Synthesis of monocarboxylic acids via direct CO2 conversion over Ni–Zn alloy catalysts

Malayil Sibi Sungkyunkwan University Deepak Verma Sungkyunkwan University Handi Setiyadi Sungkyunkwan University Paresh Butolia Sungkyunkwan University MUHAMMAD KASHIF KHAN Sungkyunkwan University Sang Kyu Kwak Ulsan National Institute of Science and Technology https://orcid.org/0000-0002-0332-1534 Kyung Yoon Chung Korea Institute of Science and Technology https://orcid.org/0000-0002-1273-746X Jeaho Park Korea Institute of Science and Technology Daseul Han Dongguk University Kyung-Wan Nam Dongguk University-Seoul https://orcid.org/0000-0001-6278-6369 Jaehoon Kim (  jaehoonkim@skku.edu ) Sungkyunkwan University https://orcid.org/0000-0001-6188-7571


Introduction
The direct conversion of CO 2 into fuels and chemicals using renewable H 2 that can be produced via water electrolysis or biomass conversion has received considerable attention because of its potential to mitigate global warming [1][2][3] . Different catalysts have been developed to effectively convert CO 2 into C 1 chemicals (e.g., methanol 4-6 , CO 7, 8 , CH 4 9, 10 , and formic acid [11][12][13] ). However, because of the inherent inertness of CO 2 (∆G 0 298K = −394.4 kJ mol − 1 ) and high energy barrier of the C-C coupling reaction 14 , it is challenging to directly synthesize long-chain hydrocarbons and oxygenated species from CO 2 , with high selectivity. Recently, cascade tandem catalysis strategies have been used to demonstrate that high-yield hydrocarbons and oxygenated species, such as lower ole ns 15, 16 , C 5 + 17, 18 , aromatics 19,20 , and longchain alcohols 21,22 , can be produced directly via the hydrogenation of CO 2 . Although many efforts have been dedicated to the direct conversion of CO 2 to long-chain hydrocarbons and alcohols, less attention has been paid to the synthesis of monocarboxylic acids, such as acetic and propionic acids (AA and PA, respectively), via direct CO 2 conversion.
Acetic acid is an important building block or solvent for producing value-added chemicals, such as vinyl acetate monomers, alkyl acetates, acetic anhydride, and terephthalic acid, which are widely used in the polymer, chemical, textile, and display industries 23 . AA has been mainly produced via methanol carbonylation over Rh-or Ir-based homogeneous catalysts using specially designed reactor systems 24 .
An alternative method for producing AA is the fermentation of ethanol 25 or other biomass-derived species 26 ; however, the contribution of the biological process to the total production of AA is small. Similar to AA, PA is an important chemical that is directly used as a preservative in the food industry or building block for the production of polymers and pharmaceuticals 26 . The main PA synthesis method consists of the hydrocarboxylation of ethylene using nickel carbonyl as catalyst 27 . The global productions of AA and PA are approximately 20 and 16 Mton/year, respectively, and their corresponding market values are approximately 690 and 2300 US$/ton, respectively (Supplementary Fig. S1) 26,28 .
The use of CO 2 as reactant to produce AA has been proposed as a sustainable and low-carbon alternative to the conventional petroleum-based route, and previously reported data are summarized in Supplementary Table 1. First, the direct C-C coupling of CO 2 and CH 4 is an attractive method because it uses low-cost CH 4 as a H source and presents an atomic e ciency of 100% 29,30 . However, the extremely high stability of CO 2 and CH 4 causes the reaction to be highly unfavorable thermodynamically (Eq. (1)).
Thus, the thermochemical conversion of CO 2 and CH 4 resulted in an extremely low CH 4 conversion and AA formation rate (0.05-0.395 mmol g − 1 h − 1 ) even in stepwise, cyclic reaction mode to alleviate thermodynamic limitations [31][32][33][34] .  32 . To mitigate the unfavorable thermodynamic C-C coupling between CO 2 and CH 4 , the use of O 2 35, 36 , methanol 37 , and plasma was proposed 35,36,38 ; however, overcoming the thermodynamic limitations is still challenging. The second approach is the reverse water-gas shift (RWGS) reaction of CO 2 or dry reforming of CH 4 to produce syngas, followed by the conversion of syngas to AA (Eqs. (2a-c)): 39 However, the RWGS reaction for the production of syngas is highly endothermic (Eq. (2a)), and therefore, it is highly energy-and cost-intensive. Third, CO 2 can be directly hydrogenated to AA and PA (Eqs. In the temperature range of 300-340 °C, the formation of AA and PA could be thermodynamically more feasible than the formation of methanol ( Supplementary Fig. 2). In 1994, Ikehara et al. 34 suggested the possibility of obtaining AA via the direct hydrogenation of CO 2 ; it was hypothesized that AA could form over Ag-promoted Rh/SiO 2 via the direct incorporation of CO 2 into methyl groups to form acetate groups.
