TLDR
(click to show/hide)
A brief introduction to abstract algebra, the ping-pong lemma (which lets you prove something about an infinitely
large group with a finite amount of work), and how I made an algorithm utilizing it in my research in the
Texas Experimental Geometry Lab (TXGL).
Today I'm going to talk about an interesting lemma in abstract algebra
and how I used it in my research at the
Texas Experimental Geometry Lab. Before we talk about the lemma, let's first establish some of the
mathematical background we'll need to know.
Abstract algebra / algebraic structures is all about the study of
groups, rings, fields, and the properties of these objects. A group
is a set along with a binary operation such that the following properties are satisfied:
-
Closure: For
, we have that
is in
-
Associativity: For
, we have a
-
Identity: There exists an
such that for
any ,
-
Inverses: For
, there is
such that
A group presentation is a collection of generators and relations such
that
. Perhaps this is best understood with a few examples:
-
is the set of all multiples of and
(such as
for any integer or
etc.)
-
The integers with the addition operation can be written as
since
(which is to say added to itself
times since the operation is addition)
will generate all of the integers.
-
is a group with the set of elements
since we have the relation which dictates that
must be the identity.
The group we will consider today is
which is the set of 2x2 real matrices with determinant 1 along with
the operation of standard matrix multiplication. Once we have the
concept of a group, we need to understand what a group action is.
Essentially, a group action is a map which takes
to
for some space and
preserves the group operation (ie.
). The space we will primarily consider is the real projective line
which is the set of equivalence classes of lines (often represented by
a circle).
Now that we have some of the terminology we need, lets talk about the
problem we want to solve. We call a group action faithful if each
element of the group other than the identity acts nontrivially on the
space , that is, there is no
such that
for all
. In general, it is very hard to show that a given group has a
faithful action on a space. However, we do know that a certain type of
group called a free group will always have a faithful action. Free
groups are groups with no relations (ie. there are no elements formed
by the generators which give us the identity). So, if we want an
approach to the problem of determining if a group will have a faithful
action on a space, we can attempt to show that the group is isomorphic
(equivalent) to a free group.
This is where the ping pong lemma comes into play. Below we will state
the lemma and see immediately how it will be useful in solving our
problem.
Ping Pong Lemma: Suppose a and b are generators for a
group G which acts on a space X. If:
-
has disjoint nonempty subsets
and
-
is contained in
and
is contained in
for all nonzero
then is isomorphic to a free group of
rank 2.
Without the ping pong lemma, we would need to check an infinite number
of conditions to see if a groups action is faithful (we would need to
see that
for all
and for all ). But here, we have a finite number of conditions to check to see
that the group is free and therefore faithful. The research done in
TXGL consisted of extending this lemma to
generators in
and coming up with an algorithm to search for intervals which will
satisfy these properties.
Avoiding diving too far into the details of the algorithm, I will talk
about the basic procedure. The idea is that the since the eigenvectors
of our generators are fixed points under the action of the generator
in
, we know that the intervals must include them. By creating epsilon
intervals around these, we have initial intervals to start the search
with. From here, we expand endpoints of the intervals, checking if
they satisfy the necessary containment properties until we find them.
There are different method for interval expansion we tried which are
discussed on the TXGL project writeup.
Here is an example of intervals in
which satisfy the necessary containment properties to show that the
respective group action is faithful: