CO Interaction with Alkali Metal Cations in Zeolites: A Density

F. Javier Torres, Jenny G. Vitillo, Bartolomeo Civalleri, Gabriele Ricchiardi, and ... Vera Bolis and Alessia Barbaglia , Silvia Bordiga, Carlo Lamber...
0 downloads 0 Views 217KB Size
9292

J. Phys. Chem. B 1997, 101, 9292-9298

CO Interaction with Alkali Metal Cations in Zeolites: A Density Functional Model Cluster Study Anna Maria Ferrari, Konstantin M. Neyman, and Notker Ro1 sch* Lehrstuhl fu¨ r Theoretische Chemie, Technische UniVersita¨ t Mu¨ nchen, D-85747 Garching, Germany ReceiVed: May 30, 1997; In Final Form: August 22, 1997X

The interaction of CO molecules with cations of alkali exchanged zeolites (Li, Na, K, Rb, and Cs forms) has been studied theoretically using a density functional (DF) method. The modification of the alkali action due to the presence of the zeolite framework has explicitly been taken into account by employing various model clusters. Inclusion of zeolite framework effetcs is shown to be important for obtaining quantitative agreement with experiment. Carbon monoxide bound to a cationic site via the oxygen center is found to exhibit an adsorption-induced red shift of the CO stretching frequency in contrast to the well-established blue shift of the common C-bound species. According to our DF results there is a certain probability (increasing with the size of the cation up to 15% at 100 K) for the presence of O-bound CO species in alkali-exchanged zeolites. The experimentally observed red-shifted satellite of the IR band of the CO stretching mode is assigned to such unusually coordinated probe molecules.

Introduction Carbon monoxide is a widely used probe molecule in surface chemistry to titrate the number and strength of acidic sites1,2 by comparing shift and intensity of the vibrational mode of adsorbed CO to gas-phase molecules. CO molecules interact with Lewis acidic centers as exposed by surface cations of metal oxides such as MgO and NiO,1 of halides,3 and with cations of metal-exchanged zeolites.4-9 The interaction of CO with the surface of ionic oxides and halides has been extensively investigated theoretically.10,11 These results provide a solid basis for the understanding of the bond nature of the probe molecule to acid sites and of the origin of the CO frequency shift. The interaction of CO with a transition metal cation implies the donation of electronic charge from the 5σ orbital of CO to a d orbital of the metal in synergy with a back-donation from the metal to the empty antibonding 2π* orbital of CO. In the case of cations without d orbitals in their valence shell no such “chemical” bonding can occur and the interaction is essentially mediated by electrostatics. Despite the great number of theoretical investigations concerning CO interacting with ionic surfaces, only very few studies have been devoted to exchanged zeolites. So far, results at the semiempirical level or SCF results obtained with minimal basis sets are available.12 It is only very recently that a copper exchanged ZSM-5 zeolite has been studied theoretically at both the SCF level13 and the density functional (DF) level14-16 employing the local spin density approximation. In these studies, the interaction of Cu species with either an isolated Al-O--Si site or with a five- or six-membered ring of silicate units has been investigated. In the past, several studies have been devoted to the interaction of CO with naked alkali cations.17-19 Although these systems constitute a rather crude representation of the real complexes of CO formed in alkali-substituted zeolites, properties of the hypothetical single cation complexes may provide a first approximation to rationalize such experimental data. It is only recently that the interaction between CO and naked alkali cations has been studied systematically at a high computational level.20 It was found that, despite the crudeness of the model, several main observed features of the interaction are reproduced quite * Corresponding author. X Abstract published in AdVance ACS Abstracts, October 15, 1997.

S1089-5647(97)01759-8 CCC: $14.00

satisfactorily: C-bound species being energetically favored, trends in the heat of formation of adsorption complexes, signs of the shifts in the CO stretching frequency, and intensity decrease of the CO stretching mode with increasing strength of the interaction. All quantities, however, were considerably overestimated because the electrostatic field generated by the naked cation is too strong compared to the real system where it is weakened by the surrounding oxygen ions. A more precise quantification of this weak adsorbate-substrate interaction has to be based on models where effects of the zeolite framework oxygen atoms on the charge balancing metal cations are taken into account. In the present work we have done so, and we have investigated the interaction of CO with the complete series of alkali cations, explicitly considering also the interaction with oxygen atoms. Both types of CO coordination, C-down and O-down, have been analyzed. The effects due to the surrounding zeolite framework have been represented in an averaged fashion by employing a standard cluster model AlH(OH)3which contains three oxygen centers in direct proximity of the alkali cations. This cluster constitutes the minimal unit which allows the simulation of effects due to surrounding oxygen centers. The same cluster, but electrically neutralized by the presence of a proton, was previously used to study the proton transfer to ammonia.21-23 The usage of this model cluster is justified since it was shown previously24 that the oxygen atoms directly connected to the aluminum atom are the most negative ones. Additional arguments for using small cluster models can be derived from a recent study on zeolite models with periodic boundary conditions;25 there, it was shown that the strong fields in zeolites are restricted to the vicinity of the metal cations and thus probably do not require the inclusion of a significant amount of the framework. With the present study we want to settle for a simple model to describe the interaction of carbon monoxide with cationic sites of alkali-exchanged zeolites that also takes into account the effect due to oxygen centers in the zeolite framework, and we want to examine to what extent this affects the interaction. Despite the model character of this study, we believe that it is possible to reproduce accurately enough such adsorption properties as binding energies and changes in the IR vibrational spectra induced by CO adsorption, and to relate them to available experimental data and to the results obtained in the naked cation approach.20 © 1997 American Chemical Society