However, the formation rate of AA was extremely low (~ 0.035 mmol g − 1 h − 1 ). In addition, the direct conversion of CO 2 to PA has never been reported.
In this study, we demonstrated that the direct CO 2 hydrogenation over Ni-Zn alloy deposited on Zn-rich Ni x Zn y O produced AA and PA with high selectivity (58.9% and 18.2%, respectively) at a CO 2 conversion of 13.4%. The catalyst was highly stable for up to 216 h on-stream and the variations in CO 2 conversion and product selectivity were negligible. The Ni-Zn alloy/Zn-rich Ni x Zn y O catalyst was synthesized via the coprecipitation of Ni and Zn precursors followed by calcination and subsequent reduction. The formation of the Ni-Zn alloy was monitored using in situ X-ray diffraction (XRD) analysis. The mechanisms and reasons for the high selectivity toward monocarboxylic acids of the CO 2 and CO hydrogenation reactions were analyzed using in situ near-ambient pressure X-ray photoelectron spectroscopy (NAP-XPS) and operando diffuse re ectance infrared Fourier-transform spectroscopy (DRIFTS). Lastly, plausible pathways for the production of AA were proposed using density functional theory (DFT) calculations.
Furthermore, this is the rst reported synthesis of PA via the direct catalytic hydrogenation of CO 2 . Even though ZnO is known to be an active catalyst for the synthesis of methanol via the hydrogenation of CO and CO 2 , the formation of methanol or higher alcohols over Ni-Zn alloy catalysts has not been reported 6,54,55 .
To analyze the effect of the calcination temperature on the CO 2 hydrogenation, we used Ni-Zn alloy catalysts calcined at different temperatures in the range of 500-1000 °C (Fig. 1a). As the calcination temperature increased from 500 to 700 °C, the CO 2 conversion slightly increased from 9. temperature range of 500-1000 °C, the C 2 -C 4 selectivity was suppressed, and ranged between 1.0-7.1%.
As discussed below, the formation of the Ni-Zn alloy, which played an important role in increasing the selectivity of the reaction toward the formation of monocarboxylic acids, occurred at calcination temperatures above 700 °C.
To further elucidate the role of Ni-Zn alloys for producing monocarboxylic acids with high selectivity, the performance of the N 1 Z n -900 (n = 2, 3, 4) catalysts was compared with that of monometallic ZnO, and the results are presented in Fig. 1b. The use of the monometallic ZnO catalyst prepared via precipitation resulted in a very low CO 2 conversion of 2.0% and high methanol selectivity of 81.3%; moreover, no monocarboxylic acids were formed. Therefore, Ni and Zn presented a strong synergistic effect during the selective conversion of CO 2 to monocarboxylic acids. To investigate the synergistic effect, Ni-Zn alloy catalysts with different Ni/Zn ratios were tested for the CO 2 hydrogenation reaction. As the Ni/Zn ratios changed from 1:2 to 1:3, the AA and CH 4 selectivities increased from 47.2-58.9% and decreased from 28.7-17.0%, respectively, which indicated that the methanation reaction over the Ni-de cient catalyst was suppressed to some extent. This suggested that the Ni/Zn ratios of the Ni-Zn alloy catalysts should be carefully adjusted to obtain catalysts with active sites that could lead to the selective production of monocarboxylic acids. As the Ni/Zn ratio further changed from 1:3 to 1:4, the changes in CO 2 conversion and product selectivity were negligible.
The N 1 Z 3 -900 catalyst presented excellent long-term stability for the CO 2 conversion reaction and remarkable selectivity toward monocarboxylic acids (Fig. 1c). The initial CO 2 conversion activity of the N 1 Z 3 -900 catalyst was maintained for up to 216 h on-stream. After 216 h on-stream, the CH 4 and AA + PA selectivities slightly increased from 17.0-21.6% and slightly decreased from 77.1-71.5%, respectively.
However, even after the long-term catalytic run, the production of methanol and long-chain alcohols were not observed.
Process optimization. The aforementioned experiments demonstrated that the use of the N 1 Z n -900 (n = 2, 3, 4) catalysts produced monocarboxylic acids directly via the hydrogenation of CO 2 with high selectivity.
For the process optimization experiments, CO 2 hydrogenation was performed under different reaction conditions to optimize the CO 2 conversion and monocarboxylic acid selectivity. First, the effect of temperature on the CO 2 conversion over the N 1 Z 3 -900 catalyst was analyzed ( Supplementary Fig. 3a Fig. 3d). The enhanced desorption of * CO adsorbed on the catalyst surface at high GHSV could be responsible for the decrease in AA + PA selectivity.
