Improved Modeling Of Turbulent Transport: From Noise In Transport Models To The Parareal Algorithm Applied To Full Turbulence Codes

Abstract

Turbulence and turbulent transport are ubiquitous in nature and are of fundamental importance in everything from the spread of pollution to confinement in fusion plasmas. In order to study this, turbulence models need to be as realistic as possible and one must also be able to evolve the turbulence and the profiles of the quantities of interest on transport (long) time scales. Improving turbulence simulations by the introduction of new techniques forms the basis of this research. One part of this work involved improving the performance of a 1D transport model by the addition of noise. On a more fundamental level, studying long time dynamics for turbulence simulations is very difficult even with the fastest computers available now or in the near future. To help overcome this difficulty, a new way of simulating turbulence has been presented, namely parallelizing in time. Time parallelization of a fully developed turbulent system is a new application. Parallelizing the space domain to computationally solve partial differential equations has been extensively used and is one of the most common forms of parallelization. In contrast, the Parareal Algorithm parallelizes the time domain and has been found to significantly reduce the computational wall time in many simpler systems. Despite its success in other less complex problems, it has not yet been successfully applied to a turbulent system (to the best of our knowledge). If efficiently applied, this algorithm will allow study of the turbulent transport dynamics on transport time scales - something that has heretofore been very difficult. In this work, the results of applying the Parareal Algorithm to simulations of drift wave turbulence in slab geometry in which the relative dominance of the polarization and E x B nonlinearities are tuned artificially, are presented. These turbulent systems are in many ways similar to neutral fluid turbulence models, so success of the Parareal scheme in them expands the prospect of a broader range of application to many other turbulent problems. This thesis also presents the results of a modification to the algorithm. A model to study and predict the parameters governing the convergence of the scheme is also explored.