Infrared Fingerprints of the CO2 Conversion into Methanol at Cu(s)/ZrO2(s): An Experimental and Theoretical Study

The methanol economy is an attractive approach to tackle the current concerns over the depletion of natural resources and the global warming intrinsically associated with the use of fossil fuels. This can be achieved by hydrogenation of carbon dioxide to produce methanol, a liquid fuel with potential use in civil transportation. In this study, we aim to pinpoint the intermediates that are involved in the catalytic CO2 conversion into methanol on pure zirconia (ZrO2), Cu and Cu/ZrO2 systems. To accomplish this, we make use of infrared (IR) spectroscopy measurements and quantum chemical simulations within the hybrid density functional theory (DFT) framework. At 250 °C and p~30 bar, the main species formed on the partially hydroxylated ZrO2 is bidentate formate, whereas the co‐production of bicarbonate is relevant upon cooling to T=25 °C. On pure Cu, the IR fingerprints of methanol and carbon dioxide indicate their presence in the gas phase and surface environment, albeit formate/formic acid and methoxy species are also detected at these experimental conditions. The production of methanol on Cu/ZrO2 is mostly dependent on the Cu catalyst, but the higher amount of the methoxy intermediate can be correlated with the consumption of formate adsorbed on ZrO2 or at the Cu/ZrO2 interface. On the Cu/ZrO2 mixture, the reaction mechanism is likely to involve formate as the main intermediate, instead of CO which would result from the reverse water‐gas shift reaction. Ultimately, the higher activity shown by the Cu/ZrO2 mixture might be associated with the extra‐production of methoxy/methanol catalyzed by ZrO2 in the presence of Cu.


Introduction
[3][4] Electric vehicles, for example, could reduce the harmful emissions of carbon dioxide and other pollutants in major cities, [5] but the climate benefit is hampered by the need to recharge the vehicle batteries in a sustainable manner.In fact, the "Statistical Review of World Energy" reported that 3/5 of the power generation relied on fossil fuels in 2021, whereas wind and solar renewable sources represented only 13 % of the world's total energy production in the same year. [6]Additionally, the emission of greenhouse gases during the large-scale production of lithium-ion batteries and other components, [2] and the low availability of extractable amounts of metals, such as lithium and cobalt, are relevant points to be considered from an environmental perspective. [7] 2005, George A. Olah, Alain Goeppert and G. K. Surya Prakash introduced the concept of a "methanol economy" supported by the use of alcohols of low-carbon chains as an alternative energy source. [4,8]Methanol is an excellent fuel and additive for combustion engines because of its high volumetric density, and it is a relevant precursor for fuel production or other materials in the chemical industry. [4,9]Its liquid nature is an additional advantage over hydrogen fuel since the storage and transportation are facilitated by using the currently available system.The synthetic route to methanol requires external power to promote the hydrogenation of carbon dioxide, particularly for the hydrogen generation from water splitting.Similar to the lithium-ion battery technology, the power source should then be renewable to provide a climatefriendly approach.In Iceland, the "George Olah Renewable Methanol Plant", for example, produces 5 million liters of methanol per year by making indirect use of geothermal energy for hydrogen generation, whereas the carbon-based source is removed from the flue gas in the geothermal power plant that is situated nearby. [3,9]In this case, the Cu/ZnO/Al 2 O 3 system is reportedly the active catalyst utilized for CO 2 conversion into methanol at an industrial scale. [10]he field of development of new catalysts that could enhance the product yield, selectivity and process cost is therefore promising for future applications. [11]In this context, Cu-based materials have been extensively investigated for the thermocatalytic CO 2 conversion into methanol throughout the years. [12,13]In 1945, unsupported copper was initially found inactive at T = 273 °C and p = 412 atm by Ipatieff & Monroe, [14] but later studies demonstrated that Cu(100), Cu(111), Cu(110) and polycrystalline Cu in fact exhibit catalytic activity in presence of CO 2 /H 2 mixture at p � 5 bar. [12,13,15,16]In another work, similar turnover frequencies for methanol production were verified in pure Cu compared to the Cu/SiO 2 system at T = 413 K, thus suggesting that the oxide support had a minor effect on the catalytic efficiency. [17]][19] Two main pathways were proposed to shed light on this point, either involving the formate species or CO as the primary precursor.In 2010, DFT calculations performed by Yang and collaborators [12] revealed that the hydrogenation of adsorbed CO is indeed more favorable from a kinetic point of view, but this route is hindered by the instability of the CO* intermediate.The formate pathway was consequently proposed as the most feasible route to explain this reaction, whereas CO would accumulate as a byproduct. [12]Wu et al. determined the correlation between the consumption of formate species and the increased amount of methoxy produced in ZrO 2 /Cu inverse catalyst with 10 % of ZrO 2 at T = 220 °C, therefore providing evidence for the formate pathway. [18]Nonetheless, the transient surface reaction experiments performed on Cu and Cu/SiO 2 by Yang et al. indicated that the adsorption of formic acidpresumably in a dissociative way to originate formate -does not lead to any methanol yield in different conditions.Based on this information, formate was considered a spectator in the hypothesis that the process does not depend on another coadsorbed species that is absent in the experiments with formic acid. [17]Thus, the uncertainty of the methanol precursor suggests that the real active species has not been identified yet. [17]he oxide support on the Cu/ZrO 2 system has a promotion role leading to an improved catalytic efficiency and selectivity towards methanol production in comparison with pure Cu. [20,21] In fact, pure ZrO 2 is unable to catalyze the methanol synthesis from CO 2 hydrogenation, instead producing stable formate, bicarbonate and/or carbonate adsorbates. [21][24] It has been reported that the monoclinic polymorph has a higher concentration of oxygen defects, which was associated with higher efficiency of the Cu/ZrO 2 system to produce methanol from CO and H 2 in comparison to tetragonal ZrO 2 as the support. [25]In this case, the insertion of CO onto the hydroxyl groups results in formate formation, and this intermediate is considered essential to generate methoxy and, ultimately, methanol. [24,25]It is also valid to mention that the surface of monoclinic ZrO 2 was associated with a higher amount of available adsorbed hydrogen. [25]imilar to pure Cu, the most relevant routes to explain the CO 2 hydrogenation on Cu/ZrO 2 system consist of either CO or formate as the main intermediates.In their work, Kattel et al. [19] evaluated the IR fingerprints of formate and methoxy on Cu/ ZrO 2 and Cu/TiO 2 and reported that the rates of consumption of formate are not correlated with methoxy production.The oxide support was further assumed to modify the reaction mechanism from formate on pure Cu to the CO pathway at the interface. [19]ey also verified a better performance for the catalytic system with the use of ZrO 2 support in comparison with TiO 2 . [19]In another work, IR and nuclear magnetic resonance (NMR) spectroscopy measurements were performed by Larmier et al. [21] indicated that 60 % of formate species were effectively converted into methoxy at the catalytic interface, also turning the discussion in favor of the formate pathway. [21]n this paper, we use IR spectroscopy and DFT methods to pinpoint the most feasible intermediates involved in the methanol production on pure ZrO 2 , Cu and Cu/ZrO 2 mixture at T = 250 °C and p~30 bar.The outcomes indicate that i) the adsorbed formate species is the most relevant product on pure ZrO 2 and remains stable upon evacuation at lower temperature and pressure conditions; ii) the formate species is involved in the methoxy production on Cu/ZrO 2 mixture; iii) Cu plays the major role in methanol production on Cu/ZrO 2 .Therefore, the formate pathway might be herein the most relevant leading to methanol production on Cu/ZrO 2 , which exhibits higher thermocatalytic activity associated with the involvement of ZrO 2 or the Cu/ZrO 2 interface in the methoxy production.

