Portal de Periódicos da CAPES

Tuning Selectivity in Catalytic Conversion of CO 2 by One-Atom-Switching of Au 9 and Au 8 Pd 1 Catalysts

2021; Chinese Chemical Society; Volume: 3; Issue: 12 Linguagem: Inglês

10.31635/ccschem.021.202000730

2096-5745

Xiao Cai, Xin Sui, Jiayu Xu, Ancheng Tang, Xu Liu, Mingyang Chen, Yan Zhu,

Carbon dioxide utilization in catalysis

Open AccessCCS ChemistryRESEARCH ARTICLE1 Dec 2021Tuning Selectivity in Catalytic Conversion of CO2 by One-Atom-Switching of Au9 and Au8Pd1 Catalysts Xiao Cai†, Xin Sui†, Jiayu Xu, Ancheng Tang, Xu Liu, Mingyang Chen and Yan Zhu Xiao Cai† School of Chemistry and Chemical Engineering, Nanjing University, Nanjing 210093 , Xin Sui† Center for Green Innovation, School of Materials Science and Engineering, University of Science and Technology Beijing, Beijing 100083 , Jiayu Xu School of Chemistry and Chemical Engineering, Nanjing University, Nanjing 210093 , Ancheng Tang School of Chemistry and Chemical Engineering, Nanjing University, Nanjing 210093 , Xu Liu School of Chemistry and Chemical Engineering, Nanjing University, Nanjing 210093 , Mingyang Chen *Corresponding authors: E-mail Address: [email protected] E-mail Address: [email protected] Center for Green Innovation, School of Materials Science and Engineering, University of Science and Technology Beijing, Beijing 100083 Shunde Graduate School, University of Science and Technology Beijing, Foshan 528000 and Yan Zhu *Corresponding authors: E-mail Address: [email protected] E-mail Address: [email protected] School of Chemistry and Chemical Engineering, Nanjing University, Nanjing 210093 https://doi.org/10.31635/ccschem.021.202000730 SectionsSupplemental MaterialAboutAbstractPDF ToolsAdd to favoritesDownload CitationsTrack Citations ShareFacebookTwitterLinked InEmail The optimization of catalysts for CO2 hydrogenation that is carried out in a traditional fixed-bed reactor predominantly focuses on pursuing various nanoparticles at the nanoscale. Much less is known about how heterogeneous catalysts can be exploited to precisely control the reaction pathways of CO2 conversion at the atomic level. Herein, we design atomically precise Au9 and Au8Pd1 clusters intercalated into montmorillonite as heterogeneous catalysts to catalyze CO2 hydrogenation performed in a fixed-bed reactor. The substitution of the central Au atom of Au9 by a Pd atom that forms Au8Pd1 enables the dynamic tuning of CO2-reduction selectivity from C1 to C2 hydrocarbons. The central Pd atom substitution also efficiently makes the cluster less fluxional to boost the catalytic activity of Au8Pd1. This work constitutes a general step toward rational design of atomic-scale catalysts for industrially important chemical reactions. Download figure Download PowerPoint Introduction Recent years have witnessed tremendous research efforts on the capture and utilization of CO2, since CO2 represents a strategic carbon resource for the generation of high value-added fuels and chemicals.1–7 Significant advances have been achieved in thermocatalytic conversion of CO2, electrochemical reduction of CO2, photosynthetic CO2 fixation, and so forth.8–18 Among these reactions, CO2 storage as hydrocarbons (e.g., alkane and olefin) or oxygenates (e.g., methanol and ethanol) via reduction with renewable hydrogen is a practical strategy for potential energy supply. However, the ability to transform CO2 into desired products in a controllable fashion via heterogeneous catalysis is quite lacking. It is an ongoing challenge to precisely cleave C–O bonds, forge C–H bonds, and couple C–C bonds in the catalytic reaction of CO2 hydrogenation driven by conventional metal-based and metal oxide-based nanoparticle catalysts.19 Following major synthetic advances on atomically precise metal nanoclusters, a new class of metal catalysts has been introduced in catalysis research.20–27 Metal nanoclusters provide unprecedented opportunities to deeply understand the structure–property relationship.28–32 They have been exploited as ideal model catalysts to address fundamental issues related to catalysis the science of catalysis, especially with regard to nanoparticle