Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Raising Matrices to power of n. (complex)

  1. Dec 18, 2008 #1
    So lets say you have a matrice that has a det of 0

    So X= [1 1; 1 1]

    So using some verification and algebra u arive at

    X^n= 2^(n-1)[1 1; 1 1]

    So I am trying to define what n can be.

    I already established n cannot be a - since one would get infinity for an answer since it is a singular matrix.

    Another limiation is 0. Since that just gives you an identity which the formula cannot express.

    But I am confused with fractions.

    Technically they should work but I do not know how to explain it.
    Same with irrational numbers.

    so can n=1/2?

    and if that is not hard enough what about imaginary numbers?????????

    For some reason I know they work with non singular matrices .
  2. jcsd
  3. Dec 19, 2008 #2
  4. Dec 19, 2008 #3
    Your formula does work for n=1/2. (If you square 2^(1/2-1) [1,1;1,1] you get back your original matrix.) In fact, the formula works for n=1/b for any positive integer b. Since you know it works for positive integers it also works for n=a/b for positive integers a/b. So it works for positive rational numbers.

    For irrational and imaginary n you have to decide what it means to raise a matrix to this type of power. Hopefully someone here will have an idea.

    Mathematica actually lets you raise your matrix to a positive irrational power and it gives back your formula. (It gave an error message when I tried to do imaginary powers). I don't know how to interpret this result, though.
  5. Dec 19, 2008 #4

    Ben Niehoff

    User Avatar
    Science Advisor
    Gold Member

    For imaginary (or in general, complex) powers, the definition for ordinary numbers (not matrices) is

    [tex]\alpha^{\beta} = e^{\beta \ln \alpha}[/tex]

    For a matrix A, this is only defined if ln(A) exists. This is not always the case (in fact, I think det(A)=0 is precisely one case for which ln(A) is not defined...the determinant is something like the "magnitude" of a matrix).

    Note: One can take arbitrary functions of matrices f(A), so long as f(A) can be defined as a convergent Taylor series. The matrix logarithm can be given by

    [tex]\ln (1+A) = A - \frac12 A^2 + \frac13 A^3 - \frac14 A^4 + ...[/tex]

    which does not converge for your matrix.
    Last edited: Dec 19, 2008
  6. Dec 19, 2008 #5
    I used matlab and it gave values for fractions. So I am confused there.

    Is there a proof for fractions?

    I want to algebriacly proove whyy fractions work. Perhaps through induction?
  7. Dec 19, 2008 #6


    User Avatar
    Science Advisor

    A proof of what for fractions? You haven't stated what it is you want to prove.

    The matrix
    [tex]A= \left[\begin{array}{cc}1 & 1 \\ 1 & 1\end{array}\right][/tex]
    has two distinct eigenvalues and so is diagonalizable. That is, there exist a matrix P such that [itex]P^{-1}AP= D[/itex] where D is a diagonal matrix having the eigenvalues 0 and 1 on the main diagonal. Then [itex]A= PDP^{-1}[/itex] and
    [tex]A^n= (PDP^{-1})(PDP^{-1})(PDP^{-1})\cdot\cdot\cdot(PDP^{-1})= PD^nP^{-1}[/tex]
    where Dn is easy- it is the diagonal matrix with 0 and 2n on the main diagonal.

    Specifically, an eigenvector corresponding to the eigenvalue 0 is <1, -1> and an eigenvector corresponding to the eigenvalue 1 is <1, 1> so we can take
    [tex]P= \left[\begin{array}{cc}1 & 1 \\ -1 & 1\end{array}\right][/tex]
    and then
    [tex]P^{-1}= \left[\begin{array}{cc} \frac{1}{2} & -\frac{1}{2} \\ \frac{1}{2} & \frac{1}{2}\end{array}\right][/tex]

    That is
    [tex]\left[\begin{array}{cc}1 & 1 \\ 1 & 1\end{array}\right]= \left[\begin{array}{cc}1 & 1 \\ -1 & 1\end{array}\right]\left[\begin{array}{cc}0 & 0 \\ 0 & 2\end{array}\right]\left[\begin{array}{cc} \frac{1}{2} & -\frac{1}{2} \\ \frac{1}{2} & \frac{1}{2}\end{array}\right][/tex]

    and so
    [tex]\left[\begin{array}{cc}1 & 1 \\ 1 & 1\end{array}\right]^n= \left[\begin{array}{cc}1 & 1 \\ -1 & 1\end{array}\right]\left[\begin{array}{cc}0 & 0 \\ 0 & 2^n\end{array}\right]\left[\begin{array}{cc} \frac{1}{2} & -\frac{1}{2} \\ \frac{1}{2} & \frac{1}{2}\end{array}\right][/tex]

    You can show that the formula works for 1/n as well by reversing it:
    [tex]\left[\begin{array}{cc}1 & 1 \\ 1 & 1\end{array}\right]= \left[\begin{array}{cc}1 & 1 \\ -1 & 1\end{array}\right]\left[\begin{array}{cc}0 & 0 \\ 0 & 2^{1/n}\end{array}\right]^n\left[\begin{array}{cc} \frac{1}{2} & -\frac{1}{2} \\ \frac{1}{2} & \frac{1}{2}\end{array}\right][/tex]
  8. Dec 19, 2008 #7
    ). I basicly want to prove that when you have a singular matrix X. And you raise it to the power of 0<n<1 one can yield results.

    Thanks for the exmplanation I get how to do it now. But I need to research more on eigenvalues and eigenvectors since in the course I am in currently we do not learn about those key properties.

    Couple more things:
    Without using eigen values can you simply prove that by saying
    Where M is the result?

    and I gota questions can you also use taylor series to explain why irrational numbers work?

    Or perhaps like a pi expansion where you put fractions for n for your required precesion?

    An why exactly doesnt ln of X work when logs work.
    Last edited: Dec 19, 2008
  9. Dec 20, 2008 #8


    User Avatar
    Science Advisor

    If X1/n= M then Mn= X, not "X1/n= Mn".

    I don't know what you mean by "why exactly doesnt ln of X work when logs work. "

    ln (the natural logarithm) works whenever any log "works". loga(X)= ln(X)/ln(a).
Share this great discussion with others via Reddit, Google+, Twitter, or Facebook