Interaction of CO with Alkali-Exchanged Zeolites The interaction of CO with alkali-exchanged zeolites can conveniently be monitored by means of IR spectroscopy.5-9 It was found that the vibrational stretching frequency of CO predominately experiences a blue shift when CO is adsorbed on a cationic site; however, a red-shifted satellite band is present as well, though of lower intensity. The origin of the red-shifted band is still disputed. Several tentative explanations have been suggested;5,9 the possibility of CO molecules interacting via their oxygen atom with a cationic site is the most promising one among them. This rationalization is supported by theoretical works10,20 that report a red shift in the CO stretching mode when CO is adsorbed on a nontransition metal cation via its oxygen end. Anyway, a CO molecule bound via its oxygen center to a more realistic cationic site model so far has not been directly investigated. Therefore, we want to characterize with high accuracy the energetics of CO adsorption for both cases, oxygen and carbon attachment, and to find out whether coordination via oxygen is a reasonable alternative from a thermodynamic point of view. In the present study, cluster size effects have been considered as well by employing an extended model cluster, AlH(OSiH3)3-, with oxygen atoms being terminated by silyl groups rather than by simple hydrogen centers; this cluster should yield a more realistic description of oxygen centers in a zeolite framework interacting with metal cations. Computational Details The model clusters AlH(OH)3- and AlH(OSiH3)3-, the CO molecule, and the lighter cations, Li+, Na+, K+, have been described by a triple-ζ valence basis set (TZV).26 The resulting contraction schemes are as follows: (62111/411) for C and O, (73211/611) for Al and Si, (311) for H, (62111) for Li+, (73211/ 51) for Na+, and (842111/631) for K+. These basis sets have been augmented by a set of two d polarization functions on the C and O atoms of the CO molecule (R ) 0.15 and 1.0) and a set of two p functions on the cations, R(Li) ) 0.06 and 0.4, R(Na) ) 0.0196 and 0.131, R(K) ) 0.0152 and 0.0417.20,26 The basis sets for Al, Si, O, and H used for the zeolite clusters have been extended by one d polarization function: R(Al) ) 0.3, R(Si) ) 0.35, R(O) ) 0.8, and R(H) ) 0.8.26 For the heavier cations, Rb+ and Cs+ nine-electron relativistic effective core potentials (ECP) as well as the corresponding contraction schemes for the valence orbitals (Rb: (4311/321), Cs: (4311/ 321)) have been adopted;27 the corresponding primitive basis sets were taken from the Internet address described in ref 28. The combination of all-electron basis sets for the lighter alkali cations, Li+, Na+, K+, and pseudopotentials on Rb+ and Cs+ has been found to furnish a well equilibrated description of the whole series of alkali cations.20 This is in line with a recent study29 which pointed out that pseudopotentials yield results of the same accuracy as all-electron descriptions provided a proper integration scheme29 and comparably flexible basis sets are employed.30,31 All calculations were carried out using the GAUSSIAN94/ DFT package32 employing the density functional option. The integration grid referred to as “FINE” in the program has been used which corresponds to 75 radial shells, up to 302 angular points for each shell, and about 7000 points all together for each atom. Becke’s three-parameter hybrid exchange functional33 in combination with the gradient corrected correlation functional of Lee, Yang, and Parr,34 generally referred to as B3LYP, has been used throughout. This functional was found to be well suited for interactions governed by electrostatics:20 B3LYP results on the interaction of CO with naked alkali cations were better than those from MP2 calculations and in good

J. Phys. Chem. B, Vol. 101, No. 45, 1997 9293

Figure 1. Clusters employed to model the zeolite cationic site: the standard cluster model AlH(OH)3-X+ (top) and the extended cluster model AlH(OSiH3)3-X+ (bottom).

agreement with results obtained using a high-quality configuration interaction scheme. Moreover, at this level of calculation (B3LYP/TZV) the CO multipole moment and molecular polarizabilty are reproduced in good agreement with experiment.20 Geometries were optimized and harmonic vibrational frequencies were evaluated with the help of analytical energy derivatives in case of all-electron calculations or with the help of finite differences in case of the ECP approach. Basis set superposition errors (BSSE) of binding energies as evaluated by the full counterpoise method35 turned out to be rather small, of the order of 0.3 kcal mol-1 only for all structures. Thermal corrections have been calculated using the standard rigid rotor/harmonic oscillator, modified to take into account that surface complexes are part of a solid which has neither rotational nor translational degrees of freedom.36 Results and Discussion Cationic Site. The standard cluster AlH(OH)3-X+ used to model the cationic center of alkali exchanged zeolites is shown in Figure 1. Its geometrical parameters and bonding characteristics are summarized in Table 1; also given are the Mulliken charges, charges from a natural bond orbital (NBO) analysis,37 and the oxygen 1s core level shifts which will be useful in the following discussion. We begin by discussing the isolated cluster, AlH(OH)3-. Two conformers have been considered: one of C3V and one of C3 symmetry. Both have been optimized within the given geometrical constraints and the C3 conformer was found to be a local minimum (no negative eigenvalues of the Hessian). It is more stable than the C3V conformer, but only by 2.3 kcal mol-1. Moreover, the charge distribution in the two conformers are quite similar as judged from the Mulliken analysis (Table 1). Thus we expect no significant differences if we employ the computationally less demanding, idealized C3V model of the cationic site. This expectation is confirmed by the adsorption characteristics of the sodium model which have been computed for both conformers (see Table 1). Except for the Al-O-H bonding angle which decreases by about 5° due to the extended steric freedom in the C3 conformation, the geometry, the charge distribution, and the energetics turn out to be rather insensitive to the small structural differences of the two cluster models. Throughout this study alkali cation attachment to the AlH(OH)3cluster has been assumed to occur in 3-fold coordination (see Figure 1). Although other coordinations are possible,16 their studies should require different cluster models and thus they have been not considered here.

9294 J. Phys. Chem. B, Vol. 101, No. 45, 1997

Ferrari et al.

