You are on page 1of 2

Eigenvalues, eigenvectors, diagonalisation

In FP1 we found out what (some) matrices looked like by plotting where they mapped
the points (1,0) and (0,1) to.

Here a computer has taken a matrix and shows where it maps a whole heap of other
points on the unit circle as well as the points (1,0) and (0,1). The v-line is what the
radius from (0,0) to (1,0) maps to, and the u-line is what the radius from (0,0) to (0,1)
maps to.

Just from the diagram, without calculations:

• draw two eigenvectors

• draw two eigenvectors

• draw two normalised eigenvectors

• draw the invariant lines

• estimate the eigenvalues.

Then calculate all those things by algebra.


This is the same diagram with the axes airbrushed out. With new axes the
transformation will be the same, but the coordinates of each point will be different, and
the matrix representing the transformation will be different.

Draw in new axes relative to which this transformation is just a scaling along the axes.

How do these axes relate to the eigenvectors?

What is the matrix of the transformation relative to these new axes?

Now calculate that new matrix by algebra

You can do the switch-to-new-axes thing as long as the eigenvalues are real numbers
(not complex), and it works nicely as long as the new axes are orthogonal (at right
angles to each other).

How do you check that vectors along axes are orthogonal?

For what sort of matrix is it guaranteed that the eigenvalues are real and the axes are
orthogonal?

You might also like