The thermal chemistry of crotonaldehyde on the surface of a polished polycrystalline copper disk was characterized by temperature-programmed desorption (TPD) and reflection–absorption infrared spectroscopy (RAIRS) and contrasted with previous data obtained on a Pt(111) single crystal substrate. A clear difference in the adsorption mode was identified between the two surfaces, highlighted by the prevalence of RAIRS peaks for the C=C bond on Cu vs for C=O on Pt. Adsorption was also determined to be much weaker on Cu vs Pt, with an adsorption energy on the former ranging from −50 kJ/mol to −65 kJ/mol depending on the surface coverage. The experimental data were complemented by extensive quantum mechanics calculations using density functional theory (DFT) to determine the most stable adsorption configurations on both metals. It was established that crotonaldehyde adsorption on Cu occurs via the oxygen atom in the carbonyl group, in a mono-coordinated fashion, whereas on Pt multi-coordination is preferred, centered around the C=C bond. The contrasting surface adsorption modes seen on these two metals are discussed in terms of the possible relevance to selectivity in single-atom alloy hydrogenation catalysis.

Selectivity has become an increasingly important criterion in catalysis, as catalytic processes can be made greener by minimizing the generation of undesirable and potentially polluting by-products.1–6 Selective catalytic conversions also require fewer reactants and avoid the need for additional separation and purification steps. In the hydrogenation of organic feedstocks, selectivity becomes an issue when the molecules to be hydrogenated contain more than one unsaturation.7–9 Several late transition metals, including Pt, Pd, and Rh, are good hydrogenation catalysts but tend to not be selective. In fact, when dealing with unsaturated aldehydes, the hydrogenation of C=C bonds often takes precedence over that of the C=O moieties. This is unfortunate, because C=C hydrogenation in those cases leads to the formation of saturated aldehydes, which are common and cheap chemicals; the production of more interesting molecules, the unsaturated alcohols, requires preferential hydrogenation of the C=O bond.

Several approaches have been proposed to selectively hydrogenate unsaturated aldehydes at the carbonyl functionality.10,11 The use of so-called single-atom alloy (SAA) catalysts in particular has recently shown some promise.12–15 The main premise in the SAA scheme is that coinage metals such as copper, in contrast with what is seen with other noble metals, may favor C=O hydrogenation steps. Unfortunately, those metals are not efficient at activating molecular hydrogen,16–18 the first step required to facilitate any hydrogenation process. For that, a diluted amount of a second metal such as Pt is added to generate atomic hydrogen on the surface;19 these H atoms are then presumed to spillover onto the main metallic phase to promote hydrogenation steps selectively.20–22 Although we have recently challenged some aspects of this model,23,24 it is nevertheless reasonable to expect that the difference in hydrogenation selectivity between coinage and other noble metals is related to differences in the adsorption mode of the reactants on the surfaces of those metals.

In this study, we have tested this hypothesis by characterizing the adsorption of crotonaldehyde, a prototypical unsaturated aldehyde, on the surfaces of both a copper polycrystalline polished disk and a Pt(111) single crystal. The aim of our work is to provide some insights into why crotonaldehyde hydrogenation may take place preferentially at the C=O bond on the Cu surfaces of Cu—Pt SAAs but not on Pt-only catalysts. The key results came from experiments using reflection–absorption infrared spectroscopy (RAIRS), which clearly indicated different adsorption modes of the crotonaldehyde molecules on the two metals. On Cu, the C=C bond does not appear to be modified in any significant way by adsorption, and bonding to the surface seems to take place primarily via the C=O moiety. On Pt, by contrast, the C=C bond is likely to be the main point of contact with the surface. This difference is certainly consistent with the differences in selectivity seen during the catalytic hydrogenation of crotonaldehyde, as the molecular functionality that binds to the surface (C=O on Cu and C=C on Pt) is in most instances the one that becomes hydrogenated first. The details of the adsorption geometry of the aldehyde on both metals were explored in more detail by carrying out complementary quantum mechanics density functional theory (DFT) calculations. The theoretical results were consistent with the experimental data and provided a more detailed molecular-level picture of the bonding involved in these systems.

The temperature-programmed desorption (TPD) and RAIRS experiments reported here were performed in a two-tier ultrahigh vacuum (UHV) chamber described in detail in previous publications.25,26 The polished polycrystalline Cu disk, 10 mm in diameter and 2 mm in thickness, was made out of an OFHC copper sheet (McMaster-Carr, 99.99% purity). This disk was chemically cleaned by sonication in 50 ml of acetone and 50 ml of 2-propanol (5 min in each solvent), rinsing with 2-propanol for 2 min, immersion and sonication (5 min) in a dilute acetic acid solution (1 ml of glacial acetic acid in 30 ml of water), and rinsing again with 2-propanol. It was then dried and immediately mounted on a long-travel manipulator by wedging two parallel thin tantalum wires on slots carved on the edge of the disk, which were then spotwelded to two perpendicular thicker Ta rods attached to copper feedthroughs. A K thermocouple was wedged on a small hole drilled on the side of the disk to follow its temperature. This arrangement afforded liquid nitrogen cooling and resistive heating of the sample to any temperature between ∼100 K and 1000 K, as described elsewhere.27 The surface was cleaned in situ before each experiment via repeated sputtering–annealing cycles, using a rare-gas (Ar) ion gun, until the surface was deemed clean by contrasting results from CO TPD with reference data.28 Gas dosing was performed by backfilling of the UHV chamber using a leak valve. The crotonaldehyde (but-2-enal, Aldrich, ≥99% purity, mostly trans, containing 0.1%–0.2% butylated hydroxytoluene—BHT—and 1% H2O as stabilizers) was purified by performing several freeze–pump–thaw cycles in the gas manifold prior to any surface dosing. Its purity was assessed ex situ by IR and in situ by mass spectrometry. Exposures are reported in Langmuirs (1 L = 1 × 10−6 Torr s), uncorrected for ion gauge sensitivity.