TABLE 1: Properties of the Cationic Site of Alkali-Exchanged Zeolites As Modeled by the Cluster AlH(OH)3-X+ a AlH(OH)3-X+ X+

AlH(OH)3-

Li+

Na+

K+

Rb+

Cs+

d(Al-O) d(Al-H) d(O-H) d(X+-Al) d(X+-O) ∠(Al-O-H) ∠(O-Al-O) ∠(H-Al-O-H) q(X+) q(O) χ(X+) χ(O) ∆1s(O)b BE(X+)

1.788 (1.788) 1.663 (1.652) 0.961 (0.962)

1.802 1.581 0.956 2.311 1.984 129.4 91.9 0.0 0.73 -0.79 0.95 -1.24 5.1 165.8

1.801 (1.803) 1.590 (1.587) 0.957 (0.961) 2.619 (2.625) 2.302 (2.325) 124.2 (119.5) 96.4 (97.1) 0.0 (45.7) 0.86 (0.84) -0.82 (-0.84) 0.97 -1.23 4.8 (4.9) 144.0 (144.3)

1.800 1.599 0.958 2.993 2.678 120.0 100.0 0.0 0.88 -0.82 0.99 -1.23 4.3 123.6

1.799 1.603 0.959 3.181 2.861 118.8 101.2 0.0 0.89 -0.81 0.99 -1.23 4.2 115.8

1.799 1.605 0.959 3.383 3.058 117.5 102.3 0.0 0.89 -0.80 1.00 -1.22 4.0 108.7

113.3 (113.0) 109.5 (109.3) 0.0 (57.7) -0.73 (-0.74) -1.19

a The data refer to the C3V symmetry model; available data for the C3 symmetry model are given in parentheses. Distances d in Å, angles in degrees, alkali ion binding energies to AlH(OH)3- BE(X+) in kcal mol-1, 1s core level shiftsb ∆1s(O) of framework oxygen centers in eV, Mulliken charges q and NBO charges χ in au. b Calculated as difference of Kohn-Sham eigenvalues relative to the bare substrate model: ∆1s(O) ) 1s(AlH(OH)3-X+) - 1s(AlH(OH)3-).

The strength of the interaction between the alkali cation and the zeolite site modeled by the AlH(OH)3- cluster decreases in the series from Li, the lightest alkali cation, down to Cs. This trend correlates quite nicely with the cation radii and the corresponding intermolecular distances X+-Al and X+-O (Table 1). One may regard this finding as a hint that electrostatics plays an important role in the zeolite-cation interaction, but actually this interaction results from the subtle interplay of three factors: a weak, but nevertheless attractive covalent bonding between the cation and the framework oxygen atoms, a much stronger attractive electrostatic bonding between the cation and the partially ionic oxygen centers, and a repulsive Coulomb interaction between X+ and the positively charged Alδ+ center. The final X+-Al and X+-O distances are the result of a delicate balance between these three contributions. The interaction with alkali cations leads to rather substantial deformations of the substrate structure around the cationic site (Table 1); most noticeable among them is the decrease of O-Al-O angle. These structural changes are expected to be overestimated because of the missing zeolite backbone in our model cluster. Nevertheless, a trend is clearly discernible: deformations are largest for Li+, decreasing steadily when going down the alkali column of the periodic table. This trend fits well with the observation that the lighter cations interact stronger with the basic oxygen centers, yielding shorter intermolecular distances. The closer the cation, the stronger is the impact of its electrostatic field on the zeolite fragment. In a metal-exchanged zeolite the cations act as Lewis acid centers38,39 while the framework oxygens, bearing a partial negative charge, are the structural basic sites of the zeolite.40,41 Experimentally, it was found that the basic strength within one family of alkali-exchanged zeolites increases within the series from Li to Cs derivatives.42,43 This trend was rationalized in terms of oxygen charges as evaluated by the Sanderson electronegativity equivalent method44,45 which shows that in alkali-exchanged zeolites the absolute value of the negative oxygen charge increases from the Li- to the Cs-exchanged species. Another interesting correlation has been reported between the oxygen 1s core level binding energy determined by XPS (X-ray photoemission spectroscopy) and the basic strength of framework oxygen centers.46-48 The O 1s binding energy decreases almost linearly with increasing Al/Si ratio (increase of the number of basic centers) for the Y, X, and A zeolites which has been attributed to a more negative oxygen charge and thus an increase of the oxygen basicity.

Since the charge of the oxygen centers may serve as a measure for the basicity strength of a zeolite, it is worth analyzing it theoretically. We therefore calculated the Mulliken charges for the AlH(OH)3-X+ cluster (Table 1). Unfortunately, they turn out to be almost constant, about -0.8 au, irrespective of the cation the oxygen atoms are interacting with. A NBO analysis leads to essentially the same conclusion, except that oxygen charges are about -1.2 au, independent of the nature of the cation (Table 1). Nevertheless, a minor increase of the oxygen negative charge does occur relative to the isolated AlH(OH)3- cluster. More sensitive to the nature of the cations are the calculated O 1s binding energies (as represented by the negative of the corresponding Kohn-Sham eigenvalues, see Table 1) whose values decrease when moving from the Li+ to the Cs+ complex in agreement with experimental findings.48 The origin of the shift in the O 1s binding energies is related to the electrostatic field and to the polarizing power of the cation: it is stronger for lighter cations because they are located closer to oxygen atoms due to their smaller sizes. If a charge transfer from an alkali cation to the zeolite substrate is prevented (by choosing a one-electron ECP description without the corresponding basis functions), then the computed O 1s binding energies agree to 0.2 eV or better with the values reported in Table 1. Considering this result together with the oxygen charges reported in Table 1 which are even slightly decreasing when moving from the Li+ to the Cs+ complexes, the correlation often reported in the literature41 between O 1s binding energies and oxygen basic strength can be questioned. However, one should keep in mind that the oxygen centers in the cluster model AlH(OH)3- belong to an Al-O-H group, a chemical entity which in principle is different from an AlO-Si group of a zeolite framework. Therefore, we have analyzed the extended cluster model AlH(OSiH3)3-X+ as well (Figure 1); its calculated characteristics are summarized in Table 2. The oxygen charge calculated at the same level of theory amounts to -0.63 au for an extended cluster AlH(OSiH3)3instead of -0.73 au obtained for the standard model cluster AlH(OH)3- (Tables 1 and 2). This result implies that the oxygen atoms effect on the alkali cations may to some extent be overestimated by the smaller cluster model. This is quite evident for the binding energies of the cations Na+ and Cs+ which are calculated lower by about 20 kcal mol-1 in the extended cluster model compared to the standard cluster model AlH(OH)3-X+. Also the X+-O bond length is computed somewhat longer with the more realistic, extended cluster model.

