Ask Your Question
1

Eigenvalues and eigenspaces of orthogonal (or rotation) matrices

asked 2020-01-10 23:05:24 +0100

anonymous user

Anonymous

updated 2020-01-11 22:20:19 +0100

Given an orthogonal transformation of finite order, e.g.

Matrix([[0,0,0,-1],[1,0,0,-1],[0,1,0,-1],[0,0,1,-1]])

Its eigenvalues are going to be of the form

exp(I*pi/5),exp(2*I*pi/5),...,exp(2*I*pi*m),...

corresponding to a splitting of the matrix into rotation (and reflection) matrices. I'd like to extract these fractions m (mod ZZ) and study the corresponding (real rotation and reflection) eigenspaces.

My impression is that Sage isn't suitable for doing this directly, but that I should use e.g. the Maxima or Mathematica interface? Any suggestions for the most suitable method?

edit retag flag offensive close merge delete

Comments

1

Well... here is a start:

sage: M=matrix(SR,[[0,0,0,-1],[1,0,0,-1],[0,1,0,-1],[0,0,1,-1]])
sage: list(map(lambda u:u.maxima_methods().polarform(), M.eigenvalues()))
[1/4*sqrt((sqrt(5) + 1)^2 - 2*sqrt(5) + 10)*e^(-I*pi - I*arctan(-2*sqrt(-1/2*sqrt(5) + 5/2)/(sqrt(5) + 1))),
 1/4*sqrt((sqrt(5) + 1)^2 - 2*sqrt(5) + 10)*e^(I*pi + I*arctan(-2*sqrt(-1/2*sqrt(5) + 5/2)/(sqrt(5) + 1))),
 1/4*sqrt((sqrt(5) - 1)^2 + 2*sqrt(5) + 10)*e^(-I*arctan(2*sqrt(1/2*sqrt(5) + 5/2)/(sqrt(5) - 1))),
 1/4*sqrt((sqrt(5) - 1)^2 + 2*sqrt(5) + 10)*e^(I*arctan(2*sqrt(1/2*sqrt(5) + 5/2)/(sqrt(5) - 1)))]
Emmanuel Charpentier gravatar imageEmmanuel Charpentier ( 2020-01-11 00:21:37 +0100 )edit

1 Answer

Sort by » oldest newest most voted
2

answered 2020-01-11 11:44:52 +0100

FrédéricC gravatar image

updated 2020-01-11 11:52:06 +0100

Like that

sage: M = Matrix([[0,0,0,-1],[1,0,0,-1],[0,1,0,-1],[0,0,1,-1]])
sage: [(z.multiplicative_order(), m) for z, m in M.charpoly().roots(QQbar)]
[(5, 1), (5, 1), (5, 1), (5, 1)]

or even better:

sage: [(p.is_cyclotomic(certificate=True), m) for (p, m) in M.charpoly().factor()]
[(5, 1)]
edit flag offensive delete link more

Comments

1

Much, much better that my attempt ...

Emmanuel Charpentier gravatar imageEmmanuel Charpentier ( 2020-01-11 13:24:38 +0100 )edit

Thanks, this is excellent! (I need the actual fractions (1/5,2/5,3/5,4/5 in the example) which I can now deduce inductively from your code using the multiplicities, and then I suppose I can obtain the eigenspaces by first comparing those eigenvalues with E(5)^i in UniversalCyclotomicField, and then taking the real and complex parts of the matching eigenvectors gives the rotation planes. In other words, this is very doable now but if you know some smart tricks for this again then please do let me know!)

Bob67846 gravatar imageBob67846 ( 2020-01-11 22:18:37 +0100 )edit

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account.

Add Answer

Question Tools

1 follower

Stats

Asked: 2020-01-10 23:05:24 +0100

Seen: 487 times

Last updated: Jan 11 '20