Calculations of the site specific stretching frequencies of CO adsorbed on Li+/ZSM-5

Interaction of the CO molecule with Li+ within ZSM-5 was investigated by means of the combined quantum mechanics/interaction potential function method.

Both, C-on and O-on species were considered.

The scaling method based on the linear correlation between CO bond length and stretching frequency has been applied to calculate CO frequencies in CO(OC)–Li+/ZSM-5 adsorption complexes.

Three types of C-on adsorption complexes with different r(CO) bond lengths, ν(CO) frequencies, and CO binding energies were identified.

The calculated IR spectra of CO adsorbed on the Li+/ZSM-5 system show three distinctive bands at about 2194 cm−1, 2187 cm−1 and 2183 cm−1 for C-on complexes and at about 2116 cm−1, 2114 cm−1 and 2104 cm−1 for O-on complexes, in excellent agreement with experimental data.

Calculated adsorption energies and CO stretching frequencies were used for the simulation of the IR spectra at various CO coverages.


Metal exchanged zeolites have been intensively studied by numerous experimental and theoretical groups due to their remarkable catalytic activity.

The properties of the catalytically active sites depend on the structure and coordination of the exchanged metal cations.

The structure of the active sites in aluminium-rich zeolites can be determined from X-ray and neutron diffraction experiments.1,2

In high silica zeolites with a low content of alkali metal cations, however, information about the metal coordination is obtained mainly from indirect experimental techniques, including the spectroscopic characterization of probe molecules interacting with various types of cationic sites.

Carbon monoxide is one of the most common probe molecules for FTIR measurements on metal exchanged zeolites due to a high sensitivity of the CO stretching frequency to the metal cation coordination with the surrounding zeolite framework.

IR spectra of CO interacting with alkali metals in zeolites show a strong adsorption band blue-shifted with respect to the frequency of gaseous CO (2143 cm−1) and a weak band with a frequency below 2143 cm−1.3,4

It was assumed that the high-frequency band is due to the CO stretching of a C-on adsorption complex, while the low-frequency band was assigned to an O-on adsorption complex.4–7

The existence of M(OC)+ species was also considered in theoretical studies of the CO interaction with alkali metal cations in zeolites.4,5

By means of variable-temperature IR measurements of CO adsorbed on Na+/ZSM-5, Otero-Arean et al8. showed that the appearance of low and high-frequency bands in the CO–M+/ZSM-5 (M+ = alkali metal) systems reflects an isomerisation equilibrium between the C-on and O-on monocarbonyl species rather than the existence of different cationic sites.

The adsorption of CO on Li+/ZSM-5 has been studied by several experimental groups.3,9,10

Various coordination complexes formed upon CO adsorption on Li+/ZSM-5 were used to explain variable-temperature IR spectroscopy measurements for different CO pressures.11

The IR bands at 2195 and 2187 cm−1 were assigned to Li+(CO) and Li+(CO)2 species, and the weak bands at 2110 and 2102 cm−1 were assigned to Li+(OC)(CO) and Li+(OC) species, respectively.

Based on the results of microcalorimetric measurements Savitz et al. doubted a formation of Li+ dicarbonyl species.9

At coverages only slightly above monolayer the observed differential heats of adsorption for CO rapidly dropped to the value measured on silicalite.

Therefore, it was concluded that in contrast to other alkali metal cations the second CO molecule is mostly unable to approach the Li+ sites.

By means of combining FTIR spectroscopic and microcalorimetric measurements Bonelli et al10. suggested that bands at 2193 and 2187 cm−1 are both due to Li(CO)+ species formed on two different types of Li+ sites.

The possibility of a new interpretation of a low-frequency region in light of the existence of these two sites was also considered.

The coordination of alkali metal ions in ZSM-5 was studied theoretically by a combined quantum mechanics/interatomic potential function method and two types of Li+ sites were found depending on the framework Al atom position:12 (i) type I sites where alkali metal ions are coordinated to 3–4 oxygen atoms of the 5- or 6-member ring on the channel wall (denoted M5 or M6 and Z5 or Z6 for sites on the wall of main and zig-zag channels, respectively) and (ii) type II sites where the metal cation is coordinated to two oxygen atoms of the single AlO4 tetrahedron on the intersection edge (denoted as I2 sites).