Heat treatment
As a first step, the air exposure of Cu-and Zr-based oxides required heat treatment to remove the water molecules incorporated into the sample.As discussed in detail in Section S1 in the Supporting Information, the dissociation of water onto the ZrO 2 surface leads to its partial hydroxylation at T � 400 °C. [26]9][30]

ZrO 2 (s) interacting with H 2 and CO 2
The experiment started at T = 250 °C in the presence of H 2 (see curve 1-in black), and proceeded with a stepwise addition of CO 2 at increased pressure (curves 2-3).The system was left unchanged for 160 min.(curve 4), followed by a decrease in temperature to T = 25 °C (curve 5).In the last curve, the outlet was opened for gas ejection.
The spectra were measured at different times during the reaction process; the resulting curves 1-6 are displayed for three separated infrared regions in Figure 1.The identified peaks are marked in the IR spectra and assigned to a mode whenever it is possible (see Table 1), whereas the relevant adsorbates on the ZrO 2 surface are illustrated in Figure 2.
The next subsections detail the mode assignment to different species: (bi)carbonates (CO 3 2À and HCO 3 À ), formate (HCOO À ) and methoxy/methanol (CH 3 O À and CH 3 OH) are the most relevant to this study, whereas the gaseous species are byproducts.
CH 4 and CO.In Figure 1 (a), the IR spectra show that small amounts of the gaseous species CH 4 are produced during the experiment, as the CÀ H stretching mode is present at 3015 cm À 1 (ṽ calc = 3015 cm À 1 ) until the chamber is opened.
Similarly, CO(g) production is identified by the peak at 2108 cm À 1 , and ṽcalc = 2110 cm À 1 , arising from its P rotational branch. [40,41]This species exhibits a dual peak with similar intensity at ~2100-2202 cm À 1 , [24,32,39,42] thus implying that the high-frequency band is out of range in Figure 1(b).As pointed out in Ref. [43-46], the most viable route to form CO on ZrO 2 consists of CO 2 adsorption at an oxygen vacancy, a step that is followed by the oxygen transferring to fill the vacant site.Upon formation on ZrO 2 surface, the primary interaction of CO occurs through the carbon atom (see Figure 2 (a)) with Zr 4 + centers, which has a CÀ O stretching mode located at 2129 cm À 1 and ṽcalc = 2167 cm À 1 .Therefore, this adsorption mode is shifted upwards in comparison with CO(g), a trend that was associated with the CO polarization after adsorption by Kondo et al. [47] Due to the acidity of Zr cations, this interaction is described as a σbond with a coordinate covalent nature. [42]This bond has a low binding energy of À 50.3 kJ/mol, and d(CÀ Zr) = 2.62 Å, as expected for 5 th -period complexes with CO ligand coordination, [48] with slight transfer of electric charge verified in the populational analysis.Maleki et al. determined similar values for the adsorption energy E ads (CÀ Zr) using ZrO 2 nanoparticles of larger size within the GGA + U/1-1-1 level of theory. [38]he right-shifted peak at 2077 cm À 1 (ṽ calc = 2080 cm À 1 ) is first attributed to loosely adsorbed CO molecules interacting with Zr cations via oxygen atoms, with d O-Zr = 2.58 Å, to nucleate the optimized structure in Figure 2 (b).With E ads = À 28.0 kJ/mol, the favorable OÀ Zr interaction weakens the CÀ O triple bond from CO(g), leading ν(CO) towards lower wavenumbers.The larger peak indicates that this is the preferential mode for CO physisorption on ZrO 2 surface, thus contradicting the thermody-Figure 1. IR spectra of ZrO 2 (s) in presence of H 2 (curve 1) and H 2 /CO 2 (curves 2-6) measured at several temperature and pressure conditions.In a), it is shown the CÀ H stretching region, 3400-2750 cm À 1 ; in b), the CO stretching region for CO, 2145-2035 cm À 1 is shown and, finally, in c) the "fingerprint" region is displayed, 1800-900 cm À 1 .The time for measurement is specified for each curve in the legend box, where the temperature of curve 5 was maintained at 25  In each structure, the nomenclature makes use of the atom connectivity (in parenthesis), the number of dentates or binding mode (bridged or tribridged) and the name of the species itself.The species described in (e) and (i) have poor agreement with the experiment and are discarded in this work as feasible configurations.