Characterization of active phases. To analyze the active sites for the effective conversion of CO 2 to monocarboxylic acids, the crystallinity and phase structure of the catalysts were examined using XRD analysis. We assumed that a new alloy-based active phase was formed during the calcination and The local chemical environments of Ni and Zn in the N 1 Z 3 -700 and N 1 Z 3 -900 catalysts were analyzed using X-ray absorption spectroscopy (XAS). The Ni K-edge X-ray absorption near edge structure (XANES) spectra of N 1 Z 3 -700 and N 1 Z 3 -900 were similar to that of a reference Ni foil, which con rmed the presence of Ni 0 in the structure of the catalysts (Fig. 2a). A close inspection of the XANES pro le revealed that the onset of the adsorption edge downshifted slightly compared to that of the Ni foil (inset of Fig. 2a). In addition, the height of the white line peak of the N 1 Z 3 -900 catalyst was lower than that of the Ni foil. These results indicated that Ni in the N 1 Z 3 -900 catalyst presented a slightly negative charge compared to that of Ni in the Ni foil, which was attributed to the electron transfer from Zn to Ni. The Fourier-transform magnitudes of the Ni K-edge extended X-ray absorption ne structure (EXAFS) spectra of the catalysts were similar to that of Ni foil (Fig. 2b). This was ascribed to the similar scattering parameters of Ni and Zn owing to the close proximity of Ni and Zn in the periodic table 61 Fig. 8a). At a low calcination temperature of 500 °C, agglomerated spherical-shape particles with an average particle size of 25 nm formed ( Supplementary Fig. 8b), and the degree of interparticle agglomeration increased at 900 °C ( Supplementary Fig. 8c). The elemental mapping and energy-dispersive X-ray spectroscopy (EDX) analysis results (Supplementary Figs. 8d-g) revealed that the Ni and Zn species were well-distributed in the co-precipitated precursor and the Ni/Zn ratio was approximately 1:3. After the reduction of the calcined catalysts at 450 °C under H 2 ow, no noticeable changes were observed in the morphology of the catalyst ( Supplementary Fig. 8).The primary particle size of the catalysts increased from 25-30 to 100-400 nm as the calcination temperature increased from 500 °C (N 1 Z 3 -500) to 900 °C (N 1 Z 3 -900). For the N 1 Z 3 -900 catalyst, the primary particles were highly agglomerated and converted to secondary micron-sized particles ( Supplementary Fig. 9e). Furthermore, as the calcination temperature increased from 500 to 900 °C, the Brunauer-Emmet-Teller  Table 3).
The diffusion of Zn into the NiO phase during calcination and formation of the Ni-Zn alloy during the reduction were analyzed using high-resolution transmission electron microscopy (HR-TEM) and highangle angular dark eld-scanning transmission electron microscopy (HAADF-STEM), and the results are presented in Fig. 3. The formation of uniform nano-sized 10-30 nm particles was observed at the surface of the 100-400 nm spherical ZnO particles. Some nano-sized particles were in close contact with each other, which suggested that the nanoparticles migrated during calcination owing to their high surface energy. The Zn and O species were well-distributed throughout the micron-and nano-sized particles. In contrast, the Ni species were predominantly distributed throughout the nano-sized particles, and only a few Ni species were distributed throughout the Zn-rich micron-sized particles. This indicated that most micron-and nano-sized particles of the N 1 Z 3 -900 catalyst were low-level Ni-doped ZnO and Ni-Zn alloy, respectively. Two different lattice parameters of the nano-sized Ni-Zn alloy particles, 7.71 and 2.29 Å, were observed in the HR-TEM images of the N 1 Z 3 -900 catalyst. The presence of the enlarged interplane spacing of 7.71 Å, which was approximately three times larger than that of pure ZnO (2.5 Å 63 ), suggested the formation of a Ni-Zn alloy 64 . In addition, the interlayer spacing of 2.29 Å of the N 1 Z 3 -900 catalyst was larger than that of pure metallic Ni (2.03 Å), which agreed well with the XRD results. To further investigate the formation of the Ni-Zn alloy, the elemental compositions of the nano-sized particles of the N 1 Z 3 -900 catalyst were analyzed using EDX line scanning ( Supplementary Fig. 11). The elemental concentrations of Zn, Ni, and O measured over the lateral distance of 40 nm con rmed that the nanosized particles on the surface of ZnO were O-de cient Ni-Zn alloy species. Unlike the Ni and Zn species in N 1 Z 3 -900, those in N 1 Z 3 -700 were well distributed throughout the catalyst (Supplementary Figs. 12a-l).
The segregation of the nano-sized Ni-Zn alloy phases was rst observed for the N 1 Z 3 -800 catalyst (Supplementary Figs. 12g-i). Therefore, highly active nano-sized Ni-Zn alloy particles were formed via calcination at the high temperatures of 800 and 900 °C followed by reduction at 450 °C.
We used the aforementioned results to propose a schematic of the Ni-Zn alloy formation (Fig. 4). During the calcination of the co-precipitated product at temperatures of up to 230 °C, Ni(CO 3 ) 3 Fig. 4b). The predicted atomic arrangement on the Ni-Zn alloy phase indicated that Nicentered pentagonal and hexagonal structures could be present in the alloy phase, and majority of the Ni-Zn alloy species presented (111) and (200) facets (Fig. 4b).