The notation for alkali metal sites in ZSM-5 introduced in ref. 12 is used in this work (Fig. 1).

The population of type I sites is expected to be larger than the population of type II sites for Li+/ZSM-.512

In this paper the results of a computational study of CO interaction with various types of Li+ sites are used for the interpretation of the IR spectra of CO/Li+/ZSM-5.

This study combined the quantum mechanics/interatomic potential function (QM-Pot)13 method with the scaling method based on the correlation between the CO bond length and the CO stretching frequency (ωCOrCO correlation).

This approach has been used recently for the interpretation of the IR spectra of CO interacting with Cu+/ZSM-.514

The ωCOrCO scaling method was developed for the assessment of highly accurate vibrational frequencies of metal ion carbonyl species interacting with complex molecular environments.

The method combines quantum chemistry approaches directly applicable to the “real” system with benchmark calculations (typically at the CCSD(T) level of theory) on selected model molecules.



The combined QM-Pot method13 has been used to study CO interaction with the Li+ sites in ZSM-5.

Within this approach, the system is divided into two parts: the inner part described at a DFT level (with BLYP exchange–correlation functional15,16) and the outer part described at the computationally less expensive interatomic potential function (IPF) level.

Dangling bonds on the inner part boundary were saturated with hydrogen atoms.

Periodic boundary conditions were applied to the unit cell consisting of 192 T-atoms (191 Si atoms and 1 Al atom) and 384 O atoms.

Rather large definitions of inner parts (described at the DFT level) were used in order to properly account for the CO interaction with framework atoms from the opposite side of the channel wall.14

The inner part consisted of a CO molecule, Li atom, AlO4 tetrahedron and from 12 to 25 SiO4 tetrahedra (13-T to 26-T models, respectively), depending on the particular Li+ site location.

Examples of inner parts used in calculations are given in Fig. 2.

The CO stretching frequencies were calculated from the ωCOrCO correlation,14 using the QM-Pot model with large inner part definition.

Calculations were performed with valence-triple-ζ-plus-polarization function basis set for C, O and Li atoms and valence-double-ζ-plus-polarization function basis set for Al, Si, and H atoms (denoted BS2).17,18

The resolution of identity approximation19,20 was used in BLYP calculations, employing the (9s2p2d1f)/[7s2p2d1f], (12s6p5d1f1g)/[5s3p2d1f1g], (12s6p5d1f)/[5s3p2d1f], (9s3p3d1f)/[7s3p3d1f], (9s3p3d1f)/[7s3p3d1f], and (4s2p1d)/[3s2p1d] auxiliary basis sets21–23 for Li, Si, Al, C, O and H atoms, respectively.

The interactions between the atoms of the outer part and the interactions between atoms of the inner and outer parts were treated at the IPF level, employing the core-shell model potentials developed previously.12,24

The interaction between the CO molecule and zeolite framework was treated with the Lennard-Jones potential with parameters derived from the universal force field.25

Atomic charges on carbon and oxygen atoms of CO were taken from the NBO (natural bond orbital) analysis obtained for the CO/Li+AlSi2O10H8 cluster (±0.4570 and ±0.5710 e for C-on and O-on complexes, respectively).

The interaction parameters for the interaction between the CO molecule with Li+ were not defined since these atoms are always included in the inner part definition (QM) which is treated at the DFT level.

The performance of the BLYP/BS2 level used in calculations on CO/Li+/ZSM-5 can be judged based on the data summarized in Tables 1 and 2 comparing BLYP, B3LYP, MP2 and CCSD(T) results.

Calculations of the CO structural parameters were carried out for 12 distinguishable framework Al positions within the orthorhombic symmetry using the numbering scheme proposed by Koningsveld et al.26

In all cases, both C-on and O-on complexes were considered.

Xn⋯Li+CO and Xn⋯Li+OC (X = H2O, F, Al(OH)4, n = 1, 2)

The CO equilibrium distances and stretching frequencies for a series of model molecules were calculated at the CCSD(T), MP2 and DFT levels.