catalysts.33,34 Despite the advantages, nanoclusters currently cannot rival traditional nanoparticles in large-scale applications, which precludes their extensive development in industry. We thus questioned whether nanoclusters might be capable of controlling reaction pathways of CO2 hydrogenation performed in a traditional fixed-bed reactor toward specific products. If successful, metal nanoclusters might not only make a breakthrough in tunability of CO2 transformation, but also hold promise for real-world applications. Herein, we develop a general approach to prepare gold-based nanoclusters confined in the layered montmorillonite (MMT) able to catalyze CO2 reaction with H2 in a fixed-bed reactor. Using this rational design, we demonstrate that the transformation of CO2 from one product to another is possible by readily altering the composition of cluster catalysts. The product change from methane to ethane can be achieved by the substitution of one Au atom of Au9 by a single Pd atom. Our experimental and theoretical studies bring insights into the reaction pathways of CO2 reduction toward distinct hydrocarbons occurring on the two catalysts. Experimental Methods and Computational Methods Catalysts synthesis Synthesis of [Au9(PPh3)8]3+ About 500 mg Au(PPh3)NO3 was added in ethanol (20 mL) in a 50 mL flask. The solution obtained by dissolving 9.1 mg NaBH4 in 10 mL ethanol was added dropwise to the above flask. After 2 h reaction at room temperature, the reaction mixture was centrifuged to remove insoluble materials and then evaporated in vacuo. The resulting solid was washed with hexane and then extracted with dichloromethane.35 Synthesis of [Au8Pd1(PPh3)8] 2+ About 142 mg Au(PPh3)Cl and 133 mg Pd(PPh3)4 were added in 12 mL ethanol in a 50 mL flask. The solution obtained by 15.2 mg NaBH4 dissolved in 5 mL ethanol was dropped into the above flask. After stirring for 30 min at room temperature, the reaction mixture was centrifuged to remove insoluble materials, and then evaporated in vacuo. The resulting solid was washed with hexane and then extracted with dichloromethane.35 Synthesis of Au9@MMT and Au8Pd1@MMT About 500 mg MMT was dispersed in 10 mL pure water and then formed a suspension by ultrasonic vibration for 15 min. About 10 mg Au8Pd1 or 10 mg Au9 clusters dissolved in 10 mL ethanol were mixed with the MMT suspension. After 6 h, the above mixture was filtered and washed with dichloromethane and water. The obtained solid was dried in a vacuum oven at 50 °C overnight. Synthesis of Au9/support and Au8Pd1/support About 10 mg Au8Pd1 or 10 mg Au9 clusters dissolved in 0.2 mL dichloromethane were mixed with 500 mg supports (such as oxides or MMT) in a mortar. After being ground for 15 min, the catalysts were dried in a vacuum oven at 50 °C overnight. Catalytic reaction The catalytic performances of the catalysts for CO2 hydrogenation were measured in a fixed bed with a U-shaped stainless steel reaction tube (an inner diameter of 3 mm). All the catalysts for evaluating were pelletized, crushed, and sieved to 20–40 mesh. About 200 mg catalyst was packed between two layers of quartz wool. First, the hydrogen gas was led over the catalyst at a flow rate of 10 mL/min, and then the fed temperature was increased to 150 °C for 90 min. Then, the hydrogen gas was changed to argon gas to purge the remaining hydrogen. After that, a mixture gas (CO2:H2 = 1:3, mole ratio) with a flow rate of 20 mL/min was led to the reactor. After the reactor pressure was increased to the desired level, the fed temperature was raised to the desired degree. The reaction products were analyzed online by GC 9680 (Shanghai Qiyang Information Technology Co., Ltd., Shanghai, China) equipped with a thermal conductivity detector (TCD) connected to a TDX-1 (Nanjing Hope Analytical Equipment Co., Ltd., Nanjing, Jiangsu, China) packed column as well as a flame-ionization detector (FID) connected to a PLOT-Q (Nanjing Jianuo Chromatography Technology Co., Ltd., Nanjing, Jiangsu, China) capillary column with argon as the carrier gas. The carbon balance results in this work are 100 ± 20%. Characterization