Copper and Iron Cooperation on Micro-Spherical Silica during Methanol Synthesis via CO 2 Hydrogenation

: A series of mono- and bi-metallic copper and iron samples were prepared by impregnation method on micro-spherical silica and used for the synthesis of methanol via CO 2 hydrogenation. Compared with conventional carrier oxides, micro-spherical silica has obvious advantages in terms of absorption capacity and optimal distribution of active phases on its surface, also exhibiting excellent heat resistance properties and chemical stability. The prepared catalysts were characterized by various techniques including XRF, XRD, SEM, TEM, H 2 -TPR and CO 2 -TPD techniques, while catalytic measurements in CO 2 hydrogenation reaction to methanol were performed in a ﬁxed bed reactor at a reaction pressure of 30 bar and temperature ranging from 200 to 260 ◦ C. The obtained results revealed that the mutual interaction of copper–iron induces promotional effects on the formation of methanol, especially on systems where Fe enrichment on the silica support favours the presence of a larger concentration of oxygen vacancies, consequently responsible for higher CO 2 adsorption and selective methanol production. Surface reconstruction phenomena rather than coke or metal sintering were responsible for the slight loss of activity recorded on the catalyst samples during the initial phase of reaction; however, with no appreciable change on the product selectivity.


Introduction
In the last number years the concentration of carbon dioxide (CO 2 ) in the atmosphere has greatly increased owing to human activities, in particular from burning fossil fuels for electricity, heat and transportation. The anthropic production of carbon dioxide increases the natural greenhouse effect, considered as the main issue related to climate change and global warming [1][2][3]. In this regard, many researchers have been focusing on the potential of carbon capture and utilization (CCU) strategies to produce high-added-value compounds such as methanol, dimethyl ether, methane, and other hydrocarbons [4][5][6][7][8][9], by realizing an ideal neutral carbon-loop when CO 2 is reacted with green hydrogen coming from renewable sources [10][11][12][13].
Being considered as one of the most efficient energy carriers in the concept of a new sustainable economy [14][15][16], methanol (MeOH) is traditionally produced from syngas by catalytic hydrogenation at high temperature and pressure, determining a serious impact on the environment. So, the potential use of CO 2 as an alternative feedstock replacing syngas in the methanol production has drawn widespread attention as an effective way to recycle carbon and reduce CO 2 emissions [17][18][19][20][21]. CuO/ZnO/Al 2 O 3 is a conventional catalyst for Figure 1. SEM micro-graphs of the "bare" ES70Y micro-spherical silica (left) and a representative Cu(x)Fe(y)-Si catalyst sample (right). In the inset the elemental mapping of the bimetallic catalyst is represented.
The list of the investigated catalysts, the relative Fe/(Cu + Fe) ratio determined from the analytical composition and the main textural properties are reported in Table 1. Table 1. List of catalysts, analytical composition, and textural properties.

Catalyst Sample Analytical Comp. (mmol/gcat) (a) Fe/(Cu + Fe) SABET (b) PV (b) Cu
Fe (at/at) (m 2 /g) (cm 3 /g) Cu(3)Fe (7) Although the metal addition leads to a slight decrease in the surface area with respect to the bare silica micro-spheres, textural data of the supported catalysts evidence that both total surface exposure and pore volume take advantage from a progressive decrease in Cu loading in the catalyst formulation, with the maximum values, respectively, of 272 m 2 /g and 3.17 cm 3 /g found for the copper-free catalyst, Fe(10)-Si. In particular, Figure 2 shows how the surface area linearly increases with the molar fraction of iron, by reaching a value as high as 272 m 2 /g at the Fe/(Cu + Fe) ratio of 1.00. Regarding the supported samples ( Figure 1-right), the SEM micro-graph of a representative bimetallic Cu(x)Fe(y)-Si sample shows how the spherical morphology of the silica support is preserved upon metal deposition, while the elemental mapping depicted as the impregnation method does not allow for a perfect homogenization of the bimetallic phase on the support, just exhibiting some patchy metal overloading at the bottom of silica spheres.
The list of the investigated catalysts, the relative Fe/(Cu + Fe) ratio determined from the analytical composition and the main textural properties are reported in Table 1. Although the metal addition leads to a slight decrease in the surface area with respect to the bare silica micro-spheres, textural data of the supported catalysts evidence that both total surface exposure and pore volume take advantage from a progressive decrease in Cu loading in the catalyst formulation, with the maximum values, respectively, of 272 m 2 /g and 3.17 cm 3 /g found for the copper-free catalyst, Fe(10)-Si. In particular, Figure 2 shows how the surface area linearly increases with the molar fraction of iron, by reaching a value as high as 272 m 2 /g at the Fe/(Cu + Fe) ratio of 1.00.
