This thesis is concerned with the computational simulation of weakly bound molecules. Two general methods of producing potential energy surfaces, Neural Networks and Gaussian processes, are described and evaluated. The Neural Network method is used to generate potential functions for HF-HCI, H<SUB>2</SUB>-HF and H<SUB>2</SUB>-HCI from good quality <I>ab initio </I>data. These surfaces are used by the diffusion Monte Carlo algorithm to solve the vibrational Schrödinger equation for the ground state of the above clusters. In addition, the DMC method is used in a size selective study of the N<SUP>+</SUP><SUB>2</SUB>-He<I><SUB>n</SUB></I> system. Good agreement is obtained with previous theoretical calculations and with the small amount of experimental data available and it is hoped that the predictions made will aid future studies of these clusters. The combined <I>ab initio-</I>Neural-DMC approach is shown to be an efficient method of studying weakly bound molecules and as such will prove to be a useful step towards understanding the structure and bonding of these systems.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:597004 |
| Date | January 1998 |
| Creators | Brown, D. F. R. |
| Publisher | University of Cambridge |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
Page generated in 0.256 seconds