High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images were acquired using FEI Talos F200X field emission transmission electron microscope (Thermo Fisher Scientific, Hillsboro, OR) equipped with energy dispersive X-ray spectroscopy. The accelerating voltage was at 200 kV, and the gun voltage was set to 3800 V. TEM images were collected by JEM-200 CX (JEOL, Tokyo, Japan) with an acceleration voltage of 100 kV. The single-crystal X-ray diffraction (XRD) data of the clusters were collected on a Bruker D8 VENTURE (Billerica, MA). Absorption spectra were recorded on a UV–vis spectrometer using SHIMADZU UV-1800 (Kyoto, Japan). The small-angle XRD (SAXRD) data of the samples were collected by a Panalytical X'Pert (Almelo, The Netherlands). Thermogravimetric (TG) spectra were recorded using a NETZSCH STA 449C (Bavaria, Germany). Time-resolved in situ Fourier transform infrared (FT-IR) experiments were performed on a Bruker Tensor 27 (Billerica, MA) equipped with a high-pressure cell. After the catalyst was heated at 150 °C for 90 min in a stream of 5% H2/95% N2, the remaining gas was purged with argon. The temperature was raised to 200 °C in an argon atmosphere, and the signal was collected as a background. After that, a mixed gas (CO2:H2 = 1:3; 0.3 MPa) was led to the high-pressure cell, and the signals were collected every 10 min. Computational methods The cluster structures for Au8Pd1 and Au9 were taken from the single-crystal structures.35 All of the calculations, including geometry optimization, vibrational frequency analysis, and transition-state search, were performed at the density functional theory (DFT) level with the Perdew–Burke–Ernzerhof (PBE) functional and with the Los Alamos National Laboratory 2 double zeta (LANL2DZ) basis and pseudopotential.36,37 The Gaussian 09 software package was used to carry out the DFT calculations.38 For the depassivation of the Au8Pd1 cluster, the single-ligand dissociation reactions were considered: [ 8 Au 1 Pd n ( 3 PP h ) ] 2 + → [ 8 Au 1 Pd ( 3 PP h ) n − 1 ] 2 + 3 + PP h , 1 ≤ n ≤ 8 To thoroughly predict the reaction pathways for the CO2 hydrogenation reaction over the Au8Pd1 and Au9 catalysts, the simplified partially depassivated [Au8Pd1(PH3)5]2+ and [Au9(PH3)5]3+ models were chosen to efficiently predict the geometry and energetics for a large number of intermediate and transition-state species. The depassivation level of the computational model was determined based on the ligand dissociation energy profile for Au8Pd1. Gibbs free energies for the reactions were obtained by correcting the electronic energies (from geometry optimization calculations) with zero-point energy, thermal, and entropy corrections (from vibrational frequency analyses). Results and Discussion The solvated [Au9(PPh3)8]3+ (abbreviated as Au9) cluster comprises an Au9 metal core, as shown in Figure 1a, which can be viewed as two Au4 units joined together via a central gold atom. The eight gold atoms in the two Au4 units are coordinated to eight PPh3 ligands with Au–P bond length ∼2.30 Å, and the gold atom located on the center is uncoordinated by ligands.35 Interestingly, the central gold atom of Au9 is substituted with a single Pd atom, and hence, the Au9 cluster is transferred to Au8Pd1 ([Au8Pd1(PPh3)8]2+; Figure 1b), in which the central Pd atom is shared by two Au4 units (Au–Pd bond length ∼2.64 Å) capped by eight PPh3 with Au–P bond length ∼2.29 Å.35 The difference between the central atoms of the two clusters is facilely identified by their spectral fingerprints ( Supporting Information Figure S1). Figure 1 | Crystal structures of (a) [Au9(PPh3)8]3+ and (b) [Au8Pd1(PPh3)8]2+. Color code: yellow = Au; blue = Pd; pink = P; gray = C. The H atoms are omitted for clarity. Download figure Download PowerPoint To evaluate the thermal-driven catalysis of the Au9 and Au8Pd1 catalysts for CO2 hydrogenation performed in the fixed-bed reactor, we initially prepared heterogeneous catalysts via loading Au9 and Au8Pd1 on oxides. From Supporting Information Figure S2, the supported Au9 catalysts showed no activity and the supported Au8Pd1 catalysts exhibited a low activity with CO and CH4 as major