aladdin Co., Ltd.) was heated in a muffle furnace to 500 °C at a rate of 10 °C min^–1 and held at this temperature for 4 h to prepare the MoO₃ precursor. The material was then crushed and sieved to a particle size of 300–450 μm. To obtain β-Mo₂C, the powder was carburized in a tubular quartz reactor under a 20% CH₄/H₂ (150 mL min^–1) flow. The temperature was ramped from room temperature to 300 °C at 5 °C min^–1, and then from 300 to 700 °C at 1 °C min^–1, and held at 700 °C for 2 h. The resulting material was cooled to room temperature under a N₂ flow and then passivated with 1% O₂/N₂ for 12 h. Before testing, Mo₂C was pretreated in 15% CH₄/H₂ at 590 °C for 2 h with a heating rate of 10 °C min^–1.

Catalyst Characterization

N₂ physisorption isotherms were measured by using an Autosorb iQ Station 2 apparatus at −196 °C. The materials were first degassed at 300 °C under N₂ flow overnight. The Brunner–Emmet–Teller surface area was determined according to the International Union of Pure and Applied Chemistry.⁸¹ Barrett–Joyner–Halenda analysis was applied to obtain pore size distributions. The carbon black model or the Harkins–Jura statistical thickness model was used for the calculation.

TEM and STEM coupled with EDX spectroscopy were performed by using a high-resolution field emission transmission electron microscope (JEM-2100F), which featured advanced drift correction capabilities. The microscope was operated at an electron acceleration energy of 200 kV, achieving a point resolution of 0.23 nm and a STEM resolution of 0.20 nm. The lattice resolution was 0.10 nm, allowing for precise energy spectrum element analysis over a broad range from beryllium (Be, atomic number 4) to uranium (U, atomic number 92). Molybdenum grids were used instead of Cu grids to avoid interference from the Cu scattering.

Powder XRD data were collected using a Bruker D8 Advance diffractometer equipped with a Bragg–Brentano high-definition mirror, operated at 60 kV and 300 mA with Cu Kα radiation (λ = 1.5406 Å). The materials were scanned over a 2θ range of 5–90° with a step size of 0.02° and a scan time of 3 s per step.

XPS was conducted on an ESCALAB Xi⁺ instrument (Thermo Fisher Scientific) within an ultrahigh vacuum chamber (<1.0 × 10^–9 mbar). A monochromatic Al target and magnetic lens mode were used with an X-ray beam spot size of 500 μm. The pass energy for the survey spectrum was set to 160 eV with a step size of 1.0 eV. For the narrow scan spectrum, the pass energy was 20 eV, and the step size was 0.05 eV. An airtight cell was utilized to transfer samples (mounted on carbon tapes) from the glovebox to the XPS chamber without exposure to air.

H₂ TPR, NH₃ TPD, and H₂ TPD were conducted by using a Micromeritics AutoChem 2920 instrument equipped with a thermal conductivity detector. Approximately 100 mg of the sample was placed in a U-shaped quartz reactor. Prior to the experiments, the sample was dried in an Ar flow at 300 °C for 30 min at a heating rate of 10 °C min^–1. H₂ TPR was performed under 5% H₂/Ar from room temperature to 500 °C at a heating rate of 10 °C min^–1. For the NH₃TPD experiment, the sample was pretreated at 500 °C under 5% H₂ for 2 h, and then exposed to a 5% NH₃/Ar flow for 30 min at 50 °C, and purged with Ar for an additional 30 min. Subsequently, the sample was heated to 800 °C at a rate of 10 °C min^–1 under an Ar flow, and the desorbed NH₃ was monitored with the thermal conductivity detector. A control experiment conducted without introduction of NH₃ was performed to establish the background. H₂TPD was conducted using a similar procedure, starting at −50 °C.

DRIFTS measurements were performed by using a Fourier-transform infrared spectrometer (Nicolet 6700) equipped with a mercury cadmium telluride detector. In situ DRIFT spectra were recorded through the collection of 32 scans at a resolution of 4 cm^–1. Prior to measurement, the catalysts were pretreated in situ with a flow of 30 mL min^–1 of 5% H₂ at 500 °C for 2 h. Background spectra were recorded at the corresponding temperatures under a N₂ flow. Subsequently, the catalysts were exposed to a gas flow with the specified composition at a consistent flow rate of 30 mL min^–1 at the required temperature. For NH₃ adsorption, the reduced sample was exposed to a 1.4% NH₃/N₂ flow (30 mL min^–1) for 20 min at room temperature. The NH₃ flow was then stopped, and N₂ was purged for 60 min. The temperature was subsequently increased to 200, 300, and 400 °C at a heating rate of 10 °C min^–1, with the corresponding spectra collected at each temperature. For CO adsorption, the in situ-reduced sample was exposed to a 5% CO/N₂ flow (30 mL min^–1) for 20 min at room temperature. Subsequently, the CO flow was stopped and N₂ was purged for 60 min. The temperature was then increased to 100, 200, 300, and 400 °C at a heating rate of 10 °C min^–1, with spectra collected at each temperature.