The XRD patterns of the bi-metallic Cu(x)Fe(y)-Si samples after calcination and reduction are shown in Figure 3.  Irrespective of the catalyst state (whether calcined or reduced), a broad ba at around 22.5° and attributable to amorphous silica is evident for all the sam calcined samples (Figure 3-left), the typical diffraction peaks of CuO well cr the form of tenorite at 35.5 and 38.7° (JCPDS 05-0661) are well visible in the C sample, while as the iron loading increases, these reflections tend to decrease till the appearance of a mixed oxide form in the Cu(3)Fe(7)-Si sample attrib cubic structure of a cuprospinel (CuFe2O4, JCPDS 25-0283). In any case, only l reflections ascribable to the cubic structure of maghemite (Fe2O3, JCPDS 39-   Irrespective of the catalyst state (whether calcined or reduced), a broad band cantered at around 22.5° and attributable to amorphous silica is evident for all the samples. In the calcined samples (Figure 3-left), the typical diffraction peaks of CuO well crystallized in the form of tenorite at 35.5 and 38.7° (JCPDS 05-0661) are well visible in the Cu(7)Fe(3)-Si sample, while as the iron loading increases, these reflections tend to decrease in intensity, till the appearance of a mixed oxide form in the Cu(3)Fe(7)-Si sample attributable to the cubic structure of a cuprospinel (CuFe2O4, JCPDS 25-0283). In any case, only low-intensity reflections ascribable to the cubic structure of maghemite (Fe2O3, JCPDS 39-1346) are detected on all the samples, consistent with the fact that Fe ions and Fe nanodomains are evidently embedded into the amorphous background of the SiO2 phase. After reduction ( Figure 3-right), the Cu(x)Fe(y)-Si samples exhibit the formation of the respective metallic phases of copper at 43° (JCPDS 04-0836) and iron at 45° (JCPDS 06-0696), whose relative intensities are proportional to the metal loading on the micro-spherical silica.
In order to determine the redox properties of the catalysts, Figure 4 shows the profiles of the temperature programmed measurements performed under hydrogen atmosphere (H2-TPR), while the quantitative results are reported in Table 2. In order to determine the redox properties of the catalysts, Figure 4 shows the profiles of the temperature programmed measurements performed under hydrogen atmosphere (H 2 -TPR), while the quantitative results are reported in Table 2. baseline drift due to H2 consumption, while the Cu(10)-Si sample displays a sharp and quite symmetric peak between 170 and 200 °C, mirroring the reduction in CuO to metallic copper. As for the Fe(10)-Si sample, two broad and poorly resolved bands are observed above 300 °C, diagnostic of the progressive reduction in Fe2O3 to Fe3O4 and then to FeO, before being followed by the reduction to metallic Fe [49]. Regarding the bi-metallic systems, the H2-TPR measurements are essentially characterized by two convoluted reduction maxima at low (Tmax, 196-216 °C) and high temperature (Tmax, 252-255 °C), related to the reduction in copper and iron oxide particles, respectively, in less or more synergy among them. In spite of a similar H2 consumption for all the bi-metallic catalysts (2.8-3.0 mmol/gcat), the molar fraction of iron significantly affects the reduction kinetics in each sample [50], so that in Cu(7)Fe(3)-Si a single peak is observed as the result of easier reduction kinetics (α = 0.99) boosted by a low iron enrichment of the sample and exemplified by a shift in the reduction onset temperature (108 °C).   In all the temperature range investigated, the bare SiO 2 sample does not show any baseline drift due to H 2 consumption, while the Cu(10)-Si sample displays a sharp and quite symmetric peak between 170 and 200 • C, mirroring the reduction in CuO to metallic copper. As for the Fe(10)-Si sample, two broad and poorly resolved bands are observed above 300 • C, diagnostic of the progressive reduction in Fe 2 O 3 to Fe 3 O 4 and then to FeO, before being followed by the reduction to metallic Fe [49].
Regarding the bi-metallic systems, the H 2 -TPR measurements are essentially characterized by two convoluted reduction maxima at low (T max , 196-216 • C) and high temperature (T max , 252-255 • C), related to the reduction in copper and iron oxide particles, respectively, in less or more synergy among them. In spite of a similar H 2 consumption for all the bi-metallic catalysts (2.8-3.0 mmol/g cat ), the molar fraction of iron significantly affects the reduction kinetics in each sample [50], so that in Cu(7)Fe(3)-Si a single peak is observed as the result of easier reduction kinetics (α = 0.99) boosted by a low iron enrichment of the sample and exemplified by a shift in the reduction onset temperature (108 • C).