The UHV vessel is equipped with a UTI quadrupole mass spectrometer, employed to carry out the TPD experiments. A linear heating rate of 10 K/s was used in the experiments reported here, driven by homemade electronics. A personal computer was employed for the data acquisition with software developed to follow the evolution of as many as 15 ions (amus) simultaneously during each individual TPD run. For the RAIRS experiments, the crystal was transferred to the second level of the UHV chamber to reach a setup similar to that in another apparatus also available in our laboratory:29,30 the IR beam from a Bruker Equinox 55 Fourier-transform infrared (FTIR) spectrometer is steered into the vacuum environment through a NaCl window and focused onto the surface at grazing incidence, and the reflected beam is then extracted through a second NaCl window and collected by a narrow-band mercury–cadmium–telluride (MCT) detector. The IR beam path outside the UHV chamber was purged with dry air to minimize light absorption by the water vapor and CO2 present in the laboratory’s atmosphere, but some residual water signal was still detected and needed to be subtracted using the reference spectra. RAIRS data were acquired using p-polarized light and ratioed against background spectra taken with the clean surface. All spectra were recorded at a resolution of 4 cm−1 by accumulating the signal from 4096 scans, a process that takes ∼8 min. The traces reported in Figs. 2 and 3 were background corrected to flatten the baseline.

For the theoretical component of this work, we used spin-unrestricted first principles calculations based on periodic density functional theory (DFT), as implemented in the Vienna Ab initio Simulation Package (VASP),31 with projector augmented waves.32 For the exchange–correlation (XC) potential, the generalized gradient approximation (GGA), with the Perdew–Burke–Ernzerhof (PBE) functional, has been employed.33 van der Waals interactions have been considered using the Grimme D3 method.34 The electronic states were expanded in plane waves with an energy cutoff of 400 eV. Either (3 × 3) or (2 × 2) surface unit cells were used to emulate adsorbate coverages of 1/9 ML and 1/4 ML, respectively; they were 4 atomic layers in thickness, with a 15 Å vacuum space added in the z direction. The Brillouin zone integration was done using a k-point grid of 3 × 3 × 1 in the (3 × 3) surface unit cell. The geometric structures were optimized by minimizing the forces on individual atoms with the criterion that all forces on each atom must be smaller than 1 × 10−3 Ry/a.u. The calculation of vibrational frequencies was performed by estimating the second derivative (Hessian matrix) of the potential energy surface using finite differences. Atoms were displaced in all directions. The dynamical space excluded the bottom layer of the substrate, which was kept frozen in order to simulate a bulk-like behavior.

The thermal chemistry of crotonaldehyde on polycrystalline Cu surfaces was explored by TPD and RAIRS. For the TPD studies, survey spectra were first acquired following several amus to identify the main desorbing products. No significant amounts of H2 (2 amu), CO (28 amu), propene (42 amu), or hydrogenation products such as crotyl alcohol (57 amu), n-butanal (72 amu), or n-butanol (56 amu) were detected in experiments with 20 L of crotonaldehyde dosed at 90 K; only molecular desorption (41 amu) was observed. This is typical of oxygenated organics on coinage metals, which usually need to be modified with a second element to activate those adsorbates.35–39 The evolution of the molecular desorption trace as a function of initial exposure is displayed in Fig. 1. Detectable desorption starts only after ∼3 L, possibly because the first molecules dosed may adsorb and fully decompose on low-coordination and other defect sites (the lack of H2 desorption detection in the TPD may be due to a sensitivity limitation with our instrumentation). A small molecular desorption peak is seen at ∼220 K after a 3 L dose, a feature that grows and shifts to lower temperatures as the exposure is increased; after a 10 L dose, the desorption peaks at around 180 K. Using the so-called Redhead equation40 and assuming a pre-exponential factor of 1 × 1015 s−1 for the reaction rate constant for crotonaldehyde molecular desorption from this Cu surface (to account for the higher entropy of the transition state of a desorbing molecule relative to its state when bound to the surface) yields values for the adsorption energy ranging from approximately Eads,lowθ ∼ −63 kJ/mol at a low coverage to Eads,ML ∼ −51 kJ/mol at monolayer saturation (the use of larger pre-exponential factors, as has been sometimes justified with larger molecules, would yield higher adsorption energy values: with A = 1 × 1017 s−1, Eads,lowθ ∼ −70 kJ/mol, and Eads,ML ∼ −58 kJ/mol). Further exposures of the surface to crotonaldehyde lead to condensation of a multilayer on the surface, the desorption of which peaks at 145 K after a 20 L dose. All the TPD traces show a long high-temperature tail, a reflection of the stickiness of the aldehyde, which makes it slow to pump out. The inset of Fig. 1 summarizes the evolution of the desorption peak area as a function of exposure. An approximately linear uptake is seen up to about 6 L, after which faster yield growth is observed. We estimate from these data that saturation of the first monolayer may occur after ∼8 L of exposure. The main trends reported here are similar to those reported previously on Cu(111).36 

FIG. 1.

