E ﬀ ect of Cu and Cs in the β -Mo 2 C System for CO 2 Hydrogenation to Methanol

: Mitigation of anthropogenic CO 2 emissions possess a major global challenge for modern societies. Herein, catalytic solutions are meant to play a key role. Among the di ﬀ erent catalysts for CO 2 conversion, Cu supported molybdenum carbide is receiving increasing attention. Hence, in the present communication, we show the activity, selectivity and stability of fresh-prepared β -Mo 2 C catalysts and compare the results with those of Cu / Mo 2 C, Cs / Mo 2 C and Cu / Cs / Mo 2 C in CO 2 hydrogenation reactions. The results show that all the catalysts were active, and the main reaction product was methanol. Copper, cesium and molybdenum interaction is observed, and cesium promoted the formation of metallic Mo on the fresh catalyst. The incorporation of copper is positive and improves the activity and selectivity to methanol. Additionally, the addition of cesium favored the formation of Mo 0 phase, which for the catalysts Cs / Mo 2 C seemed to be detrimental for the conversion and selectivity. Moreover, the catalysts promoted by copper and / or cesium underwent redox surface transformations during the reaction, these were more obvious for cesium doped catalysts, which diminished their catalytic performance.


Introduction
The study of CO 2 transformations has increased significantly due to the need to diminish its atmospheric emissions. The possibility of using CO 2 as a raw material is a convenient way from an economic and industrial point of view, therefore conversion to fuels such as methane, or syn-gas, which could be later used in the Fisher-Tropsch process for the synthesis of fuels or other chemicals, have received renewed attention. Among them, CO 2 hydrogenation to methanol is an appealing alternative since it can be used directly in fuel cells [1].
Despite being an interesting approach, the main challenge of these reactions is the chemical inertness of CO 2 . The most studied system is Cu-ZnO for which copper has proven to be critical when it is loaded onto oxide supports such as ceria or alumina. Unfortunately, these catalysts suffer deactivation due to the oxidation and sintering of copper upon reaction conditions [2,3].
Transition metal carbides (TMCs) have received a lot of attention since they display excellent catalytic behavior in transformations such as steam reforming of methanol, dry reforming of methane or CO hydrogenation [4]. Some transition metal carbides can adsorb CO 2 and favor the C-O scission

Results
We have studied four catalysts based on Mo 2 C with 1 wt.% loading of each metal and they are labelled as follows: β-Mo 2 C, Cu/Mo 2 C, Cs/Mo 2 C and Cu/Cs/Mo 2 C. The catalysts were tested in CO 2 hydrogenation at 150 • C and 20 bar for 16 h and the conversion and selectivity profiles are included in Figure 1. In regard to the conversion, it can be observed that while β-Mo 2 C catalysts showed a quite stable profile, the other samples displayed some changes along time on stream. The conversion obtained was in the range 3%-5% and the values followed the trend Cu/Mo 2 C > Cu/Cs/Mo 2 C > β-Mo 2 C > Cs/Mo 2 C, which are in fair agreement with the reported conversion levels for this challenging reaction [15]. Moreover, the main compound obtained was methanol and the selectivity reached 50%-65%, following a slightly different trend to that of the conversion: Cu/Mo 2 C > β-Mo 2 C > Cu/Cs/Mo 2 C > Cs/Mo 2 C. Unlike the conversion behavior, the selectivity profiles were quite stable for all the catalysts except for the sample Cu/Cs/ Mo 2 C. The other detected products were CO and CH 4 , with minor contributions of ethanol, which was below 0.2% except for the catalyst Cs/Mo 2 C that was around 2% and is included in Figure 1. The selectivity of the main secondary products, i.e., CH 4 and CO, over β-Mo 2 C was quite different with values of 30% and 5%, respectively. However, the selectivity to those products obtained for each catalyst was more similar for Cu/Mo 2 C, Cs/Mo 2 C and Cu/Cs/Mo 2 C, for which the values differ between 3%-5% when comparing CO and CH 4 selectivity of every catalyst.