Interaction of CO with Alkali-Exchanged Zeolites

J. Phys. Chem. B, Vol. 101, No. 45, 1997 9295

TABLE 2: Properties of the Adsorption Complexes AlH(OSiH3)3-X+ and AlH(OSiH3)3-X+/CO with CO Interacting via Its Carbon Enda AlH(OSiH3)3-X+ AlH(OSiH3)3-X+/CO X+

AlH(OSiH3)3

d(Al-O) d(Si-O) d(X+-Al) d(X+-O) (Al-O-Si) (O-Al-O) d(CO) d(X+-C) q(X+) q(O) ∆1s(O)b BE(X+) BE(CO) BE(CO;BSSE) ν(CO) I(CO)

1.776 1.617

-

145.0c 108.0

-0.63

Na+

Cs+

Na+

Cs+

1.798 1.649 2.640 2.353 141.9 97.9

1.791 1.639 3.484 3.149 140.7 102.5

1.796 1.647 2.657 2.373 141.5 98.2 1.1218 2.658

1.790 1.638 3.493 3.160 140.3 102.6 1.1239 3.832

0.90 -0.78 4.0 122.0

0.93 -0.75 3.4 89.8 5.1 4.8 2265 59

1.6 1.4 2241 70

a The data refer to the C3V symmetry model. Distances d in Å, angles in degrees, alkali ion binding energies to AlH(OSiH3)3- BE(X+) and CO binding energies to the substrate BE(CO), the latter also BSSE corrected BE(CO;BSSE), in kcal mol-1, core level shifts ∆1s(O) of framework oxygen centers in eV, Mulliken charges q in au, frequency ν(CO) of the CO stretching mode in cm-1 and its intensity I(CO) in km mol-1. b Calculated as difference of Kohn-Sham eigenvalues relative to the bare substrate model: ∆1s(O) ) 1s(AlH(OSiH3)3-X+) - 1s(AlH(OSiH3)3-). c Fixed during optimization

TABLE 3: Properties of the AlH(OH)3-X+/CO Adsorption Complexes with CO Interacting via Its Carbon Enda AlH(OH)3-X+/CO X+

Li+

Na+

K+

Rb+

Cs+

d(Al-O) d(Al-H) d(O-H) d(X+-Al) d(X+-O) ∠(Al-O-H) ∠(O-Al-O) ∠(H-Al-O-H) d(C-O)b d(X+-C) BE(CO) BE(CO;BSSE) ∆H° (0) ν(CO)b I(CO)b ∆ν(CO) ∆ν(CO)expc ∆ν(CO)expd ∆ν(CO)expe

1.800 1.583 0.956 2.331 2.099 129.0 92.4 0.0 1.1222 2.286 5.7 5.4 4.7 2263 71 45 45 45

1.800 (1.802) 1.592 (1.589) 0.957 (0.961) 2.632 (2.639) 2.319 (2.343) 124.0 (120.4) 96.7 (97.5) 0.0 (46.3) 1.1225 (1.1224) 2.681 (2.681) 4.4 (4.4) 4.2 (4.2) 3.6 (3.6) 2259 (2260) 63 (63) 41 (42) 35 34 29

1.800 1.600 0.958 3.001 2.688 119.8 100.1 0.0 1.1234 3.233 2.7 2.4 2.1 2248 67 30 23 20 16

1.799 1.604 0.959 3.186 2.869 118.7 101.3 0.0 1.1239 3.523 2.1 1.8 1.5 2244 69 26 19 16 11

1.799 1.607 0.959 3.388 3.065 117.5 102.4 0.0 1.1243 3.854 1.5 1.4 1.1 2239 72 21 14 12 2

a The data refer to the C3V symmetry model; available data for the C3 symmetry model are given in parentheses. Distances d in Å, angles in degrees, CO binding energies BE(CO) and BSSE corrected binding energies BE(CO;BSSE) in kcal mol-1, and CO adduct formation entalphies ∆H°(0), computed at 0 K and corrected for the zero-point vibrational energy, in kcal mol-1, frequency ν(CO) (and shift ∆ν(CO) with respect to free CO) of the CO stretching mode in cm-1 and its intensity I(CO) in km mol-1. b Computed values for free CO: d(CO) ) 1.1267 Å, ν(CO) ) 2218 cm-1, I(CO) ) 76 km mol-1. c Reference 5. d Reference 6. e Reference 50.

Interaction with a CO Probe Molecule. We now turn to discussing the interaction of CO probe molecules with cationic sites. We start with the common type of coordination where the CO probe interacts with the alkali cation via its carbon atom. The calculated properties of the resulting adsorption complex AlH(OH)3-X+/CO are summarized in Table 3. As expected, the structure of the acidic site does not change significantly upon CO adsorption because of the rather weak interaction (cf. Tables

Figure 2. Correlation between the calculated binding properties of CO adsorbed at a cationic site of a zeolite as computed for the model cluster AlH(OH)3-X+ vs the correspondent ones computed for the simple naked cation X+ model (from ref 20): frequency shift ∆ν of the CO stretching mode (top panel) and alkali ion binding energies BE (bottom panel).

1 and 3). The CO bond distance is shortened with respect to the free molecule, by 0.0045 Å for Li+ down to 0.0024 Å for Cs+. The X+-C distance, of course, increases with increasing radius of the cations X+. More interesting in the present context is the energetics of the adsorption complexes. The binding energies, BE(CO), of the probe molecule CO exhibit a distinct trend to decrease along the series of alkali cations from Li+ to Cs+. This trend is not altered by taking into account the basis set superposition error (BSSE) which is small anyway (see above). The CO stretching frequency is enhanced upon adsorption, with the blue-shift values again decreasing from Li+ to Cs+ (see Table 3). In Table 3, we also report the frequency shift measured for various alkali exchanged zeolites.5,6,50 Agreement between theory and experiment is quite satisfactory, especially for zeolites with low cation loading,5,6 a situation which is closer to that described by our models with a single, isolated cationic site. Except for Li+ the predicted IR intensities of the CO stretching mode exhibit a uniform trend as well: they increase with decreasing blue shift, in agreement with experiment.9 Apart from this single exception of the oscillator strength of the CO stretching mode in AlH(OH)3- Li+/CO, all results calculated for the standard cluster model follow the same trends (though different in magnitude) as obtained previously using naked alkali cations as models.20 In particular, if we compare the CO binding energies to the substrate and the blue shift of the CO stretching mode as calculated for the two models, AlH(OH)3-X+ and naked cation X+, very similar linear correlations and deviations are observed (see Figure 2). The interaction between a CO molecule and naked cations was found to be essentially electrostatic in nature.20 The similarity of the linear relationship shown in Figure 2 indicates that the same mechanism also governs the CO interaction with the model cluster AlH(OH)3-X+. The electrostatic field generated by the AlH(OH)3-X+ cluster model, as calculated from the selfconsistent electron density, is reduced by a constant factor (to about one-half) as a consequence of (a) the additional electro-

