3 Matching Results

Search Results

Reducing the Computational Cost of Ab Initio Methods

Description: In recent years, advances in computer technology combined with new ab initio computational methods have allowed for dramatic improvement in the prediction of energetic properties. Unfortunately, even with these advances, the extensive computational cost, in terms of computer time, memory, and disk space of the sophisticated methods required to achieve chemical accuracy - defined as 1 kcal/mol from reliable experimental data effectively - limits the size of molecules [i.e. less than 10-15 non-hydrogen atoms] that can be studied. Several schemes were explored to help reduce the computational cost while still maintaining chemical accuracy. Specifically, a study was performed to assess the accuracy of ccCA to compute atomization energies, ionization potentials, electron affinities, proton affinities, and enthalpies of formation for third-row (Ga-Kr) containing molecules. Next, truncation of the correlation consistent basis sets for the hydrogen atom was examined as a possible means to reduce the computational cost of ab initio methods. It was determined that energetic properties could be extrapolated to the complete basis set (CBS) limit utilizing a series of truncated hydrogen basis sets that was within 1 kcal/mol of the extrapolation of the full correlation consistent basis sets. Basis set truncation for the hydrogen atom was then applied to ccCA in the development of two reduced basis set composite methods, ccCA(aug) and ccCA(TB). The effects that the ccCA(aug) and ccCA(TB) methods had upon enthalpies of formation and the overall percent disk space saved as compared to ccCA was examined for the hydrogen containing molecules of the G2/97 test suite. Additionally, the Weizmann-n (Wn) methods were utilized to compute the several properties for the alkali metal hydroxides as well as the ground and excited states of the alkali monoxides anion and radicals. Finally, a multi-reference variation to the correlation consistent Composite Approach [MR-ccCA] was presented and utilized in the computation ...
Date: August 2008
Creator: Mintz, Benjamin
Partner: UNT Libraries