Molecular TPD of crotonaldehyde adsorbed on a polished polycrystalline Cu surface at 100 K as a function of initial exposure. The inset summarizes the evolution of the peak areas.

FIG. 1.

Molecular TPD of crotonaldehyde adsorbed on a polished polycrystalline Cu surface at 100 K as a function of initial exposure. The inset summarizes the evolution of the peak areas.

Close modal

RAIRS data vs exposure are reported in Fig. 2. Like in the TPD results, no evidence for organic adsorbates is seen after exposures below 3 L, supporting our assumption of the initial total decomposition at defect sites. At 5 L, however, a few distinct features are clearly seen in the spectra at 964 cm−1, 1164 cm−1, 1425 cm−1, 1618 cm−1, and 1650 cm−1. On the basis of previous reports and our own DFT calculations, to be discussed below, we assign those peaks to a combination of methyl rocking and internal C—H bond bending, to an internal C—C stretching, to an asymmetric methyl deformation, and to two C=C stretching vibrations, respectively.41–43 Some of these peaks shift slightly with increasing surface coverage, and once monolayer saturation is reached (after an 8 L exposure), additional features are seen at 1303 cm−1 and 1377 cm−1, corresponding to an in-plane vinyl C—H deformation and the symmetric umbrella mode of the terminal methyl group, respectively. Importantly, no signal is seen for the C=O stretch, at ∼1700 cm−1, until multilayers start building up. This is significant because the peak becomes by far the most intense in the spectra of the multilayer, and it also dominates the spectra of crotonaldehyde in gas and liquid phases. The absence of this feature in the case of the monolayer adsorbed on Cu strongly suggests bonding to the metal with a C=O moiety oriented parallel to the surface: an IR selection rule establishes that on metal surfaces, only vibrational modes with dynamic dipoles with a component perpendicular to the surface plane are visible in the spectra.44–46 On the other hand, it appears that the C=C bond interacts only minimally with the Cu surface. It is conceivable that the preferential bonding of unsaturated aldehydes via their C=O bond on Cu surfaces makes such metal a good catalyst for the selective hydrogenation of those molecules.

FIG. 2.

RAIRS of crotonaldehyde adsorbed on a polished polycrystalline Cu surface at 100 K as a function of initial exposure.

FIG. 2.

RAIRS of crotonaldehyde adsorbed on a polished polycrystalline Cu surface at 100 K as a function of initial exposure.

Close modal

The adsorption mode of crotonaldehyde on Cu surfaces has proven to be quite different from that seen on Pt. To illustrate this point, Fig. 3 displays relevant RAIRS data, namely, spectra for low and saturation coverages on the Cu disk and on Pt(111) (from previous work from our laboratory),42 as well as reference traces for adsorption on supported Cu catalysts (Cu/SBA-15)23 and for liquid and condensed crotonaldehyde. In contrast to the case of the adsorption on Cu, on Pt, the predominant feature in the RAIRS traces is a peak at ∼1670 cm−1. This peak is associated with the carbonyl group of adsorbed crotonaldehyde, but because it is significantly red shifted from the value of the same vibrational mode in pure crotonaldehyde (∼1720 cm−1), it indicates significant rehybridization, with more sp3 character of the C=O bond. Also to note is the virtual absence of the signal from the C=C stretching modes at low coverages, an indication that such a bond is oriented parallel to the Pt surface and a hint at the adsorption mode in this system most likely involving pi coordination of this double bond with metal surface atoms. Some C=C signal is seen after monolayer saturation but with frequencies lower than in the free molecule, pointing again to rehybridization (in this case of the C=C bond). An additional difference between the spectra of crotonaldehyde on Cu and Pt is that while the internal C—C single bond is strengthened in both cases, as manifested by the increases seen in the frequency for its stretching mode (∼1165 cm−1 on Cu, ∼1175 cm−1 on Pt, and 1158 cm−1 in condensed crotonaldehyde), there is a noticeable change in the vinyl twist vibrational frequency only on Cu (∼965 cm−1) but not on Pt (973 cm−1). All this suggests that the adsorption of crotonaldehyde on Pt surfaces involves multiple bonding, primarily via the C=C bond. We speculate that because of this multiple coordination mode of adsorption on Pt surfaces, both C=C and C=O bonds are easily accessible for the addition of adsorbed hydrogen atoms, making pure Pt catalysts non-selective in the hydrogenation of unsaturated aldehydes.

FIG. 3.

Comparison of crotonaldehyde RAIRS data on polycrystalline Cu vs Pt(111) at half and full monolayer coverages. Additional data are provided for adsorption on a Cu/SBA-15 catalyst and for condensed multilayers and liquid crotonaldehyde.

FIG. 3.

Comparison of crotonaldehyde RAIRS data on polycrystalline Cu vs Pt(111) at half and full monolayer coverages. Additional data are provided for adsorption on a Cu/SBA-15 catalyst and for condensed multilayers and liquid crotonaldehyde.

