Stefan Cruceanu

Advisor: Dr. Gene Allgower and Dr. Simon Tavener

Degree Conferred: Spring 2007

After Graduation: Applied Math Institute of the Romanian Academy

Website: Not Available

Thesis Title:

Numerical Solutions of Nonlinear Systems Derived From Semilinear Elliptic Equations

Abstract:

The existence and the number of solutions for N-dimensional nonlinear boundary value problems has been studied from a theoretical point of view, but there is no general result that states how many solutions such a problem has or even to determine the existence of a solution. Numerical approximation of all solutions (complex and real) of systems of polynomials can be performed using numerical continuation methods. In this thesis, we adapt numerical continuation methods to compute all solutions of finite difference discretizations of boundary value problems in 2-dimensions involving the Laplacian. Using a homotopy deformation, new solutions on finer meshes are obtained from solutions on coarser meshes. The issue that we have to deal with is that the number of the solutions of the complex polynomial systems grows with the number of mesh points of the discretization. Hence, the need of some filters becomes necessary in this process.
   We remark that in May 2005, E. Allgower, D. Bates, A. Sommese, and C. Wampler used in \cite{Allgower-Bates-Sommese-Wampler} a similar strategy for finding all the solutions of two-point boundary value problems in 1-dimension with polynomial nonlinearities on the right hand side. Using exclusion algorithms, we were able to handle general nonlinearities.
   When tracking solutions sets of complex polynomial systems an issue of bifurcation or near bifurcation of paths arises. One remedy for this is to use the gamma-trick introduced by Sommese and Wampler in \cite{Sommese-Wampler}. In this thesis we show that bifurcations necessarily occur at turning points of paths and we use this fact to numerically handle the bifurcation, when mappings are analytic.