At the DFT level, the BLYP and B3LYP exchange–correlation functionals15,16,27 were employed.

At the MP2 and DFT levels the geometries were fully optimized using the BS2 basis set defined above.

The CCSD(T) calculations were carried out with the correlation consistent valence-quadruple-ζ basis set with polarization functions28,29 (cc-pVQZ), denoted as BS1.

Three geometry parameters (Al–Li, Li–C and C–O distances) were varied in the CCSD(T) optimization for Al(OH)4⋯Li+CO and Al(OH)4⋯Li+OC clusters (1-T model), all other parameters were held fixed at the corresponding BLYP/BS2 equilibrium values.

The CCSD(T) frequencies were evaluated using the two-dimensional (Li–C, C–O) stretching Hamiltonian (see ref. 14 for more details).

Charge distribution analysis was carried out at the natural bond orbital (NBO) level.30

The ab initio calculations of Xn⋯Li+CO and Xn⋯Li+OC clusters were performed using MOLPRO (CCSD(T)) and Gaussian (MP2, DFT) program suites.31,32

The calculations on CO/Li+/ZSM-5 were carried out with QM-Pot program13 which makes use of the TurboDFT19 and Gulp33 programs for DFT and IPF calculations, respectively.


CO interaction with small molecular ions

The CO equilibrium distances and stretching frequencies of C-on and O-on complexes for Li+CO, H2O⋯Li+CO, (H2O)2⋯Li+CO, F⋯Li+CO and (F)2⋯Li+CO model molecules calculated at the CCSD(T), DFT (employing the BLYP and B3LYP exchange correlation functional), and MP2 levels are summarized in Table 1.

The ωCOrCO correlation derived from calculated data covers the CO frequency range 2108–2224 cm−1 and 2060–2157 cm−1 for C-on and O-on complexes, respectively.

Throughout this work we assume the constant anharmonic correction for the CO stretching vibration of 29 cm−1 calculated previously for monocarbonyl species in various molecular environments.14,34

The Δω correction introduced in eqns. (1) and (2) in ref. 14 was estimated from a comparison of the scaled and CSSD(T) frequencies calculated for the 1-T model which possesses the most important features of the cationic sites in the CO/Li+/ZSM-5 system.

Parameters of the ωCOrCO correlation (see ref. 14) for Li+-exchanged zeolites are given in Table 3.

The scaled harmonic frequencies are evaluated from the following equationωCO[cm−1] = arCO [Å] + b + Δω,where rCO is the equilibrium CO distance obtained from DFT or MP2 geometry optimization.

The Δω correction in eqn. (1) is defined as the difference between the CO harmonic frequency calculated for the 1-T cluster model at the CCSD(T) level and the corresponding scaled ωCO.

The parameters a, b, and Δω are basis set dependent and, therefore, they have to be used only with the BS2 basis employed in our calculations on the CO/Li+/ZSM-5 system.

As can be seen from Table 2, the BSSE corrected CO binding energies for the 1-T model calculated at the B3LYP and BLYP levels agree with CCSD(T) values to within 0.3 kcal mol−1, slightly worse agreement was found for MP2 (0.6 kcal mol−1).

The DFT/BLYP level used in this work seems to underestimate binding energies for monocarbonyl species by 0.2–0.3 kcal mol−1.

The calculated zero-point energy (ZPE) corrections are essentially the same at the DFT and MP2 levels.

The results for Li+(CO)2, Li+(OC)CO and Li+(OC)2 species (1-T cluster model) are also summarized in Table 2.

The CO stretching modes in dicarbonyls are only weakly coupled and therefore, a single band is expected to appear in the FTIR spectra.11

The calculated splitting of a1 and b2 CO frequencies (1-T model, C2v symmetry) is less than 2 cm−1 at all levels of theory.

The fact that geminal dicarbonyls behave as independent oscillators was also reported for Na+(CO)2 species in Na-Y zeolite.35

The relative positions of the dicarbonyl C-on and O-on bands with respect to their monocarbonyl counterparts can be estimated by comparing the results for the 1-T model (Table 2).

