Silica accelerates the selective hydrogenation of CO2 to methanol on cobalt catalysts

The reaction pathways on supported catalysts can be tuned by optimizing the catalyst structures, which helps the development of efficient catalysts. Such design is particularly desired for CO2 hydrogenation, which is characterized by complex pathways and multiple products. Here, we report an investigation of supported cobalt, which is known for its hydrocarbon production and ability to turn into a selective catalyst for methanol synthesis in CO2 hydrogenation which exhibits good activity and stability. The crucial technique is to use the silica, acting as a support and ligand, to modify the cobalt species via Co‒O‒SiOn linkages, which favor the reactivity of spectroscopically identified *CH3O intermediates, that more readily undergo hydrogenation to methanol than the C‒O dissociation associated with hydrocarbon formation. Cobalt catalysts in this class offer appealing opportunities for optimizing selectivity in CO2 hydrogenation and producing high-grade methanol. By identifying this function of silica, we provide support for rationally controlling these reaction pathways.

Note: Methane is usually formed on the cobalt catalysts in CO2 hydrogenation, but the methane and CO selectivities depend on the multiple factors (e.g. CO2 conversion, reaction temperature, pressure, catalyst amount, and the support) 1-10 . In our experiments, we found that the Co/SiO2 and CoOx samples with metallic cobalt species exhibited higher methane selectivity than the Co@Six catalyst in CO2 hydrogenation. In these cases, the methane selectivity is still slightly lower than CO (Figure 2a), which should be due to the employed reaction conditions, and similar phenomena have been observed previously 1,4,7,10 .
We explored the methane selectivity as a function of reaction temperature in the Co/SiO2 and CoOx catalyzed CO2 hydrogenation. As shown in Supplementary Figures 2 and 10, the methane selectivity remarkably increased with the reaction temperatures. For example, the Co/SiO2 gave CH4 selectivity at 24.7% at 300 °C. By increasing the reaction temperature to 380 °C, the CH4 selectivity raised to 54.9%. For the CoOx catalyst, the CH4 selectivity was 35.4% at 300 °C and then increased to 70.2% at 380 °C. Further increasing the reaction temperature to 500 °C gave lower CH4 selectivity at 52.2% on the Co/SiO2 catalyst (Supplementary Figure 8a). This might be attributed to the strong exothermic nature of the CO2 methanation, leading to thermodynamically favorable CO formation via reverse water-gas shift reaction and decrease of CH4 selectivity 11 . These data suggest that the reaction temperatures have a significant effect on the CH4 and CO selectivity 4,5,10 . In addition to the reaction temperature, we also explored the influences of reaction pressure, feed gas composition, and catalyst amount to the methane selectivity in CO2 hydrogenation ( Supplementary Figures 8b-d). The higher reaction pressure, more hydrogen feed, and larger catalyst amount (low GHSV) are regarded to be the factors for enhancing the methane selectivity.