To gain insight regarding the synthesis mechanisms of monocarboxylic acids, the chemical state of the N 1 Z 3 catalyst under CO 2 hydrogenation conditions was monitored using NAP-XPS, and the results are presented in Fig. 5 Fig. S13d) 70,71 . When the catalyst was exposed to a mixture of H 2 and CO 2 at 25 °C, an additional peak at 292.7 eV appeared, which could be attributed to the gaseous CO 2 in the in situ cell (Fig. 5d) 72 ; this peak almost disappeared after the cell was evacuated. In the presence of H 2 and CO 2 at 325 °C, several adsorption peaks of reaction intermediates, which were produced via CO 2 hydrogenation, were observed. The peaks at 291.4 and 288.2 eV corresponded to the adsorbed CO and formate species, respectively. 73 . In addition, a carboxylate peak was observed at 291.1 eV 74 .
To investigate the interactions between the surface Ni species and ZnO, the hydrogen temperatureprogrammed reduction and desorption (H 2 -TPR and H 2 -TPD, respectively) pro les of the N 1 Z 3 catalysts were analyzed. The N 1 Z 3 catalysts prepared at temperatures in the range of 500-800 °C exhibited three peaks at low, medium, and high temperatures (100-200, 200-500, and > 500 °C, respectively) in the H 2 -TPR pro le (Supplementary Fig. 14a). Typically, the H 2 -TPR pro le of bulk NiO particles, in the absence of supports, presents an asymmetric peak in the temperature range of 250-500 °C, which corresponds to the overlapped reduction of Ni 3+ to Ni 2+ and Ni 2+ to Ni 0 75-78 . When NiO interacts strongly with a support (e.g., γ-Al 2 O 3 or ZnO), the di culty in reducing the NiO phase generates broad high-temperature peaks above 500 °C 67,79 . The high-temperature reduction peaks could also be ascribed to the reduction of Zn 2+ to metallic Zn in the presence of Ni, although bulk ZnO is barely reducible 60,80 . For highly-dispersed NiO nanoparticles that interact weakly with their support, the reduction temperature range was lower (200-250 °C) than that of the bulk NiO particles 79,81 . Thus, it was reasonable to assign the low-, medium-, and high-temperature peaks of the N 1 Z 3 catalysts to the reduction of highly-dispersed NiO nanoparticles that interacted weakly with ZnO, bulk NiO phases, and NiO phases that interacted strongly with ZnO, respectively. The low-temperature peak at 230-290 °C, which were observed in the H 2 -TPR pro les of the  Fig. 14b). The N 1 Z 3 catalysts prepared at calcination temperatures above 700 °C exhibited similar H 2 desorption behavior.
To gain insight into the CO 2 adsorption behavior and active sites for the CO 2 conversion reaction, the suggested the formation of more easily desorbed species on the surface of catalysts prepared at higher calcination temperatures. Conversely, the high-temperature peaks in the temperature range of 360-385 °C, which were the major desorption peaks in the O 2 -TPD pro les of the N 1 Z 3 -500, N 1 Z 3 -600, and N 1 Z 3 -700 catalysts calcined, were less prominent in the O 2 -TPD pro le of the N 1 Z 3 -800 catalyst. A remarkable change in the desorption of O was observed in the O 2 -TPD pro le of the N 1 Z 3 -900 catalyst; the removal of from the metal oxide lattice was signi cantly increased from 0.033-0.059 mmol g -1 (for the N 1 Z 3 -600, N 1 Z 3 -700, and N 1 Z 3 -800 catalysts) to 0.135 mmol g -1 (for the N 1 Z 3 -900 catalyst), as listed in Supplementary The CO 2 temperature-programed desorption (CO 2 -TPD) pro les of the catalysts were used to examine the adsorption behavior of CO 2 on the catalyst surfaces, and the results are illustrated in Supplementary   Fig. 16 and Supplementary Table 6. The desorption peaks in the CO 2 -TPD pro les of the catalysts could be categorized into three regions: <250, 250-600, and >600 °C, which represented weak, medium, and strong basic sites. The main peak in the CO 2 -TPD pro les of the N 1 Z 3 -500, N 1 Z 3 -600, and N 1 Z 3 -700 catalysts occurred in the temperature range of 535-555 °C and did not change signi cantly as the calcination temperature of the catalysts increased, which indicated that strong interactions were maintained between the adsorbed CO 2 and catalyst surfaces. However, the main desorption temperature of the N 1 Z 3 -900 catalysts shifted to 432 °C, which suggested that the surface basicity of the N 1 Z 3 -900 catalyst was signi cantly lower than those of the N 1 Z 3 -500, N 1 Z 3 -600, and N 1 Z 3 -700 catalysts. The favorable CO 2 adsorption on the surface of the N 1 Z 3 -900 catalyst at low temperatures could be attributed to the increase in the number of sites at the calcination temperature of 900 °C. The amount of CO 2 adsorbed at the medium basic sites of the N 1 Z 3 -900 catalyst was 0.158 mmol g −1 , which was signi cantly higher than those of the N 1 Z 3 -800 catalyst (0.102 mmol g −1 ) and other lower-calcinationtemperature catalysts (0.018-0.037 mmol g −1 ). The high CO 2 hydrogenation activity of the N 1 Z 3 -900 catalyst suggested that the CO 2 adsorbed at the medium basic sites effectively participated to the reaction and the CO 2 adsorbed at the strong basic sites was less suitable for the subsequent conversion reaction. This hypothesis was supported by the strong interactions between the CO 2 molecules adsorbed at the medium basic sites and the catalyst surface at the reaction temperatures and facile desorption of the reaction products from the catalyst surface after the reaction.