products. These disappointing results were mainly attributed to the drastic sintering of clusters on the oxide supports into larger nanoparticles ( Supporting Information Figure S3). We next redirected our attention to the survival of atomically precise clusters under severe conditions, the key issue to be addressed in the heterogeneous catalysis applications of clusters. Based on the charge states of clusters, we thus designed an ion-exchange strategy to intercalate the clusters into the layered MMT, which is selected as a matrix to stiffen the clusters, because it is known that its framework is layered by stacking silicate flakes through electrostatic action, and there are many counter cations between the layers.39 The ion exchange was carried out using positively charged [Au8Pd1(PPh3)8]2+ and [Au9(PPh3)8]3+ clusters with metal cations in MMT (Figure 2a), thereby obtaining confined catalysts (e.g., Au9@MMT and Au8Pd1@MMT). Figure 2b delineates the results of cation exchange between layers by SAXRD. The peak at 5.8° is assigned as (001) crystal phase of MMT. After ion exchange, the peak at (001) shifts to a smaller angle (3.9°), and the peak width is broadened. This indicates that the interlayer spacing of MMT is widened and its order degree of interlayer is decreased after cluster insertion. According to Bragg's law, the interlayer distance of MMT is estimated to be about 1.5 nm, and the interlayer distances of both Au8Pd1@MMT and Au9@MMT are about 2.2 nm. Au and Pd crystal phases were not detected by the powder XRD diffraction ( Supporting Information Figure S4). HAADF-STEM images of Au8Pd1@MMT and Au9@MMT illustrate that the clusters are uniformly distributed in the MMT, and the size of the clusters is about 1 nm (Figures 2c and 2d). Figure 2 | (a) Design of Au8Pd1 and Au9 intercalated into MMT. (b) SAXRD patterns of Au9@MMT and Au8Pd1@MMT. HAADF-STEM images of (c) Au9@MMT and (d) Au8Pd1@MMT. Download figure Download PowerPoint We next sought to explore the catalytic performance of the Au9@MMT and Au8Pd1@MMT as heterogeneous catalysts for CO2 hydrogenation. As displayed in Figure 3, we discovered that only one-atom-switching of Au9 and Au8Pd1 can rouse significant differences in catalytic activity and selectivity. In Figure 3, the space-time conversion of CO2 on the Au8Pd1@MMT catalyst was higher than that of the Au9@MMT catalyst under identical reaction conditions. When the pressure of the reaction gas (CO2:H2 = 1:3) was <2 MPa or the reaction temperature was <200 °C, Au9@MMT had little activity, while Au8Pd1@MMT showed convincing activity (Figures 3a–3d). Perhaps most importantly, the major product transformation occurred from methane on the Au9@MMT to ethane on the Au8Pd1@MMT. This transformation was hardly subject to the reaction conditions, as shown in Figures 3a and 3c, except that a small amount of ethylene was produced on the Au8Pd1@MMT at low reaction pressures (1–1.5 MPa) or low reaction temperatures (150–175 °C; note that the reaction temperature was not <150 °C considering that MMT could not be stable below that temperature, as shown in Supporting Information Figure S5). Meanwhile the Au9@MMT always gave methane product with a much lower activity (Figures 3b and 3d). Figures 3e and 3f showed that Au8Pd1@MMT exhibited higher durability than Au9@MMT in the reduction processes of CO2, even though the catalytic stability of Au8Pd1@MMT was not excellent, and the selectivity for C1 or C2 hydrocarbons over the two catalysts was almost constant. The spent Au8Pd1@MMT catalyst showed no significant aggregation after the catalytic reaction ( Supporting Information Figure S6), though the cluster aggregation was to some extent observed, compared with the Au8Pd1/MMT sample prepared via loading Au8Pd1 on MMT. The Au8Pd1/MMT catalyst had very low activity under identical reaction conditions ( Supporting Information Figure S7) due to the drastic sintering of the clusters during the reaction ( Supporting Information Figure S6). Figure 3 | Effect of reaction temperature on CO2 hydrogenation over (a) Au8Pd1@MMT and (b) Au9@MMT under 2 MPa reaction gas (CO2:H2 = 1:3) with a SV of 6000 mL/g/h. Effect