Numerical approximation of mean curvature flow of graphs and level sets (Q4424463)
From MaRDI portal
!
This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:
scientific article; zbMATH DE number 1975242
| Language | Label | Description | Also known as |
|---|---|---|---|
| default for all languages | No label defined |
||
| English | Numerical approximation of mean curvature flow of graphs and level sets |
scientific article; zbMATH DE number 1975242 |
Statements
15 December 2003
0 references
mean curvature flow
0 references
level stes
0 references
degenerate equation
0 references
algorithms
0 references
viscosity solution
0 references
Evans-Spruck regularization
0 references
numerical examples
0 references
finite element method
0 references
semi-implicit time discretization
0 references
error estimates
0 references
convergence
0 references
0.8858568668365479
0 references
0.8856277465820312
0 references
0.8463164567947388
0 references
0.8463160991668701
0 references
Numerical approximation of mean curvature flow of graphs and level sets (English)
0 references
Mean curvature flow means the motion of a surface or an interface with a velocity, which is its mean curvature. For graphs the problem can be written as a differential equation for the height function: \[ \partial _t u - \sqrt{1+| \nabla u| ^2} \nabla \cdot {\nabla u\over\sqrt{1+| \nabla u| ^2}} =0 \text{ in } \Omega \times (0,T) \] \[ u=u_0 \text{ on } \partial \Omega \times(0,T) \cup \Omega\times \{ 0\} \] Under some asspumtion on \( u_0\) and the domain \( \Omega\) existence of a unique solution is known. Some energy estimates are proved. Two numerical algorithms are proposed; the first is a semidiscrete approximation based on the finite element method and the second one involves also time discretization as explicit, fully implicit or semiimplicit approach. The existence of the numerical solution for the semidiscrete and the fully discrete scheme is proved and error estimates of optimal order in space and time are presented for any space dimension, too. The mean curvature flow of level sets allows the modeling of topological changes during the flow where the interface is written as the zero level set of a scalar function \( u(x,t) \) which is the solution of equation \[ \partial _t u - | \nabla u| \nabla \cdot {\nabla u\over| \nabla u| } =0 \text{ in } \mathbb R^n \times (0,T) \] \[ u(\cdot ,0)=u_0 \text{ in } \mathbb R ^n. \] This equation is degenerate because it is not defined where the gradient of \(u\) vanishes and moreover the equation is not uniformly parabolic. There exists a unique viscosity solution of this problem under some assumption on the function \(u_0.\) Using the so-called Evans-Spruck regularization we obtain the problem \[ \partial _t u^{\varepsilon} - \sqrt{\varepsilon+| \nabla u^\varepsilon| ^2} \nabla \cdot {\nabla u^\varepsilon\over\sqrt{\varepsilon+| \nabla u^\varepsilon| ^2}} =0 \text{ in } \mathbb R^n \times (0,T) \] \[ u^\varepsilon(.,0)=u_0 \text{ in } \mathbb R^n, \] the solution of which converges to the viscosity solution. If we restrict the computation of this problem to some finite domain \( \Omega\) we obtain the solution \(u_{\varepsilon}\). A semidiscrete and a fully discrete scheme for this problem are proposed and using the transformation \( U:= {u_\varepsilon\over \varepsilon} \), one can use the results obtained in the first part of article to obtain convergence and error estimates for the numerical solution in regularized form of the level set equation. Numerical test examples for the regular case (nonzero gradient) but also for a highly singular case are included together with tables of errors and experimental orders of convergence. For the entire collection see [Zbl 1013.00023].
0 references