Table 1. IR data for H
2 and CO 2 interacting with ZrO 2 .The experimental and theoretical data from this work are presented in the first five columns, alongside the band assignment, whereas the reported values are described in the last three columns.The mode assignment from the literature refers to the same adsorbed species from the present study when it is not indicated otherwise.Notes:1) ν and δ stand for stretching and bending vibrational modes, respectively.2) The value indicated by A + B = C represents a combination band.3) The boldface abbreviations stand for monodentate (m.), bidentate (b.), tridentate (tr.), bibridged (br.) and tribridged (tribr.).4) In the third column, each peak has intensity described as weak (w), medium (m) or strong (s) at 250 and 25 °C (first and second part of the experiment, respectively), whereas the arrow indicates whether the peak becomes stronger/weaker throughout the experiment.The dashed line indicates that the peak did not exist at the specified temperature, or it is not easily distinguishable.
In Figure 1 (b)-curve 6, all peaks have totally or mostly disappeared when the chamber is opened due to the weak interactions or the presence of gaseous species.
CO 2 adsorption.The associative adsorption of CO 2 on ZrO 2 is usually an activated process leading to the formation of a carbonate-based species. [22,23]However, the presence of oxygen vacancies can facilitate molecular adsorption, where the adsorbate oxygen atoms bind to different Zr sites distanced by 2.03-2.07Å to form a bridged configuration (see Figure 2 (c).With E ads = À 146.1 kJ/mol, and H ads = À 146.7 kJ/mol at standard conditions, this interaction is considered strong, and it is accompanied by a change in the OCO angle from the original linear geometry to 110.2°.
The fingerprints of this species are found at ṽcalc = 962 and 1143 cm À 1 , frequencies that correspond to the CO and COO symmetric stretching, respectively.Table 1 shows that these modes are respectively assigned to the peaks centered at 1007 and 1137 cm À 1 .From this point, bridged CO 2 could undergo decomposition to fill in the oxygen vacancy and form physisorbed CO.The activation barrier of 52 kJ/mol (B3LYP/6-31G (d,p) level of theory) is likely to hinder the process at room temperature. [43]However, at T = 250 °C, the kinetic simulations performed by Kalered et al. in Ref n°[ [43] ]. indicate that the decomposition is completed in just microseconds.This explains why the aforementioned peaks only exist at low temperatures (T = 25 °C), since the adsorbate (CO 2 ) would need a higher energy amount to proceed towards CO formation.HCO 3 À .Bicarbonate is a frequent intermediate/byproduct species during the CO 2 conversion catalyzed by metal oxides, [39,49] that usually originates from the CO x reactant insertion into the À OH bond from hydroxylated surfaces in an activated process. [20,22,24,34,39,49,50]Therefore, it is important to determine the availability of this species during the CO 2 hydrogenation on ZrO 2 and Cu/ZrO 2 catalytic systems under the experimental conditions of interest.
The HCO 3 À adsorbate interacts with Zr sites through oxygen atoms to form a bidentate or bibridged configuration, which are distinguished by the presence of a chelating interaction with the surface in the bibridged arrangement.Figure 2(d) and S4 illustrate different possibilities for these configurations, although the bidentate species in Figure 2(d), and Figure S4(4), is the most thermodynamically stable in this work.Nonetheless, the small difference of 19.7 kJ/mol also makes the configuration (3) plausible in this experiment.
In Figure 1(c), the peak located at 1638 cm À 1 (ṽ calc = 1591 cm À 1 ) has sole contribution from the coupled mode between the COO antisymmetric stretching and the OH bending in HCO 3 À .For this reason, this mode allows the identification of bicarbonate species only at T = 25 °C (see curves 5 and 6), [22,23] since its stability decreases with the temperature. [23]Korhonen et al. previously reported that this species is stable up to ~175 °C. [22]The symmetric stretching of COO group in HCO 3 À , with the highest contribution from the  983-1010 [22] δ(CH) out [22]   Footnotes: a-If present, it is likely to be in small quantities.*-Minor contributions.
CO stretching, appears at 1454 cm À 1 (ṽ calc = 1408 cm À 1 ).In this case, this peak is also verified at high temperatures because it contains contributions from tribridged carbonate, as shown in Table 1.Finally, ṽ = 1230 cm À 1 (ṽ calc = 1213 cm À 1 ) corresponds to the OH bending mode with slight contribution from the COO antisymmetric stretching, which is more evident at the conditions where HCO 3 À is present, i. e. in curves 5-6.CO 3 2À .Carbonates are common intermediates upon activated CO 2 adsorption on ZrO 2 surface at room temperature, typically adopting mono-and bidentate configurations in monoclinic ZrO 2 . [23,50]Based on DFT calculations, the tribridged arrangements were also found feasible by Korhonen et al. [22] On the other hand, Pokrovski and collaborators verified preferential bi-and tribridged configurations on tetragonal ZrO 2 . [23]he herein evaluated adsorption modes for the carbonate species correspond to i) bidentate, ii) bibridged and iii) tribridged configurations, as displayed in Figure S5.Initially, the activated adsorption was considered on a dehydroxylated surface, but the interaction with hydroxyl groups (from water) was later added to improve the experimental-theoretical agreement.The monodentate species was not found stable in this work, possibly due to the open coordination of Zr sites.At 0 K, the bidentate mode has E ads = À 157.4 kJ/mol, which is 29.3 and 1.5 kJ/mol more favorable than ii) and iii), respectively.For matters of comparison, previous calculations on ZrO 2 (À 111) found E ads = À 31.0 kJ/mol, a difference that might be driven by the methodology or hydroxylation effect. [22]revious reports have verified two experimental peaks at 1550-1575 cm À 1 and 1310-1335 cm À 1 resulting from different CO group stretching modes in the bidentate carbonate species. [22,23,34]Nonetheless, the bidentate species from Figure S5 (1) and (2) do not explain these peaks, instead displaying frequency modes at 1733-1789 cm À 1 , from ν(CO) stretching, that are not present in our IR spectrum.For the monodentate species, the IR fingerprint at 1490 cm À 1 ( [50]) is also absent here.These facts suggest that monodentate and bidentate carbonate species are not present in this experiment, especially considering that their stability would enable their identification.The details supporting this analysis are available in Table S3.
Another possibility is that these species have an underlying influence of hydrogen bonding from hydroxyl groups, water molecules or even bicarbonate species.For instance, Korhonen and collaborators performed DFT calculations on ZrO 2 (À 111), [22] still finding a largely overestimated ṽcalc = 1667-1791 cm À 1 , but the approach slightly improved the agreement in the first case, as the carbonate could interact with OH and other O sites and weaken the C=O bond responsible for the high frequency.
