Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Good day!

Given a characteristic polynomial $P$ of matrix $A$ I need to show that the characteristic polynomial $O$ of $A^2$ can't have more different real roots than $P$.

I know that the characteristic polynomial for both cases can be calculated like this:
$P = |A - \lambda I| = 0$
$O = |A^2 - \lambda^2 I| = 0$
But in a general case with $n*n$ matrices they become way too complicated.

Can anyone guide me in the right direction?

share|cite|improve this question
That's wrong, isn't it? $A = \begin{pmatrix} 0 & -1 \\\ 1 & 0 \end{pmatrix}$ has charakteristic polynomial $P(t) = t^2 + 1$ and $A^2 = -\mathrm{Id}$, which has $Q(t) =(t+1)^2$? – martini May 10 '12 at 13:00
Shouldn't it be $Q(t) = (t^2 + 1)^2$? In this case both $P(t)$and $Q(t)$ have 0 real roots. – fikster May 10 '12 at 13:13
Because $Q(t) = A^2 - t^2I = \begin{vmatrix} -1 - t^2 & 0\\ 0 & -1 - t^2 \end{vmatrix} = (1+t^2)^2$ – fikster May 10 '12 at 13:21
In your equation for the characteristic polynomial of $A^2$ you shouldn't have $\lambda^2$, you should just have $\lambda$. It is true though that the square of any eigenvalue for $A$ will automatically be an eigenvalue of $A^2$. – rschwieb May 10 '12 at 13:21
It's plain to see that $A^2=\begin{pmatrix} -1 & 0 \\\ 0 & -1 \end{pmatrix}$ has $-1$ as an eigenvalue. You can't go wrong: every vector is an eigenvector for that particular $A^2$! – rschwieb May 10 '12 at 13:30

This is false, for somewhat trivial reasons. Let $A$ be a matrix with eigen values $\lambda_1,\dots,\lambda_n$ then note that the eigenvalues of $A^2$ are $\lambda_1^2,\dots,\lambda_n^2$ since if $v_i$ is an eigenvector of $\lambda_i$ then

$$A^2v_i=A(\lambda v_i)=\lambda Av_i=\lambda^2v_i.$$

For instance if the eigenvalues are purely imaginary the number of real eigenvalues increases. To take an example from the comments

$$A=\begin{pmatrix}0 & -1 \\ 1 & 0\end{pmatrix}$$

then $\mathsf{char}(A)=x^2+1$ which has imaginary roots $\pm i$ but $\mathsf{char}(A^2)=(x+1)^2$ has real repeated root $-1$.

Are there some more conditions on your matrix perhaps?

share|cite|improve this answer
I only assumed it was true, since I didn't think about using complex roots. What if I changed the question from real roots to complex roots? – fikster May 10 '12 at 13:34
Most plausibly, the question should have been "can't have less" rather than "can't have more". Or, "can't have more complex roots" rather than "can't have more real roots". – rschwieb May 10 '12 at 13:35
@fikster If you change the question to complex roots, there's not much to prove. Since every eigenvalue of a real matrix is going to be complex. I suppose you could consider complex eigenvalues that aren't purely real, but that seems a bit silly. Do you have some sort of context for this question? Are you trying to prove something else and this is a step in that proof? – JSchlather May 10 '12 at 13:38
The question was about different (unique) roots, so if I changed it to complex roots, then: $char(A) = x^2 +1$ has 2 unique complex roots, while $char(A^2) = (x+1)^2$ has one unique (complex) root, which is less. – fikster May 10 '12 at 13:46

Since the roots of these polynomials are eigenvalues, let's just think in terms of eigenvalues.

The eigenvalues of $A^2$ are just the eigenvalues of $A$, squared (right? It's tempting to believe, and scribbling on a napkin made it seem so but I could be overlooking something silly).

In that case, the real eigenvalues of $A$ would stay real after squaring, and some complex eigenvalues might become real after squaring.

This would show that $A^2$ has at least as many real eigenvalues as $A$. (This is not the original question, but it seems like this might have been the intended question.)

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.