RT Dissertation/Thesis
T1 GAFit : a computational tool kit for parameterizations of potential energy surfaces
A1 Rodríguez Fernández, Roberto
A2 Universidade de Santiago de Compostela. Facultade de Química. Departamento de Química Física. Centro Singular de Investigación en Química Biolóxica e Materiais Moleculares (CIQUS), 
K1 PES (Potential Energy Surfaces)
K1 GAFit
K1 MOPAC
K1 genetic algorithm
K1 potential energy
K1 surface
AB The potential energy surface of a molecular system governs many of its chemical properties, and particularly, the dynamics, that is, the spatial evolution of nuclei with time.Most of the chemical dynamics simulations performed nowadays involve integration of the classical equations of motion, calculating the forces on atoms at each step either directly by electronic structure calculations (i.e., “on-the-fly” or direct dynamics) or from analytical PES (Potential Energy Surfaces). In principle, the direct dynamics approach may be the preferred option for simulations of reactive systems that include a small number of atoms, because one avoids the construction of the analytical surface.The use of analytical PES, however, has a clear advantage in terms of CPU-time costs, being mandatory in molecular dynamics simulations of systems composed by thousands of atoms. Even for small-size systems, the use of an analytical surface may be a convenient choice. If it is developed with care, it may be almost as accurate as the exact surface corresponding to the electronic structure method used as a reference for its construction.The development of analytical PES may be facilitated by using optimization methods, and many research groups have been using them for their particular purposes. However, to our knowledge, there is not a general code that allows users to parametrize analytical surfaces in a relatively easy way. The aim of the present work was to write a suite of programs to assist users in developing analytical surfaces. This suite of programs is called GAFit.GAFit is a package of programs initially developed to facilitate fittings of intermolecular potentials and reparametrizations of semiempirical Hamiltonians. However, it can be easily adjusted for other purposes in which fittings of a series of parameters are needed. The core of the package is the genetic algorithm developed by Marques and co-workers (Marques, J M C; Prudente, F V; Pereira F B; Almeida M M; Maniero A M and Fellows C E. “A new genetic algorithm to be used in the direct fit of  potential energy curves to ab initio and spectroscopic data”. In: Journal of Physics B: Atomic, Molecular and Optical Physics 41.8 (2008), p. 085103.).The functionality of the package was extended separating the core itself from the fitting targets. Now, users can choose, upon their programming skills, from introducing their custom potentials directly into code, use an easy pre-coded potential template to do so, or for those with no programming knowledge at all, use an analytical expression or the most used potentials coded just ready to use. A complete set of tools were added to the package to facilitate the creation and configuration of input files.In addition, an external interface was developed to interact with external programs.Using this interface the tools needed to use GAFit to parametrize the MOPAC program were developed. A further MOPAC enhanced interface permits running parallel copies of MOPAC to speed up calculations and face up some problems encountered during the first stage development.
YR 2014
FD 2014-11-05
LK http://hdl.handle.net/10347/11771
UL http://hdl.handle.net/10347/11771
LA eng
DS Minerva
RD 24 abr 2026