New structures were then prepared considering the hydroxylation of ZrO 2 surface.In the case of the bidentate species, the stretching CÀ O mode (ν(CO)) coupled with H 2 O bending is seen at ṽcalc = 1569-1605 cm À 1 , with antisymmetric and symmetric counterparts exhibiting calculated frequencies of 1200 and 1044 cm À 1 , respectively.For the bibridged configuration, ν(CO) is slightly shifted to 1534-1683 cm À 1 , whereas the other modes remain almost unchanged.The better agreement with the experiment (see the possible peak assignments in Table S3) makes the bidentate carbonate species the most likely among the tested configurations.However, the stabilization of monoor bidentate carbonates is possibly hindered by the formation of HCOO À , as discussed below.
The tribridged configuration displayed in Figure 2(g) is thermodynamically plausible due to the similar adsorption energy to bidentate carbonate.That species has fingerprints at 1454 cm À 1 (ṽ calc = 1452 cm À 1 ) resulting from CO stretching, whereas the antisymmetric and symmetric stretching modes appear at 1230 cm À 1 (ṽ calc = 1255 cm À 1 ) and 1052, 1044 cm À 1 ( ṽcalc = 1033 cm À 1 ).Therefore, this structure presents better agreement with the experimental outcomes, although both structures are thermodynamically favorable.According to Pokrovski et al., [23] the tribridged carbonate could decompose at higher temperatures, which would partly explain the decrease in intensity of the correlated peaks.
Figure 1a-c shows that the aforementioned peaks remain after the chamber is opened, thus suggesting that small amounts of tribridged carbonates, as well as CO 2 , are chemisorbed in a thermodynamic stable fashion at room temperature on ZrO 2 surface.
HCOO À .The primary route for HCOO À formation on ZrO 2 consists of a reaction between gaseous CO and terminal OH groups. [22,24,34,39]In this case, the formation of this species would then depend on the amount produced of CO(g) as an intermediate.Another possibility is that formate is generated from carbonate or bicarbonate during the process. [51][24] As further explained, HCOO À is indeed formed at T = 250 °C and stabilized during the cooling process to room temperature.
The most stable configurations found for HCOO À are shown in Figure 2(h-i), differing by the shared Zr center in both ZrÀ O bonds in Figure 2 (i).From a thermodynamic viewpoint, both configurations could be present during the experiment, since the configuration (i) is only 19 kJ/mol more stable than (h).Similar to carbonates, the monodentate configuration was not found stable in our calculations.In another work, DFT calculations actually predicted the existence of this structure on ZrO 2 (À 111), although displaying much lower stability than the bidentate configuration. [22]n Figure 1 (c), an important IR fingerprint from formate species is identified at 1564 and 1575 cm À 1 , where ṽcalc = 1523-1533 cm À 1 , mainly arising from the antisymmetric stretching mode of COO in this species.These values are in good agreement with previous experimental reports of 1560-1610 cm À 1 , [24] and theoretical predictions in the range of 1534-1561 cm À 1 . [22,33]It is also noticeable in Figure 1 (b) that the intensity of the initial peak decreases in curves 5 and 6, possibly due to decomposition.
Other IR fingerprints from the HCOO À species include: i) the stretching mode of CH at 2885-2970 cm À 1 ; and ii) the COO symmetric stretching, which is present as a sole peak at < 1390 cm À 1 , a combination mode (ν s + ν as ) at 2885-2970 cm À 1 and coupled to the in-plane bending mode of CH at 2738-2742 cm À 1 .The out-of-plane bending mode of CH has almost negligible contribution to the peak localized at 1007 cm À 1 and ṽcalc = 1043 cm À 1 .In Table 1, it is clear that the theoretical calculations predict the experimental measurements well for the bidentate species, and clarify the details of the coupled modes.On the other hand, the chelating species in Figure 2 (i) displays δ(CH) at ṽcalc = 1252 cm À 1 , and is in poor agreement with the experimental assessment.This suggests that the bidentate species would be more frequent during this experiment.
CH 3 OH and CH 3 O À .Guglielminotti reported the formation of methoxy groups (CH 3 O À ) on ZrO 2 from an H 2 /CO mixture through the peaks located at 2837, 2936 and 1150 cm À 1 in the range T = 423-573 K. [34] Later on, the distinction between the mono-and bidentate CH 3 O À was established by the lowfrequencies at 1149-1154 and 1047-1052 cm À 1 , respectively. [19,52]To complete the identification, the CÀ H stretching mode should be present as a low-intensity peak at 1460 cm À 1 . [19]In another report, the presence of Cu was found necessary to enable the methoxy production on ZrO 2 . [21]he peak centered at 1137 cm À 1 in curves 5 and 6 (see Figure 1 (c)) could be assigned to CO stretching, with values reported at 1149-1154 cm À 1 .On Cu/ZrO 2 , [19] the formation of methoxy was verified after 5 min. of reaction time, whereas 20 min.were necessary for its formation on pure ZrO 2 using an H 2 /CO mixture. [34]Therefore, the peak at 1137 cm À 1 cannot be unambiguously assigned to the methoxy species, whereas the CH 3 stretching mode from CH 3 O* is not seen at 2820-2837 cm À 1 . [19,34]In accordance with Larmier et al., [21] the CO 2 hydrogenation does not lead to methoxy formation on pure zirconia, instead the presence of metallic Cu is required.Similarly, CH 3 OH(g) fingerprints are not identified in the IR spectrum, namely at ñ = 2977, 2844, 1340 and 1034 cm À 1 . [53]astly, the presence of CH 3 OH in adsorbed form was evaluated (see Figure 5 (j)).With E ads = À 156.0 kJ/mol, this adsorbate could strongly interact with ZrO 2 through the oxygen atom acting as a Lewis base in contact with Zr acidic centers.Table 1 shows that all the fingerprints are present in the spectrum but cannot be unambiguously assigned to this species, i. e. 2955, 1455, 1376 and 1052 cm À 1 .In either case, the nearly constant amount of CO and C=O species (formate or carbonate) suggests that the production of bound CH 3 OH on ZrO 2 would be very low, if existent.The assessment of the thermodynamics of this process could help to clarify this point, but it is out of the scope of this study.
Intermediate Conclusions.At 250 °C, the most relevant species on ZrO 2 in an H 2 /CO 2 environment is formate (HCOO À ).The initial peak at 1564 cm À 1 is slightly changed with the reduction of temperature, but this species remains stable until the end of the experiment.From DFT calculations, this peak does not have contributions from bidentate and bibridged carbonates, indicating that these species would quickly react upon formation.On the other hand, tribridged carbonates could be stabilized in small quantities.The production of bicarbonate (HCO 3 À ) is relevant at lower temperatures, whereas methoxy/methanol product is not clearly identified.Based on this information, we can affirm that pure ZrO 2 is not able to catalyze the carbon dioxide conversion into methanol under these conditions, instead producing predominantly formate.

