Ab Initio Study of the Adducts of Small Molecules with the Isolated Hydroxyl of Silica and the Brønsted Site in Zeolites: A Comparison between B3-LYP and MP2 Methods B. Civalleri, E. Garrone, and P. Ugliengo* Dipartimento di Chimica Inorganica, Chimica Fisica e Chimica dei Materiali, UniVersita ´ di Torino, Via P. Giuria 7, 10125 Torino, Italy ReceiVed: July 14, 1997; In Final Form: NoVember 17, 1997 A set of adducts of H 3 SiOH (SIL) and H 3 Si(OH)AlH 3 (BRO), minimal cluster models for the isolated hydroxyl species in silica and the Brønsted site in zeolites, respectively, have been studied by the B3-LYP method using various basis sets, including aug-cc-pVDZ and aug-cc-pVTZ. Molecules considered were H 2 (in two configurations), CO, H 2 O, and NH 3 with SIL and CO only with BRO: results were compared with similar calculations run at MP2 as well as with experimental results for amorphous silica and proton exchanged zeolites. B3-LYP results were as good as MP2, though a larger basis set dependence is seen for the former. As the interactions studied span a sizable range of energy, a number of correlations are drawn, similar to those usually drawn for experimental data. The case of SIL/NH 3 , treated including anharmonicity of the OH motion, shows that SIL is definitively less acidic than the isolated hydroxyl of silica. 1. Introduction A reliable measure of the acidity of hydroxyls on surfaces or in zeolites is the shift in frequency of the O-H stretching mode caused by the interaction with suitable molecules. 1 This may then be related to other observables, such as the proton affinity of the molecule, the deprotonation energy of the hydroxyl species, etc., in agreement with the general knowledge on H-bonding. 2 For these reasons, several authors have studied the interaction of small molecules with acidic sites in catalysts mainly by means of IR spectroscopy. On the computational side, the interaction between a base molecule and a surface acidic hydroxyl is usually treated by adopting a cluster approach in the description of the acidic center, 1 whereby the cluster is terminated by hydrogen atoms. The acidic centers most commonly investigated are the isolated hydroxyl at the surface of amorphous silica and the Brønsted acid site in zeolites, referred to hereafter as SiOH and Si(OH)- Al, respectively. Various cluster sizes may be chosen for both, but the minimal ones are at the moment still mandatory when a high-level treatment is requested, including for example the study of the anharmonicity in the O-H motion. Minimal-size clusters correspond to the two molecules H 3 SiOH and H 3 Si- (OH)AlH 3 , which are not chemically stable as such: the former rapidly condenses to siloxane and may only be studied as a vanishing intermediate, while the latter is unknown. The two molecules are described in detail below and are referred to in the following as SIL and BRO, respectively. Previous computational work has concerned, on one hand, SIL alone 3 and in interaction with ammonia, 4 water, 5 formal- dehyde, 6 hydrogen, 7,8 carbon monoxide 9-12 and N 2 O. 13 As far as BRO is concerned, computational results concern the cluster molecule alone 14 and the interaction with CO, 11,12 C 2 H 2 ,C 2 H 4 , CH 3 -CC-CH 3 , and CH 3 -CC-H. 15 All these calculations have been run at the SCF level (self- consistent field) or at the MP2 level (perturbative Møller- Plesset method truncated at the second term 16 ), that is, with no allowance for electron correlation in the former case and with partial allowance in the latter. An overall agreement between experimental data and computational results has been found. However, computational results show a systematic underevalu- ation of both heats of adsorption and OH frequency shift as caused by the adsorption process when compared with the corresponding measured quantities. This may be due to an intrinsically weaker acidity of the cluster model with respect to the acidic centers at the surface, both with SIL and BRO, but it could also be related to the computational approaches adopted. To check such a possibility, in the present work we resort to the newly developed, highly fashionable density functional (DF) techniques, 17 which account for electron correlation by means of an a priori well-defined exchange correlation functional. The ability of DF techniques, and in particular of the functional adopted in the present work, referred to hereafter as B3-LYP (see methods section), to deal with weak and medium-strength H-bonds has been proved recently. 18,19 The systems investigated are SIL alone and in interaction with ammonia, water, carbon monoxide, and hydrogen and BRO alone and in interaction with CO. All such systems have been chosen because their experimental counterparts have been characterized to some extent. As these cases span a sizable range of interaction energies, a few correlations among observ- ables may be drawn. The case of weakest interaction (SIL/H 2 ) is crucial because previous HF and MP2 calculations yielded conflicting results as to the assumed configuration. 2. Methods All the calculations have been performed at ab initio level, using the program GAUSSIAN-94: 20 the option for the integra- tion grid for the density functional calculations was set to FINE. The DF method adopted is the one proposed by Becke, 21 based on a three-parameter formula as far as the exchange part is concerned, to include some Hartree-Fock exchange, and on the functional proposed by Yang, Lee, and Parr 22 for correlation. The corresponding MP2 computations have also been carried out, when not already available. Basis sets adopted are Dunning’s double- plus polarization functions hereafter referred to as DZP; 23 standard Pople 6-31G(d,p) augmented with a set 2373 J. Phys. Chem. B 1998, 102, 2373-2382 S1089-5647(97)02281-5 CCC: $15.00 © 1998 American Chemical Society Published on Web 03/11/1998