TEM micro-graphs of the calcined and reduced Cu(5)Fe(5)-Si are displayed in Figure 5, showing how the copper and iron phases distribute onto the silica carrier. Despite a not ideal homogenization of the supported active phase determined by the adopted preparation method, it is evident that the size regularity of silica spheres allows the generation of a metallic phase with homogeneous distribution of particles between 13 and 19 nm. TEM micro-graphs of the calcined and reduced Cu(5)Fe(5)-Si are displayed in Figure  5, showing how the copper and iron phases distribute onto the silica carrier. Despite a not ideal homogenization of the supported active phase determined by the adopted preparation method, it is evident that the size regularity of silica spheres allows the generation of a metallic phase with homogeneous distribution of particles between 13 and 19 nm. Considering that the catalytic behaviour is primarily controlled by CO2 activation, CO2 desorption measurements were performed to determine the relative abundance of the surface adsorption sites on the various investigated catalysts, with profiles and quantitative data reported in Figure 6 and Table 3, respectively. So, despite an irrelevant CO2 coverage on the bare SiO2 (2 mol/gcat), it can be observed that for all the bi-metallic Cu(x)Fe(y)-Si samples, the CO2 uptake concentrates on the temperature range between 100 and 300 °C, as the result of similar convoluted desorption bands, more or less shouldered on the high-temperature side and associated with two main CO2 populations of weak-medium (below 175 °C) or medium-strong sites (above 175 °C), respectively. In particular, in line with the desorption patterns of the mono-metallic Cu(10)-Si and Fe(10)-Si catalysts, in the bi-metallic samples, not only the CO2 capacity increases with the iron loading but also the temperature of desorption maxima shift at a lower temperature, so that on Cu(3)Fe (7)-Si the CO2 uptake as high as 50 mol/gcat is mainly to put in relation to the largest population of weak-medium sites (0.56) associated with surface hydroxyl groups or metal-oxygen pairs [51], while on Cu(7)Fe(3)-Si (24 molCO2/gcat), a prevailing concentration of medium-strong sites (0.69) is diagnostic of oxygen vacancies generated on the catalyst surface [52,53]. Considering that the catalytic behaviour is primarily controlled by CO 2 activation, CO 2 desorption measurements were performed to determine the relative abundance of the surface adsorption sites on the various investigated catalysts, with profiles and quantitative data reported in Figure 6 and Table 3, respectively.  So, despite an irrelevant CO 2 coverage on the bare SiO 2 (2 µmol/g cat ), it can be observed that for all the bi-metallic Cu(x)Fe(y)-Si samples, the CO 2 uptake concentrates on the temperature range between 100 and 300 • C, as the result of similar convoluted desorption bands, more or less shouldered on the high-temperature side and associated with two main CO 2 populations of weak-medium (below 175 • C) or medium-strong sites (above 175 • C), respectively. In particular, in line with the desorption patterns of the mono-metallic Cu(10)-Si and Fe(10)-Si catalysts, in the bi-metallic samples, not only the CO 2 capacity increases with the iron loading but also the temperature of desorption maxima shift at a lower temperature, so that on Cu(3)Fe (7)-Si the CO 2 uptake as high as 50 µmol/g cat is mainly to put in relation to the largest population of weak-medium sites (0.56) associated with surface hydroxyl groups or metal-oxygen pairs [51], while on Cu(7)Fe(3)-Si (24 µmol CO2 /g cat ), a prevailing concentration of medium-strong sites (0.69) is diagnostic of oxygen vacancies generated on the catalyst surface [52,53].
