Application of Hybrid Functionals to the Modeling of NO Adsorption on Cu−SAPO-34 and Co−SAPO-34: A Periodic DFT Study

The structural and chemical properties of Cu- and Co-exchanged silicoaluminophosphates with a chabazite-type structure (SAPO-34) and the adsorption of NO molecules on the extraframework cations have been studied using periodic density functional (DFT) calculations using both convential gradient-corrected exchange-correlation functionals and hybrid functionals mixing exact (i.e., Hartree−Fock) and DFT exchange. To assess the importance of the cation-support interaction, comparative studies of NO adsorption on isolated metal atoms and ions have been performed. We find that whereas the geometrical properties for stoichiometric aluminophosphate (ALPO-34) and protonated silicoaluminophosphate (H−SAPO-34) calculated using both types of functionals are essentially on par, the hybrid functionals predict much wider energy gaps in better agreement with experiment. Significant differences arise for the coordination of extraframework cations in the metal-exchanged silicoaluminophosphates (Cu−SAPO-34 and Co−SAPO-34) and, in particular, for the electronic eigenstates of the cations. The different metal−framework coordination and the differences in the electronic properties lead to markedly different predictions of the chemical reactivities: hybrid functionals predict a weaker binding of molecular adsorbates such as NO. The reduction of the adsorption energy is more pronounced for the magnetic Cu(II) and Co(II) cations than for the nonmagnetic Cu(I) cation. This is a consequence of a much larger exchange splitting of the d states of the cation and also of an enhanced ligand-field splitting due to the binding to the framework. The occupied d states of the transition-metal cations are shifted to larger binding energies, and the empty d states are located in the upper part of the energy gap. Hybrid functionals tend to overestimate the N−O stretching frequency even for the isolated molecule and for gas-phase nitrosyls, whereas conventional functionals predict eigenvalues that are too low. This tendency is also reflected in the vibrational spectra of the Cu(Co)−NO adsorption complexes, whereas the adsorption-induced frequency shifts predicted by both types of functionals are in reasonable agreement with experiment. Significant differences are found in the prediction of the spin states of the adsorption complexes. Altogether, we conclude that conventional gradient-corrected DFT functionals describe these complex materials with good accuracy, whereas the inclusion of exact exchange in the hybrid functionals does not lead to improved performance for the metal-exchanged systems.