At the BLYP level the CO stretching frequencies of C-on and O-on dicarbonyls are shifted by about −14 cm−1 and +5 cm−1, respectively.

The binding energies for the second CO ligand in the dicarbonyl species are significantly lower than for the corresponding monocarbonyls (BLYP binding energies are −2.8 kcal mol−1 and −1.5 kcal mol−1 for C-on and O-on species, respectively).

Similar results were obtained at the B3LYP level.

Interaction of CO with Li+/ZSM-5 system

Two types of Li+ sites were identified in ZSM-5 by Kučera et al.12

It was shown that the relative stability of the type I and type II sites depends on the location of the framework Al atom.12

When the framework Al is at T4, T8, T10, or T11 only the channel wall sites (type I) were found.

However, for all other positions of framework aluminium both type I and type II sites coexist (.ref. 12)

The stability of type I and type II sites is roughly the same when Al is at T2, T3, T6, or T12.

The type I site is slightly more stable (about 3 kcal mol−1) than the type II site when Al is at T1, T5, T7, T9 (see Supplementary data of .ref. 12)

The interaction of CO with both Li+ site types was considered in this study.

Interaction energies (including BSSE and ZPE), r(CO) bond lengths, and CO frequencies are summarized in Table 4 for both C-on and O-on adsorption complexes.

Based on the interaction energies and vibrational frequencies the CO/Li+/ZSM-5 complexes can be divided into three groups: (i) intersection complexes (type II complexes) where upon interaction with CO the Li+ ion stays coordinated to just two oxygen atoms of a single AlO4 tetrahedron located at the intersection edge, (ii) channel wall complexes where an AlO4 tetrahedron is located on the intersection edge (type Ia complexes), and (iii) channel wall complexes where an AlO4 tetrahedron is located on the channel wall (type I complexes).

Type I complexes can be found only when framework aluminium is at T4, T8, T10, or T11 positions.

Type Ia complexes are always adjacent to intersection sites and these complexes can be viewed as channel wall complexes in the vicinity of a channel intersection.

For type I and type Ia complexes the Li+ ion is coordinated to 3–4 oxygen atoms of five- or six-member ring containing AlO4 tetrahedron.

As noted above for a framework Al atom located on the intersection edge both type I and type II complexes can be found.

Examples of all CO/Li+/ZSM-5 complex types are depicted in Fig. 2.

The Li+ ion is considered to be coordinated to a framework oxygen atom if the Li+–O distance is shorter than 2.33 Å.12

Coordination of the Li+ ion remains the same upon interaction with the CO molecule and the Li+ ion positions are only slightly influenced by the CO adsorption.

The Li–O bonds typically do not lengthen by more than 0.02 and 0.1 Å upon the CO adsorption for intersection and channel wall complexes, respectively (Table 5).

The only exception is the M7/T7 site (Li+ located at M7 site/Al atom at T7 position), where Li+ was originally coordinated to 3 framework oxygen atoms of the six-member ring and upon CO adsorption the Li+ ion loses the coordination to the non-AlO4 oxygen atoms and is coordinated as at other intersection sites.

The C-on OC–Li+/ZSM-5 complex is typically more stable by about 2 kcal mol−1 than the O-on CO-Li+/ZSM-5 complex for a particular Li+ site.

In a few cases this difference is smaller and in one case (I2/T5 site) the O-on complex is more stable by 0.4 kcal mol−1 than the C-on complex.

However, this may be an artefact of the model used (large partial charges on C and O atoms used for O-on complexes, for discussion see the next section).

First we discuss the C-on adsorption complexes.

The CO molecule is most strongly bound at intersection (type II) sites.

For these adsorption complexes the CO stretching vibration shows the largest shift to higher frequency compared to the gas phase CO (blue-shift of 50 cm−1).

The interaction energies are from −6.1 to −7.6 kcal mol−1 and νCO are in the range 2191–2197 cm−1.

Slightly weaker CO interactions with Li+/ZSM-5 (–5.0 to –5.7 kcal mol−1) and smaller CO frequencies (2184–2190 cm−1) were found for type Ia sites.