Close modal

IR spectra are also provided in Fig. 3 for the adsorption of crotonaldehyde on a catalyst made out of Cu nanoparticles dispersed on a SBA-15 silica mesoporous material23 to help relate the data on model surfaces with more realistic systems. The trace obtained at 300 K shows peaks at 1380 cm−1, 1399 cm−1, 1448 cm−1, 1630 cm−1, 1658 cm−1, 1642 cm−1, and 1692 cm−1, with shapes and relative intensities similar to those obtained with either condensed or liquid crotonaldehyde. Our interpretation of this result is that, although the data were acquired after pumping out the compound from the gas phase, some remained trapped in the pores of the oxide support. However, a very different spectrum was obtained at 450 K. In fact, this trace does not resemble any of the others in Fig. 3, suggesting the presence of different surface species. An intense peak is seen at 1665 cm−1, most likely from a hybridized C=O moiety, and an additional, weaker feature is seen at 1645 cm−1, pointing to the retention of the C=C bond as well. The terminal methyl groups clearly display their asymmetric (1468 cm−1) and symmetric (1382 cm−1) deformation modes, and the vinyl C—H bond its in-plane deformation motion (1446 cm−1). Together, these observations may be interpreted as the result of molecular or nearly molecular adsorption with a partially bent adsorption geometry, but additional species may be present on the surface, possibly responsible for the additional broad features seen in the spectrum between 1500 cm−1 and 1620 cm−1. It is nevertheless noticeable that many molecular features can be seen for the adsorbed species in this case even though adsorption was carried out at room temperature. One possible explanation for this is that in the case of real supported catalysts the surface of the metal may be partially oxidized23 and that the oxygen/hydroxo species associated with that modified surface may stabilize the adsorption of the crotonaldehyde (as mentioned above).35–39 A summary of the IR data is provided in Table I.

TABLE I.

Summary of IR data for crotonaldehyde adsorption on Pt and Cu surfaces. All frequencies are in cm−1.

Pt (111)aCu diskCu/SBA-15b
ModeLiquidcCondensedd0.5 ML (2 L)1 ML (4 L)0.5 ML (5 L)1 ML (8 L)300 K450 K
ν(C=O) 1718 1696 (vs) 1672 (s) 1670 (vs)   1692 (vs) 1665 (s) 
ν(C=C) 1653, 1640 1644 (m) 1640 (m) 1638 (s) 1650 (s), 1649 (vs), 1658 (m), 1645 (m) 
     1618 (m) 1620 (s) 1642 (s)  
ν(C=C)hyb   1592 (m) 1592 (s),   1630 (m) 1605 (w), 
    1575 (m)    1570 (m), 
        1550 (w) 
δa(CH31452 1447 (m) 1440 (w) 1438 (m) 1425 (w) 1437 (m)  1468 (w) 
δa(CH31447      1448 (s) 1446 (m) 
δip(CH)ald 1390 1397 (w)     1399 (m) 1398 (w) 
δs(CH31381 1376 (w)  1375 (m)  1377 (w) 1380 (m) 1382 (w) 
δip(CH)vinyl 1303 1300 (vw)    1303 (w)   
δip(CH)vinyl 1252, 1290        
ν(C—C) 1149 1158 (s) 1174 (m) 1175 (s) 1164 (m) 1167 (s)   
ν(C—CH31072 1083 (m) 1080 (w) 1082 (m)     
ρ(CH3), τ(C2H2973 977 (s)  973 (s) 964 (s) 967 (vs)   
δoop(CH)ald 928 934 (m)       
Pt (111)aCu diskCu/SBA-15b
ModeLiquidcCondensedd0.5 ML (2 L)1 ML (4 L)0.5 ML (5 L)1 ML (8 L)300 K450 K
ν(C=O) 1718 1696 (vs) 1672 (s) 1670 (vs)   1692 (vs) 1665 (s) 
ν(C=C) 1653, 1640 1644 (m) 1640 (m) 1638 (s) 1650 (s), 1649 (vs), 1658 (m), 1645 (m) 
     1618 (m) 1620 (s) 1642 (s)  
ν(C=C)hyb   1592 (m) 1592 (s),   1630 (m) 1605 (w), 
    1575 (m)    1570 (m), 
        1550 (w) 
δa(CH31452 1447 (m) 1440 (w) 1438 (m) 1425 (w) 1437 (m)  1468 (w) 
δa(CH31447      1448 (s) 1446 (m) 
δip(CH)ald 1390 1397 (w)     1399 (m) 1398 (w) 
δs(CH31381 1376 (w)  1375 (m)  1377 (w) 1380 (m) 1382 (w) 
δip(CH)vinyl 1303 1300 (vw)    1303 (w)   
δip(CH)vinyl 1252, 1290        
ν(C—C) 1149 1158 (s) 1174 (m) 1175 (s) 1164 (m) 1167 (s)   
ν(C—CH31072 1083 (m) 1080 (w) 1082 (m)     
ρ(CH3), τ(C2H2973 977 (s)  973 (s) 964 (s) 967 (vs)   
δoop(CH)ald 928 934 (m)       
a

Reference 42.

b

Reference 23.

c

Reference 43.

d

Intensities (in parentheses): vs = very strong, s = strong, m = medium, and w = weak.