9296 J. Phys. Chem. B, Vol. 101, No. 45, 1997

Ferrari et al.

static field generated by the substrate cluster AlH(OH)3- and (b) the framework oxygen induced reduction of the cation charge, as indicated by the Mulliken charges of the cations in AlH(OH)3-X+ of about 0.88 au for Na+ to Cs+ (Table 1). The case of Li+ is somewhat different. Due to its smaller radius, Li+ can interact stronger with the zeolite cationic site, resulting in a net atomic charge of Li roughly 0.1 au lower than for the other cations (Table 1). This deviation may be taken as direct evidence for a higher degree of covalency in the Li+ bonding to the substrate. In this context, it is worth mentioning that quite often properties of lithium compounds differ far more from those of the other alkali compounds than the latter among each other.49 The occurrence of covalency in lithium compounds is rather common:49 many organolithium derivatives appear to be covalent and often exhibit polymeric structures while the other alkali elements give rise to essentially ionic compounds. Furthermore, lithium reacts readily with oxygen, forming the normal monoxide while other alkali elements predominantly form peroxides and superoxides.49 In line with this peculiarity of lithium the electric field generated by the complex AlH(OH)3-Li+ is lower than expected, thus leading to a reduction of the CO binding energy and the CO stretching frequency as discernible in Figure 2. Despite this finding, the CO interaction with the cationic sites is qualitatively similar for Li+ and the other alkali cations (Figure 3). In fact, if one regards the vibrational frequency shift ∆ν(CO) and the CO binding energy BE as functions of the electric field F acting at the center of mass of the CO molecule, Li+ is found to fit very nicely the general relationship obtained for the other cations: a linear correlation between the blue shift of the CO stretching mode and the electrostatic field F and an approximately quadratic increase of the binding energy with the strength of the electric field F (see Figure 3). Different experimental shifts of the CO stretching frequency have been observed in various types of zeolites.5-7,9,50 If one plots the experimental frequency shifts ∆νexp of the CO stretching mode as a function of the corresponding calculated values ∆νcalc from the standard cluster model, a linear relationship is obtained (see Figure 4). This linear correlation suggests that the electrostatic mechanism which governs the interaction between the CO molecule and the AlH(OH)3-X+ cluster model also holds for the CO adsorption in a real system. Distinct slopes and offsets are found for the different types of zeolites as a consequence of the more complex origin for the electric field experienced by a CO probe in zeolite cavities. The electric field in a zeolite cage, Fzeo, as experienced by an adsorbed CO probe molecule at a distance d(X+-CO), can be described by means of a simple electrostatic model:51

Fzeo ) (qsite/4π0)/[d(X+-CO)]2 + Fenv

(1)

where qsite is the formal charge of the site due to the cation and its charged neighboring oxygen centers. Fenv represents the contribution to the local electric field generated by the remaining zeolite framework, polarized by the presence of the cation. The effective charge qsite of the active site and the environment field Fenv depend on structural and electronic properties of a zeolite and determine the different slopes and offsets of the curves displayed in Figure 4. For the M-Y zeolite, both Fenv and qsite are larger compared to other types of zeolites because of the significantly higher Al/Si ratio in M-Y. As already mentioned, the standard cluster AlH(OH)3-X+ tends to overestimate the effect of framework oxygen atoms to some extent. For instance, the CO frequency shifts are about twice as large in the naked cation model (Na+ 78 cm-1; Cs+ 43 cm-1),20 compared to the standard model (Table 3). Since

Figure 3. Correlation between the calculated properties, frequency shift ∆ν of the CO stretching mode (top panel) and the binding energy BE (bottom panel) of CO adsorbed on the AlH(OH)3-X+ model, and the electric field F generated by the cluster at the center of mass of the adsorbed CO molecule.

Figure 4. Correlation between the experimentally observed frequency shift of the stretching mode of a CO molecule adsorbed on different zeolites (see also Table 2) and the corresponding calculated shift for a CO molecule adsorbed on the standard AlH(OH)3-X+ cluster model.

it has been demonstrated that the interaction of the CO probe depends essentially on the strength of the local electrostatic field, one expects that the adsorption characteristics of the extended cluster AlH(OSiH3)3- lie between those of the standard cluster model and the simple naked alkali cations approach. Indeed, the binding energies (for Na+ only, for Cs+ the difference is too small) and the frequency shifts are larger in the extended cluster model than in the standard cluster model (see Tables 2 and 3). Also the IR intensities exhibit the same trend: half as large in the naked cation model (Na+ 32 km mol-1; Cs+ 48 km mol-1)20 as in the standard one, they decrease again in the extended cluster model (see Tables 2 and 3), in the order of about 10%, and are in line with the change in the charge distribution on the oxygen centers as quantified in a Mulliken analysis (see above). Now, we turn to the alternative type of CO coordination: adsorption via the oxygen atom (Table 4). As mentioned above for the CO interaction via the C end, the geometry of the cationic site is only slightly deformed by the CO interaction. Again, the intermolecular X+-OC distances increase with the cation radii but they are shorter than d(X+-CO) because of the smaller van der Waals radius of oxygen compared to carbon. The

Interaction of CO with Alkali-Exchanged Zeolites

J. Phys. Chem. B, Vol. 101, No. 45, 1997 9297

TABLE 4: Properties of the AlH(OH)3-X+/OC Adsorption Complex with CO Interacting via Its Oxygen Enda AlH(OH)3-X+/OC X