Cu(s) Interacting with H 2 and CO 2
Here we have some differences compared to the experiment with ZrO 2 powder: i) additional data was recorded (curves 3 to 8) during the dwell time at a constant temperature, i. e. T = 250 °C; ii) the measurement in which half of the CO 2 (g) content is injected was skipped, and iii) shorter waiting times were applied to generate the curves 10 and 11.In Figure 3, the peaks were assigned to a frequency mode in accordance with the optimized structures from Figure 4, as well as gas phase molecules or ions that are not shown for brevity.The large broad peak at 1173 cm À 1 is due to the equipment and is not considered further.
CH 4 .The CÀ H stretching mode from this species is localized at 3015 cm À 1 (see Figure 3 (a)) and ñcalc = 3025 cm À 1 .This peak exhibits oscillations from ~3175-3040 cm À 1 and 3000-2850 cm À 1 corresponding to the combination of vibrational and rotational modes of CH 4 (g).Here, the large peak points out for a more substantial CH 4 (g) production compared to the use of ZrO 2 catalyst.Similarly, the large peak at 1304 cm À 1 (ñ calc = 1297 cm À 1 ) is attributed to bending modes of CH 4 (g).As indicated by curve 11, small amounts of CH 4 still remain inside the chamber after it is opened, albeit it is unlikely to be bound to the catalyst surface.
CO.The typical C=O stretching peaks from CO(g) are seen at 2173 and 2110 cm À 1 (ṽ calc = 2144 cm À 1 ) in Figure 3 (b).At T = 250 °C, these peaks have an increasing intensity with exposure time as a result of the reaction progress.In this case, the preference for remaining at the gas phase suggests that the adsorbed state is unfavorable or has low binding energy.At 25 °C (see curve 10), CO(g) production is clearly favored by cooling and a drop in pressure.
Loosely bound CO can be divided into C-and O-bound species to Cu sites, as depicted in Figure 4 (a) and (b).In the former case, on-top and hollow positions are found stabilized upon optimization, although the former is only 9.42 kJ/mol more favorable, with E ads = À 71.0 kJ/mol.Other studies have estimated the on-top adsorption energy of CO on Cu(100), Cu(110) and Cu(111) between À 27 and À 66 kJ/mol based on DFT methodology. [60]The IR fingerprint for the on-top site at ñcalc = 2058 cm À 1 , whereas ñcalc = 1821 cm À 1 for the hollow site, points out for an on-top preferential adsorption.According to the literature, that fingerprint should be assigned to ñ = 2094 and 2078 cm À 1 in Figure 3 (b).Nonetheless, the peak at 2057 cm À 1 has a better theory-experiment agreement and contains contributions from this mode.Similar behavior to CO(g) is seen for this case, with production accentuated with exposure time and upon cooling.
For the O-bound species, the positive value of E ads = + 31.2 kJ/mol indicates a thermodynamically unfavorable local minimum adsorption.For this reason, its attachment requires that energy is provided to the system, apparently a condition that could be satisfied at room temperature.As expected, the C=O bond is stronger, leading to a blue-shift in its stretching mode to ñcalc = 2103 cm À 1 , which is also assigned to the peaks located at 2094 and 2078 cm À 1 .Most of the CO species appear to leave the chamber upon opening, as depicted in curve 11, since it is present in the gas phase and weakly adsorbed to Cu in prior steps.
HCOO À and HCOOH.Upon adsorption on the Cu surface, HCOO À is likely to adopt the bibridged configuration in Figure 4 (c), since the monodentate structure is not stable on the clean surface.The calculated frequencies for this species consist of 2934, 1537, 1363, 1268 and 1033 cm À 1 , but not all modes can be clearly identified in the IR spectrum.For instance, the first ñcalc overlaps with the combined vibrational and rotational modes from CH 4 (g), whereas the C=O symmetric stretching, at ñcalc = 1537 cm À 1 , is not resolved from other peaks.Nonetheless, the frequency arising from CÀ H in-plane bending mode (δ(CH), ñcalc = 1363 cm À 1 ) is assigned to the peaks centered at 1354 and 1340 cm À 1 , therefore allowing the identification of the formate species adsorbed on Cu. [18] To complement the assessment, the mode located at ñcalc = 1268 cm À 1 is assigned to the peaks at 1269 and 1261 cm À 1 , and corresponds to the symmetric stretching of the COO group.
The production of HCOOH is first considered in the gas phase, but the low E b of À 18.5 and + 3.9 kJ/mol for the monodentate species, depicted in Figure 4 (d) and (e), indicates that the physisorption would be feasible upon formic acid production.For these species, the high-frequency modes at ñcalc = 3551-3586 cm À 1 , 2916-2953 cm À 1 and 1718-1809 cm À 1 arising from the stretching modes of OH, CH and CO bonds, respectively, cannot be clearly distinguished in the IR spectra.The other modes calculated at 1373-1381 cm À 1 , 1147-1216 cm À 1 and 994-1037 cm À 1 are properly assigned in Table 2.In particular, δ(CH) coupled to δ(OH) holds an excellent agreement with the IR experiment, where ñ = 1373 cm À 1 , and evidences the formic acid production in this work.Here, it is noticeable that the higher frequency for ν(CO) in comparison with HCOO À , as a result of the stronger CO-bond in HCOOH, and therefore, a weaker connection to Cu.Most of these peaks vanish when the chamber is opened, which implies that formate groups and formic acid do not stick to the Cu surface as hard as they did to the ZrO 2 surface.
CH 3 OH and CH 3 O À .The production of gaseous methanol at T = 250 °C is evidenced by the IR fingerprints located at 1340-1331 cm À 1 (ñ calc = 1346 cm À 1 ) and 1261-1269 cm À 1 (ñ calc = 1260 cm À 1 ) in Figure 3 -curves 2-9, which are attributed to the in-plane δ(OH) and rocking δ(CH 3 ) modes, respectively.The antiand symmetric δ(CH 3 ) modes estimated at 1455-1477 cm À 1 cannot be discriminated from other modes in the IR spectrum, but their presence will be clarified when the Cu/ZrO 2 mixture is analyzed below.As these fingerprints remain intact at room temperature (in curve 10), we can infer that 10 min. of time exposure at this temperature is not enough to condensate CH 3 OH(g).
Additional peaks corresponding to the R-and Q-branches of ν(CO) are located at 1033-1050 cm À 1 (ñ calc = 1034-1060 cm À 1 ).The original work by Falk & Whalley [53] revealed that the combination mode at 2078 and 2056 cm À 1 is also visible for this species, probably arising from the CO stretching multiplied by 2. These frequencies, alongside adsorbed CO species, are assigned to the peaks located at 2078 and 2055 cm À 1 (ñ calc = 2068 cm À 1 ) in Table 2.
With E b = À 27.5 kJ/mol, CH 3 OH could also be found in the adsorbed form (see Figure 4 (e)).In this case, the interaction with Cu sites occurs via an oxygen atom, thus weakening its internal bonds in CH 3 OH.As a result, δ(OH) has a red-shifted frequency in CH 3 OH(ads) calculated as 1314 cm À 1 , whereas ν(CO) is shifted to 990 cm À 1 and can possibly be assigned to ñ = 976 cm À 1 .
In either case, i. e., in adsorbed form or gas phase, the removal of this species can be achieved by opening the chamber, with the most cited fingerprints disappearing from the IR spectrum (see Figure 3-curve 11).The only exception is ν(CO), whose peak has decreased but remains with contributions from other stabilized species.
The presence of methoxy (CH 3 O À ) is identified through the mode assignment of ñcalc = 1140 and 1001 cm À 1 to the experimental peaks at ñ = 1135 and 1033-1050 cm À 1 , respectively, that arise from the CH 3 group (δ(CH 3 ) twist ) and CÀ O bond.The stability of these fingerprints is consistent with the chemisorption that characterizes this surface-adsorbate interaction.Similar to several species, the frequency modes calculated at 1425-1434 cm À 1 and 2874-3037 cm À 1 overlap with other modes.The optimized structure is depicted in Figure 4 (f).
Intermediate Conclusions.On Cu system, the hydrogenation of CO 2 leads to the production of methoxy/methanol, formate/formic acid and gaseous products.Therefore, the selectivity of this material is decreased with favorable CH 4 production.It is also interesting to notice that the IR fingerprints from methoxy/methanol and formate/formic acid desired products tend to disappear when the outlet is open, thus suggesting that these products are mostly gaseous or interact weakly with the catalytic surface.

