Undergraduate research project
One of the most remarkable theorems of mathematics is the
Riemann mapping theorem which says that any simply connected domain
in the plane can be mapped conformally onto the unit disc. These maps
are used in a variety of applications in physics and engineering.
In a few special cases the map can be found explicitly, but in most
applications one must compute a numerical approximation.
The zipper algorithm constructs an approximation to the conformal map by
composing a large number of simple maps.
The goal of this project is to learn about conformal maps, the zipper
algorithm and (for the ambitious) develop a much faster version of the
zipper algorithm.
Depending on the student's interest, this project could last for a
single semester or several semesters.
The first semester would be spent learning about conformal maps,
how the zipper algorithm works and applying this algorithm to some examples.
There is a nice description of the zipper algorithm in a recent
preprint by Stephen Rohde and Don Marshall.
Some pictures of conformal maps and Fortran code implementing the
zipper algorithm may be downloaded from
Don Marshall's web site.
The code for the zipper algorithm can be downloaded and run
relatively easily to produce some nice pictures of conformal maps.
Ambitious student(s) can continue after one semester to develop
a faster zipper algorithm. The zipper algorithm is related to
the Loewner differential equation which also produces conformal maps.
In the last few years, there have been advances which greatly
speed up finding solutions to the Loewner equation.
These same ideas should lead to a much
faster zipper algorith. The ultimate goal would be to produce
code with a faster zipper algorithm that can be made available to
the public (as is the current version of the zipper algorithm).
The algorithm begins by approximating the boundary of the simply
connected domain by a sequence of N points. The time the algorithm
takes grows with N as N^2.
The algorithm for the Loewner differential equation also
takes a time that is of order N^2 to produce N points.
However, the recent advances give an algorithm whose time is only of
order N^1.4. It is expected that an equally dramatic improvement is
possible for the zipper algorithm.
prerequisites:
Familiarity with complex variables, e.g.,
Math 424 would be helpful, would is not essential.
Conformal maps are not typically covered
in our undergraduate courses, so it is expected that the student will
begin by learning about these beautiful maps.