DFT calculations were carried out to complement the experimental study and help develop a molecular picture of the adsorption of crotonaldehyde on Cu and Pt surfaces. For reference, the energetics of crotonaldehyde in the gas phase was estimated first. In the gas phase, there are four major crotonaldehyde isomers, as shown in Fig. S1 (supplementary material): E-(s)-trans, E-(s)-cis, Z-(s)-trans, and Z-(s)-cis. The E-(s) isomers differ from the Z-(s) by the relative orientation of the methyl and formyl groups, whereas trans and cis isomers differ by the orientation of the C=O bond. The calculated relative energies of the other configurations with respect to the E-(s)-trans isomer are summarized in Table S1 (supplementary material). It can be seen that the E-(s)-trans isomer is the most stable, followed by the E-(s)-cis, Z-(s)-trans, and Z-(s)-cis structures. These results are in good agreement with experiments and previous calculations.47 Table S1 also reports the main structural parameters of each isomer. All bond lengths are similar in all four isomers.

Since our results show that the E-(s) isomers are the most stable, and given that they are also the most common isomers among the commercially sold crotonaldehyde, in the following discussion, we only consider the adsorption of those isomers. We have carried out calculations for two surface coverages, θ1 = 1/9 ML and θ2 = 1/4 ML (1 ML = 1 molecule per metal surface atom), which we call low and high coverages, respectively. These were set by placing a single molecule in two different unit cells, with (3 × 3) and (2 × 2) dimensions, respectively. In Fig. 4, we summarize the results obtained for the energetics of adsorption on both Pt(111) and Cu(111) surfaces at the low coverage. Several adsorption configurations were considered and optimized in each case, from which five are reported in the figure, going from a single coordination via the oxygen atom (η1–O) to full coordination via the four atoms in the C=C and C=O bonds (η4). It is clear from those data that the coordination mode makes a big difference in the adsorption energy on Pt(111) surfaces, with multiple coordination being greatly favored for all isomers. According to our DFT calculations, for the θ1 = 1/9 ML coverage reported in the figure, the η4-trans-OC1C2C3 bonding is the most stable, with an adsorption energy of approximately Eads = −170 kJ/mol. This is consistent with previous calculations.47 It should be indicated, however, that η4-cis-OC1C2C3, η3-cis-OC2C3, η2-trans-C2C3, and η2-cis-C2C3 all display similar energetics and cannot be ruled out. On the other hand, low-coordination species are considerably less stable and unlikely to form on Pt surfaces, at least at low to intermediate coverages. For instance, the adsorption energy of the η1-O species is ∼95 kJ/mol lower in absolute terms than that for the η4-trans-OC1C2C3 bonding mode. Higher coverages (θ2 = 1/4 ML) result in increases in instability for the species that occupy larger surface areas (i.e., the η3 and η4 coordinations) but barely affect the energetics of the η1 and η2 species (results not shown), which display structures with the molecular axis standing up from the surface. It appears that on Pt(111), crotonaldehyde prefers to adsorb via multiple bonding and with a particularly strong interaction through the C=C double bond.

FIG. 4.

Calculated energies of adsorption for crotonaldehyde on Pt(111) and Cu(111) vs coordination mode. The data correspond to the θ1 = 1/9 ML coverage, the (3 × 3) cell. The red dot corresponds to the energy of the η1-O species on the Cu(111) surface at θ2 = 1/4 ML.

FIG. 4.

Calculated energies of adsorption for crotonaldehyde on Pt(111) and Cu(111) vs coordination mode. The data correspond to the θ1 = 1/9 ML coverage, the (3 × 3) cell. The red dot corresponds to the energy of the η1-O species on the Cu(111) surface at θ2 = 1/4 ML.

Close modal

The story is quite different with the Cu(111) surface. First of all, adsorbed crotonaldehyde molecules on Cu(111) are much less stable than on Pt(111) in all cases. Moreover, at low coverages (θ1 = 1/9 ML), all of the configurations considered in this study display similar adsorption energies: the difference between the least stable η1-O-cis and the most stable η4-OC1C2C3 configurations is less than 20 kJ/mol. The energy of adsorption of the trans isomer in the most likely η1-O configuration was calculated to be Eads,lowθ,DFT ∼ −59 kJ/mol, a value close to that measured experimentally (Eads,lowθ,Exp ∼ −63 kJ/mol, see above). At higher coverages (θ2 = 1/4 ML), only the configurations that involve Cu—O bonds were found to be stable (data not shown), with the η1-O species being the most stable (red dot in Fig. 4). In contrast to the adsorption on Pt(111), it appears that on Cu(111) crotonaldehyde prefers single coordination to the surface, via the oxygen atom. This difference is consistent with what was seen experimentally, based on the RAIRS results summarized in Fig. 3, and helps explain the differences in selectivity observed experimentally with catalysts based on Pt vs Cu during the hydrogenation of unsaturated aldehydes. Again, the comparison is made between adsorption on the Cu surfaces of Cu—Pt SAA catalysts (since conversion is assumed to occur on Cu; the Pt single atoms are presumably added only to help with H2 dissociation) vs pure Pt catalysts. We argue that the multiple coordination of crotonaldehyde on Pt leads to indiscriminate hydrogen addition steps on Pt-only catalysts, whereas the specific bonding through the oxygen atom in the carbonyl group on Cu surfaces may direct hydrogenation reactions toward that functional group with SAA catalysts.

