Logo Logo
Help
Contact
Switch language to German
Development and application of efficient ab initio molecular dynamics simulations of ground and excited states
Development and application of efficient ab initio molecular dynamics simulations of ground and excited states
Ab initio molecular dynamics reflect the movement of nuclei on a potential energy surface generated by ab initio methods. These simulations give access to an entire series of chemically relevant properties, such as vibrational spectra and free energies, and have thus become indispensable in, for example, biochemistry and materials sciences. They are, however, computationally demanding, due to the expensive quantum-chemical calculations that are required at every step. In order to overcome some of the limitations, this thesis presents steps towards efficient but still accurate \textit{ab initio} molecular dynamics simulations, combining recent progress in different fields of computational chemistry. The time-consuming two-electron integral evaluations are conducted on graphics processing units. Their massively parallel architecture leads to speed-ups (with respect to calculations on central processing units) and strong scaling is observed. Expensive electronic structure calculations are circumvented using parametrized methods, such as the corrected small basis set Hartree-Fock method or simplified time-dependent density functional theory. From the field of molecular dynamics, the extended Lagrangian method is adopted to stabilize the trajectories and to accelerate the convergence of the self-consistent field algorithm. Finally, couplings between electronic states are approximated from a finite differences approach to avoid the time-consuming analytical evaluations at the time-dependent density functional theory level. As a result of these approaches, large molecular systems become accessible at comparably low computational cost. This is demonstrated for several illustrative applications. Excited-state dynamics are used to explore the relaxation pathway of the rhodopsin protein and four newly designed rotary molecular motors using the same Schiff base motif. Ground-state simulations deliver vibrational spectra of medium-sized molecules and liquid water. They are used in addition to determine free energy differences of molecular transformations, for which a novel scheme is introduced delivering deeper insights into the underlying process.
Not available
Peters, Laurens Dirk Marga
2020
English
Universitätsbibliothek der Ludwig-Maximilians-Universität München
Peters, Laurens Dirk Marga (2020): Development and application of efficient ab initio molecular dynamics simulations of ground and excited states. Dissertation, LMU München: Faculty of Chemistry and Pharmacy
[thumbnail of Peters_Laurens_D_M.pdf]
Preview
PDF
Peters_Laurens_D_M.pdf

13MB

Abstract

Ab initio molecular dynamics reflect the movement of nuclei on a potential energy surface generated by ab initio methods. These simulations give access to an entire series of chemically relevant properties, such as vibrational spectra and free energies, and have thus become indispensable in, for example, biochemistry and materials sciences. They are, however, computationally demanding, due to the expensive quantum-chemical calculations that are required at every step. In order to overcome some of the limitations, this thesis presents steps towards efficient but still accurate \textit{ab initio} molecular dynamics simulations, combining recent progress in different fields of computational chemistry. The time-consuming two-electron integral evaluations are conducted on graphics processing units. Their massively parallel architecture leads to speed-ups (with respect to calculations on central processing units) and strong scaling is observed. Expensive electronic structure calculations are circumvented using parametrized methods, such as the corrected small basis set Hartree-Fock method or simplified time-dependent density functional theory. From the field of molecular dynamics, the extended Lagrangian method is adopted to stabilize the trajectories and to accelerate the convergence of the self-consistent field algorithm. Finally, couplings between electronic states are approximated from a finite differences approach to avoid the time-consuming analytical evaluations at the time-dependent density functional theory level. As a result of these approaches, large molecular systems become accessible at comparably low computational cost. This is demonstrated for several illustrative applications. Excited-state dynamics are used to explore the relaxation pathway of the rhodopsin protein and four newly designed rotary molecular motors using the same Schiff base motif. Ground-state simulations deliver vibrational spectra of medium-sized molecules and liquid water. They are used in addition to determine free energy differences of molecular transformations, for which a novel scheme is introduced delivering deeper insights into the underlying process.