Eigenvalues and Eigenvectors
Summary
We're considering the transformation . Eigenvectors provide the ideal basis for when considering this transformation. Their images under the transformation are simple scalings.
Eigenstuff: An eigenvector of is a nonzero vector such that . The scalar is called the eigenvalue of A corresponding to . There may be several eigenvectors corresponding to a given .
The idea is that an eigenvector is simply scaled by the transformation, so the actions of a transformation are easily understood for eigenvectors. If we could write a vector as a linear combination of eigenvectors, then it would be easy to calculate its image: if there are n eigenvectors , with n eigenvalues , then if
then
Nice, no?
If is an eigenvalue of matrix A corresponding to eigenvector , then
This means the
which is equivalent to
So is in the null space of . If the null space is trivial, then is the zero vector, and is not an eigenvalue. Alternatively, all vectors in the null space are eigenvectors corresponding to the eigenvalue .
As for determining the eigenvectors and eigenvalues, there is some cases in which this is extremely easy:
The eigenvalues of a diagonal matrix are the entries on its diagonal. More generally,
Theorem 1: The eigenvalues of a triangular matrix are the entries on its main diagonal.
Theorem 2: If are eigenvectors corresponding to distince eigenvalues of an matrix A, then the set is linearly independent.
The eigenvectors and difference equations portion of this section can be illustrated with the example of the Fibonacci numbers transformation: recall that the Fibonacci numbers are those obtained by the recurrence relation
and and .
where
The eigenvalues of this matrix are approximately and -0.618033988749894. is the so-called ``golden mean'', which is a nearly sacred number in nature, well approximated by the ratio of consecutive Fibonacci numbers.
An eigenvector corresponding to the golden mean (normalized to have a norm of 1) is approximately
so that