Post-graduate theses
Current Record: 70 of 125
|
Identifier |
000419298 |
Title |
Development of coarse-grained models for the dynamics of molecular systems, through atomistic simulation data analysis. |
Alternative Title |
Ανάπτυξη αδροποιημένων μοντέλων για τη δυναμική των μοριακών συστημάτων μέσω της ανάλυσης δεδομένων ατονιστικής προσομείωσης |
Author
|
Μπαξεβάνη, Γεωργία
|
Thesis advisor
|
Χαρμανδάρης, Ευάγγελος
|
Reviewer
|
Καλλιγιανάκη, Ευαγγελία
Πανταζής, Ιωάννης
|
Abstract |
Molecular dynamics simulations is a field of science that studies interactions of atoms at a very small order of magnitude. Many natural phenomena are characterized by a range of interactionsand processes that cover a wide range of length and time scales. Many chemical and biologicalprocesses can be studied using the tool of molecular dynamics simulations. Today, scientists areable to simulate important biological systems such as proteins or DNA to study their behavior.
The diffculties due to the temporal and spatial limitations of simulations, lead us to average outthe details of the atomistic level, to the molecular level. The method that combines microscopicand mesoscopic models is called Coarse Graining (CG). Defining the new effective coarse-grainedsystem, means to find that coarse-grained model which represents best the reference atomisticsystem ideally both structure and dynamic properties.
Hierarchical CG approaches of atomistic molecular systems at equilibrium has been an intensiveresearch topic over the last few decades. Several methods to obtain parametric and non-parametricCG models have been developed and successfully applied to different molecular systems underequilibrium conditions. These methods are optimization problems that match chosen properties of the atomistic system, such as the forces (the force-matching method), the pair correlation function (the inverse Boltzmann method), or the probability density (the relative entropy minimization method).
However, many interesting phenomena are inherently in non-equilibrium. In this thesis, our
approach lies in the derivation of path-space force matching, which is applicable to systems under both equilibrium and non-equilibrium conditions. We first set up a path-space variational inference problem, using the relative entropy between distributions on the path-space. We study the information-based method for high-dimensional stochastic processes driven by Langevin dynamics. Specifically, we apply the path-space force matching method to a methane liquid system at the transient dynamics regime, as well as at equilibrium dynamics. We are able to validate the path-space force matching and retrieve the potential of mean force.
|
Language |
English |
Subject |
Force -matching |
|
Langevin dynamics |
|
Variational inference |
|
Πρόβλημα βελτιστοποίησης |
|
Στοχαστικές διαδικασίες |
Issue date |
2018-11-23 |
Collection
|
School/Department--School of Sciences and Engineering--Department of Mathematics and Applied Mathematics--Post-graduate theses
|
|
Type of Work--Post-graduate theses
|
Permanent Link |
https://elocus.lib.uoc.gr//dlib/b/d/c/metadata-dlib-1542975539-565565-12858.tkl
|
Views |
401 |