PhD thesis – Christian Goll
Christian Goll
Parallel algorithms for molecular dynamics simulations in fluids [Full text via OPUS]
finished 2010-12
supervised by Klaus Mecke
The subject of this thesis is the implementation of parallel time driven molecular dynamic code for the use on the nano scale. Problems and details of the implementation are presented in Part I. In Part II, three applications of the code are shown. Dynamics on the nano scale have attracted a great interest in the last few years. This work does not concentrate on technical application, but on the exploration of processes within the nano scale. As tool for this exploration computer simulations are utilized.
Unfortunately in the last years the development in the computer technology did not lead to faster processors, but to the availability of more computer cores especially to the development of multicore processors. In order to utilize these architecture for scientific computations, parallel programming techniques must be used. Part I describes the problems with the programming technique and the implementation details. A comparison between existing solutions and the code of this work is presented. With this techniques it is possible to simulate more than one million particles, which corresponds to a cubic unit cell of an ideal gas under normal conditions with a side length of approximately 200 nm.
Three different applications of molecular dynamics simulations on the nano scale are presented in Part II. The first application is the simulation of a rectified Brownian motion in Chapter 3. In this model, the relaxation of an asymmetric particle into thermal equilibrium is used for a directed motion. Although in this simple two dimensional case only a trivial parallelization was used, further investigations on this theme will require an explicit parallelization. In Chapter 4, the initial generation of sound waves in a fluid after the hit of fast proton is analyzed. Several models for the hit of the fast proton are presented. Finally, in Chapter 5, the diffusion of gas in different nano-sized pore is analyzed. Two different pore kinds are used, which are minimal surfaces and a random Boolean pattern. For these simulations, massive computer power was used in order to simulate the gas in the pores with a diameter a magnitude greater than the components of the fluid.