To elucidate the reaction mechanism for the formation of monocarboxylic acids, in situ DRIFTS experiments were performed to analyze the evolution of the reaction intermediates over Ni-Zn alloy catalysts. To identify the chemical species derived from the adsorption of CO 2 on the surface of the N 1 Z 3 -900 catalyst, DRIFTS pro les were obtained by owing CO 2 at 3.0 MPa through the DRIFTS cell that contained an in situ reduced N 1 Z 3 -900 catalyst sample, as presented in Fig. 6a. The detailed peak assignment is listed in Supplementary Table 7. As the temperature of the cell increased from 30 to 325 °C, the intensities of the infrared (IR) bands of CH 4 at 3015 and 1305 cm − 1 increased; however, further extending the adsorption time to 60 min at 325 °C led to a decrease in the intensity of CH 4 bands.
The reaction between the residual pre-adsorbed H 2 on the catalyst surface after reduction and adsorbed peak was ascribed to the formyl species produced via the hydrogenation of CO 95 . In addition, the IR bands at 1080, 1050, 977, and 950 cm − 1 were associated with the bending of the = C-H groups and those at 1540 and 1340 cm − 1 were ascribed to the carboxylate species 92,96 . The intermediates and products formed via the adsorption of CO 2 on the reduced N 1 Z 3 -900 catalyst (Fig. 6a) were different from those formed on the reduced N 1 Z 3 -500 catalyst (Fig. 6b). First, the formation of a large amount of CH 4 was observed at 150 °C over the N 1 Z 3 -500 catalyst; however, the formation of CH 4 was suppressed on the N 1 Z 3 -900 catalyst, and these results were in good agreement with the CO 2 hydrogenation data (Fig. 1a).
This indicated that the metallic Ni particles on the N 1 Z 3 -500 catalyst surface facilitated methanation 97,98 , and the formation of the Ni-Zn alloy suppressed the methanation reaction. Second, the formation of CO 2 -adsorbed intermediates (i.e., carbonate, bicarbonate, and formate) on the surface of the N 1 Z 3 -500 catalyst was less favored than on the surface of the N 1 Z 3 -900 catalyst. This implied that the adsorption of CO 2 on the Zn-rich Ni x Zn y O phase was more favorable than that on the ZnO phase.
H 2 at 325 °C and 3.0 MPa was used as ow gas in the DRIFTS cell immediately after CO 2 ow, and the Fourier-transform infrared (FT-IR) spectra of the N 1 Z 3 -900 catalyst were collected to investigate its hydrogenation behavior (Fig. 7a). As the H 2 ow time increased to 15 min, the formation of CH 4 began to be favored, and the intensity of the IR band associated with the = C-H groups decreased signi cantly. In addition, the amounts of CO 2 -adsorbed intermediate species, RWGS-formed gaseous CO (2200-2100 cm − 1 ), and water (1653 cm − 1 ) 99 , which could be produced via the re-adsorption of OH from the decomposition of formate species 100, 101 , began to increase. This indicated that progressive CO 2 hydrogenation occurred after 15 min of H 2 ow, after changing the ow gas from CO 2 to H 2 . After 140 min of H 2 ow, the peaks of gaseous CO and l-CO almost disappeared from the FT-IR spectra of the N 1 Z 3 -900 catalyst, and the intensity of the CH 4 peak began to decrease. Moreover, residual CO 2intermediate species were present on the surface of the N 1 Z 3 -900 catalyst. The CO 2 hydrogenation behavior over the N 1 Z 3 -500 catalyst was similar (Fig. 7b).
A distinct difference in the catalytic behaviors of the N 1 Z 3 -900 and N 1 Z 3 -500 catalysts was observed using their operando DRIFTS pro les collected utilizing a mixture of H 2 and CO 2 with a H 2 /CO 2 ratio of 2:1 at 325 °C and 3.0 MPa as ow gas. After 15 min of H 2 and CO 2 ow, the formation of gaseous CO and CH 4 on the surface of N 1 Z 3 -500 was highly favored and the formation of CO and CH 4 on the surface of N 1 Z 3 -500 was suppressed ( Supplementary Fig. 17). As presented in Fig. 8, the time evolution of the formation of CO indicated that the amount of gaseous CO produced during the initial stage of reaction over the N 1 Z 3 -900 catalyst continuously decreased, which suggested that CO was actively transformed into next-stage hydrogenated products. After 90 min of reaction, the amount of gaseous CO remained almost unchanged, which indicated that the rates of the hydrogenation of CO and formation of CO reached steady-state. In contrast, the amount of gaseous CO reached a maximum after 30 min of reaction over the N 1 Z 3 -500 catalyst and progressively decreased as the reaction continued for 300 min.