Catalysts 2020, 10, x FOR PEER REVIEW 3 of 9 around 2% and is included in Figure 1. The selectivity of the main secondary products, i.e., CH4 and CO, over β-Mo2C was quite different with values of 30% and 5%, respectively. However, the selectivity to those products obtained for each catalyst was more similar for Cu/Mo2C, Cs/Mo2C and Cu/Cs/Mo2C, for which the values differ between 3%-5% when comparing CO and CH4 selectivity of every catalyst. CO2 hydrogenation to CH3OH can take place through direct hydrogenation, or following an intermediate path producing CO through the reverse water gas shift reaction, which eventually can be transformed into the alcohol. Alternatively, formate and formaldehyde intermediate species can be formed on the surface followed by subsequent hydrogenation to CH3OH [16].
The selectivity profiles obtained for β-Mo2C suggest that the RWGS path plays a key role, and for this reaction molybdenum carbide has been reported to be highly active [17].
The improved catalytic performances of Cu/Mo2C confirm that the addition of copper benefits both the conversion and selectivity to methanol on the molybdenum carbide system, as already described. It was reported that this improvement is related to the different paths followed when the copper-molybdenum carbide interface exists since the carbide can modify the electronic density of the metal [16,18].
The elemental analyses showed that the concentration of Mo and C agrees quite well, within the experimental error, with the Mo2C stoichiometry. In addition, the mapping of the Cs-doped sample, in Figure 2A, confirms the interaction between Cu, Cs and Mo2C.
Additionally, the positive effect of copper can also be related to its ability to activate hydrogen, which eventually reacts with CO2 adsorbed on the molybdenum carbide.
We expected to observe some improvement in the catalytic performance by adding an alkali as dopant since it has been shown to improve the activation of the CO2 molecule [8]. However, the opposite tendency was indeed observed, and both conversion and selectivity to methanol are diminished, which for the catalyst Cs/Mo2C are even lower than the values achieved with the bare β-Mo2C. CO 2 hydrogenation to CH 3 OH can take place through direct hydrogenation, or following an intermediate path producing CO through the reverse water gas shift reaction, which eventually can be transformed into the alcohol. Alternatively, formate and formaldehyde intermediate species can be formed on the surface followed by subsequent hydrogenation to CH 3 OH [16].
The selectivity profiles obtained for β-Mo 2 C suggest that the RWGS path plays a key role, and for this reaction molybdenum carbide has been reported to be highly active [17].
The improved catalytic performances of Cu/Mo 2 C confirm that the addition of copper benefits both the conversion and selectivity to methanol on the molybdenum carbide system, as already described. It was reported that this improvement is related to the different paths followed when the copper-molybdenum carbide interface exists since the carbide can modify the electronic density of the metal [16,18].
The elemental analyses showed that the concentration of Mo and C agrees quite well, within the experimental error, with the Mo 2 C stoichiometry. In addition, the mapping of the Cs-doped sample, in Figure 2A, confirms the interaction between Cu, Cs and Mo 2 C. It has been suggested that different crystal phases or Mo/C ratios can influence the selectivity of CO2 hydrogenation [19]. However, in the studied system, we only detected one carbide phase, therefore the reason must be found in another feature.
In order to understand the results, we evaluated the XRD patterns of the fresh and spent catalysts. The diffractograms showed that all the fresh catalysts, as shown in Figure 3a, displayed the characteristics peaks of β-Mo2C with hexagonal closet packing crystal structure (JCPDS 35-0787) at 2θ of 34.4°, 38°, 39.4°, 52.1°, 61.5°, 69.6° and 74.6°, and that no copper or cesium is observed, likely due to the small percentage of these elements. Additionally, the absence of the characteristic peaks of molybdenum oxides confirmed the successful carburization. Moreover, the alkali-doped samples also displayed peaks at 2θ of 40.6° and 58.7°, which corresponds to metallic Mo (JCPDS 42-1120) and whose formation has already been explained as a result of the negative charge transfer of Cs to molybdenum [14]. As can be observed, the relative contribution of Mo 0 is more relevant on the sample Cs/Mo2C. Hence, the presence of this metallic molybdenum phase could explain the worse catalytic performance of Cs-doped samples since Mo 0 has a poorer activation ability towards the CO2 molecule. It has been reported that when Mo 0 adsorbs CO2 HO2C-Mo species are formed, which do not favor 20 30 40 50 60 70 80 Additionally, the positive effect of copper can also be related to its ability to activate hydrogen, which eventually reacts with CO 2 adsorbed on the molybdenum carbide.
