Benchmarking semiempirical and DFT methods for the interaction of thiophene and diethyl sulfide molecules with a Ti(OH)4(H2O) cluster.

Semiempirical methods pm6 and pm7 as well as density functional theory functionals exchange LSDA, exchange-correlation PW91 and PBE, hybrid B3LYP1 and PBE0 were compared for energy and geometry of thiophene, diethyl sulfide (DES) molecules and their binding to a frozen Ti(OH)4(H2O) complex having one coordinatively unsaturated Ti5C site representing small fragment of TiO2 anatase (001) surface. PBE0/6-31G(d) with DFT-D3 dispersion correction was the best method for description of thiophene and DES molecules geometries as comparison with experimental data demonstrated. Semiempirical methods pm6 and pm7 resulted in only three of four possible binding configurations of thiophene with the Ti(OH)4(H2O) complex while pm7 described correctly the enthalpy and all configurations of DES binding with the Ti(OH)4(H2O) complex. SBKJC pseudopotential and LSDA with and without dispersion correction produced flawed results for many configurations. PBE0 and PBE with and without dispersion correction and PW91 with 6-31G(d) basis set systematically produced dependable results for thiophene and DES binding to the Ti(OH)4(H2O) complex. PBE0-D3/6-31G(d), B3LYP1-D3/6-31G(d), and PBE-D3/6-31G(d) gave best match of binding energy for thiophene while PBE0/6-31G(d) gave best match of DES binding energy as comparison with CCSD(T) energy demonstrated. On the basis of the superior results obtained with PBE0/6-31G(d), it is the recommended method for modeling of adsorption over TiO2 surfaces. Such a conclusion is in agreement with recent literature.