This indicated that the N 1 Z 3 -500 catalyst could not convert CO to next-stage hydrogenated products during the initial reaction stage. The CO 2 -adsorbed reaction intermediates (i.e., carbonate, bicarbonate, and formate) and double-bonded hydrocarbons reached steady-state peak intensities after 15 min of reaction over both catalysts ( Supplementary Fig. 17).
To further elucidate the reaction behavior of the N 1 Z 3 -900 catalyst, CO adsorption on the in situ reduced catalyst followed by CO hydrogenation using a ow of H 2 were performed. As presented in Supplementary Fig. 18a To further elucidate the adsorption behavior of hydrocarbon species formed on the N 1 Z 3 catalysts, we used CH 4 temperature-programed desorption (CH 4 -TPD) experiments; the results are presented in Supplementary Fig. 19 and Supplementary Reaction mechanism. Based on the TPD pro les of the N 1 Z 3 catalysts, in situ NAP-XPS, and in situ DRIFTS, a plausible reaction pathway was proposed (Fig. 9a) 102 might not be responsible for the production of monocarboxylic acids.
To further examine the possibility of direct C-C coupling of CO 2 and the surface adsorbed * CH 3 species, we used DFT simulation (Fig. 9b). The detailed steps of the hydrogenation of CO 2 to AA are presented in Supplementary Fig. 20. Based on the experimental XRD data ( Supplementary Fig. 4a), the (111) plane of as bidentate con gurations with bond lengths of 1.910 and 2.038 Å for Ni-C and Zn-O, respectively. The * COOH species were subsequently dissociated into CO and hydroxyl groups, which were adsorbed on the Ni and Zn sites, respectively. Subsequently, CO was hydrogenated to form * HCO by combining H with C or O atoms, and then, it was further hydrogenated and deoxygenated to form * CH x species. During the hydrogenation of CO, C was shifted from the Ni center to a Zn site, and the subsequent hydrogenation of Zn-CH x produced Zn-CH 3 species. The bond length of Zn-C in the Zn-CH 3 species was 1.956 Å. This was slightly longer than that of −[ZnCH 3 ] + (1.93 Å) 103 , which suggested that the Zn-C bonds at the Ni 4 Zn 22 catalyst sites were slightly more activated. For the adjacent Zn-CH 3 species, CO 2 was adsorbed on the Zn sites and the adsorbed CO 2 was tilted toward the Zn-CH 3 bonds, and therefore, the Zn-CH 3 bond length increased to 2.015 Å. Subsequently, the activated CO 2 was inserted into the Zn-C bond of the Zn-CH 3 species to produce surface acetate species (Zn-OOCCH 3 ). Afterward, the surface Zn-OOCCH 3 species were hydrogenated to produce AA, which desorbed from the catalyst surface. The total Gibbs free energy for the formation of AA from CO 2 was − 1.03 eV, which indicated that the reaction was thermodynamically feasible on the surface of the Ni 4  Stability of the N 1 Z 3 -900 catalyst. The stability of the crystal structure of the N 1 Z 3 -900 catalyst during the CO 2 hydrogenation reaction was tested using in situ XRD, and the results are depicted in Supplementary Fig. 21. The phase structure of the N 1 Z 3 -900 catalyst did not change as the temperature increased up to 400 °C under a H 2 /CO 2 ow, which indicated that the Ni-Zn alloy and Zn-rich Ni x Zn y O phases were stable during the CO 2 hydrogenation reaction. This suggested that the water molecules formed via the RWGS reaction did not alter the phase structure of the catalyst. The XRD pattern of the spent N 1 Z 3 -900 catalyst, which was collected after a 216 h on-stream test, was almost identical to that of the fresh catalyst. This demonstrated the robust structure of the N 1 Z 3 -900 catalyst under the reaction conditions ( Supplementary Fig. 21b). In addition, the Ni-Zn alloy structure on the surface of the Zn-rich Ni x Zn y O phase in the spent catalyst was maintained after the long-term test ( Supplementary Fig. 21c).
The high stability of the N 1 Z 3 -900 catalyst under the reaction conditions caused the CO 2 conversion and product selectivity of the catalysts to remain unchanged (Fig. 1c).