The smallest CO interaction energies (−3.4 to −4.2 kcal mol−1) were found for type I sites where also the νCO frequencies are the smallest (2182–2183 cm−1).

The differences in νCO and the interaction energies between intersection and channel wall complexes are due to the large changes in the Li+ ion coordination.

However, in both channel wall complexes (type I and Ia) the Li+ is three- or four-coordinated to the zeolite framework atom.

The differences in νCO and interaction energies are mainly due to the changes in Li+⋯Of distances that are shorter in type I sites than in type Ia sites.

The details of Li+ ion coordination at selected sites and details of the Li+ ion coordination at corresponding OC–Li+/ZSM-5 complexes are summarized in Table 5.

The relative stability of O-on complexes increases in the same order as found for C-on complexes (I < Ia < II).

However, the CO stretching frequencies increase for O-on complexes in the opposite order to the C-on complexes.

This seemingly surprising behavior can be understood in terms of the character of occupied molecular orbitals of CO (see next section for details).


The calculated CO stretching frequencies for C-on OC–Li+/ZSM-5 complexes are in the range 2182–2197 cm−1, in excellent agreement with experimental data.3,10,11

The calculated blue-shift in CO frequencies is due to two completely independent effects: first, the interaction of CO with alkali metal cations causes a change in polarization of the bonding molecular orbitals (this will be discussed below) and, second, the oxygen atom of the CO molecule (OCO) has a repulsive interaction with the opposite side of the channel wall.

Comparing the CO frequencies calculated using small models (not accounting for the repulsion with the opposite side of the channel wall) and using large models (properly describing the CO interaction with the zeolite framework) we conclude that the effect of the channel wall on ν(CO) is 11–13 cm−1.

This effect does not depend on the Li+ coordination number nor on the structure of the OC–Li+/ZSM-5 adsorption complex.

This is significantly less than found recently for the OC–Cu+/ZSM-5 system14 where an effect as large as 20 cm−1 was observed.

In fact, in the OC–Cu+/ZSM-5 system the blue-shift of the CO frequencies upon the adsorption was found to be purely due to the repulsive interaction with the channel wall.

The relatively small effect of the channel wall repulsion on ν(CO) found for the OC–Li+/ZSM-5 system seemingly contradicts the results of the charge distribution analysis on the C and O atoms (the NBO charges are ±0.3565 and ±0.4570 e for the OC–Cu+/ZSM-5 and OC–Li+/ZSM-5 systems, respectively).

The large CO polarization found for the OC–Li+/ZSM-5 system implies a strong electrostatic repulsion between the OCO and framework oxygen atoms (Of) and, therefore, a large blue-shift of ν(CO) due to this repulsion could be anticipated.

To understand the relative magnitude of the channel wall effect for the Li+/ZSM-5 and Cu+/ZSM-5 systems it is important to compare the nature of the CO interaction in these systems.

In order to bind CO efficiently the Cu+ ion changes its coordination with the zeolite framework and stays coordinated to just two oxygen atoms of the single framework AlO4 tetrahedron upon the interaction with the CO molecule.

In the OC–Cu+/ZSM-5 adsorption complex the Cu+ ion has trigonal planar hybridization with the CO bond in the Of–Cu–Of plane (Of framework atom in coordination with Cu+).

In this system it is energetically advantageous to keep the trigonal planar hybridization on Cu+ at the expense of a small increase of OCO repulsion with the zeolite channel wall.

On the contrary, the Li+ coordination with the framework remains unchanged upon the formation of the OC–Li+/ZSM-5 adsorption complex.

The interaction of CO with Li+/ZSM-5 is almost entirely electrostatic in nature, therefore, it is only important to keep the Li+⋯CO coordination linear (in order to maximize the monopole–dipole interaction) and the orientation of Li+⋯CO with respect to the zeolite framework is relatively energy independent.

Thus, the OC–Li+/ZSM-5 complex structure is driven by the minimization of the OCO⋯Of repulsion.