Cu/ZrO 2 (s) interacting with H 2 and CO 2
As described in the experimental section, a mixture of ZrO 2 and CuO powder was used as a catalyst for CO 2 hydrogenation.This mixture was first exposed to H 2 gas, leading to CuO reduction to metallic copper (Cu), while ZrO 2 remained unaffected.The final catalytic system was, therefore, a mixture of ZrO 2 and Cu(s) interacting through weak forces.Thereafter, the measurements were performed in a similar fashion to the previous experiments to result in the IR spectra in Figure 5, from which the detected peaks are detailed in Table 3. From bottom to top, the curves were measured before CO 2 was inserted into the chamber (curve 1), during the dwell time (2-10), after the cooling process (11) and after the chamber was emptied (12).
Most peaks identifying the gaseous CH 3 OH and its adsorbed form were previously discussed in this work.Nonetheless, the peak located at ~1470 cm À 1 , and extended towards lower frequencies, is now depicted in the IR spectrum from Figure 5 (c).This peak has contributions from the antisymmetric and symmetric ν(CO) modes of CH 3 OH on Cu and ZrO 2 (see Table 3) and completes the identification of this species.In addition, this peak has a contribution from the methoxy species adsorbed on ZrO 2 or Cu, which explains its stability in curve 11.This species has also IR fingerprints lying in the range 1154-1052 cm À 1 , [19,52] Table 2. IR data for H 2 and CO 2 interacting with Cu(s).The experimental and theoretical data from this work are presented in the first four columns, alongside the band assignment, whereas the reported values are described in the last two columns.The mode assignment from the literature refers to the same adsorbed species of the present study when it is not indicated otherwise.Notes:1) ν and δ stand for stretching and bending vibrational modes, respectively.2) The value indicated by A + B = C represents a combination band.3) The boldface abbreviations stand for monodentate (m.), bidentate (b.), tridentate (tr.), bibridged (br.) and tribridged (tribr.).4) In the third column, each peak has intensity described as weak (w), medium (m) or strong (s), whereas the arrow indicates whether the peak becomes stronger/weaker throughout the experiment.
The quantitative analysis of the Cu/ZrO 2 spectrum, in comparison with sole Cu(s), reveals that the fingerprints at 1331-1342 and 1261-1269 cm À 1 have their intensity increased in about ~0.12 a.u.This fact indicates that i) the methanol production is mostly occurring on Cu(s), and ii) the addition of ZrO 2 is responsible for improving the catalytic efficiency of this material.Furthermore, the CO and CH 4 fingerprints are weakened by ~0.1 a.u., thus suggesting that ZrO 2 addition enhances the material selectivity in comparison with the sole Cu catalyst.Such improvement in selectivity against CO production was also found by Kattel et al. by replacing TiO 2 with ZrO 2 as the oxide support for Cu. [19]terestingly, the formation of the methoxy species is accompanied by the decreased intensity of the peak located at 1575-1563 cm À 1 from curve 1.This IR fingerprint corresponds to the ν as (COO) mode from HCOO À in ZrO 2 , with a decrease that is more evident in curve 11 (orange).In this case, the HCOO À decomposition to form bicarbonate is a plausible reaction that also affects the peak intensity, but it is clear that the formate groups become more reactive in the presence of Cu.These outcomes suggest that the decreased amount of formate groups is linked to the production of methoxy and methanol on the ZrO 2 surface in the presence of Cu or Cu/ZrO 2 interface.Such conversion would then be responsible for the increase in catalytic efficiency of the material.
Scheme 1 illustrates the possible reaction routes for the methoxy/methanol formation on Cu/ZrO 2 based on the exper- Here, the measurements from curves 1-8 were taken at p = 24.75bar.In a), it is shown the CÀ H stretching region, 3200-2800 cm À 1 ; in b), the CO stretching region for CO, 2220-2035 cm À 1 and, finally, in c) the "fingerprint" region is displayed in the range 1800-900 cm À 1 .The time for measurement is specified for each curve in the legend box.imental and theoretical assessment in this work.A first assessment of the reaction thermodynamics is shown in Scheme 2, where the Gibbs free energies for the carbon dioxide conversion on pure ZrO 2 are given at T = 25 °C.The low temperature was chosen as the final step of the experiment.
The carbon dioxide adsorption to form CO 3 2À (a 23 , where a xy is a matrix element in Schemes 1 and 2 ) is a barrierless step with Δ r G = À 115.2 kJ/mol.At high temperatures, the reaction with H 2 to form HCOO À (a 34 ) is a downhill step with Δ r G = À 260.2 kJ/mol.This step is expected to have a low kinetic barrier, since the bidentate carbonate species is not verified in the IR spectrum at T = 250 °C, whereas the tridentate carbonate peak is not significant.From this point, the production of CH 3 O À (a 24 ) and CH 3 OH require the presence of Cu to occur, but the assessment of the thermodynamics on pure ZrO 2 indicates that only a low amount of energy is required to accomplish it (< 30 kJ/mol).For this reason, the kinetic barriers should be assessed to provide a better understanding of the role of Cu in this process, but this is out of the scope of the present work.
Since the experimental temperature goes from high (T = 250 °C) to lower (T = 25 °C), the HCOO À present (a 34 ) may gradually convert to HCO 3 À (a 13 ), thus reducing the amount of Table 3. IR data for H 2 and CO 2 interacting with Cu/ZrO 2 system.The experimental and theoretical data from this work are presented in the first four columns, alongside the band assignment, whereas the reported values are described in the last two columns.The mode assignment from the literature refers to the same adsorbed species of the present study when it is not indicated otherwise.Notes:1) ν and δ stand for stretching and bending vibrational modes, respectively.2) The value indicated by A + B = C represents a combination band.
3) The boldface abbreviations stand for monodentate (m.), bidentate (b.), tridentate (tr.), bibridged (br.) and tribridged (tribr.).4) In the third column, each peak has intensity described as weak (w), medium (m) or strong (s), whereas the arrow indicates whether the peak becomes stronger/weaker throughout the experiment.An absent arrow means that was not possible to distinguish the peak throughout the experiment.
The assignment of the CH 3 O À species is given to the stable peaks according to the chemisorption.
formate available for the methoxy production.If instead, the experiment started at a low temperature going to higher, then HCO 3 À would be a plausible intermediate in the process.