In summary, we demonstrated that a Ni-Zn alloy/Zn-rich Ni x Zn y O catalyst could be used to produce AA and PA with high selectivity (58.9% and 18.2%, respectively) via direct CO 2 hydrogenation at a CO 2 conversion of 13.4% and by suppressing the selectivity toward CH 4 (17.0%) and C 2 -C 4 (5. The presence of CO 2 -adsorbed species (i.e., carbonate, bicarbonate, and formate) and gaseous CO indicated that the RWGS reaction converted CO 2 to CO. Subsequently, CO was hydrogenated to formyl species and then to surface-adsorbed ( * CH 3 ) n species. AA and PA were produced via the direct C-C coupling of CO 2 to the surfaceadsorbed ( * CH 3 ) n species. The CO 2 conversion and monocarboxylic acid selectivity were maintained for up to 216 h on-stream, and the morphology and phase structure of the Ni-Zn alloy/Zn-rich Ni x Zn y O catalyst were maintained during the long-term stability test, which demonstrated the robust nature of the Ni-Zn alloy catalyst. Considering that, to date, direct CO 2 conversion has mostly been used for the production of liquid fuels, methanol, and light ole ns, the possibility of producing value-added monocarboxylic acids from CO 2 demonstrated the novelty and usability of the Ni-Zn alloy catalyst.

Materials And Methods
Catalyst preparation. Ni-Zn alloy catalysts with different Ni/Zn ratios were synthesized via co- tube with an inner diameter of 1.5 mm, which was xed on both sides using quartz wool. Both ends of the quartz tube were connected with gas inlet and outlet lines, and air (99.999% purity, Shinyang Sanso Co., South Korea) was passed through the tube at a ow rate of 10 mL min − 1 . The quartz tube was subsequently mounted on the thermal stage of the XRD instrument and diffraction data were collected as the temperature was increased from 27 to 900 °C at a ramping rate of 2 °C min − 1 .The temperature was decreased to 30 °C, the gas was switched to H 2 (99.999% purity, Shinyang Sanso Co.) with a ow rate of 10 mL min − 1 , and then, diffraction data were collected as the temperature was increased from 27 to 450 °C at a ramping rate of 2 °C min − 1 . To study the phases that formed during the CO 2 hydrogenation, the reduced catalyst sample was exposed to a ow of 75% H 2 (99.999% purity, Shinyang Sanso Co.) and 25% CO 2 (99.999% purity, Shinyang Sanso Co.); ow rate of 10 mL min − 1 , as the temperature was increased from 27 to 400 °C at a ramping rate of 2 °C min − 1 .
The N 2 adsorption-desorption isotherms of the catalysts were obtained using a Belsorp-mini II (BEL Inc., Japan) apparatus at the liquid N 2 temperature of − 196 °C. The speci c surface areas of the catalysts were calculated using the multipoint BET method, and the average pore diameters (D p , nm) and pore size distributions were calculated using the Barrett-Joyner-Halenda method and isotherm data. The external surface area and volume of the micropores were determined using the t-plot method. The morphology of the catalysts was observed using an S-4100 (  Catalyst evaluation. The hydrogenation of CO 2 was conducted in a high-pressure continuous-ow xedbed reactor setup. The reactor consisted of gas feeding lines, mass ow controllers, a stainless-steel xed-bed catalytic reactor with an inner diameter of 10 mm, a back-pressure regulator, and a condenser. For a typical run, 500 mg of catalyst was mixed with 500 mg of silicon carbide (S/0365/60, Fisher Chemicals, USA) which was used as thermal diluent, and the mixture was charged in the reactor. The catalyst was placed in the tubular reactor between the quartz wool beds. Subsequently, the catalyst was reduced by owing pure H 2 in the reactor at a ow rate of 50 mL min − 1 for 10 h at 450 °C at a ramping rate of 10 °C min − 1 . Afterward, the catalysts were cooled to the experimentally desired temperatures of 280-350 °C, and the gas ow was switched to H 2 /CO 2 (molar ratios of 1:1, 2:1, and 3:1). During CO 2 hydrogenation, the product stream was passed through the condenser, and the temperature of the condenser was maintained at 10 °C to liquefy the liquid products; furthermore, the residual gaseous stream was directed toward an Clarus 580 GC-Model Arnel 1115PPC (PerkinElmer, USA) re nery gas analyzer-gas chromatograph (RGA-GC). The RGA-GC was equipped with a TCD for the analysis of CO 2 , CO, and H 2 , and a ame ionization detector for the quantitative measurement of the C 1 -C 6 gaseous hydrocarbons. The RGA-GC speci cations are presented in detail in the literature 109 . The liquid products collected in the condenser were diluted with high-performance liquid chromatography (HPLC)-grade water and were directly injected into an Alliance, model e2695 (Waters, USA) HPLC instrument equipped with a ultraviolet-visible (UV-Vis) detector and an Aminex HPX-87H (Bio-Rad, USA) ion exclusion column (300 mm × 7.8 mm). The column and detector temperatures were maintained at 60 and 50 °C, respectively. The mobile phase consisted of a 0.15 N aqueous H 2 SO 4 solution with a ow rate of 0.5 mL min − 1 . A standard calibration curve was used to quantify the products.
