edoc

Probing the characteristics of potential energy surfaces

Schäfer, Bastian. Probing the characteristics of potential energy surfaces. 2015, Doctoral Thesis, University of Basel, Faculty of Science.

[img]
Preview
PDF
8Mb

Official URL: http://edoc.unibas.ch/diss/DissB_11543

Downloads: Statistics Overview

Abstract

The potential energy surface of multi-atomic systems encodes important aspects such as thermodynamic and dynamic properties or the equilibrium geometries. Collections of low-energy minima and the reaction pathways that connect the minima with each other can be key elements in the study of potential energy surfaces and their properties. The extension of the minima hopping (MH) global optimization method to the minima hopping guided path search method (MHGPS) forms the heart of this thesis. MHGPS is a MH based approach for finding complex reaction pathways that connect the local minima in an efficient, automatised and unbiased fashion. Also, in this context, novel stabilized quasi-Newton local optimizers for the computation of minima and saddle points are developed. These optimizers are designed for robustness to the noisy forces delivered by density functional codes. Using benchmarks, the MHGPS method as well as the stabilized quasi-Newton optimisers are found to compare favorably with existing algorithms. Using the MHGPS method, novel results are presented for previously extensively studied Lennard-Jones clusters. Besides that, an ab-initio structure prediction study using the MH global optimization method is presented for the neutral and anionic gold clusters with 26 atoms. Finally, computationally efficient methods for a qualitative characterization of potential energy surfaces are discussed. In this context, MHGPS is applied at the density functional level of theory to the potential energy surface of Si20.
Advisors:Goedecker, Stefan
Committee Members:Gross, Axel
Faculties and Departments:05 Faculty of Science > Departement Physik > Physik > Physik (Goedecker)
UniBasel Contributors:Goedecker, Stefan
Item Type:Thesis
Thesis Subtype:Doctoral Thesis
Thesis no:11543
Thesis status:Complete
Number of Pages:182 S.
Language:English
Identification Number:
edoc DOI:
Last Modified:22 Jan 2018 15:52
Deposited On:22 Dec 2015 11:11

Repository Staff Only: item control page