Note:
The lower C2+ selectivity than the general Fischer-Tropsch process should be due to the abundant CO2 and scarce CO in the mixed gas. Note: The Co3O4 without silica species showed the reduction peaks at 387 and 494 ºC, which are assigned to the reduction of Co 3+ to Co 2+ and Co 2+ to Co 0 . When the cobalt oxide was modified by silica, these peaks were remarkably shifted to higher temperature and the Co 2+ to Co 0 signal was weakened. For example, the Co3O4@Si0.52 showed the Co 3+ to Co 2+ signal at 402 ºC and weak Co 2+ to Co 0 signal at 533 ºC. The sample with lower Co/Si ratio, such as the Co3O4@Si0.95 sample, exhibited even higher Co 3+ to Co 2+ signal at 410 ºC and almost undetectable Co 2+ to Co 0 signal. These data demonstrate the different silica amount strongly influences the oxidation state of cobalt nanoparticles. Note: In-situ Raman spectra of catalysts were performed under H2 reduction treatment, which provided clear identification of the interaction between Co and O species. Co3O4, Co3O4@Si0.95, Co3O4@Si1.87, and Co3O4/SiO2 all gave typical Raman peaks at 196, 474, 518, 616 and 682 cm -1 assigning to the Co-O species 16,17 . By reduction at higher temperatures, these peaks were reduced in intensity and even undetectable on Co3O4 and Co3O4/SiO2 after reaction at 600 °C . But the Co@Si0.95 and Co@Si1.87 samples still exhibited these peaks under the equivalent treatments. These phenomena indicate the silica modification improves the anti-reduction ability of Co-O species, which is in good agreement with the results of XPS spectra. The M-O-SiOx interface have been widely investigated in hydrogenation reactions. Although the role of such interface has not been fully understood, the synergy of the metallic phase and oxide phase are regarded to be crucial. Generally, the metal catalyzed hydrogenation processes involve the steps including (i) dissociation of hydrogen, (ii) adsorption of unsaturated compounds, (iii) stepwise hydrogenation with H atoms. According to such knowledge, the metallic phase (e.g. Co 0 ) might activate H2 and the oxide phase (e.g. Co δ+ species) adsorbs CO2. This proposed pathway is in good agreement with the general knowledge on the Cu-O-SiOx catalysts 18,19 . In addition, the previous study has revealed that the Cu-O(H)-SiOx interface accelerate the hydrogen activation/splitting 20 . In order to identify whether the Co@Six catalysts have similar feature, we removed part of the silanol groups by treating the sample with NaOH. The resulted Co@Si0.95-Na catalyst exhibited remarkably reduced hydrogenation activity than the untreated Co@Si0.95 (Supplementary Figure 36), suggesting the important role of Co-O-SiOx interface for the reaction. Possibly, both the metallic Co 0 and interfacial Co δ+ -O-SiOx activated the hydrogen and accelerate the hydrogenation reactions. The interfacial Co δ+ -O-SiOx species stabilized the crucial reaction intermediates (e.g. *CH3O) and improved the methanol selectivity, while the metallic Co 0 would catalyze the *CH3O decomposition to form CO and methane. By optimizing the composition, the best catalyst was realized as Co@Si0.95. Note: The Co 0 signals in Co 2p XPS spectra were weakened by introducing CO2 to Co@Si0.95 catalyst and reduced by feeding hydrogen, suggesting the adsorption and hydrogenation of CO2 taking place on the Co sites. The signal of *HCOO appeared at 533.0 eV and was almost unchanged even under H2 purging, confirming the highly stable *HCOO intermediate. Note: The CO2 hydrogenation process was identified by in-situ XPS. The Co/SiO2 sample gives peaks at 293.0, 290.6, 288.4, 287.2 and 285.1 eV in C 1s XPS spectra after CO2 adsorption treatment, assigned to gaseous CO2, CO3 2-, CO2 δ-, HCO3and *CHx species, respectively 24 . When a slight amount of hydrogen was introduced over the sample (CO2:H2 at 3), signals of CO3 2-, CO2 δ-, and HCO3visibly changed and the *HCOO signal generated. Notably, a weak peak at 286.3 eV appeared by introducing more hydrogen, assigned to the *CH3O species. The *CH3O and *CHx signals were raised by continuously increasing the hydrogen content in the feed gas. A hydrogen purging treatment caused the elimination of almost all the peaks, because of their further hydrogenation and desorption from the catalyst surface. But the *HCOO signal still existed, confirming the highly stable *HCOO might not be the reaction intermediate. Note: The Co@Si0.95 catalyst was further studied in the atmosphere of mixed gas with low hydrogen pressure (1.0 mbar CO2 and 0.1 mbar H2), giving C 1s XPS peaks at 293.0, 290.6, 288.4 and 287.2 eV, assigned to gaseous CO2, CO3 2-, CO2 δand HCO3species, respectively (Supplementary Figure 47c). Then the CO2 and H2 gases were slowly extracted from the chamber, giving the strong signals at 289.2 and 286.3 eV, assigned to *HCOO and *CH3O species, respectively. Although these signals were slightly reduced with the extracting treatment, because of their desorption or further transformation, they still exhibited obvious signals confirming their high stability on the catalyst surface. Finally, by purging with hydrogen, the *CH3O peak immediately disappeared, suggesting its easy transformation in hydrogen, while the *HCOO was still stable, in good agreement with the abovementioned results.

Supplementary
The equivalent treatment was performed over the Co/SiO2 catalyst ( Supplementary  Figure 48c), which gave the C 1s XPS peaks at 293.0, 290.6, 287.2, and 285.1 eV assigned to gaseous CO2, CO3 2-, HCO3and *CHx species, respectively. When extracting the feed gases in the chamber, the *HCOO and *CHx (289.2 and 285.1 eV) appeared dominantly with almost undetectable *CH3O. In the end, the hydrogen purging process eliminated other intermediates but *HCOO species was still unchanged. This phenomenon demonstrates the poor stability of *CH3O species, which disappeared without further hydrogen introduction, suggesting its transformation might proceed the direct decomposition rather than direct hydrogenation (direct hydrogenation is a key step for methanol formation). This result is different from that on the Co@Si0.95 catalyst, suggesting the important role of Co-O-SiOn linkage for stabilizing the *CH3O intermediate to avoid decomposition and benefiting the direct hydrogenation to form methanol. Note: We performed pressure-dependent tests and showed the reaction orders of CO2 and H2. An approximately first-order dependence on the H2 partial pressure is observed for methanol synthesis, giving 0.80 and 0.91 over Co@Si0.95 and Co/SiO2 catalysts, respectively. This result indicates that the H2 positively influences the methanol formation in the CO2 hydrogenation [25][26][27][28][29][30][31] . The apparent CO2 reaction order in methanol synthesis is close to zero (0.09 for Co@Si0.95 and 0.11 for Co/SiO2), suggesting that the catalyst surface is saturated by CO2 and/or the reaction intermediates, and higher hydrogen pressure could efficiently accelerate the whole reactions.