+

Li

+

Na

+

K+

Rb+

TABLE 5: Thermodynamic Characteristics of the Exchange Reaction ZO-X+/CO T ZO-X+/OC at 77 K As Computed for the Standard Cluster AlH(OH)3-X+a AlH(OH)3-X+

Cs+ X

Li

∆H°(0 K) ∆(∆H°(77 K)) -T∆S°(77 K) ∆G°(77 K) p(ZO-X+/OC)

1.51 -0.26 -0.02 1.23 0.0

+

d(Al-O) d(Al-H) d(O-H) d(X+-Al) d(X+-O) ∠(Al-O-H) ∠(O-Al-O) d(C-O)b d(X+-O) BE(OC) BE(OC;BSSE) ∆H°(0) ν(CO)b I(CO)b ∆ν(CO)

1.800 1.583 0.956 2.327 2.003 129.2 92.3 1.1294 2.127 3.5 3.2 3.2 2199 139 -19

1.800 1.591 0.957 2.629 2.314 124.1 96.6 1.1297 2.500 2.9 2.5 2.3 2195 130 -23

1.800 1.600 0.958 2.999 2.685 119.9 100.1 1.1294 2.993 2.0 1.6 1.4 2197 117 -21

1.799 1.603 0.959 3.185 2.866 118.8 101.3 1.1291 3.272 1.5 1.2 0.9 2199 112 -19

1.799 1.607 0.959 3.389 3.063 117.5 102.3 1.1288 3.590 1.1 0.8 0.6 2201 109 -17

a

The data refer to the C3V symmetry model. Distances d in Å, angles in degrees, OC binding energies BE(OC) and BSSE corrected binding energies BE(OC;BSSE) in kcal mol-1, and OC adduct formation entalphies ∆H°(0), computed at 0 K and corrected for the zero-point vibrational energy, in kcal mol-1, frequency ν(CO) (and shift ∆ν(CO) with respect to free CO) of the CO stretching mode in cm-1 and its intensity I(CO) in km mol-1. b Computed values for free CO: d(CO) ) 1.1267 Å, ν(CO) ) 2218 cm-1, I(CO) ) 76 km mol-1.

energetic trend is also as for the X+-CO species: the adsorbate binding energies BE(OC) decrease steadily from Li+ to Cs+. However, the binding energy for CO interacting via its oxygen end is somewhat lower than for the corresponding “carbondown” configuration, a result which is well established provided electron correlation is taken into account.10 On the other hand, the frequency of the CO stretching mode in the unusual coordination is red-shifted in contrast to the blue shift obtained for C-bound CO; it is also in line with the computed CO bond elongation compared to free CO. The vibrational red shift as well as the CO bond distance decrease when the probe interacts with heavier cations. However, at variance with the C-bound complexes, the IR intensity of the stretching mode decreases as the red shift decreases. All these trends are characteristic for an interaction essentially governed by electrostatics. As before, the case of Li+ represents an exception: the CO distance is shorter and the vibrational shift is lower than for Na+ complexes and the trend of the IR intensity is broken. A weak IR absorption signal was detected in the region of 2120-2124 cm-1 for CO interacting with alkali-exchanged ZSM-5 zeolites, although it is only very faint for the Li-ZSM5 and Na-ZSM5 species.5,9 A similarly weak IR peak has been reported for alkali-exchanged mordenite.6 Several tentative explanations have been suggested for this IR spectroscopic feature: (i) CO molecules interacting simultaneously through both ends with a pair of Lewis acid sites which could be either charge balancing cations or other extraframework sites;52,53 (ii) CO molecules adsorbed on structural defects;54,55 or (iii) CO molecules interacting with one metal cation through the oxygen end.9,56,57 Recently, it was found that simultaneous interaction with a pair of cations does not lead to the observed red shift in the CO vibrational frequency.58 The additional red-shifted IR band has been observed in all kinds of zeolites, independent of their preparation and structure;5,7,9 such a finding would be hardly expected for defect-induced IR shifts. On the other hand, there is some theoretical evidence for the possibility that O-bound adsorption complexes are formed. Consider the formal exchange reaction AlH(OH)3-X+/CO T AlH(OH)3-X+/OC. The calculated reaction enthaphies ∆H°(0 K) of the two configurations, C-bound and O-bound (see Tables

+

Na

+

1.33 -0.20 -0.02 1.11 0.0

K+

Rb+

Cs+

0.69 -0.11 -0.09 0.49 4.1

0.59 -0.09 -0.08 0.42 6.0

0.50 -0.08 -0.07 0.35 9.2

a Energies in kcal mol-1. The percentage p(ZO-X+/OC) of the O-bound CO species at thermal equilibrium is reported.

3 and 4), decrease along the series from Li+- to Cs+-exchanged zeolites. Moreover, these values come closer to each other the heavier the cation (see Table 5). Taking thermal corrections into account, the changes in the reaction enthalpies ∆H°(77 K) and the changes in entropy ∆S°(77 K) yield free reaction enthalphy changes ∆G°(77 K) which range from 0.49 kcal mol-1 for K+ down to only 0.35 kcal mol-1 for Cs+. Thus, these calculated quantities indicate a nonvanishing statistical probability to find O-bound species in the zeolite (see the last row of Table 5). In thermal equilibrium at 77 K, 9.2% of the CO molecules adsorbed on Cs+ should be attached via their oxygen end. Taking into account that the sample temperature should be somewhat higher than that of the heat bath (due to IR irradiation) the probability of the O-bound species may be even larger (e.g. up to 15% Cs+-OC species for a local temperature of 100 K). The probabilities given in Table 5 have to be understood as rough guidance only, because minor variations of the computational details may easily affect the calculated population distribution. However, it is crucial for the present argument that the energy differences under consideration are inevitably very small. Anyway, an assignment of the observed red-shifted satellite of the main CO stretching mode peak in the IR spectrum to CO molecules adsorbed on alkali-exchanged zeolites via their oxygen end is strongly suggested by the present DF results. This assignment is further supported by the fact that the intrinsic intensity of the CO stretching mode of the O-bound species is calculated to be more than 50% higher than that of the C-bound species (Tables 3 and 4) and that the temperature enhanced probability for O-bound adducts is in line with experimental findings.9 Conclusions The interaction of CO probe molecules with cationic sites of the whole series of alkali-exchanged zeolites has been studied in a systematic way with the help of DF model cluster calculations. In extension of an earlier study20 which employed a simple naked cation model to simulate the probe-substrate interaction, various cluster models have been used here to account for the effects due to the zeolite framework, and especially due to the oxygen centers in the immediate surrounding of the cationic center. We were able to demonstrate that the concept of an essentially electrostatic interaction between the probe molecule and the alkali ion, as derived from the naked cation model, is also valid for models where a fragment of the zeolite framework is present. The present more realistic cluster models which take into acount a part of the zeolite surrounding achieve quantitative agreement with experiment for the vibrational properties. They also confirm the general trends predicted by the naked cation model. Already the model AlH(OH)3-X+ of the cationic site turned out to yield quite satisfactory results. The electrostatic field is only slightly overestimated by this