To verify and quantify the possible generation of oxygen vacancies directly linked to the CO 2 adsorption capacity brought by the iron enrichment, in Figure 7 the XPS analysis performed over the bimetallic reduced samples is displayed. The survey scans show the presence of copper in the metallic state at 932.9 eV (Cu 2p3/2), while iron is present at Fe(III), Fe(II) and Fe(0) oxidation states as evidenced by complex multiplet-split Fe 2p spectra at 710.9, 709.6 and 706.6 eV, respectively, with their typical satellite features (Fe and Cu multiplex peak data reported in the Supplementary Material Figure S1a,b). In quantitative terms, all the samples exhibit a surface enrichment of iron over the surface, considering that the Fe/(Cu + Fe) surface ratio is always greater (0.85 for Cu(3)Fe(7)-Si, 0.62 for Cu(5)Fe(5)-Si and 0.54 for Cu(7)Fe(3)-Si) than the respective bulk composition (see Table 1). The O 1s core level spectra were deconvoluted according to three main peaks, associated with lattice oxygen of metal oxides (O L ), oxygen vacancies (O V ) and chemisorbed oxygen on the surface (O C ) from low to high binding energies [54]. The relative concentrations of oxygen species were reported in Table 4 through quantitative analysis based on peak area. Table 4. Relative concentration of oxygen species calculated from the O 1s peak area. As a rule, the percentage of oxygen vacancies generated over the samples linearly increases with the iron loading, from a minimum value of 4% over the Cu-rich sample until a maximum of 24% over the Cu(3)Fe(7)-Si sample at higher Fe loading, in a strong consistency with the previously reported results. Being offset by a decrease in lattice oxygen, these findings clearly suggest that the generation of oxygen vacancies upon the reducing treatment also proceeds according to a mechanism of migration of oxygen species from inside the bulk to the catalyst surface [54,55]. As a rule, the percentage of oxygen vacancies generated over the samples linearly increases with the iron loading, from a minimum value of 4% over the Cu-rich sample until a maximum of 24% over the Cu(3)Fe(7)-Si sample at higher Fe loading, in a strong consistency with the previously reported results. Being offset by a decrease in lattice oxygen, these findings clearly suggest that the generation of oxygen vacancies upon the reducing treatment also proceeds according to a mechanism of migration of oxygen species from inside the bulk to the catalyst surface [54,55].

Cu(7)Fe(3)-Si Cu(5)Fe(5)-Si Cu(3)Fe(7)-Si
The catalytic behaviour of the prepared samples was investigated under CO2 hydrogenation conditions and in Table 5 the results obtained at 30 bar in the temperature range of 200-260 °C are reported, in terms of carbon dioxide conversion and product selectivity. As a rule, for all the samples a progressive increase in the CO2 conversion is recorded with temperature, while methanol selectivity exhibits a clear drop, offset by a regular rise in CO and CH4 selectivity, in line with thermodynamic predictions [26]. In particular, by inspecting the activity-selectivity pattern of the mono-metallic systems, a superior functionality is observed over the Fe(10)-Si sample, with a maximum CO2 conversion of 4.6% attained at 260 °C, resulting in MeOH, CO and CH4 selectivity of 18.0, 54.6 and 27.4%, respectively.  The catalytic behaviour of the prepared samples was investigated under CO 2 hydrogenation conditions and in Table 5 the results obtained at 30 bar in the temperature range of 200-260 • C are reported, in terms of carbon dioxide conversion and product selectivity. As a rule, for all the samples a progressive increase in the CO 2 conversion is recorded with temperature, while methanol selectivity exhibits a clear drop, offset by a regular rise in CO and CH 4 selectivity, in line with thermodynamic predictions [26]. In particular, by inspecting the activity-selectivity pattern of the mono-metallic systems, a superior functionality is observed over the Fe(10)-Si sample, with a maximum CO 2 conversion of 4.6% attained at 260 • C, resulting in MeOH, CO and CH 4 selectivity of 18.0, 54.6 and 27.4%, respectively. Table 5. Conversion/selectivity data (X CO2 /S MeOH /S CO /S CH4 ) in the CO 2 hydrogenation reaction at different temperatures (P R , 30 bar; GHSV: 8800 NL/kg cat /h).

S MeOH /S CO /S CH4 (%) X CO2 /S MeOH /S CO /S CH4 (%) X CO2 /S MeOH /S CO /S CH4 (%) X CO2 /S MeOH /S CO /S CH4 (%)
Cu (3) As for the bi-metallic samples, the X CO2 values in the range of 200-260 • C regularly rise from 1.8 to 11.6%, while the S CH3OH decreases in the range 66.6-29.7%. Over these samples, the CH4 selectivity also increases with temperature, at 260 • C resulting as low as 0.7% on Cu(5)Fe(5)-Si, while 6.5-7.0% on the other bi-metallic samples, values anyhow lower than that obtained under the same temperature over the mono-metallic copper-free sample (S CH4 , 27.4%). However, these results confirm a poorer catalytic performance with respect to benchmark CZA or CZZ systems tested under similar conditions [26,27]; furthermore, they clearly put in evidence the beneficial cooperative effect between copper and iron atoms in driving the production of methanol via direct CO 2 hydrogenation. Yet, both the metal loading and the iron fraction depict a determinant control on the activity-selectivity pattern, so that thorough data normalization is necessary for a full understanding of the catalytic behaviour. In this respect, by elaborating the rate of CO 2 conversion as a function of the number of oxygen vacancies, in turn determined from the CO 2 adsorbed on mediumstrong sites, the "turnover frequency" over each catalyst sample was calculated and put in relation to the atomic iron fraction.