of reaction pressure on CO2 hydrogenation over (c) Au8Pd1@MMT and (d) Au9@MMT at 200 °C with a SV of 6000 mL/g/h. Catalytic stability of (e) Au8Pd1@MMT and (f) Au9@MMT for CO2 hydrogenation at 200 °C and under 2 MPa reaction gas with a SV of 6000 mL/g/h. n.d. denoted as not detected, five-point stars denoted as CO2 conversion, columns denoted as product selectivity. Download figure Download PowerPoint To obtain insight into the potential pathways of CO2 hydrogenation on the two catalysts, time-resolved in situ FT-IR spectroscopy experiments were performed and the intermediate species produced by the interaction between the catalysts, and the reaction gas (CO2:H2 = 1:3) were monitored. For FT-IR spectra obtained from Au8Pd1@MMT (Figure 4a) system, the bands at 2931, 2861, and 1457 cm−1 were assigned to *CH2 species corresponding to asymmetrical stretching, symmetrical stretching, and twisting vibrations of C–H,40,41 respectively. The C = O in *CHO as well as C = O and COO signals in HCOO* were respectively observed at 1744, 1685, and 1545 cm−1.42–44 Asymmetrical stretching vibrations of COO in *COOH appeared at 1623 cm−1.42 The bands at 2067, 1700, and 1511 cm−1 were attributed to adsorbed carbon monoxide (*CO), carboxylate (CO2−), and monodentate carbonate (m-CO32−),42,45,46 respectively. From the FT-IR spectra for the Au9@MMT system (Figure 4b), besides the *CO and HCOO* signals, the asymmetrical (2954 cm−1) and symmetrical stretching (2898 cm−1) vibrations of *CH3, and gaseous CH4 signal (3014 cm−1) were also observed.44 In comparison, the FT-IR spectra of Au8Pd1/MMT and Au9/MMT systems suggested that Au8Pd1/MMT had no selectivity for ethane and Au9 had no activity ( Supporting Information Figure S8), which was consistent with the catalytic results of the two catalysts ( Supporting Information Figure S7). Combining the above observations, it was suggested that the essential intermediates adsorbed onto the two catalysts distinctly signified the CO2 reduction event to direct reaction route and product selectivity. It is worth pointing out that CO2 was not directly adsorbed onto the Au sites of the two catalysts, but it can be bound to the catalysts with assistance of hydrogen activated on the Au sites ( Supporting Information Figure S9). Figure 4 | Time-resolved in situ FT-IR spectra and corresponding contour maps of reaction intermediate obtained on (a) Au8Pd1@MMT and (b) Au9@MMT catalysts with the reaction gas (CO2:H2 = 1:3). Download figure Download PowerPoint To reveal the molecular-level mechanisms of CO2 hydrogenation routes toward C1 and C2 hydrocarbons on the Au8Pd1 and Au9 catalysts, DFT calculations were performed. First, the ligand dissociation reactions of the Au8Pd1 are investigated using the full structural model [Au8Pd1(PPh3)8]2+ to predict the in situ depassivation states of the catalyst ( Supporting Information Figure S10). The results show that all of the PPh3 removal reactions are endergonic for Au8Pd1 at 298 K. The endergonicity for removing a PPh3 essentially increases as the cluster becomes more depassivated. The endergonicities for the removal of the first three PPh3 ligands of [Au8Pd1(PPh3)8]2+ are <20 kcal/mol, whereas the succeeding removal reactions are found to be notably more endergonic. We opine that with proper thermal pretreatment, the removal of PPh3 is controllable, in order that the cluster does not degrade and contains sufficiently exposed metal sites for possible multisite reactions. Based on these conclusions, simplified [Au8Pd1(PH3)5]2+ and [Au9(PH3)5]3+ models were chosen for the prediction of the CO2 hydrogenation reaction pathways over the pretreated Au8Pd1 and Au9 catalysts. In [Au9(PH3)5]3+, the Au9 core retains the geometry of an inflated hexagon as in [Au9(PH3)8]3+, whereas in [Au8Pd1(PH3)5]2+, the Au8Pd1 core turns into an inflated hexagon (like Au9 of [Au9(PH3)5]3+) from square antiprism in the original [Au8Pd1(PH3)8]2+. The nature of transition from square antiprism to inflated hexagon is structural flattening, which only requires two pairs of Au–Au antiprism side edges rotating to align with the antiprism equator and thus can be considered