As a result the blue-shift in ν(CO) is smaller in OC–Li+/ZSM-5 than in the OC–Cu+/ZSM-5 system.

Based on the CO stretching frequencies and interaction energies three groups of OC–Li+/ZSM-5 complexes can be distinguished.

The largest ν(CO) and largest CO interaction energies were found for the OC–Li+/ZSM-5 adsorption complexes on the channel intersection (type II) while the smallest ν(CO) and smallest interaction energies were found for complexes on the channel wall sites (type I complexes).

The differences in CO frequencies are not due to the effect of the channel wall repulsion (see above), instead, it can be explained by the differences in Li+ coordination in particular site types.

The almost purely electrostatic interaction of CO with Li+/ZSM-5 depends on the partial charge on the Li+ ion in a particular site in ZSM-5.

The NBO analysis shows a larger partial charge on Li+ for both bare Li+/ZSM-5 and OC–Li+/ZSM-5 models, shorter r(CO) distance, and larger Wiberg bond index for type II sites on the channel intersection than for type I sites on the channel wall.

The charge on the Li+ ions becomes smaller with an increasing number of framework oxygen atoms in coordination with Li+.

In addition, the partial charge on Li+ increases with the increasing average Li–Of bond length.

The average Li–Of bond length is larger in type Ia sites than in type I sites.

It can be concluded that with the increasing coordination of Li+ with the framework the charge on Li+ becomes smaller and, therefore, the CO interaction with Li+/ZSM-5 becomes weaker, CO bond polarization smaller, and CO frequency smaller.

The experimentally determined IR bands at 2195 and 2187 cm−1 were originally assigned to mono- and di-carbonyl complexes with Li+/ZSM-.511

However, this interpretation is somewhat doubtful in terms of the results of microcalorimetric measurements.

Savitz et al. showed that the differential heats of adsorption rapidly drop above the monolayer coverage.9

The most recent IR experiments for the CO/Li+/ZSM-5 system were interpreted in terms of site-specific CO stretching frequencies for two types of Li+ sites in ZSM-.510

The experimental IR bands at 2193 and 2187 cm−1 were assigned to two types of Li+ sites characterized by different CO adsorption energies.

The shift of the 2187 cm−1 band to lower frequencies for CO coverage approaching monolayer was interpreted as a consequence of the formation of dicarbonyl species characterized by the IR band at 2185 cm−1.

We have attempted to estimate the CO frequency for dicarbonyl species from a simple 1-T model (Table 2).

The CO frequency calculated for dicarbonyl species from the ωCOrCO correlation (2185 cm−1) thus supports the experimental interpretation of Bonelli et al.10

However, the calculated interaction energy for the second CO molecule is small compared to the interaction energy of the first CO molecule.

From the 1-T model calculations (Table 2) it appears that the second CO molecule interaction energy is only about 2–3 kcal mol−1 (including BSSE and ZPE corrections).

Our computational results predict three bands at about 2194, 2187 and 2183 cm−1 in the IR spectra of CO adsorbed on Li+/ZSM-5, all of them due to the monocarbonyl C-on species.

These results offer an alternative interpretation of the experimentally observed frequency shifts in the IR spectra.

Since the sites with the least favorable binding energy show the lowest CO stretching frequency, changes in the CO coverage should result in corresponding changes in the spectra.

It can be expected that at high CO equilibrium pressure when nearly all the Li+ sites are populated the lower frequency component should peak at about 2185 cm−1.

Since the type I and Ia complexes represent a major fraction of Li+ sites, the lower frequency band should also predominate in the IR spectra at very high coverage.

Therefore, the experimental spectra can be interpreted in terms of monocarbonyl species only, without the need to invoke the existence of dicarbonyls.

The IR spectra of the CO/Li+/ZSM-5 system (Fig. 3) were calculated in order to compare theoretical and experimental results.

Calculated spectra are based on a simple model (see Appendix A) and it should be stressed that this model is by no means quantitative and it should only demonstrate that all the features of the experimental spectra can be attributed to the monocarbonyl species only.

At low CO coverage the spectra are dominated by the 2194 cm−1 band (type II sites on the channel intersection) where the CO interaction with Li+/ZSM-5 is the strongest.