Figure 5 displays the optimized structures determined by DFT for all of the configurations reported in Fig. 4, and Tables S2 (Pt) and S3 (Cu) summarize the calculated relative bond lengths (supplementary material). On Pt (Table S2), the C=O bond in the η2 configurations retains the same distance as in the gas phase. This is understandable since those atoms are not involved in the bonding to the surface, but it is inconsistent with the changes in C=O stretching frequency seen in RAIRS upon crotonaldehyde adsorption (Fig. 3 and Table I). A slight C=O bond lengthening is seen in configurations η1-O and η3-OC2C3, but it is in the η4 configurations that this bond stretching is maximized, approaching values close to those seen in C—O single bonds. Similarly, the C2=C3 bond is mostly unaffected in configurations η1-O and η2-OC1 but shows increased sp3 character (longer bond distance) in all other adsorption modes. Given that both C=O and C=C bond stretching frequencies of crotonaldehyde adsorbed on Pt(111) are red shifted in the RAIRS data compared to the gas-phase values (Fig. 3 and Table I), it is reasonable to conclude that the most likely adsorption configurations in this case are either η3-OC2C3 or η4 coordinations. It is also worth noticing that the C1—C2 and C3—C4 bond lengths do not vary in any significant way in any of the adsorption modes considered here. This, again, is consistent with the fact that the ν(C—C) stretching frequencies of crotonaldehyde do not change much upon bonding to the Pt(111) surface (Fig. 3 and Table I).

FIG. 5.

Calculated structures for crotonaldehyde adsorbed on Pt(111) and Cu(111) vs coordination mode. The data correspond to the θ2 = 1/4 ML coverage, the (2 × 2) cell.

FIG. 5.

Calculated structures for crotonaldehyde adsorbed on Pt(111) and Cu(111) vs coordination mode. The data correspond to the θ2 = 1/4 ML coverage, the (2 × 2) cell.

Close modal

The case of crotonaldehyde adsorbed on Cu(111) is again somewhat different (Table S3). It is particularly curious that the C=O bond remains at about the same length in all adsorption configurations, including those where the oxygen atom is involved in bonding to the surface (η1-O and η2-OC1). This is a reflection of the low energy of adsorption in this case, that is, of the weak overall bonding of the molecule to the surface. Unfortunately, due to the surface selection rule of RAIRS on metals, the stretching frequency of this bond is not visible in the spectra and cannot therefore be compared to the DFT results. Nevertheless, because of the absence of the ν(c=O) peak in the RAIRS traces due to its orientation parallel to the surface, it is reasonable to assume that crotonaldehyde binds to Cu in a η1-O or η2-OC1 configuration; on the basis of the other data provided in this report, we favor the former option. According to the DFT calculations, the length of the C=C bond is equally unaffected by the adsorption regardless of the bonding mode to the surface, and indeed, the ν(C=C) stretching frequency of crotonaldehyde on Cu is similar to that in the gas phase.

Finally, an attempt was made to calculate the vibrational frequencies of the adsorbed species using the VASP code and to match those to the experimental data. For reference, the vibrational spectra of gas-phase crotonaldehyde were calculated first, for its four main isomers; the results are reported in Table S4 (supplementary material). Our results quite nicely match the previous calculations reported by the Sautet group (reported in parentheses in Table S4): all our calculated frequencies match theirs (for the E-trans and E-cis isomers) within a couple of wavenumbers.47 The match with the experimental values43 is also quite good, although large deviations are seen there. It should be noted, however, that the quantum mechanics calculations seem to underestimate most frequencies, sometimes by as much as 20 cm−1–30 cm−1. We attribute these differences to either the perfect periodicity in the theoretical models and/or to differences in the surface coverage (in the case of the adsorbed species).

The calculations of the vibrational modes for the adsorbed species, reported in Tables S5 (Pt) and S6 (Cu) (supplementary material), display larger discrepancies and failed to provide definitive answers with respect to the interpretation of the experimental RAIRS data. Nevertheless, some interesting information deriving from those calculations is worth highlighting. For the case of crotonaldehyde adsorption on Pt(111) (Table S5), our values again compare reasonably well with those reported by Haubrich et al.47 The calculated frequencies for the C=O and C—CH3 stretching modes of the η2-C2C3-trans species are also quite close to those seen experimentally (1675 vs 1670 cm−1 and 1081 vs 1082 cm−1, respectively). The vibration of the C=C stretching mode is expected to be much lower than in the gas phase, around 1100 cm−1–1120 cm−1 according to our calculations, but the mode is invisible in the experimental RAIRS and cannot therefore be checked. The frequencies of the asymmetric and symmetric (umbrella) methyl group deformation modes are larger in the experiments (1438 cm−1 and 1375 cm−1, respectively), but still within a reasonable range of those calculated by DFT (1431 cm−1 and 1348 cm−1). A few other vibrations can be accounted for by our calculations, which therefore support bonding of the crotonaldehyde molecule to Pt via the C=C double bond.