The bell-shaped curve shown in Figure 8 visibly evidences the need of an optimal relative concentration of iron atoms to maximize the rate of CO 2 conversion, indirectly disclosing how a large number of oxygen vacancies is necessary to favour the adsorption of carbon dioxide, but not sufficient to lower the energy barrier associated with the cleavage of C=O bonds and then to activate CO 2 under the adopted conditions. sample (SCH4, 27.4%). However, these results confirm a poorer catalytic performance w respect to benchmark CZA or CZZ systems tested under similar conditions [26,27]; fu thermore, they clearly put in evidence the beneficial cooperative effect between copp and iron atoms in driving the production of methanol via direct CO2 hydrogenation. Y both the metal loading and the iron fraction depict a determinant control on the activi selectivity pattern, so that thorough data normalization is necessary for a full understan ing of the catalytic behaviour. In this respect, by elaborating the rate of CO2 conversion a function of the number of oxygen vacancies, in turn determined from the CO2 adsorb on medium-strong sites, the "turnover frequency" over each catalyst sample was calc lated and put in relation to the atomic iron fraction.
The bell-shaped curve shown in Figure 8 visibly evidences the need of an optim relative concentration of iron atoms to maximize the rate of CO2 conversion, indirec disclosing how a large number of oxygen vacancies is necessary to favour the adsorpti of carbon dioxide, but not sufficient to lower the energy barrier associated with the clea age of C=O bonds and then to activate CO2 under the adopted conditions. Despite the complexity in finding a basic relationship comprehensively describi the catalytic behaviour, it is, nevertheless, observable that the Cu(3)Fe(7)-Si sample ch acterized by a Fe/(Cu + Fe) ratio of 0.67 exhibits the highest turnover frequency of CO suggesting that the CO2 hydrogenation reaction over Cu(x)Fe(y)-Si catalysts takes a vantage from the presence of a prevailing concentration of metallic Fe (see Figure 3 right), evidently acting as the active site for the primary formation of CO via rWGS [5 Despite the complexity in finding a basic relationship comprehensively describing the catalytic behaviour, it is, nevertheless, observable that the Cu(3)Fe(7)-Si sample characterized by a Fe/(Cu + Fe) ratio of 0.67 exhibits the highest turnover frequency of CO 2 , suggesting that the CO 2 hydrogenation reaction over Cu(x)Fe(y)-Si catalysts takes advantage from the presence of a prevailing concentration of metallic Fe (see Figure 3-right), evidently acting as the active site for the primary formation of CO via rWGS [56]. Then, the CO adsorbed on metallic Fe can be further hydrogenated till methanol, likely owing to a phase change occurred in the Fe sites in the form of carbides as generated on the Fisher-Tropsch catalysts [57]. On the opposite site, the Cu-rich catalyst compositions, like in Cu(7)Fe(3)-Si, depress the kinetics of the rWGS route, rather addressing a formate pathway very common on Cu-based catalysts under CO 2 -to-MeOH hydrogenation conditions [58]; the active site being essentially associable to the existence of a large metallic copper phase in strong interaction with a poorly reducible iron oxide phase. Still, on catalysts characterized by an equimolar concentration of copper and iron sites, a "pure" specific mechanism (whether rWGS or formate route) is not really dominant, so that the extent of interaction between the copper oxide and the iron oxide phases requires a certain period for the complete sample reduction and the possible conversion of metallic Fe into carbides via rWGS. Indeed, the lowest selectivity to methane at 260 • C on the Cu(5)Fe(5)-Si sample (S CH4 , 0.7%) is the result of the lack of a really "pure" mechanism, evidently due to the co-existence of mixed phases on the micro-spherical silica.
The catalytic stability evaluated in terms of CO 2 conversion as a function of time-onstream is shown in Figure 9.
like in Cu(7)Fe(3)-Si, depress the kinetics of the rWGS route, rather addressing a fo pathway very common on Cu-based catalysts under CO2-to-MeOH hydrogenation c tions [58]; the active site being essentially associable to the existence of a large m copper phase in strong interaction with a poorly reducible iron oxide phase. Still, o alysts characterized by an equimolar concentration of copper and iron sites, a "pure cific mechanism (whether rWGS or formate route) is not really dominant, so that t tent of interaction between the copper oxide and the iron oxide phases requires a c period for the complete sample reduction and the possible conversion of metallic F carbides via rWGS. Indeed, the lowest selectivity to methane at 260 °C on the Cu(5) Si sample (SCH4, 0.7%) is the result of the lack of a really "pure" mechanism, evident to the co-existence of mixed phases on the micro-spherical silica.