as a minor structural change driven by thermodynamics. The CO2 hydrogenation pathways involving formate (HCOO*), carboxyl (*COOH), dihydroxymethylidene [*C(OH)2], and formyl (*CHO) are thoroughly considered computationally for Au8Pd1 and Au9. The predicted thermodynamics for the pathways suggest that the reactions over the edge Au sites are more favorable than the reactions over the central Pd or Au site, as the former sites have lower coordination numbers than the latter sites ( Supporting Information Figure S11). The maximal endergonicity at 298 K for the carboxyl–formyl pathway that leads to the formation of methylene (*CH2) and methyl (*CH3) is 16.1 and 15.3 kcal/mol for reactions over the edge Au sites of Au8Pd1 and Au9, respectively, whereas the *C(OH)2 and HCOO* pathways have maximal endergonicity over 40 kcal/mol. Therefore, for both Au8Pd1 and Au9, it is most likely that the carboxyl–formyl pathway (Figures 5a and 5b, pathway 1) is most favorable. Figure 5 | Gibbs free energy profile for the predicted CO2 hydrogenation pathways at the most favorable sites of the pretreated (a) Au8Pd1 catalyst using the [Au8Pd1(PH3)5]2+ model and (b) Au9 catalyst using the [Au9(PH3)5]3+ model at the PBE/LANL2DZ level. The values are Gibbs free energies at 298 K. The molecular structures of the reaction intermediates and transition states can be found in Supporting Information Figures S11 and S12. Color code: orange = gold, blue = Pd, pink = P, black = C, red = O, and green = H. Download figure Download PowerPoint To further understand the CO2 conversion toward C2 hydrocarbons on Au8Pd1, the transition states along the investigated reaction pathways for the Au8Pd1 catalyst were sought (Figure 5a and Supporting Information Figure S12). Despite the formation of HCOO* (with activation energies ΔG‡ = 27.1 kcal/mol 298 K, i.e., the Gibbs free energy difference between the transition state and the reactant state) is slightly more favorable than the formation of *COOH (ΔG‡ = 36.3 kcal/mol), the further hydrogenation of *COOH [ΔG‡ = 13.2 and 10.6 kcal/mol toward *CO and *C(OH)2 respectively] is much more facile than the further hydrogenation of HCOO* (ΔG‡ = 62.3 kcal/mol). This implies that *COOH is likely transient, consumed by further hydrogenation steps, whereas HCOO* is a more persistent species. The predicted endergonicity of 3.6 kcal/mol for forming HCOO* implies a small portion of CO2 will be converted into the form of HCOO* and remain that way. This is in good agreement with the in situ infrared spectra of Au8Pd1@MMT where a significantly higher amount of HCOO* was detected than *COOH (Figure 4). Between the two pathways succeeding the formation of *COOH, the carboxyl–formyl pathway (via *CO and *CHO) is found to be much more favorable than the carboxyl–hydroxyl pathway (via *C(OH)2 and *COH), because the latter pathway requires high activation energy of 39.2 kcal/mol and overall endergonicity of 50.7 kcal/mol for one of the steps [*C(OH)2 → *COH]. The DFT results predict that the carboxyl–formyl pathway is the most favorable reaction pathway (Figure 5a). The CO2 → *COOH and *CO → *CHO steps have the highest activation energies along the carboxyl–formyl pathway of 36.3 and 38.5 kcal/mol, respectively, and are projected to be the rate-limiting steps. This is consistent with the measured time evolution of the infrared spectra for the Au8Pd1@MMT catalytic reaction where the *CO appears to be a persistent species with the strongest presence (Figure 4). Along the carboxyl–formyl pathway, CH2O* can potentially lead to CH3O* by adding a H on C or to *CH2OH by adding a H on O. The formation of *CH2OH (ΔG‡ = 23.4 kcal/mol) is much more favorable than the formation of CH3O* (ΔG‡ = 50.9 kcal/mol). This rules out further reactions of the CH3O• radical that lead to CH3OH, and so forth. The hydrogenation and dehydration of *CH2OH yield *CH2, and the further hydrogenation of *CH2 leads to *CH3, both with low activation barriers (ΔG‡ = 7.7 kcal/mol toward *CH2 and ΔG‡ = 11.4 kcal/mol toward *CH3). This suggests that alkenes, such as C2H4, and alkanes, such as CH4 and C2H6, are possibly the end products for CO2 