Let $A$ be a diagonalizable real matrix such as $A^3=A$. Prove that $\mbox{rank}(A) = \mbox{tr}(A^2)$.

My way so far:

I know that if $B$ is simillar to $A$ then $B^3=B$.

$A$ is diagonalizable so there is a diagonal matrix $D=P^{-1}AP$

$rank(A)=rank(D)$ and $trace(A)=trace(D)$

$rank(A^2)=rank(D^2)$ and $trace(A^2)=trace(D^2)$

$rank(A^2)\le rank(A)$


But how can I continue from here?


Since $D^3=D$ so the eigenvalues are either $0$ or $-1$ or $1$. If we denote the number of nonzero eigenvalues by $p$ and the dimension of matrix by $n$ then since $A^2$ is equivalent to $D^2$ it has $p$ eigenvalues with value $1$ and $n-p$ zeros. therefore $$tr(A^2)=\sum{\lambda_i}=p$$ also the rank of a matrix is number of nonzero eigenvalues since the zero ones make linear dependence between columns or rows. Therefore $$rank(A)=p$$which leads us to $$rank(A)=tr(A^2)$$

| cite | improve this answer | |
  • $\begingroup$ Why the rank of a matrix should be the number of its non-zero eigenvalues? If we pick $A= \begin{bmatrix} 0 & 1 \\ 0 & 0\\ \end{bmatrix} $ then A has no eigenvalues but zero and $rk(A)=1$ $\endgroup$ – edo1998 Jan 31 '18 at 22:07
  • $\begingroup$ But this is not diagonalizable $\endgroup$ – Mostafa Ayaz Jan 31 '18 at 22:11
  • $\begingroup$ Of course you are right, sorry for the mistake $\endgroup$ – edo1998 Jan 31 '18 at 22:18
  • $\begingroup$ It is OK....... $\endgroup$ – Mostafa Ayaz Jan 31 '18 at 22:21

One may use the fact that the trace of a projection of a finite-dimensional vector space over a field of characteristic $0$ equals its rank. Observe that $A^2$ is a projection (or an idempotent linear transformation). Thus, $\text{rank}(A^2)=\text{trace}(A^2)$. It is an easy exercise to prove that $\text{rank}(A)=\text{rank}(A^2)$.

In general, let $n$ and $k$ be nonnegative integers such that $n\geq k$. For a linear transformation $A$ on a finite-dimensional vector space $V$ over a field of characteristic $0$, the condition $A^{n+k}=A^k$ implies that $\text{rank}(A^k)=\text{trace}(A^n)$.

| cite | improve this answer | |

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.