We expected to observe some improvement in the catalytic performance by adding an alkali as dopant since it has been shown to improve the activation of the CO 2 molecule [8]. However, the opposite tendency was indeed observed, and both conversion and selectivity to methanol are diminished, which for the catalyst Cs/Mo 2 C are even lower than the values achieved with the bare β-Mo 2 C.
It has been suggested that different crystal phases or Mo/C ratios can influence the selectivity of CO 2 hydrogenation [19]. However, in the studied system, we only detected one carbide phase, therefore the reason must be found in another feature.
In order to understand the results, we evaluated the XRD patterns of the fresh and spent catalysts. The diffractograms showed that all the fresh catalysts, as shown in Figure 3a, displayed the characteristics peaks of β-Mo 2 C with hexagonal closet packing crystal structure (JCPDS 35-0787) at 2θ of 34.4 • , 38 • , 39.4 • , 52.1 • , 61.5 • , 69.6 • and 74.6 • , and that no copper or cesium is observed, likely due to the small percentage of these elements. Additionally, the absence of the characteristic peaks of molybdenum oxides confirmed the successful carburization. Moreover, the alkali-doped samples also displayed peaks at 2θ of 40.6 • and 58.7 • , which corresponds to metallic Mo (JCPDS 42-1120) and whose formation has already been explained as a result of the negative charge transfer of Cs to molybdenum [14]. As can be observed, the relative contribution of Mo 0 is more relevant on the sample Cs/Mo 2 C. It has been suggested that different crystal phases or Mo/C ratios can influence the selectivity of CO2 hydrogenation [19]. However, in the studied system, we only detected one carbide phase, therefore the reason must be found in another feature.
In order to understand the results, we evaluated the XRD patterns of the fresh and spent catalysts. The diffractograms showed that all the fresh catalysts, as shown in Figure 3a, displayed the characteristics peaks of β-Mo2C with hexagonal closet packing crystal structure (JCPDS 35-0787) at 2θ of 34.4°, 38°, 39.4°, 52.1°, 61.5°, 69.6° and 74.6°, and that no copper or cesium is observed, likely due to the small percentage of these elements. Additionally, the absence of the characteristic peaks of molybdenum oxides confirmed the successful carburization. Moreover, the alkali-doped samples also displayed peaks at 2θ of 40.6° and 58.7°, which corresponds to metallic Mo (JCPDS 42-1120) and whose formation has already been explained as a result of the negative charge transfer of Cs to molybdenum [14]. As can be observed, the relative contribution of Mo 0 is more relevant on the sample Cs/Mo2C. Hence, the presence of this metallic molybdenum phase could explain the worse catalytic performance of Cs-doped samples since Mo 0 has a poorer activation ability towards the CO 2 molecule. It has been reported that when Mo 0 adsorbs CO 2 HO 2 C-Mo species are formed, which do not favor the scission of the C-O bond [20]. In contrast, dissociation of CO 2 over β-Mo 2 C is relatively easy and it may take place through a previous hydrogenation to CO followed by formation of HCO followed by hydrogenation to CH 3 OH. Furthermore, since the adsorption and activation of CO 2 molecules can occur on both Cu and Mo 2 C surfaces, the negative effect of the Mo 0 phase was not so relevant in the catalytic performance of the Cu/Cs/Mo 2 C as it is on the copper free sample, Cs/Mo 2 C [21].