Solid-state NMR experiments were conducted by using a JEOL 600 MHz JNM-ECZ600R spectrometer equipped with a 3.2 mm MAS triple resonance probe and employing the single-pulse method. The MAS frequency was set to 12 kHz, and cross-polarization (¹H–¹³C) was utilized as the primary technique, with a ¹H excitation frequency of 100 kHz. For ¹H–¹³C HETCOR experiments, a contact time of 0.5 ms was used, and DUMBO homonuclear (¹H–¹H) decoupling was applied during the t₁ period. DD-MAS NMR experiments were also performed.

XAS data for the Cu K-edge and Zn K-edge were collected in the transmission mode using a TableXAFS-500A instrument (Specreation Instruments Co., Ltd.). The electron beam energy was set to 20 keV (top-up), with an energy resolution ranging from 0.5 to 1.5 eV. The samples were ground and pressed into tablets with a diameter of 13 mm. Air-sensitive catalyst samples were sealed in capillaries within a glovebox and analyzed without exposure to air. Cu XAS spectra were obtained at the K-edge in transmission mode using a silicon drift detector, with continuous scanning from 8879 to 9429 eV. Zn XAS spectra were collected at the K-edge in transmission mode using a silicon drift detector, with continuous scanning between 9469 and 10019 eV. The data were processed with Athena software (version 0.9.26),⁸² and EXAFS fittings were performed using Artemis software (version 0.9.26).⁸²

CO₂ Hydrogenation Reaction

The CO₂ hydrogenation reactions were conducted in a tubular fixed-bed reactor (ID = 6 mm) with 0.05 g of catalyst loaded in the isothermal zone (Figure S1). Prior to each test, the Cu/ZnO/Al₂O₃ catalyst was reduced at 500 °C for 2 h (heating rate of 10 °C min^–1) under a 5% H₂/N₂ flow (60 mL min^–1). The reaction was conducted with gas mixtures of H₂ and CO₂ in ratios of 1:1, 2:1, or 3:1, with N₂ as the balancing gas (1/3). The products were analyzed online using a Gasboard-3100 gas analyzer (Hubei Ruiyi Co., Ltd.). Different contact times were achieved through variation of the gas flow rate from 15 to 60 mL min^–1. The product formation rate, CO₂ conversion, and selectivity to each product were calculated using eqs 1–5:

1

2

3

4

5

where N_x,out is the outlet flow rate of species x [mol h^–1]; C_x,out is the outlet gas fraction of species x; N_x,in is the inlet flow rate of species x [mol h^–1]; r_x is the formation rate of species x [mmol h^–1 g_cat^–1]; m_cat is the mass of catalyst used in the reaction [g]; X_CO2 is the conversion rate of CO₂; S_x is the selectivity of species x; t is the contact time [s g mL ^–1]; and V_gas is the gas flow rate [mL s^–1].

Computational Details

The DFT calculations were performed using the Vienna Ab Initio Simulation Package code.⁸³ The exchange-correlation function was described through the generalized gradient approximation method in the Perdew–Burke–Ernzerhof⁸⁴ form. van der Waals interactions were accounted for using the DFT-D3 method.⁸⁵ A cutoff energy of 450 eV was employed, and a vacuum space of 15 Å in the z-direction was set to avoid periodic effects. Γ-centered Monkhorst–Pack K points of 1 × 1 × 1 and 5 × 5 × 1 were employed for geometry optimization and electronic property calculations. In addition, the convergence criteria for energy and force were set to 10^–5 eV/atom and 0.02 eV/Å, respectively, throughout the calculations. Atomic charges were computed using the atom-in-molecule scheme proposed by Bader.⁸⁶ The transition states in the chemical reactions were located through the climbing nudged elastic band (Cl-NEB) method, with a force convergence criterion of 0.05 eV/Å.⁸⁷

For the relative energy, ΔE_DFT represents the difference between the two states. The Gibbs free energies can be calculated using the following equation:

where ΔE_DFT is the energy difference calculated by DFT; Δ_ZPE is the difference in zero-point energy between reactants and products; TΔS and ∫C_pdT represent the contributions of entropic and enthalpic energies at 400 °C (T = 673.15 K). This step analysis was facilitated by the VASPKIT postprocessing software.⁸⁸

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacsau.4c01097.

• Additional experimental details, data, and analysis, including photographs of the experimental setup (PDF)

Author Contributions

H.Z. and X.B. conceived the research project. X.B. planned the experimental work. S.Z. prepared the Ni/CeO₂ catalysts. J.H. helped with the charge density analysis. X.Y. assisted the catalyst characterization. Q.L. and Yg.Z. helped with the CO₂ hydrogenation reactions. Yy.Z. designed and supervised the DFT calculations. Y.P. and X.W. performed the DFT calculations. Data were discussed among all coauthors. R.E.P. assisted in revising the manuscript. Y.P. and Yy.Z. wrote the calculation section, while X.B. and H.Z. wrote the remaining sections with contributions from all authors.

The authors declare no competing financial interest.