9298 J. Phys. Chem. B, Vol. 101, No. 45, 1997 model as judged by the extended model AlH(OSiH3)3-X+ which also comprises complete Al-O-Si groups in the immediate vicinity of the cation. The interaction of Li+ with the zeolite framework was found to differ somewhat from the other alkali ions as judged indirectly by the CO bonding characteristics (see Figure 2 for details). This is in line with the well-known peculiarity of Li to enter covalent bonds much easier than the other alkali ions. On the other hand, as far as the interaction characteristics with a CO probe are concerned, Li+ show no significant differences to the other alkali ions in the sense that Li+ fits well the correlations between electrostatic field and adsorption properties found for alkali ions (see Figure 3). We have examined in detail both types of CO probe interaction, C-attached and O-attached to the cationic site. O-bound CO probe molecules may very well give rise to the experimentally observed red-shifted satellite band in the IR spectra.5,6,9 The difference in adsorption entalphy (at 0 K) between the two coordination modes was calculated to at most 1.5 kcal mol-1, getting smaller the larger the exchanged cation. Computed values for the free enthalphy change of the exchange reaction ZO-X+/CO T ZO-X+/OC revealed that there is a nonvanishing probability (up to 15% at 100 K) for O-bound CO species to be present in alkali-exchanged zeolites under equilibrium conditions. Despite the applicability of the present results to all kinds of zeolites, the characteristics of particular zeolites, like the structure of the cationic sites and the local electrostatic field near the cationic position, may affect the delicate equilibrium ZO-X+/CO T ZO-X+/OC and therefore the relative intensity of the red-shifted IR C-O band. Acknowledgment. We thank Dr. U. Birkenheuer for critically reading this manuscript and for many useful suggestions. This work has been supported by Deutsche Forschungsgemeinschaft, Bayerischer Forschungsverbund Katalyse, and Fonds der Chemischen Industrie. References and Notes (1) Zaki, M. I.; Kno¨zinger H. J. Catal. 1987, 119, 311. (2) Kno¨zinger, H. Adsorption on Ordered Surfaces of Ionic Solids and Thin Films; Springer Series in Surface Science, Vol. 33; Freund H.-J., Umbach E., Eds.; Springer-Verlag: Berlin, 1993; p 257. (3) Richardson, H. H.; Baumann, C.; Ewing, G. E. Surf. Sci. 1987, 15, 185. (4) Angell, C. L.; Schaffer, P. C. J. Phys. Chem. 1966, 70, 1413. (5) Zecchina, A.; Bordiga, S.; Lamberti, C.; Spoto, G.; Carnelli, L.; Otero Area´n, C. Phys. Chem. 1994, 98, 9577. (6) Bordiga, S.; Lamberti, C.; Geobaldo, F.; Zecchina, A.; Turnes Palomino, G.; Otero Area´n, C. Langmuir 1995, 11, 527. (7) Bordiga, S.; Garrone, E.; Lamberti, C.; Zecchina, A.; Otero Area´n, C.; Kazansky, V. B.; Kustov, L. M. J. Chem. Soc., Faraday Trans. 1994, 90, 3367. (8) Bolis, V.; Fubini, B.; Garrone, E.; Giamello, E.; Morterra, C. Structures and ReactiVity of Surfaces; Morterra, C., Zecchina, A., Costa, G., Eds.; Elsevier: Amsterdam, 1989. (9) Katoh, M.; Yamazaki, T.; Ozawa, S. Bull. Chem. Soc. Jpn. 1994, 67, 1246. (10) Sauer, J.; Ugliengo, P.; Garrone, E.; Saunders, V. R. Chem. ReV. 1994, 94, 2095. (11) Colbourn, E. A. Surf. Sci. Rep. 1992, 15, 281. (12) Vetrivel, R.; Catlow, R. C. A. Modelling of Structure and ReactiVity in Zeolites; Catlow, R. C. A., Ed.; Academic Press: London, 1992; p 217. (13) Zhanpeisov, N. U.; Nakatsuji, H.; Hada, M.; Nakai, H.; Anpo, M. Catal. Lett. 1996, 42, 173. (14) Hess, K. C.; Schneider, W. F. J. Phys. Chem. 1996, 100, 9292. (15) Schneider, W. F.; Hess, K. C.; Ramprasad, R.; Adams, J. B. J. Phys. Chem. 1996, 100, 6032. (16) Trout, B. L.; Chakraborty, A. K.; Bell, A. T. J. Phys. Chem. 1996, 100, 4173. (17) Dixon, D. A.; Gole, J. L.; Komornicki, A. J. Phys. Chem. 1988, 92, 1378. (18) Ikuta, S. Chem. Phys. Lett. 1984, 109, 550. (19) Ikuta, S. Chem. Phys. 1985, 95, 235.

