[Back]


Diploma and Master Theses (authored and supervised):

M. Brunner:
"Local error analysis for generalised splitting methods";
Supervisor: W. Auzinger; Institut für Analysis und Scientific Computing, 2020; final examination: 2020.



English abstract:
This thesis introduces the basic theory of splitting methods for evolution equations. A
symmetrised version of the defect is discussed and the defect is established as an asymptotically
correct local error estimator. The general background for high order splitting such as the
Baker-Campbell-Hausdorff formula and symmetrised versions thereof are treated and order
conditions for high order splittings are extracted. In particular, we take a closer look at
skew-hermitian matrices. In addition, we cover a ´dual´ approach - the Zassenhaus splitting -
and discuss the main ingredients Magnus provided for the analysis of the Zassenhaus splitting.
A symmetrisation of Magnus´ approach is made. Next, we introduce inner symmetrised defects
and elaborate on its Taylor expansion. This is the key component to the more basic approach.
We focus on the error expansion of the Strang splitting - our basic case of the more general
Zassenhaus type setting. The systematic treatment of the general case offers ideas for further
generalisations and provides a basis for a good understanding of the high level theory results.
It is based on the Faà di Bruno identity and Bell polynomials play a key role when generalising
the Lie expansion formula. We use Feynman diagrams for a compact and clear picture of the
derivatives we will encounter. In the end, we have successfully recovered the order condition
previously seen in the BCH formula by using the Taylor approach. We will conclude the thesis
with an application of the order conditions to a physical problem.

Keywords:
evolution equations, splitting methods, symmetrised defect, order conditions, local error estimators, Zassenhaus splitting; Bell polynomials, Faà di Bruno formula

Created from the Publication Database of the Vienna University of Technology.