Problemas diretos e inversos no estudo de estrutura de sistemas líquidos e gasosos

• Ano de defesa: 2014
• Autor(a) principal: Cayo Emílio Monteiro Gonçalves
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; UFMG
• Palavras-chave em Português: Problemas inversos; Rede neural de hopfield; Função de distribuição radial; Dinâmica molecular; Físico-química; Problemas inversos (Equações diferenciais); Dinamica molecular; Redes neurais (Computação)
• Link de acesso: http://hdl.handle.net/1843/SFSA-9J3TC6

This work has as main objective the study of microscopic properties oftwo systems, clusters of noble gases and pure liquids. In both the property aimed is the internal structure, but in each case it is approached differently. In the first one will be emphasized the mechanism of growth and the structure optimization for the cluster ArnNO2 (n=1-22), of atmospheric interest. To this, the theoretical foundations for the development of the study will be presented, specifically the Molecular Dynamics, a classic and deterministic method of analyzing potential energy surfaces to remove, mainly, information about the system structure. The potential energy surface used was a non-rigid,which allows studies about the relaxation of the structure, that is not taken in account in the commercial programs. The second differences on the binding energies were calculated and the structures of relative stability are discussed. The second study deals with the structure of liquids, usually described by the radial distribution function, that is a function which counts the number of atoms distributed radially around a central atom. This property has a functional relationship with the scattering intensity of x-rays or neutrons, this property macroscopic and measurable. To retrieve internal informations about the system from experimental measures is said to be an Inverse Problem. Due to the fact that measures have inherent experimental error, the problem may not have one or more of three singularities, continuity, existence and unicity. When this occur, the problem is classified as ill posed, and requires especial numerical methods for their solution, and three different ones were applied to obtain the radial distribution function of liquid argon. They are the Singular Value Decomposition, the Tikhonov regularization and the Hopfield Neural Network. A comparison is made of the effectiveness of these methodologies and is presented an easier path to the study of complex liquid systems, which is use as initial information the radial distribution function of the gas phase.