The product selectivity was based on the carbon mole percentages (C-mol%) of all tested catalysts. After the gas and liquid products were quanti ed, the carbon balance obtained during the catalytic performance test for 24 h on-stream was 92-98%. The CO 2 conversion, monocarboxylic acid selectivity, and CO selectivity were calculated as follows: and where CO 2 in , CO 2 out , and CO out are the mole fractions of CO 2 at the inlet, CO 2 at the outlet, and CO at the outlet, respectively.
DFT calculations. The crystal structure of Ni 4 Zn 22 was determined using the Re ex plus program of the Material Studio software. The powder diffraction pattern was indexed using the TREOR90 program inside the Re ex Powder Indexing module followed by the PowderSolve module to solve the structure using the simulated annealing method 110,111 . The nal structure was obtained using a powder solve module based on the Monte Carlo simulated annealing procedure 112 . The similarity between the experimental and calculated XRD patterns was con rmed using Rwp values. The optimized structure solution was used as an initial structural model for the Rietveld re nement to obtain the nal crystal structure solution and nal Rwp value, which was calculated as follows: 113 Then the nal result of the Ni 4 Zn 22 spectrum were mixed with the ZnO spectrum and re ned using the QPA module to nd the nal rwp value and the ratio between both the structures 114,115 . The reaction pathway of CO 2 to AA was determined using the CASTEP module of the Materials Studio 116,117 . Two layers of the (111) surface of Ni 4 Zn 22 were constructed and a 20 Å vacuum space was created to avoid self-interaction. The rst layers were constrained and the top surface was unconstrained to facilitate the conversion reaction ( Supplementary Fig. 22). The generalized-gradient approximation with the Perdew-Burke-Ernzerhof 118 type exchange-correlation functions were used to calculate the interactions between atoms. The energy tolerance, maximum force tolerance, and maximum displacement tolerance were 10 − 5 eV atom − 1 , 0.03 eV atom − 1 , and 1 × 10 − 3 Å, respectively. The core electrons were described using ultrasoft pseudopotentials in conjunction with a cut-off energy of 300 eV. The k-point sampling of the Brillouin zone was performed according to the Monkhorst-Pack scheme using a 2 × 2 × 1 k-points mesh. The reaction energy was calculated as follows: An adsorption locator module was used to nd the adsorption site of the * CO 2 and H * on the surface of Ni 4 Zn 22 (111) plane. The Universal force eld was employed to estimate the forces between the atoms within the lattice. The location was set to surface region de ned by atom set to specify the region around the substrate in which adsorbate con gurations was sampled. The most stable adsorption sites were then used to calculate the rest of the reaction pathway including the subsequent adsorption of * CO 2 and H * . MeOH denote formic acid, acetic acid, propionic acid, butyric and valeric acids, and methanol, respectively. Figure 2 (a) Normalized Ni K-edge X-ray absorption near edge structure (XANES) spectra and (b) Fouriertransforms (FTs) of the normalized Ni K-edge extended X-ray absorption ne structure (EXAFS) spectra (κ3χ(k)) of N1Z3-700 and N1Z3-900. (c) Normalized Zn K-edge XANES spectra and (d) FTs of the normalized Zn K-edge EXAFS spectra (κ3χ(k)) of N1Z3-700 and N1Z3-900. Here, N1Z3-700 and N1Z3-900 are catalysts with Ni/Zn ratios of 1:3 calcined at 700 and 900 °C, respectively.    In situ diffuse re ectance infrared Fourier-transform spectroscopy (DRIFTS) analysis of CO2 hydrogenation after the adsorption of CO2 on the (a) N1Z3-900 and (b) N1Z3-500 catalysts with Ni/Zn ratios of 1:3 calcined at 900 and 500 °C, respectively. Prior to hydrogenation, the catalysts were reduced at 450 °C for 6 h under a H2 ow of 50 mL min−1. After reduction, the DRIFTS cell was evacuated and purged with a N2 ow of 30 mL min−1 for 2 h to remove H2, was and naturally cooled to 30 °C.
Subsequently, the DRIFTS cell was pressurized with CO2 at 3.0 MPa and 50 mL min−1 as the temperature was increased from 30 to 325 °C. The CO2 ow of 50 mL min−1 was maintained at 325 °C for 2 h, then the ow gas was switched from CO2 to H2 at a ow rate of 50 mL min-1 at 325 °C and 3.0 MPa. Here, l-CO and b-CO denote linear and bridged CO, respectively, and g-CO denotes gaseous CO. Time-dependent diffuse re ectance infrared Fourier-transform spectroscopy (DRIFTS) pro le of the N1Z3-900 catalyst with a Ni/Zn ratio of 1:3 calcined at 900 °C. Prior to hydrogenation, the catalyst was reduced at 450 °C for 6 h under a H2 ow of 50 mL min−1. After the reduction, the DRIFTS cell was evacuated and purged with N2 for 2 h to remove H2 at 325 °C. Subsequently, the ow gas was switched from N2 to a H2/CO2 mixture with a ratio of 2:1, the pressure of the DRIFTS cell was increased to 3.0 MPa at 325 °C, and the spectra were obtained.