Conclusions
In this work, we employed experimental and theoretical methods to evaluate the carbon dioxide hydrogenation into methanol on pure ZrO 2 , Cu and Cu/ZrO 2 mixture at T = 250 °C and p = 29-33 bar.Our outcomes indicate that: i) Pure ZrO 2 might not be able to thermocatalyze the conversion of CO 2 into methoxy/methanol, in agreement with the literature data.Instead, it promotes the formation of formate groups that remain attached to the surface upon evacuation, alongside bicarbonate groups that are produced after cooling the reaction chamber.ii) Cu is suitable to promote methoxy/methanol formation, although CO and CH 4 appear as undesirable byproducts that decrease the overall catalytic efficiency.In this case, it is likely that the process depends on the production of formate species to generate methanol.iii) On Cu/ZrO 2, it is evident that the selectivity for methanol is improved, with a larger amount of this product that is likely associated with Cu.Moreover, we verify a correlation between the amount of formate species on ZrO 2 and methoxy formation, thereby suggesting that formate groups possibly interacting with the material interface are involved in the catalytic process.Naturally, the catalytic activity rises for Cu/ZrO 2 in comparison with the pure catalysts, although the formate pathway is also likely to prevail in this case.

IR spectroscopy measurements
The experiment was initially performed with the separate powders, where zirconium (IV) oxide (ZrO 2 , 99 % trace metals basis, Sigma Aldrich) and copper (II) oxide (CuO, purum, � 99 % (RT), Sigma-Aldrich) were used as received.Approximately 10 mg of the material was reduced under H 2 flow (100 mL/min) at T = 20-400 °C for 151 min (ZrO 2 ) and 153 min for CuO.Thereafter, the heterogeneous mixture composed by CuO and ZrO 2 (1 : 1 ratio) was heated at the same temperature for 175 min.
The temperature was then decreased to 250 °C, and CO 2 and H 2 were fed into the cell at a molar ratio of CO 2 /H 2 = 1/3, and a total pressure of 33 bar.The purity of CO 2 and H 2 was slightly higher than the catalysts (99.9997 %, AGA).The reactions continued at 250 °C for 162 min on ZrO 2 , 893 min on CuO and 2181 min on ZrO 2 /CuO mixture, thereafter the temperature was decreased to 25 °C (all temperature changes were done ramp-wise with 30 °C every 5 min.).Finally, the cell was evacuated by opening the outlet.Here, it is valid to point out that CuO is converted into metallic Cu under the reducing atmosphere, thus implying that the active catalysts were ZrO 2 , Cu and Cu/ZrO 2 mixture.
Each process was then studied with in situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS).The measurements were conducted in a Bruker Vertex 70 spectrometer equipped with an LN2 cooled broad band MCT detector and a water-cooled glowbar MIR-source.The samples were placed in a 13 ml high pressure-high temperature DRIFTS-cell (Harrick Scientific) situated in the spectrometer and connected to a custom-made gas distribution system.Using KBr powder for the background measurement, repeated DRIFT spectra were collected at 4 cm À 1 resolution and each spectrum was averaged over 128 scans.In each spectrum, the logarithm of IR light reflected from the rough surface corresponds to the vibrations from a particular functional group.Hence, the peak area gives the quantitative information about the functional group.As explained in the next subsection, the calculation of the vibrational frequencies (ṽ calc ) allowed us to confirm or predict the mode frequency.

