Abstract
dc:descriptionIn the present Thesis, diffusion processes and subsequent properties are studied in a variety of media, going from simple 1D chain lattices to complex Scale Free networks with an emphasis on biased diffusion mechanisms. Chapter 1 is an introduction to diffusion problems in complex systems. In chapter 2, memory effects after equilibrium in a 2D and a 3D lattice fluid of nearest neighbor interacting particles, at temperatures slightly above the critical one, are studied. Starting from the analysis of Monte Carlo simulations data of the Self-Intermediate Scattering Function, we study the coverage and temperature dependence of the diffusion processes through a general form of the tracer diffusion coefficient function, the one that depends on both the wave vector and time. Semi-analytical expressions are derived from the simulations. Anisotropic aspects and memory effects are shown through the relaxation processes of the various parameters. In chapter 3, the trapping process of non classically diffusing particles in a lattice with a concentration of traps is studied theoretically and by means of computer simulations, with an emphasis on biased diffusion where the bias depends inversely on the distance from the attractive traps. In the case of a simple expression of the attraction, in a 1D chain lattice, an exact expression of the mean trapping time as a function of the trap concentration is derived from the recurrent equation. The survival probability, which depends directly on the number of distinct sites visited is a much more complex function. Therefore it is studied mostly through fitting functions of numerical Monte Carlo simulation results. In chapter 4, a variant of the previous theoretical model is considered to study abnormally fast diffusion processes observed recently between metastable and stable nano-islands in Pb/Si epitaxial growth systems. Through the analysis of realistic simulations, we show that in contrast with the classical diffusion model of isotropic random walks, the biased diffusion model succeeds in reproducing the main characteristics of the experimental results, such as the observed anisotropy, the very abrupt coarsening process, the unusually fast decay and growth of unstable and stable islands, respectively. In chapter 5, the concept of biased diffusion is furthermore extended and applied to study diffusion processes in complex networks. In the first part, special attention is given on mean first passage time (MFPT) problems when a packet/particle travels from a randomly chosen node to a randomly chosen target according to a bias. The bias parameter p is expressed as a probability to stay on the shortest path. An exact analytical solution is found for the case of Random Regular (RR) networks and it is shown, by mean field considerations and simulation data, that the solution is also valid for Erdos-Renyi (ER) networks. It is found that, at a threshold value of the bias parameter, there is an abrupt transition of the scaling of the MFPT with the size N of the network, from a power law to a logarithmic function. For Scale Free (SF) networks, an upper bound of the MFPT is found and it is shown that it scales at most as a power of log(Ν). In the second part, the previous model is extended to a trapping problem. Special attention is given to the transmission efficiency of biased random walks in the presence of malfunctioning nodes. For a given concentration of traps, it is found that there is a minimum value of the bias parameter, for which the transmission is unaffected by the presence of traps and practically all the particles/packets reach the target. This minimum value can be directly evaluated when the MFPT without traps is known, i.e. for RR and ER networks. For SF networks, a secure value of the minimum bias can be derived from the upper bound expression of the MFPT, and thus it can be applied to avoid the loss of information in real communication networks. Using a simple approximation of the mean number of nodes visited before reaching the target, an analytical expression of the transmission efficiency ιs proposed which is accurate for a sufficiently large bias. Also, the efficiency of the process in SF networks after failure of the most connected node is also investigated, i.e. when there is only one trap on the hub. In the case of the unbiased random walk, an analytical expression is found.
Degree
thesis:*- Grantor dc:publisher
- Aristotle University Of Thessaloniki (AUTH)
- Year dc:date
- 2014
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Skarpalezos, Loukas Nikolaos Markos
- Σκαρπαλέζος, Λουκάς Νικόλαος Μάρκος
Subjects
dc:subject × 10Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/34555