Let $A \in \mathbb{R}^{n \times n}$ be a diagonalizable matrix with $1$ or $-1$ as its only eigenvalues. Prove that $A^{2} = I_{n}$.

Could someone help me on this one? I have no idea how to start.

  • 1
    $\begingroup$ If $A$ is diagonalizable then $A = Q \Lambda Q^{-1}$, hence $A^2 = Q \Lambda Q^{-1} Q \Lambda Q^{-1} = Q \Lambda^2 Q^{-1}$. Since your $\Lambda$ is diagonal with $\pm 1$ on its diagonal, $\Lambda^2 = I_n$, hence $A^2 = Q Q^{-1} = I_n$. $\endgroup$ – Florian Jan 10 '17 at 12:06
  • $\begingroup$ Right, I get it, thank you! $\endgroup$ – simp Jan 10 '17 at 12:11
  • $\begingroup$ Interesting. This question is a partial converse of Is the matrix $A$ diagonalizable if $A^2=I$ $\endgroup$ – BCLC Oct 24 '18 at 5:51

You simply have to use the hypothesis that $A$ is diagonalizable : $A=PDP^{-1}$ where $D$ is a diagonal matrix with non-zero coefficients which are $1$ or $-1$ (and $P$ is invertible). From this you can easily compute $A^2$ and find $I_n$.


Since $A$ is a diagonalizable matrix, by definition, it can be written as $PDP^{-1}$ where $D$ is an $n \times n$ diagonal matrix (only non-zero elements are on the diagonal). Moreover, the entries on the diagonal of $D$ can be comprised only of eigenvalues of $A$ (in this case, $1$ or $-1$).


$ \begin{align*} A^2 &= (PDP^{-1})^2 \\&= (PDP^{-1})\cdot(PDP^{-1}) \\&= PD (P^{-1}P)DP^{-1} \\&= PD( I_n)DP^{-1} \\&= PD^2P^{-1}. \end{align*} $

Now, let's think about what possible matrices $D^2$ could be. Since its entries (remember, only on the diagonal) must be either $1$ or $-1$, and $1^2 = (-1)^2 = 1$, $D^2$ will always be $I_n$ (the $n \times n$ identity matrix). So,

$ \begin{align*} PD^2P^{-1} &= PI_nP^{-1} \\&=PP^{-1} \\&= I_n. \end{align*} $

Thus, $A^2 = I_n$.


Another proof: we have

$$ \mathbb R^n=ker(A-E) \oplus ker(A+E).$$

If $x \in \mathbb R^n$, then $x=y+z$ with $Ay=y$ and $Az=-z$. We get

$Ax=Ay+Az=y-z$, therefore



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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