However, it is apparent that even for a low CO coverage the spectra are not symmetrical showing a weak band at lower frequencies.

This lower frequency band becomes more populated with increasing CO coverage.

At about θ = 0.7 ML both bands have similar maximum.

At higher CO coverage the low frequency band maximum shifts to lower frequencies.

This shift is not due to the formation of dicarbonyl, instead it is due to the increasing population of type I sites relative to type Ia sites (Table 4).

Experimental IR spectra show qualitatively the same behavior (see Fig. 3 in .ref. 10)

The weak band in the 2100–2120 cm−1 region was assigned to O-bonded CO–Li+/ZSM-5 species.10

Calculated IR spectra for the O-on CO–Li+/ZSM-5 complexes show three types of Li+/ZSM-5 adsorption sites characterized by CO stretching bands at 2104, 2114 and 2116 cm−1 corresponding to type II, type Ia and type I Li+ sites, respectively.

Similarly as for C-on complexes the CO interaction energies for O-on complexes decreases from type II sites to type I sites.

For the O-on complexes the CO frequencies are smallest for type II sites and largest for type I sites, opposite to what was found for C-on complexes.

This behavior is in agreement with the IR experiments.

For a low CO coverage the experimental spectra exhibit a dominant peak at the lower energy part of the 2100–2120 cm−1 region while for a high CO coverage this spectral region is dominated by the higher energy band.

Based on the calculations the 2104 cm−1 band (type II site) predominates in the IR spectra at low coverage, while the 2116 cm−1 and 2114 cm−1 bands gain intensity with increasing equilibrium CO pressure.

The blue-shifting of the higher frequency band maximum may be difficult to observe because of very small differences in the CO frequency (2 cm−1) and low intensity.

Our estimate of the CO stretching frequency for dicarbonyl species calculated from 1-T model is about 2109 cm−1.

The opposite dependence of νCO on Li+ coordination (and thus on the partial charge on Li+) observed for C-on and O-on complexes can be qualitatively understood in terms of the polarization changes of bonding molecular orbitals in CO. The positive charge on the Li+ cation for the C-on complex partially reduces the polarization of σ and π bonding orbitals of the CO molecule and the CO bond becomes stronger.

On the contrary, the positive charge on the Li+ cation for the O-on complex increases the polarization of the bonding orbitals of CO, thus, it reduces the CO bond strength.36

In addition to σ polarization the effect of the opposite channel wall also contributes to the overall CO frequency.

Due to the positive sign of the partial charge on the carbon atom the interaction of carbon with the channel wall is attractive.

The estimated effect of the channel wall is a 12–14 cm−1 red-shift.

Both effects (orbital polarization and channel wall repulsion) shift the CO frequency towards the lower values for O-on species.

Interaction energies calculated for C-on complexes of type I and II are 4 and 7 kcal mol−1 in good agreement with available calorimetric data.9,10

The calculated interaction energies for O-on complex are lower by about 2–3 kcal mol−1, in good agreement with the estimate from variable-temperature IR experiments (2 kcal mol−1).37

However, for the I2/T5 site we found that the O-on complex is slightly more stable than the C-on complex.

The analysis of the model convergence (increasing the size of the inner part) shows that the interaction energies of O-on complexes depend on the model size more than C-on interaction energies or νCO of C-on or O-on complexes.

We believe that CO frequencies and interaction energies for C-on complexes are well converged with respect to the further enlargement of the inner part of the combined QM-Pot model.

However, interaction energies for O-on complexes should be calculated with even larger models than used in this study.

First, the partial charges on C and O atoms used for the O-on complexes are larger than those used for the C-on complexes (both based on NBO analysis) and, second, the carbon atom directing into the channel bears a positive partial charge, thus, the interaction of the C atom with framework oxygen atoms is attractive.

Calculations with the inner part consisting of 30–50 framework T-atoms could provide more reliable interaction energies for O-on complexes, however, due to the computational requirements of such calculations these were not carried out.

It should be stressed that reported O-on interaction energies should be interpreted only in a qualitative way.