The catalytic stability evaluated in terms of CO2 conversion as a function of tim stream is shown in Figure 9. It can be observed that, irrespective of the differences in the absolute values rec on the bi-metallic Cu(x)Fe(y)-Si catalysts, the activity pattern in all the cases reach almost steady state after about 600 min from the start of the experiments, then not ing any further decay for over 4500 min.
The "post reaction" analysis performed on the catalyst samples proves that no (see Figure S2 in the Supplementary Materials) or metal sintering (crystallite size a unchanged with respect to the fresh sample) can be invoked to justify the observed catalytic activity during the first phase of reaction, rather the presence of a peak refl at 45° in the XRD pattern confirms the formation of an iron carbide phase (Fe3C, J 06-0670) during reaction, envisaging a structural modification due to an incipient iro burization (see Figure 10). Moreover, a clear signal centred at around 38° and asso with the generation of a spinel CuFe2O4 phase on Cu(5)Fe(5)-Si (not visible in the "used" bi-metallic samples) clearly corroborates the insight into the existence of phases on the micro-spherical silica support, being well known that a poorly red phase (i.e., spinel-like) leads to a lower surface coverage of adsorbed H species, s venting the complete hydrogenation till methane [59]. It can be observed that, irrespective of the differences in the absolute values recorded on the bi-metallic Cu(x)Fe(y)-Si catalysts, the activity pattern in all the cases reaches an almost steady state after about 600 min from the start of the experiments, then not showing any further decay for over 4500 min.
The "post reaction" analysis performed on the catalyst samples proves that no coke (see Figure S2 in the Supplementary Materials) or metal sintering (crystallite size almost unchanged with respect to the fresh sample) can be invoked to justify the observed loss of catalytic activity during the first phase of reaction, rather the presence of a peak reflection at 45 • in the XRD pattern confirms the formation of an iron carbide phase (Fe 3 C, JPCSD 06-0670) during reaction, envisaging a structural modification due to an incipient iron carburization (see Figure 10). Moreover, a clear signal centred at around 38 • and associated with the generation of a spinel CuFe 2 O 4 phase on Cu(5)Fe(5)-Si (not visible in the other "used" bi-metallic samples) clearly corroborates the insight into the existence of mixed phases on the micro-spherical silica support, being well known that a poorly reducible phase (i.e., spinel-like) leads to a lower surface coverage of adsorbed H species, so preventing the complete hydrogenation till methane [59].

Catalyst Preparation
A commercial micro-spherical silica ES70Y (SABET: 295 m 2 /g; PV: 1.6 cm 3 /g) sup by PQ Corporation was used as the support of bimetallic Cu(x)Fe(y)-SiO2 catalytic ples. Cu(NO3)2·3H2O and Fe(NO3)3·9H2O (supplied by Sigma Aldrich) were together solved in a suitable amount of water, prior to being impregnated on the silica suppor changing the Cu:Fe atomic ratio from 2:1 to 1:2. Monometallic Cu/SiO2 and Fe/SiO2 lysts were prepared by dry impregnation method starting from the aqueous solutio the nitrate precursors. After impregnation, the catalysts were dried in air at 110 °C f h and calcined in static air at 400 °C for 4 h.

Characterization
The analytical composition of catalysts was determined by X-ray fluorescence a sis, using a Bruker AXS-S4 Explorer spectrometer (Karlsruhe, Germany), equipped w rhodium X-ray source (Rh anode and 75 μm Be window), a LiF 220 crystal analyzer a 0.12° divergence collimator. Samples were analysed at the solid state, taking into acc the emission value of Cu-Kα1 and Fe-Kα1 transitions.
The specific surface area and the pore volume data were obtained by elaboratio the nitrogen adsorption isotherm (−196 °C) according to the BET and BJH methods measurements were performed in a Micromeritics ASAP 2020 instrument (Norcross, USA), upon reduction in the samples in flowing H2 at 400 °C, being then passivated °C in a 2% O2/He flow.
The sample structure was analysed by a Philips X-Pert diffractometer (Amsterdam Netherlands), operating with a Ni β-filtered Cu Kα radiation (λ = 1.5406 Å ) at 40 kV 30 mA and a scan step of 0.05° s −1 .