The diffractograms of the spent catalyst, as shown in Figure 3b, showed that Cu/Mo 2 C preserved the β-Mo 2 C crystal structure and surprisingly it also showed the typical diffractions of Mo 0 , which were not observed in the fresh catalysts. This is an interesting finding and there is not an obvious reason for the formation of such a phase under the reaction conditions used. As observed by previous characterization by XRD and XPS of these samples, the fresh catalyst does not display metallic molybdenum, therefore we can tentatively rule out that these peaks appear as a consequence of the sintering of molybdenum during the reaction [14]. Another possibility is that the reaction conditions could somehow favor the formation of metallic molybdenum. The reduction of metals by Mo 2 C has been already reported to occur. Some authors have observed that metal precursors of Pt, Pd and Cu were spontaneously reduced when deposited onto unpassivated molybdenum carbide [22,23]. This was explained by a redox phenomenon on the surface, which probably implied the oxidation of molybdenum, since it is in a highly reduced state. However, this is not observed in most of the literature because molybdenum carbide is frequently passivated, and this surface layer might prevent the oxidation-reduction reactions [24]. A plausible explanation is that the same happens with molybdenum, therefore redox transformations during the reaction could be the reason for the chemical state of the spent catalyst. Nevertheless, the possibility of a molybdenum oxide passivation layer on Mo 0 nanoparticles of the fresh catalysts cannot be totally disregarded. However, conventional ex situ XPS cannot be used to assess the effect of the reaction conditions on the crystal phase changes, and further in situ studies would be required, which are beyond this proof-of-concept work.
Finally, no diffractions ascribed to MoO 2 or MoO 3 are observed, this indicates a good stability towards oxidation under the tested conditions.
On the other hand, the XRD patterns of spent catalysts Cu/CsMo 2 C and Cs/Mo 2 C display diffractions at 2θ of 26.0 • , 37.1 • and 53.3 • , which correspond to the MoO 2 crystal phase (JCPDS 32-671). Furthermore, the diffractions due to the Mo 2 C phase mostly disappear and just a broad hump with maximum at 2θ of 39.4 • can be envisaged on the catalyst Cs/Mo 2 C. This broad peak probably includes just the contributions of Mo 2 C and MoOx, since no diffraction at 2θ of 58 • corresponding to Mo 0 is observed. On the contrary, the spent catalyst Cu/Cs/Mo 2 C, despite having mostly lost the typical diffractions of Mo 2 C, still displayed the diffractions ascribed to metallic Mo.
The conversion and selectivity profiles can be partially explained by the XRD patterns. The catalyst β-Mo 2 C, despite displaying lower conversion and selectivity than the catalyst Cu/Mo 2 C, showed a more stable catalytic performance, in agreement with the structure preservation observed by XRD. In addition, it can be observed that Cu/Mo 2 C, Cs/Mo 2 C and Cu/Cs/Mo 2 C catalysts are activated at the beginning of the reaction as the increase in conversion indicates. However, while for Cu/Mo 2 C conversion was then stable after the first 4 h, this is not true for Cu/Cs/Mo 2 C and Cs/Mo 2 C for which an apparent deactivation occurred, this effect being more pronounced for the Cs/Mo 2 C catalyst. Nevertheless, the conversion at the end of the experiment was still higher than that observed at the start of the reaction. The reason for this behavior can be that at the beginning of the reaction, the hydrogen flow of the reactants feed is able to reduce the oxide layer of copper, which could subsequently activate hydrogen to carburize molybdenum oxycarbide species, MoO x C y, and/or to reduce Mo oxide layers [24,25]. Then, for Cu/Cs/Mo 2 C and Cs/Mo 2 C, oxidation of the molybdenum carbide took place, as the XRD shows. Furthermore, the extent of oxidation seems to be in agreement with the deactivation slope, which is more obvious for Cs/Mo 2 C.
The oxidation of the alkali-doped samples could be explained by the reverse water-gas-shift reaction (RWGS) (CO 2 + H 2 ↔ CO + H 2 O) taking place simultaneously. The occurrence of RWGS would not only decrease the selectivity, but it would also increase the amount of water on the reactant stream, which could eventually oxidize the molybdenum carbide phase as the XRD of these spent catalysts shows. The reason for the RWGS reaction being favored by Cu/Cs/Mo 2 C and Cs/Mo 2 C is in agreement with previous literature reporting the effect of alkalis on the RWGS [14,26,27].