The case for Cu (Table S6) is less straightforward. The C=O stretching mode is clearly seen in this case in the experimental RAIRS data in two peaks at 1649 cm−1 and 1620 cm−1. These values are quite high and do not match any of the calculated spectra, but the second peak can be justified by the values of 1617 cm−1, 1625 cm−1, 1652 cm−1, and 1622 cm−1 seen for the η1-O-trans, η2-C2C3-trans (a), η2-C2C3-trans (b), and η2-C2C3-cis bonding modes, respectively. The asymmetric and symmetric methyl deformation modes, detected experimentally at 1437 cm−1 and 1377 cm−1, respectively, are also best matched by the corresponding calculated values for the same configurations: η1-O-trans (1425 cm−1 and 1349 cm−1), η2-C2C3-trans (a) (1421 cm−1 and 1337 cm−1), η2-C2C3-trans (b) (1426 cm−1 and 1342 cm−1), and η2-C2C3-cis (1420 cm−1 and 1336 cm−1). Perhaps the peaks that better help discern among these options are those for the in-plane C—H vinyl deformation and C—C stretching: the experimental values for their frequencies, 1303 cm−1 and 1167 cm−1, respectively, are best matched by the results for the η1-O-trans adsorption configuration (1284 cm−1 and 1155 cm−1). It is unfortunate that the C=C stretching mode is not apparent in the RAIRS traces, but that, we believe, is because the bond is oriented close to parallel to the surface. In any case, it is worth noticing that in the calculations, the vibrational frequency of the bond tends to be higher than on Pt(111), reflecting the overall weaker character of the adsorption on Cu. All together, although the match is not ideal, the calculations of the vibrational spectra do support adsorption of crotonaldehyde via the oxygen atom in a η1-O configuration. As with the Pt(111) surface, a better agreement between experiments and calculations is found by shifting the calculated frequencies by ∼20 cm−1 to 40 cm−1.

The adsorption of crotonaldehyde on Cu surfaces was characterized and contrasted with that on Pt. TPD data indicate the desorption being mainly molecular with minimal, if any, surface decomposition. The molecular desorption peak drifts from 220 K at low coverages to 180 K after monolayer saturation, values that correspond to an adsorption energy varying from Eads,lowθ ∼ −63 kJ/mol to Eads,ML ∼ −51 kJ/mol. Vibrational data obtained by RAIRS for the adsorbed species are dominated by peaks at 973 cm−1, 1164 cm−1, and 1653 cm−1 at low coverages, corresponding to methyl rocking and internal C—C and C=C stretching modes, respectively. Additional features develop through monolayer saturation for the terminal methyl deformation modes (1377 cm−1 and 1437 cm−1), but conspicuously absent is any signal from the C=O stretching vibration, strongly suggesting that bonding to the surface occurs through that bond and that its axis is oriented close to parallel to the surface.

The conclusions reached here regarding the adsorption of crotonaldehyde on Cu contrast those previously reported on Pt.42 Specifically, the RAIRS recorded on Pt(111) surfaces is dominated by a broad peak at 1672 cm−1, characteristic of the C=O moiety, at low coverages; it is the C=C stretching mode that is not detected in this case. Adsorption on Pt is therefore concluded to take place mainly via bonding between the C=C double bond and the surface, with possible additional interactions leading to multiple coordinated adsorbed species. It is this significant difference in adsorption geometry for crotonaldehyde on the Cu vs Pt surface that we believe may explain the differences in selectivity seen during its catalytic hydrogenation. It should be noted, however, that realistic copper-based catalysts may exhibit oxidized surfaces23 and adsorption on those substrates may be different, as suggested by the different appearance of the transmission IR traces obtained for adsorbed crotonaldehyde in those systems.

Quantum mechanics calculations were also carried out on both Pt(111) and Cu(111) surfaces to complement the experimental information and help develop a molecular model for crotonaldehyde adsorption on those surfaces. On Pt, adsorption was shown to clearly become stronger upon an increase in the coordination number, with η4-trans-OC1C2C3 bonding being the most stable at low coverages; the calculated adsorption energy for this state is Eads ∼ −170 kJ/mol, close to that reported in previous publications.47 On Cu, by contrast, adsorption is always weak and not greatly affected by the extent of coordination to the surface. In the end, we determined that the most stable adsorption configuration of crotonaldehyde on Cu(111) is η1-O, independent of coverage. The adsorption energy of crotonaldehyde on Cu(111) was estimated at between Eads,DFT = −50 kJ/mol and −70 kJ/mol (depending on the configuration and coverage), consistent with the measured values. Additional calculations were carried out to determine the adsorption geometry and the vibrational spectra. The results are consistent with an adsorption mode involving a single coordination via the oxygen atom with the C=O bond parallel to the surface plane.

See the supplementary material for the additional figures and tables cited in the text.

Financial support for this project was provided by a grant from the U.S. National Science Foundation, Division of Chemistry (Grant No. NSF-CHE1953843). N.T. and J.G.-S. acknowledge DGAPA-UNAM Project Nos. IN101019 and IA100920 and Conacyt Grant No. A1-S-9070 for partial financial support. The calculations were performed in the DGCTIC-UNAM Supercomputing Center, projects LANCAD-UNAM-DGTIC-051 and LANCAD-UNAM-DGTIC-368. We also acknowledge the technical support provided by Aldo Rodriguez Guerrero.

The data that support the findings of this study are available from the corresponding author upon reasonable request.

