Avaliação do uso de métodos mecânico-quânticos na obtenção de dados termodinâmicos de sistemas reacionais visando a produção de hidrogênio

Abstract

Process optimization requires knowledge of equilibrium constants. From these, the mole fractions of species present at equilibrium can be determined, allowing the calculation of the maximum theoretical conversion of a given system. However, many experimental thermodynamic data are not available. This is due to experimental complexity and the high cost involved in obtaining such data. Thus, the use of computational resources for the acquisition of these data proves to be of great interest. In recent decades, quantum-mechanic methods have been used coupled with computational techniques for the prediction of thermodynamic properties. Thus, the present work provided a comparison between the values of thermodynamic quantities obtained by some methods based on Quantum Theory with data calculated from experimental energies of formation at 298.15 K. For that, the ab-initio methods Hartree-Fock (HF), Gaussian-1 (G1), Gaussian-2 (G2), Gaussian-3 (G3), Gaussian-4 (G4) and CBS-QB3, as well as the approach based on Density Functional Theory, represented by the hybrid functional B3LYP, were used. The experimental data of standard heat of formation (fH0 f) and entropy (fS0) of the chemical species in the desired temperatures were extrapolated by integrating the specific heat at constant pressure as a function of temperature. From these quantities, the Gibbs free energy for the reactions involved in the systems of steam reforming of methane and of Liquefied Petroleum Gas (LPG) was obtained, and consequently, their equilibrium constants. As previously expected, the HF method provided the largest deviations from values based on experimental quantities. The hybrid functional B3LYP showed better results compared to the HF method, although with high deviations from the extrapolated experimental data. It was noticed a remarkable improvement on the predictive capability of DFT approach when diffuse and polarized functions were added to the basis set employed. The compound method G1 proved to be unsatisfactory, since it has improved slightly compared to the DFT approach with no diffuse and polarized functions on the basis set and showed a significantly higher computational cost. Statistical analysis proved that the results provided by G2 and CBS-QB3 methodologies are equivalent, in the present context. Even with a considerable computational cost, smaller deviations between the experimental and theoretical quantities were obtained by using the G2, G3, and G4 methods.