hydrogenation over Au8Pd1@MMT. The low activation barrier for the *CH2 → *CH3 conversion suggests that the alkene products might be minor compared with the alkane products. The reactions toward C2H4, C2H6, and CH4 following the formation of *CH2 and *CH3 over the Au8Pd1 catalysts are further investigated at the DFT level (Figures 6a and 6b). The adsorbent Au atom of the Au-CH3 active site can further adsorb a H atom to form *[CH3]H with a stepwise endergonicity of 6.8 kcal/mol, leading to the barrierless hydrogenation to form CH4, or host the formation of another CH3 to yield *[CH3]2 with an exergonicity of 7.9 kcal/mol, leading to the recombination of two CH3 to form C2H6 with a low activation barrier of 15.0 kcal/mol (Figure 6b). Therefore, the formation of CH4 from *[CH3]H has a small barrier but a low reactant concentration, whereas the formation of C2H6 from *[CH3]2 has a relatively large barrier but a high reactant concentration. The rates for CH4 and C2H6 formation reactions as functions of the temperature are compared with an approximate and qualitative emulation, by assuming an established equilibrium between *CH3, *[CH3]H, and *[CH3]2, and assuming the reaction rate has a first-order dependency on *[CH3]H or *[CH3]2 concentration. The results ( Supporting Information Figure S13) show that the C2H6 is likely the dominant product at low temperatures, and increasing the temperature can increase the production of CH4 as a minor product. Figure 6a shows that the *CH2 could react with another CH2 adsorbate to form *[CH2]2, or react with a CH3 adsorbate to form *[CH3][CH2]. *[CH2]2 yields C2H4 with ΔG‡ = 49.8 kcal/mol at 298 K, which is not easy to achieve. This is because CH2 preferentially forms strong adsorption at the AuAu bridge site, which hinders the recombination both structurally and thermally. Experimentally, the hydrogenation of CO2 over Au8Pd1@MMT yields a small portion of C2H4 at 150 and 175 °C, and does not yield C2H4 at higher temperatures. Our DFT results show the difficulty in producing C2H4 from CO2 hydrogenation with the [Au8Pd1(PH3)5]2+ model; the reason for the low yield of C2H4 at relatively low temperature is still unclear. *[CH3][CH2] can facilely recombine to form *C2H5 with ΔG‡ = 12.8 kcal/mol at 298 K, which undergoes a barrierless hydrogenation reaction to yield C2H6. The pathway to C2H6 via *[CH3][CH2] is likely less favorable than the pathway to C2H6 via *[CH3]2 under most conditions, as the conversion from *CH2 to *CH3 is quite facile. Figure 6 | (a) The predicted reaction pathways for the hydrogenation reactions following the formation of CH2* over the pretreated Au8Pd catalyst using the [Au8Pd1(PH3)5]2+ model. The predicted reaction pathways for the hydrogenation reactions following the formation of CH3* over the (b) pretreated Au8Pd1 catalyst using the [Au8Pd1(PH3)5]2+ model and (c) pretreated Au9 catalyst using the [Au9(PH3)5]3+ model. Color code: orange = gold, blue = Pd, pink = P, black = C, red = O, and green = H. Download figure Download PowerPoint In our computations, we find that the Au9 cluster undergoes drastic structural rearrangement when the *CH2OH and successive species along the carboxyl–formyl pathway are formed (Figure 5b, pathway 1). During the catalytic structural rearrangement of Au9, the original hexagonal-corner Au active site of the original Au9 transitions into a triangular-corner Au site as a consequence of Au migration. Such rearrangement is consistent with structural motifs and trends of previously reported theoretically predicted global minima for the unprotected pure and Pd-doped Aun clusters: the triangular corner shapes are more common in pure Aun than in Pd-doped Aun−1Pd clusters, n < 1047–49. For example, Au7 has the shape of a planar edge-capped equilateral triangle, whereas Au6Pd has the shape of a planar regular hexagon. The geometries of the metal cores of PPh3-protected Au8Pd1 and Au9 are structurally closer to the hexagon-motif pure clusters (inflated hexagon or square antiprism). For the Au9 catalyst, the triangular geometry, which is akin to that of the unprotected cluster, gains favorability as the catalyst is depa