The measurements of reducibility under hydrogen atmosphere (TPR) were formed in a customized micro-plant by using a linear quartz micro-reactor (i.d., 4 mm with a 5.6 vol.% H2/Ar mixture at the flow rate of 60 STP mL/min. The experiments

Catalyst Preparation
A commercial micro-spherical silica ES70Y (SA BET : 295 m 2 /g; PV: 1.6 cm 3 /g) supplied by PQ Corporation was used as the support of bimetallic Cu(x)Fe(y)-SiO 2 catalytic samples. Cu(NO 3 ) 2 ·3H 2 O and Fe(NO 3 ) 3 ·9H 2 O (supplied by Sigma Aldrich) were together dissolved in a suitable amount of water, prior to being impregnated on the silica support, by changing the Cu:Fe atomic ratio from 2:1 to 1:2. Monometallic Cu/SiO 2 and Fe/SiO 2 catalysts were prepared by dry impregnation method starting from the aqueous solutions of the nitrate precursors. After impregnation, the catalysts were dried in air at 110 • C for 12 h and calcined in static air at 400 • C for 4 h.

Characterization
The analytical composition of catalysts was determined by X-ray fluorescence analysis, using a Bruker AXS-S4 Explorer spectrometer (Karlsruhe, Germany), equipped with a rhodium X-ray source (Rh anode and 75 µm Be window), a LiF 220 crystal analyzer and a 0.12 • divergence collimator. Samples were analysed at the solid state, taking into account the emission value of Cu-Kα1 and Fe-Kα1 transitions.
The specific surface area and the pore volume data were obtained by elaboration of the nitrogen adsorption isotherm (−196 • C) according to the BET and BJH methods. The measurements were performed in a Micromeritics ASAP 2020 instrument (Norcross, GA, USA), upon reduction in the samples in flowing H 2 at 400 • C, being then passivated at 25 • C in a 2% O 2 /He flow.
The sample structure was analysed by a Philips X-Pert diffractometer (Amsterdam, The Netherlands), operating with a Ni β-filtered Cu Kα radiation (λ = 1.5406 Å) at 40 kV and 30 mA and a scan step of 0.05 • s −1 .
The measurements of reducibility under hydrogen atmosphere (TPR) were performed in a customized micro-plant by using a linear quartz micro-reactor (i.d., 4 mm) fed with a 5.6 vol.% H 2 /Ar mixture at the flow rate of 60 STP mL/min. The experiments were carried out in the range 25-800 • C, with a heating rate of 12 • C/min. The hydrogen consumption was monitored by a thermal conductivity detector, calibrated by the peak area of a known amount of CuO. TPR data resulted in being very reproducible both in terms of maximum position (±3 • C) and extent of H 2 consumption (±3%).
TEM micro-graphs of the "fresh" and "used" hybrid catalysts were acquired and elaborated by a Philips CM12 instrument equipped with a high-resolution camera. Powdered samples were dispersed in 2-propanol under ultrasound irradiation and the resulting suspension put drop-wise on a holey carbon-coated support grid.
SEM analysis was carried out to study the morphology of the prepared catalysts by using a Philips XL-30-FEG instrument (Eindhoven, The Netherlands). An EDAX analyzer (Oxford, model 6587) was used to determine the atomic composition, with specimens of catalyst samples deposited as powders on pin flat stubs.
Measurements of temperature-programmed desorption of carbon dioxide (CO 2 -TPD) were performed in the experimental setup used for TPR to determine the surface concentrations of base sites. Before CO 2 -TPD experiments, the catalyst samples (~100-200 mg) were reduced in a linear quartz micro-reactor (l., 200 mm; i.d., 4 mm) at atmospheric pressure, by flowing hydrogen (100 STP mL/min) from room temperature to 400 • C (heating rate of 10 • C/min). After an isothermal step of 60 min at 300 • C under hydrogen flow, followed by purging with helium, the samples were saturated in a gas mixture of 20 vol.% CO 2 /He (flow rate of 50 STP mL/min) for 60 min. Then, the samples were cooled down to 100 • C in He flow until a constant baseline level was maintained. The desorption measurements were carried out in a range from 100 to 500 • C, at a heating rate of 10 • C/min, using helium as the carrier flow (50 STP mL/min). The CO 2 (m/z, 44) desorption process was monitored by a Pfeiffer Vacuum, ThermoStar ® quadrupole mass spectrometer (Asslar, Germany) equipped with a heated (150 • C) fast-response inlet capillary system, quantitatively calibrated by known pulses of CO 2 .
