| Geometry optimization has long been an active research area in theoretical chemistry. Many algorithms currently exist for the optimization of minima (reactants, intermediates, and products) on a potential energy surface. However, determination of transition state structures (first order saddle points) has been an ongoing problem. The computational technique of genetic algorithms has recently been applied to optimization problems in many disciplines. Genetic algorithms are a type of evolutionary computing in which a population of individuals, whose genes collectively encode candidate solutions to the problem being solved, evolve toward a desired objective. Each generation is biased towards producing individuals which closely resemble the known desired features of the optimum. This thesis contains a discussion of existing techniques for geometry optimization, a description of genetic algorithms, and an explanation of how the genetic algorithm technique was applied to transition state optimization and incorporated into the existing ab initio package Mungauss. Results from optimizing mathematical functions, demonstrating the effectiveness of the genetic algorithm implemented to optimize first order saddle points, are presented, followed by results from the optimization of standard chemical structures used for the testing of transition state optimization methods. Finally, some ideas for future method modifications to increase the efficiency of the genetic algorithm implementation used are discussed. |