Supplementary
In the CO2 hydrogenation to methanol, it is generally recognized that there are two major reaction pathways based on experimental observation and theoretical calculations, including the RWGS + CO hydrogenation pathway and the formate pathway 1,23,25,32−38 . Different rate-control steps have been suggested for methanol pathway from CO2 and H2 via these pathways. For example, in the formate pathway, hydrogenation of *HCOO (*HCOO + H → *HCOOH) and *CH3O (*CH3O + H → *CH3OH) intermediates are the rate-control steps 1,23,32 . As reported previously, if the hydrogenation of *HCOO is the rate-control step, the maximum attainable reaction orders according are 1 for both CO2 and H2. Generally, H* and HCOO* are always abundant on the catalyst surface 25,26,29 , the apparent CO2 and H2 reaction orders would be much smaller than 1, which is inconsistent with our results (H2 reaction order is close to 1). These data confirm that the H2 or CO2 activation should not be the control step. Our in-situ DRIFTS and XPS results demonstrate that the *CH3O was abundantly detected on the catalyst surface in CO2 and H2 atmosphere, and the further transformation of *CH3O was slow, suggesting the *CH3O hydrogenation should be the rate-control step. The low reaction order of CO2 (0.09-0.11) can be attributed to the slow step of *CH3O hydrogenation to CH3OH as a rate-control step 25,26 . Similarly, if CO2 hydrogenation follows the RWGS + CO hydrogenation pathway, the *CH3O is also the crucial intermediate for CH3OH formation, and *CH3O hydrogenation to CH3OH is the rate-control step 1,23,32 . These kinetic data demonstrate that methanol production over Co@Si0.95 and Co/SiO2 catalysts follows similar route that *CH3O hydrogenation to CH3OH is the rate-control step.
The apparent Ea for methane and CO production on Co@Si0.95 are 135.5 and 53.2 kJ mol -1 , respectively. These values are remarkably higher than those over the Co/SiO2 catalyst (96.4 kJ mol -1 for methane and 43.1 kJ mol -1 for CO production). These data confirm that easier methane and CO formation on Co/SiO2 catalyst compared with Co@Si0.95. In addition, the Co@Si0.95 exhibited apparent Ea for methanol production at 58.2 kJ mol -1 , lower than 62.4 kJ mol -1 on Co/SiO2 catalyst, which suggests easier methanol production on Co@Si0.95 catalyst.
The Co@Si0.95 catalyst exhibits slightly higher CO2 conversion than that of traditional Co/SiO2 or CoOx catalysts. Although the former is unfavorable for H2 dissociation because of the lower metallic Co content. Actually, similar phenomenon has been observed previously in different reaction systems, where the catalysts with lower ability for reactant activation might exhibit faster transformation rate [39][40][41][42] . This phenomenon is explained by the balanced rates for the production and further transformation of key intermediates, which have been studied in various reaction systems previously [41][42][43][44][45] . Note: CO is usually formed in the cobalt-catalyzed CO2 hydrogenation reactions. With regard to CO formation, it is recognized that there are two major reaction pathways for the CO2-to-CH3OH transformation based on experimental observation and theoretical calculations, including the RWGS + CO hydrogenation pathway and the formate pathway 1,23,25,32−38 . In the RWGS + CO hydrogenation pathway, the *CO intermediate is firstly produced from the RWGS reaction via the *HOCO intermediate, or the direct C-O bond cleavage of *CO2. The *CO intermediate is further hydrogenated to *CH3O, which could be transformed into CH3OH. Simultaneously, the *CO species could also desorb from the catalyst directly to form CO, which have been reported on the Co-based catalysts 2,4,5,7,10 . The formate pathway proceeds the *HCOO species by the primary hydrogenation of CO2, which produces *CH3O via the C-O bond cleavage of the *HxCOOH intermediates, and is eventually hydrogenated to the CH3OH. According to such proposed pathway, it seems the reaction does not produce CO 1 . However, we cannot exclude the CO production from the formate route because the formic acid is easily decomposed into CO and water on the metal catalysts 46−48 . In addition, the *CH3O and methanol species could be decomposed into CO species. The sample was obtained after CO2 hydrogenation over 100 h. N, coordination number; R, distance between absorber and backscatterer atoms; σ 2 , disorder term (EXAFS Debye-Waller factor). Amplitude factor S0 2 = 0.83. For Co/SiO2, k-range: 3.0-14.0 Å -1 , Rrange: 1.0-3.0 Å, E0 = 9.26 ± 0.21 eV; for Co@Six, k-range: 3.0-13.9 Å -1 , R-range: 1.0-3.2 Å, E0 = -3.01 ± 0.33 eV. Estimated EXAFS error bounds: N, ±20%; R, ±0.01 Å; ∆σ 2 , ±20%.