Eigenvalue of 3D rotation matrix

[tickle_monste]

It was pretty cool to stumble upon Euler's formula as the eigenvalues of the rotation matrix.

det(Rot - kI) = (cos t - k)² + sin²t
=k²-2(cos t)k + cos²t + sin²t
=k²-2(cos t)k + 1

k = {2cos t +/- \sqrt{4cos^2(t) - 4}}/2
k = cos t +/- \sqrt{cos^2(t) - 1}
k = cos t +/- \sqrt{cos^2(t) - cos^2t - sin^2(t)}
k = cos t +/- \sqrt{-sin^2(t)}
k = cos t +/- i sin t = e^(+/-)it

I was wondering what the eigenvalues are for the rotation matrix in 3D, and if there's a 3D equivalent to Euler's formula.

 

[CompuChip]

Actually, I suppose it is not that strange when you look at a rotation as complex multiplication.
If you write a two-dimensional vector v = (x, y) as z = x + iy, then rotation over an angle t can be written either as R(t) v, where R is the 2d rotation matrix. But you can also write it as e^it z.

A 3D rotation matrix, in the appropriate basis, looks like
R_3(t) = \begin{pmatrix} R_2(t) & 0 \\ 0 & 1 \end{pmatrix}
where R₂(t) is the 2-dimensional rotation matrix and the z-axis is the rotation axis (i.e. rotation in the (x,y) plane). So, by expansion along the last row or column,
det(R₃ - k I) = (1 - k) det(R₂ - k I)
which you can work out in terms of your previous result. (And then add that an arbitrary change of basis does not alter the eigenvalues).

 

[Zaphos]

In addition to the two imaginary eigenvalues, the most intuitive one (for me) is the real valued one -- it's 1, with eigenvector along the axis of rotation. You can see that one just by noting that rotating won't change a vector along the axis of rotation.

edit:

if there's a 3D equivalent to Euler's formula.

Quaternions could be considered an extension of Euler's formula ...