Determinação de propriedades termodinâmicas de reações de esterificação de ácidos graxos a partir da modelagem molecular

• Ano de defesa: 2017
• Autor(a) principal: Silva, Marcos Vinícius Domingues
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Uberlândia; Brasil; Programa de Pós-graduação em Engenharia Química
• Palavras-chave em Português: Engenharia química; Esterificação (Química); Biodiesel; Modelagem molecular; B3LYP; Esterificação; Solvatação; Molecular modeling; Esterification; Solvation; CNPQ::ENGENHARIAS::ENGENHARIA QUIMICA
• Link de acesso: https://repositorio.ufu.br/handle/123456789/18892; http://dx.doi.org/10.14393/ufu.te.2017.81

The esterification reactions of fatty acids, in the context of biodiesel production, are a relevant chemical system to be studied, given the very importance of biodiesel as an alternative, renewable and low polluting fuel, and also a diesel fuel of efficiency comparable to petroleum diesel. The thermodynamic study of these reactions is scarce in the literature. Moreover, the thermochemical data of the most commonly molecules found in raw materials for biodiesel production are equally scarce. Molecular modeling allows the accurate calculation of thermochemical quantities of molecules in general; but there is a dichotomy between accuracy and computational cost; the higher the accuracy, the great the cost. For molecules of many atoms, like most fatty acids, accurate calculations require a great calculation time, and sometimes the calculation is not even possible because of the lack of computer resources. In order to overcome these difficulties, this work presents some semi-empirical techniques for calculating enthalpies of formation in the gas phase for large molecules, such as fatty acids and esters. These techniques consist of defining parametric models that fit well-known experimental data of the proposed set of molecules, which are the fatty acids and the methyl esters formed from the esterification reaction in methanol. These models aim to add corrections to the enthalpies of formation calculated by the B3LYP/6-311+G(d,p) model, whose parameters, such as the number of carbons, hydrogens and double bonds of the molecules are chosen to account for the effect of increasing systematic errors of the B3LYP method when the size of the simulated molecules is increased. The models were adjusted to available experimental data by applying two different techniques: least squares method and neural networks. With the corrected enthalpies of formation, we can calculate the Gibbs free energy and the equilibrium constant of the reactions, to determine the information about the viability and the energetic conditions required by such reactions. The proposed correction models decreased significantly the deviation between the experimental data and the B3LYP calculated enthalpies of formation, and the required precision of 1 kcal mol-1 was achieved. Thus, the application of these models enabled the accurate calculation of enthalpy of formation with reasonable computational cost. The neural network correction method made possible the calculation of enthalpy of formation with higher precision than the least squares method correction method. The application of the corrected values of enthalpy of formation enabled to verify the expected behavior of esterification reactions for biodiesel production, which is the favoring of the reaction by the temperature increase. In addition, for a better description of the esterification reaction, the SMD solvation method with the M06-2X/cc-pVTZ model were used to simulate the reaction condition in solution, with the methanol reagent as the solvent, which is usually used in excess to conduct esterification reactions and to shift the equilibrium towards of ester formation. The application of the proposed model showed that the esterification of acetic acid is not favored by the temperature increase but it is favored by the methanol excess. In a general sense, the molecular modeling proved to be an important tool, and in spite of the limitations of the available computational resources, it provided, together with semi-empirical correction techniques, reliable results regarding the studied systems.