Untersuchung von MD-Simulationen im Rahmen von Hybrid-Schemata zur Beschreibung komplexer Flüssigkeiten
Date issued
Authors
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
License
Abstract
Several types of simulations permit an insight into the dynamics and the intrinsic properties of a system, according to the underlying algorithms and simulated time and length scales. Depending on the type of simulation and the coarse graining of the physical system, we gain access to a specific set of dynamics, and therefore a specific set of physical properties. Each type of simulation has its respective advantages and disadvantages.
To profit from the advantages and simultaneously compensate the disadvantages of certain simulation types, the Heterogenous Multiscale Method combines the molecular accuracy of a particle-based Molecular Dynamics simulation with the computational efficiency of a continuum-type method, Computational Fluid Dynamics. In the present thesis, we present a multiscale method to accurately describe the flow behaviour of soft matter fluids on large length scales.
For every time step of the continuum simulation, one or more time-intensive simulations at the micro scale are initialized to obtain the necessary information. Thus, the micro-scale simulations typically slow down the whole method, constituting a bottleneck to the hybrid scheme. To avoid this situation or at least to reduce the needed time to conduct such a simulation, the goal of this study is to optimize the necessary simulation parameters and underlying algorithms at the micro scale to accelerate the hybrid scheme calculations.
We investigate a variety of thermorstat candidates for temperature control and their respective parameter space. To determine suitable values for the parameters, we study the pressure and the viscosity, which can be derived from the stress tensor. For the Lowe-Andersen thermostat, the parameter choice is critical, since it has a major influence on the dynamics of the system, the effects on the viscosity are surveyed in detail. With careful choice of the thermostat parameters, one can simulate a fluid with higher viscosity and thus control the dynamics, avoiding the use of a complex hydrodynamic simulation scheme, such as Multiple Particle Collision Dynamics.
The isokinetic thermostat is also studied in detail. With help of this temperature controlling algorithm, the impact and influence of simulational parameters such as the timestep and the simulation box length on pressure and viscosity are studied. We report that several small-scale simulations are preferable to a large-scale system with complex parallelisation.
Both simple Weeks-Chandler-Andersen particles and small polymer chains using a Kremer-Grest model are simulated. The latter differ in their viscous behaviour depending on the strain applied on the system, while the monomer system exhibit Newtonian behaviour. Parameters such as the monomer density of the polymer system as well as the chain length and their influence on the viscosity are examined.
Apart from the results of the investigations into the influence of the parameters, the procedures to refine the data and the fitting methods are presented. These algorithms are applied to the results extracted from the MD simulation. By application of these methods, the amount of required simulations can be reduced. In addition, first results based on the complete Heterogeneous Multiscale Method are shown.