Model creation
The ZrO 2 surface was modelled using clusters of various sizes that were cut from the monoclinic structure of the bulk material. [62]In this sense, the Zr 9 O 18 cluster was chosen as the surface model for its convenient small size, while still providing results that are consistent with a larger cluster, as seen in Table S3. Figure 6(a) shows that Zr and O atoms have coordination numbers of 4-6 and 2-4, respectively, therefore assuming different spatial dispositions across the cluster.
The CuO material was modelled using a Cu 12 O 12 cluster, as shown in Figure 6(b), that presents a distorted geometry with a coordination number of 3 for both Cu and O atoms.This cluster has high multiplicity (S = 13) arising from an unpaired electron on each metallic center, as evidenced by the stability of ~401 kJ/mol reached in comparison with the closed-shell structure (see Table S4).Figure 6 (c) shows the Cu-based cluster that is herein utilized to model the CuO material after exposure to a reducing environment (H 2 ), then forming Cu.This material has metallic behavior and a closed-shell structure, with 13 copper atoms ordered in a closed packed icosahedral arrangement.
The Cu/ZrO 2 mixture was modelled using the clusters for pure ZrO 2 and Cu, since the mixture of these powders does not lead to the formation of a new structure.Instead, these catalysts are expected to interact with each other through weak forces.

Quantum Chemical Calculations
The adsorbate-surface interactions were investigated through quantum chemical calculations by making use of density functional theory (DFT).More specifically, the B3LYP [63] hybrid method and the 6-31G(d,p) [64] (for O, H atoms) and CEP-6-31G [65] (for Cu, Zr atoms) basis sets were chosen to perform the simulations in Gaussian 09. [66]Other method/basis set combinations were also tested during the validation step, with results found available in Table S5 in the Supplementary Material.
After structural optimization, the theoretical IR data were obtained from the vibrational normal modes, i. e., from the second derivatives of the energy with respect to the nuclear coordinates and dipole moment derivatives, as given by the quantum chemical program.Thereafter, the values were scaled by an appropriate factor according to the chemical bond that originated from a certain compound for further comparison with the experiment.For more details, see Tables S6-S8 in the Supplementary Material.
The adsorption energy E ads was calculated considering the difference of energy between the catalyst system in the presence of an adsorbate and the reference system based on the pure catalyst and the pure adsorbate in the gas phase.The binding energy E b was calculated similarly, but it was used to estimate the strength of the chemical bond between the catalyst and the adsorbate.Some basic definitions for the adsorption conformations or modes are presented in Figure 7. Parentheses are sometimes used as prefixes in dentate configurations as well to clarify bonding conditions, e. g., "(CÀ Cu) monodentate CO" means that the C atom in CO is bonded to one Cu atom at the surface.
The Gibbs free energy in gas phase G for each species is defined similarly to the description given in ref. [48,67]

°C and 30
bar for 70 min.

Figure 2 .
Figure 2. Optimized geometries of the adsorption modes of CO (a, b), CO 2 (c), bicarbonate (d) carbonates (e-g), formates (h-i) and methanol (j) on the Zr 9 O 18 surface.In each structure, the nomenclature makes use of the atom connectivity (in parenthesis), the number of dentates or binding mode (bridged or tribridged) and the name of the species itself.The species described in (e) and (i) have poor agreement with the experiment and are discarded in this work as feasible configurations.

Figure 3 .
Figure 3. IR spectra of Cu(s) in presence of sole H 2 (curve 1) and H 2 /CO 2 (curves 2-11) measured at different temperature and pressure conditions.The curves 1-8 were measured at p = 24.75bar.In a), it is shown the CÀ H stretching region, 3200-2700 cm À 1 ; in b), the CO stretching region for CO, 2220-2035 cm À 1 and, finally, in c) and d) the "fingerprint" region is displayed, 1380-850 cm À 1 .The time for measurement is specified for each curve in the legend box.For the measurement given by curve 10, the cell was maintained at T = 25 °C for 10 min.

Figure 4 .
Figure 4. Optimized geometries of the adsorption modes of CO (a, b), formate (c), formic acid (d-e) methanol (f) and methoxy (g) on the Cu surface.In each structure, the nomenclature makes use of the atom connectivity (in parenthesis), the number of dentates or binding mode (bridged) and the name of the species itself.

Figure 5 .
Figure 5. IR spectra of Cu/ZrO 2 mixture in presence of H 2 (curve 1) and H 2 and CO 2 (curves 2-12) measured at different temperature and pressure conditions.Here, the measurements from curves 1-8 were taken at p = 24.75bar.In a), it is shown the CÀ H stretching region, 3200-2800 cm À 1 ; in b), the CO stretching region for CO, 2220-2035 cm À 1 and, finally, in c) the "fingerprint" region is displayed in the range 1800-900 cm À 1 .The time for measurement is specified for each curve in the legend box.

Scheme 1 .
Scheme 1. -Three different possibilities for the CO 2 adsorption on ZrO 2 : insertion on OH bond (A), adsorption on O site (B), and involving oxygen vacancies (C).In A. the adsorption depends on the water dissociation to enable OH groups to be formed, ultimately generating bound bicarbonate (HCO 3 À ) at room temperature.In B., carbonate (CO 3 2À )* is first formed upon CO 2 adsorption and converted into formate under H 2 exposure.Finally, in C., the exposure to H 2 environment leads to the production of water molecules, which are trapped on the surface to form oxygen vacancies and OH groups.The subsequent CO 2 adsorption near to the vacancy leads to its dissociation into CO and O, thus regenerating the defect.The CO species could then attack the OH groups to form formate. From this point, methoxy and methanol are subsequently formed in presence of Cu in pathways B. and C. Note (*): The carbonate species was found in this study in a tribridged configuration, indicating that the bridged case could be easily transformed into formate in case it is formed.

Scheme 2 .
Scheme 2. -Gibbs free energies of reaction (Δ r G, in kJ/mol) at T = 298.15K for the intermediates of CO 2 conversion into methanol on ZrO 2 , as described in Scheme 1.The formate species (HCOO À ) has Δ r G = À 260.2 kJ/mol.Notes:* stands for adsorbed species, whereas ** corresponds to ZrO 2 in presence of an oxygen vacancy.If Scheme 1 is described as a (4x4) matrix, then each element (intermediate) is given by a xy (in blue).