A means to take advantage of molecular similarity to lower the computational cost of electronic structure theory is explored in which parameters are embedded into a low-?cost, low-?level (LL) ab initio theory and adjusted to obtain agreement with a higher-level (HL) ab initio theory. Results will be presented for a model that is trained on 320 structures in which between zero and two substituents (selected from NH2, OH, F) are attached to ethane and ethylene, and the geometry is randomly distorted (over a range of �0.2 � for bond lengths, �10 degrees for bond angles and full rotation about the central C-C bond). The LL and HL models are, respectively, minimal basis and split-valence basis Hartree-Fock. A wide variety of approaches to embedding the parameters into the LL model have been explored. The principle means used to select an embedding scheme is the performance on molecules not included in the fitting of the parameters. Tests on substituted ethane and ethlyene molecules, not included in the fitting of the model, show agreement between the trained model and the HL model of about 2.3 kcal/mol for the change in energy due to geometric distortion, 0.05 Debye for dipole, and 0.12 eV for HOMO/LUMO orbital energies. Transfer of the parameters to substituted methane molecules increases the errors to 3.4 kcal/mol, 0.04 Debye and 0.8 eV. Obtaining this level of agreement requires the use of parameters that are functions of the bond lengths, atomic charges, and bond orders. The extension of this approach to more diverse molecules and properties will be discussed.

Embedding parameters in ab initio theory to develop well-?controlled approximations based on molecular similarity

248th National Meeting (2014)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.