1.
F.
Zaera
,
J. Phys. Chem. B
106
,
4043
(
2002
).
2.
G. J.
Hutchings
,
Philos. Trans. R. Soc. B
363
,
985
(
2005
).
3.
G. A.
Somorjai
and
J. Y.
Park
,
Angew. Chem., Int. Ed.
47
,
9212
(
2008
).
4.
F.
Zaera
,
Acc. Chem. Res.
42
,
1152
(
2009
).
5.
F.
Zaera
,
J. Phys. Chem. Lett.
1
,
621
(
2010
).
6.
S. H.
Pang
and
J. W.
Medlin
,
J. Phys. Chem. Lett.
6
,
1348
(
2015
).
7.
H. U.
Blaser
 et al,
Adv. Synth. Catal.
345
,
103
(
2003
).
8.
F.
Zaera
,
ACS Catal.
7
,
4947
(
2017
).
9.
L.
Zhang
 et al,
Chem. Rev.
120
,
683
(
2020
).
10.
M.
Luneau
 et al,
Chem. Rev.
120
,
12834
(
2020
).
11.
X.
Lan
and
T.
Wang
,
ACS Catal.
10
,
2764
(
2020
).
12.
P.
Aich
 et al,
J. Phys. Chem. C
119
,
18140
(
2015
).
13.
G.
Giannakakis
,
M.
Flytzani-Stephanopoulos
, and
E. C. H.
Sykes
,
Acc. Chem. Res.
52
,
237
(
2019
).
14.
J.
Han
 et al,
Chin. J. Chem.
37
,
977
(
2019
).
15.
R. T.
Hannagan
 et al,
Chem. Rev.
120
,
12044
(
2020
).
16.
J. M.
Campbell
and
C. T.
Campbell
,
Surf. Sci.
259
,
1
(
1991
).
17.
M.
Johansson
,
O.
Lytken
, and
I.
Chorkendorff
,
J. Chem. Phys.
128
,
034706
(
2008
).
18.
L.
Álvarez-Falcón
 et al,
Surf. Sci.
646
,
221
(
2016
).
19.
H.
Thirumalai
and
J. R.
Kitchin
,
Top. Catal.
61
,
462
(
2018
).
20.
G.
Kyriakou
 et al,
Science
335
,
1209
(
2012
).
21.
Q.
Fu
and
Y.
Luo
,
J. Phys. Chem. C
117
,
14618
(
2013
).
22.
F. R.
Lucci
 et al,
J. Phys. Chem. C
119
,
24351
(
2015
).
23.
Y.
Cao
 et al,
ACS Catal.
9
,
9150
(
2019
).
24.
Y.
Cao
 et al,
ACS Catal.
10
,
3431
(
2020
).
25.
D.
Chrysostomou
,
J.
Flowers
, and
F.
Zaera
,
Surf. Sci.
439
,
34
(
1999
).
26.
X.
Qin
and
F.
Zaera
,
J. Phys. Chem. C
122
,
13481
(
2018
).
27.
Q.
Ma
and
F.
Zaera
,
J. Vac. Sci. Technol., A
33
,
01A108
(
2015
).
28.
A.
Kokalj
,
T.
Makino
, and
M.
Okada
,
J. Phys.: Condens. Matter
29
,
194001
(
2017
).
29.
A.
Tilekaratne
 et al,
ACS Catal.
2
,
2259
(
2012
).
30.
F.
Zaera
,
Chem. Soc. Rev.
43
,
7624
(
2014
).
31.
G.
Kresse
and
J.
Furthmüller
,
Phys. Rev. B
54
,
11169
(
1996
).
32.
P. E.
Blöchl
,
Phys. Rev. B
50
,
17953
(
1994
).
33.
J. P.
Perdew
,
K.
Burke
, and
M.
Ernzerhof
,
Phys. Rev. Lett.
77
,
3865
(
1996
).
34.
S.
Grimme
 et al,
J. Chem. Phys.
132
,
154104
(
2010
).
35.
M. E.
Chiu
 et al,
Angew. Chem., Int. Ed.
45
,
7530
(
2006
).
36.
M. E.
Chiu
 et al,
Chem. Commun.
1283
(
2006
).
37.
M.
Boronat
,
M.
May
, and
F.
Illas
,
Surf. Sci.
602
,
3284
(
2008
).
38.
K.
Xiong
,
W.
Wan
, and
J. G.
Chen
,
Surf. Sci.
652
,
91
(
2016
).
39.
K.
Xiong
and
J. G.
Chen
,
Catal. Today
339
,
289
(
2020
).
40.
P. A.
Redhead
,
Vacuum
12
,
203
(
1962
).
41.
H.-J.
Oelichmann
,
D.
Bougeard
, and
B.
Schrader
,
J. Mol. Struct.
77
,
179
(
1981
).
42.
J. C.
de Jesús
and
F.
Zaera
,
Surf. Sci.
430
,
99
(
1999
).
43.
R.
Lindenmaier
 et al,
J. Phys. Chem. A
121
,
1195
(
2017
).
44.
R. G.
Greenler
,
J. Chem. Phys.
44
,
310
(
1966
).
45.
Y. J.
Chabal
,
Surf. Sci. Rep.
8
,
211
(
1988
).
46.
F.
Zaera
,
Int. Rev. Phys. Chem.
21
,
433
(
2002
).
47.
J.
Haubrich
 et al,
J. Phys. Chem. C
113
,
13947
(
2009
).

Supplementary Material