The chemical environment of oxygen species was determined by X-ray photoelectron spectroscopy, using a Physical Electronics (PHI) 5800-01 spectrometer. The samples were previously outgassed overnight and then reduced "in situ" at 400 • C in an environmental chamber prior to being entered by a fork into the analysis chamber. A monochromatic Al-Kα X-ray source was used at a power of 350 W. Spectra were obtained with pass energies of 11.75 eV. The Ag 3d5/2 peak of an argentum foil was taken after 30 min of argon sputtering for checking the calibration of the binding energy (BE) scale. XPS data have been interpreted by using the online library of oxidation states implemented in the PHI MULTIPAK 6.1 software and the PHI Handbook of X-ray photoelectron spectroscopy. The spectra have been deconvoluted by using a linear combination of three Gaussian-Lorentzian model functions, superimposed to a Shirley background, as exemplified by the overlapping of three oxygen species associated with oxygen lattice (O L ) at 532 eV, oxygen vacancies (O V ) at 534 eV and oxygen chemisorbed (O C ) at 535 eV, respectively, with FWHM less than 2 eV.

Testing
The catalytic measurements were carried out in a fixed bed stainless steel reactor (i.d., 4 mm; l., 200 mm) jacketed within a stainless steel rod (o.d., 14 mm; l., 180 mm) to maintain an effective control of temperature during the run. The catalytic data were achieved at 30 bar, in a range of temperature between 200 and 260 • C, with a feed mixture CO 2 /H 2 /N 2 at a volumetric ratio of 3/9/1, operated at a space velocity of 8800 NL/kg cat /h. For each test, the 40-70 mesh fraction was chosen as a suitable compromise between the pressure drop along the reactor and the mechanical resistance of catalyst particle. All the catalysts were pre-reduced in situ at 400 • C for 1 h under a "pure" hydrogen flow at atmospheric pressure. The composition of the reaction stream was online, analysed by means of a gas chromatograph connected to the reactor outlet by means of two lines heated to 150 • C and equipped with two sampling valves and two injection systems connected to two detectors: TCD (thermoconductivity detector) and FID (flame ionization detector), for the analysis of permanent gases and carbon compounds (methane, methanol, . . . ), respectively. The separation systems consisted of a molecular sieve column in series with a Porapak Q column for permanent gas analysis and a HayeSep ® packed column for methanol and methane analysis.
In addition, the analytical system included the use of a shear column for water, with associated flow reversal. Chromatographic analysis was conducted under isothermal conditions at 80 • C. Both internal standard and mass-balance methods were adopted for the calculation of conversion-selectivity data, with an accuracy of ±3% and carbon balance close to 100%.

Conclusions
A silica sample, characterized by a uniform distribution of regular spheres, was used as support either for mono-metallic Cu and Fe samples or for bi-metallic Cu-Fe systems at variable Cu/Fe atomic ratio, finally allowing for the generation of a metallic phase with particle size between 13 and 19 nm.
TPR measurements emphasized how the Cu/Fe ratio controls the synergy between the metal oxides particles, significantly affecting the reduction kinetics in each sample. Two main CO 2 adsorption sites were identified by TPD measurements, corresponding to metal-oxygen pairs (weak-medium sites) and oxygen vacancies (medium-strong sites), respectively. A clear relationship between the CO 2 adsorption capacity and the relative concentration of oxygen vacancies highlighted that the adsorbed hydrogen can generate oxygen vacancies on the sample surface or induce their migration from inside the bulk oxide to the catalyst surface, as prompted on the systems where an Fe enrichment on the silica support was realized.
The catalytic activity measured at 30 bar in the temperature range 200-260 • C during the hydrogenation of CO 2 to methanol showed, for all the samples, a progressive increase in the CO 2 conversion with temperature, while methanol selectivity exhibited a clear drop, offset by a regular rise in CO and CH 4 selectivity. On the whole, the catalytic results put clearly in evidence the beneficial cooperative effect between copper and iron atoms in driving the production of methanol, the highest methanol yield of 4.1% being achieved on the Cu(3)Fe(7)-Si sample at a maximum CO 2 conversion of 11.6%. The rationalization of the catalytic disclosed how a large number of oxygen vacancies is a necessary but not sufficient condition to lower the energy barrier associated with the cleavage of C=O bonds and then to activate CO 2 under the adopted conditions. Anyhow, an optimal relative concentration of iron atoms resulted in being crucial for maximizing the rate of CO 2 conversion, addressing a specific mechanism of methanol formation via rWGS, contrarily to the formate route mainly suggested for Cu-based catalyst compositions under CO 2 hydrogenation conditions. Without evidence for a catalytic deactivation, due to coke formation or metal sintering, the initial loss of activity during the stability test was put in relation to the formation of an iron carbide phase during reaction, mirroring an incipient surface reconstruction. The appearance of a spinel CuFe 2 O 4 phase on the "used" Cu(5)Fe(5)-Si sample clearly demonstrated that the catalyst composition controls the formation of mixed phases (i.e., spinellike), so determining a specific surface coverage of adsorbed H species and affecting the final product selectivity.