Application of the scaling method developed for the assessment of highly accurate vibrational frequencies of metal ion carbonyl species interacting with the zeolite framework to the CO–Li+/ZSM-5 system demonstrates that the ωCOrCO scaling approach is capable of predicting FTIR spectra of CO adsorbed on metal exchanged zeolites basically with CCSD(T) accuracy (to within a few wavenumbers).

The method is computationally favorable since it only requires performing the geometry optimization at the DFT level of theory.

The large reduction of CPU demands is achieved by the use of resolution of identity approximation.

To obtain reliable CO stretching frequencies one has to use a sufficiently large cluster model for the inner (QM) part definition of the QM-Pot computational scheme, namely interaction of the probe molecule with the opposite channel wall must be included.

Calculated CO stretching frequencies are in very good agreement with experimental data.3,10,11

Computational results support the conclusion of Bonelli et al10. that the CO frequencies are site specific for the CO/Li+/ZSM-5 system.

Bludský et al. recently used the same computational approach (combined QM-Pot approach together with a ωCOrCO scaling method) in the investigation of the CO interaction with Cu+ exchanged ZSM-514 and concluded that CO frequencies are not site-specific in that system in agreement with experimental results.38

Similarly it has been concluded from the combination of computational and experimental study of the CO/Cu+/FER system that the CO stretching frequencies are not site-specific.39

The site-specificity of CO stretching frequencies found for the CO/Li+/ZSM-5 system by Bonelli et al. and theoretically confirmed here together with the interpretation of spectral features at the atomic scale level offers a variety of interesting applications.

Further exploration of this site-specificity for other high-silica zeolites where the information about the cation coordination and localization are difficult to obtain is in progress.


The calculated CO stretching frequencies for CO adsorbed on the Li+/ZSM-5 system show three distinctive bands at about 2194 cm−1, 2187 cm−1 and 2183 cm−1 for C-on complexes and at about 2116 cm−1, 2114 cm−1 and 2104 cm−1 for O-on complexes.

Theoretical predictions of IR spectra of the CO/Li+/ZSM-5 system are in excellent agreement with the available experimental data.

The formation of dicarbonyl species cannot be fully ruled out, however, the corresponding CO bands are likely to overlap with the spectral features due to the monocarbonyl species.

The site-specificity of CO frequencies in the CO/Li+/ZSM-5 system suggested by Bonelli et al10. is confirmed.

From a combination of experimental variable pressure IR spectra and the theoretical interpretation of individual spectral features information about the coordination, localization, and distribution (population) of extra-framework metal cations can be obtained.

The results obtained for CO/Li+/ZSM-5 together with the recently published results for CO/Cu+/ZSM-5 offer a new insight into the understanding of site-specificity of vibrational frequencies of the CO probe molecule.

CO frequencies in the CO/M+/ZSM-5 system are driven by CO interaction with the opposite site of the channel wall (repulsive interaction for the C-on complex resulting in the blue-shift in CO frequencies) and by the CO interaction with the metal cation.

The strong CO interaction with the Cu+ sites results in the changes in Cu+ coordination with the framework oxygen atoms.

The coordination and hybridization of the Cu+ ion is the same for all cationic sites upon interaction with CO and, as a consequence, the CO stretching frequencies are not site-specific in this system.

In addition, the CO⋯Cu+ interaction is stabilized by two effects, σ-donation (polarization) and π-backdonation, each of them having an opposite effect on the CO frequency.

As a consequence, CO adsorption energies and CO frequencies do not correlate.

On the contrary, the Li+ ion coordination does not change upon interaction with CO. The charge on the Li+ ion decreases with increasing Li+ coordination with the framework.

Both, the CO interaction energy for a particular Li+ site and the CO frequencies depend on the partial charge on the Li+ ion.

Therefore, the CO stretching frequencies for the CO/Li+/ZSM-5 system are site specific, probing the information about the Li+ coordination within the zeolite matrix.

In addition, the CO frequency correlates with the CO interaction energies.

It can be anticipated that the same rationalization will hold for other monovalent cations and other zeolite frameworks.