Ferrari et al. (20) Ferrari, A. M.; Ugliengo, P.; Garrone, E. J. Chem. Phys. 1996, 105, 4129. (21) Teunissen, E. H.; van Duijneveldt, F. B.; van Santen, R. A. J. Phys. Chem. 1992, 96, 366. (22) Teunissen, E. H.; van Santen, R. A.; Jansen, A. P. J.; van Duijneveldt, F. B. J. Phys. Chem. 1993, 97, 203. (23) Van Santen, R. A.; Kramer, G. J. Chem. ReV. 1995, 95, 637. (24) Van Santen, R. A.; van Beest, B. W. H.; de Main, A. J. M. Guidelines for Mastering the Properties of Molecular SieVes; NATO ASI Series, Ser. B: Physics, Vol. 221; Barthomeuf, D., Derouane, E. G., Ho¨lderich, W., Eds.; Plenum Press: New York, 1990; p 201. (25) White, J. C.; Nicholas, J. B.; Hess, A. C. J. Phys. Chem. 1997, 101, 590. (26) Scha¨fer, A.; Huber, C.; Ahlrichs, R. J. Chem. Phys. 1994, 100, 5929. (27) Hay, P. J.; Wadt, W. R. J. Chem. Phys. 1985, 82, 299. (28) Internet address: http://www.emsl.pnl.gov:2080/forms/basisform.html. Extensible Computational Chemistry Environment Basis Set Database, Version 1.0; Molecular Science Computing Facility, Environmental and Molecular Sciences Laboratory, Pacific Northwest Laboratory, P.O. Box 999, Richland, WA 99352. (29) Russo, T. V.; R. L. Martin, R. L.; Hay J. J. Phys. Chem. 1995, 99, 17085. (30) Glendening, E. D.; Feller, D.; Thompson, M. A. J. Am. Chem. Soc. 1994, 116, 10657. (31) Feller, D.; Glendening, E. D.; Woon, D. E.; Feyereisen, M. W. J. Chem. Phys. 1995, 103, 3256. (32) Gaussian 94, Revision D.4; Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Gill, P. M. W.; Johnson, B. G.; Robb, M. A.; Cheeseman, J. R.; Keith, T.; Petersson, G. A.; Montgomery, J. A.; Raghavachari, K.; Al-Laham, M. A.; Zakrzewski, V. G.; Ortiz, J. V.; Foresman, J. B.; Cioslowski, J.; Stefanov, B. B.; Nanayakkara, A.; Challacombe, M.; Peng, C. Y.; Ayala, P. Y.; Chen, W.; Wong, M. W.; Andres, J. L.; Replogle, E. S. Gomperts, R.; Martin, R. L.; Fox, D. J.; Binkley, J. S.; Defrees, D. J.; Baker, J.; Stewart, J. P.; Head-Gordon, M.; Gonzalez, C.; Pople, J. A. Gaussian Inc.: Pittsburgh, PA, 1995. (33) Becke, A. D. J. Chem. Phys. 1993, 98, 5648. (34) Lee, C.; Yang, W.; Parr, R. G. Phys. ReV. B 1988, 37, 785. (35) Boys, S. F.; Bernardi, F. Mol. Phys. 1970, 19, 553. (36) Sauer, J.; Zahradnı´k, R. Int. J. Quantum Chem. 1984, 26, 793. (37) Carpenter, J. E.; Weinhold, F. J. Mol. Struct. (THEOCHEM) 1988, 169, 41. (38) Ward, J. W. Zeolite Chemistry and Catalysis; Rubo, J., Ed.; American Chemical Society Monograph No. 171; American Chemical Society: Washington, DC, 1976; p 226, 233. (39) Ward, J. W. J. Catal. 1968, 10, 34. (40) Gibbs, G. V.; Meagher, E. P.; Smith, J. V.; Pluth, J. J. ACS Symp. Ser. 1977, No. 40, 19. (41) Barthomeuf, D. Catal. ReV. 1996, 38, 521. (42) Scokart, P. O.; Rouxhet, P. G. Bull. Soc. Chem. Belg. 1981, 90, 983. (43) Scokart, P. O.; Rouxhet, P. G. J. Chem. Soc., Faraday Trans. 1 1980, 76, 1476. (44) Sanderson, R. T. Chemical Bonds and Bond Energy; Academic Press: New York, 1976. (45) Sanderson, R. T. J. Am. Chem. Soc. 1983, 105, 2259. (46) Barr, T. L.; Lishka, M. A. J. Am. Chem. Soc. 1986, 108, 3178. (47) Okamoto, Y.; Ogawa, M.; Maezawa, A.; Inanaka, T. J. Catal. 1988, 112, 427. (48) Okamoto, Y. Zeoraito 1993, 10, 195. (49) Setzer, W. N.; von Schleyer, P. AdV. Organomet. Chem. 1985, 24, 353. (50) Weitkamp, J.; S. Ernst, S.; Hunger, M.; Ro¨ser, T.; Huber, S.; Schubert, V. A. Thomasson, P.; H. Kno¨zinger, H. Stud. Surf. Sci. Catal. 1996, 101, 731. (51) Lamberti, C.; Bordiga, S.; Geobaldo, F.; Zecchina, A.; Otero Area´n, C. J. Chem. Phys. 1995, 103, 3158. (52) Bordiga, S.; Scarano, D.; Spoto, G.; Zecchina, A.; Lamberti, C.; Otero Area´n, C. Vib. Spectrosc. 1993, 5, 69. (53) Bordiga, S.; Escalona Platero, E.; Otero Area´n, C.; Lamberti, C.; Zecchina, A. J. Am. Chem. Soc., Faraday Trans. 1992, 137, 179. (54) Zecchina, A.; Bordiga, S.; Spoto, G.; Marchese, L.; Petrini, G.; Leofanti, G.; Padovan M. J. Phys. Chem. 1992, 46, 4985. (55) Zecchina, A.; Bordiga, S.; Spoto, G.; Marchese, L.; Petrini, G.; Leofanti, G.; Padovan M. J. Phys. Chem. 1992, 46, 4991. (56) Bo¨se, H.; Fo¨rster, H.; Schumann, M. Proceeding of the 6th International Zeolite Conference; Olsan, D., Bisio, A., Eds.; Butterworth: Reno, NV, 1983; p 201. (57) Kustov, L. M.; Kazansky, V. B.; Beran, S.; Kubelkova, L.; Jiru, P. J. Phys. Chem. 1987, 91, 7833. (58) Ugliengo, P.; Garrone, E.; Ferrari, A. M.; Zecchina, A.; Otero Area´n, C. To be published.