Much progress has been made towards the development of viable catalysts for CO 2 hydrogenation to methanol. Some recent works related with the performances and reaction conditions of these catalysts have been summarized in Table 1. In industry, methanol is produced from synthesis gas mixtures at elevated pressures P (50 to 100 bar) and temperatures T (200 • C to 300 • C) over Cu/ZnO/Al 2 O 3 catalysts, with a worldwide demand of~50 Mt year −1 [28]. Therefore, Cu/ZnO/Al 2 O 3 series still attract numerous attention [29]. As shown in Table 1, the CO 2 conversion and methanol selectivity towards Cu/ZnO/Al 2 O 3 highly depend on the temperatures. The CO 2 conversion and CH 3 OH selectivity reached 5.19% and 67%, respectively, at 200 • C. When the temperature increased to 250 • C, the corresponding values changed to 13.4% and 13%, respectively. In addition, other combinations of Cu and metal/metal oxides have been explored extensively in the methanol synthesis field, such as Cu/Mg/Al [30], Cu/ZnO/ZrO 2 [31,32]. For comparison's sake, the performances of other Mo 2 C related catalysts have also been listed in Table 1 [15,33]. As can be seen, although Cu-based catalysts are suitable for CO 2 hydrogenation, an efficient catalyst exhibiting high selectivity, conversion, and functional stability towards methanol synthesis at low pressure (less than 50 bar) has not been developed yet [31]. In summary, the results show that for the studied system, alkali doping with cesium is not beneficial for the synthesis of methanol in the studied catalytic system and that in situ carburization would be required. This did not happen during the reverse water-gas-shift reaction, for which re-carburization of the active phase was observed during reaction conditions [14]. Hence, the absence of the re-carburization of catalysts in the methanol synthesis is probably due to the different reaction conditions used. The conditions used in ref. 14 for RWGS, i.e., atmospheric pressure and 400-750 • C, could allow the re-carburization of molybdenum during the reaction, in contrast to what is observed in the present work at 20 bar and mild temperature, 150 • C.

Materials and Methods
The β-Mo 2 C catalyst was obtained through TPC procedure using ((NH 4 ) 6 Mo 7 O 24 ·4H 2 O Sigma-Aldrich) as reported before [28]. Briefly, the precursor was calcined and heated at a rate of 5 • C/min up to 500 • C and maintained for 4 h to get MoO 3 . Then, the oxide precursor was placed under an atmosphere composed of 20% CH 4 and 80% H 2 and submitted to thermal treatment at a heating rate of 5 • C/min up to 300 • C. Then, at a rate of 2 • C/min, the temperature was raised up to 700 • C and kept for 2 h. The Cs/Mo 2 C, Cu/Mo 2 C and Cu/Cs/Mo 2 C catalysts were prepared using the corresponding amounts of the same molybdenum precursor and Cu(NO 3 ) 2 or Cu(NO 3 ) 2 /Cs 2 CO 3, using the co-precipitation method as described previously [14]. The nominal loading of copper and cesium was 1 wt% for each element. Then, the same TPC process was used to prepare.
The catalytic tests were done in a stainless steel fixed-bed flow reactor of 3/8" internal diameter. Around 0.25 g of catalyst was used as prepared. Before the measurements, 20 bar was reached in the reactor with a reactant mixture of (CO 2 :H 2 :He = 3:9:18 vol) and a total flow of 30 mL/min. The compounds were analyzed by gas chromatographs (Varian CP 3400) with FID and TCD detectors and fitted with a column SupelQ Plot and 60/80 Carboxen-1000 column, respectively. The carbon balance was over 95% in all cases. The conversion and products selectivity were obtained according to the following equations: (4) n i : number of carbon atoms of i. mol i : number of moles i. mol CO 2 -un: mol of unreacted CO 2 .
X-ray diffraction (XRD) patterns of the catalysts were acquired using a Polycristal X'Pert Pro PANalytical diffractometer with Ni-filtered Cu Kα radiation (λ = 1.54 Å) operating at 45 kV and 40 mA. The 2θ range was between 4 • and 90 • with a step of 0.04 • /s. XEDS-mapping analysis was performed in STEM mode with a probe size of~1 nm using the INCA x-sight (Oxford Instruments) detector. To prepare the samples, small amounts of the samples were ground and dissolved in acetone solution under sonication. Small drops were added to the gold grid (Aname, Lacey carbon 200 mesh) and the acetone was evaporated at room temperature before introducing the microscope. Funding: This research was funded by Fundación General del CSIC ComFuturo programme, and the APC was funded by iLink project Nº 20211 from CSIC. The team at Surrey acknowledges the funding from CO 2 ChemUK via the EPSRC grant EP/P026435/1.

Conflicts of Interest:
The authors declare no conflict of interest.