8
$\begingroup$

For given $n\times n$ matrix $A$ singular matrix, prove that $\operatorname{rank}(\operatorname{adj}A) \leq 1$

So from the properties of the adjugate matrix we know that $$ A \cdot \operatorname{adj}(A) = \operatorname{det}(A)\cdot I$$

Since $A$ is singular we know that $\operatorname{det}(A) = 0$, thus $$ A \cdot \operatorname{adj}(A) = 0$$

This is where I'm getting lost, I think I should say that for the above to happen one of the two, $A$ or $\operatorname{adj}(A)$ would have to be the $0$ matrix, but if $A = 0$ then $\operatorname{adj}(A) = 0$ for sure, which means I said nothing.

A leading hint is needed.

$\endgroup$
  • $\begingroup$ Hmm, when $A$ is singular, isn't one of its eigenvalues zero, therefore $\det(A)=0$... $\endgroup$ – draks ... Jun 3 '13 at 18:52
  • $\begingroup$ we haven't learned eigenvalues yet so I believe we're not supposed to use it... $\endgroup$ – Georgey Jun 3 '13 at 18:55
12
$\begingroup$

Since $A$ is singular, $\mbox{rank}A\leq n-1$.

Case 1: $\mbox{rank}A\leq n-2$. Then $A$ contains no invertible submatrix of order $n-1$. So every minor of order $n-1$ is zero. What can you conclude about $\mbox{adj}(A)$?

Case 2: $\mbox{rank}A= n-1$. By rank-nullity, we get $\dim\ker A=1$. Now $A\cdot \mbox{adj}(A)=0$ means that the range of $\mbox{adj}(A)$ is contained in $\ker A$. So...

$\endgroup$
  • $\begingroup$ Why can you tell that because $A$ is singular $\rightarrow \space\space rankA \leq n-1$? $\endgroup$ – Georgey Jun 3 '13 at 19:18
  • $\begingroup$ @Georgey rank $A=n$ iff $A$ is surjective iff $A$ is invertible. $\endgroup$ – Julien Jun 3 '13 at 19:19
  • $\begingroup$ I don't even know what a surjective matrix embarrassed $\endgroup$ – Georgey Jun 3 '13 at 19:27
  • $\begingroup$ @Georgey Don't you know, at least, that a square $n\times n$ matrix is invertible iff it has full rank $n$? So if it is not invertible (i.e. singular) it has rank $\leq n-1$. $\endgroup$ – Julien Jun 3 '13 at 19:30
  • 1
    $\begingroup$ @Georgey Great. It is helpful to understand the rank of an $n\times m$ matrix as the dimension of the range of the linear map induced by $A:K^m\rightarrow K^n$ on the column vectors. Then you see that the rank of $A$ is $n$ iff the range of $A$ is the whole of $K^n$, that is $A$ is surjective. In the square $m=n$ case, this is equivalent to $A$ being invertible by the rank-nullity theorem. $\endgroup$ – Julien Jun 3 '13 at 20:00
1
$\begingroup$

This is an old question, but I would like to add a solution that only uses properties of matrices i.e. no abstract algebra.
Sylvester's rank inequality states that for two matrices $X,Y \in \mathcal{M_n(\mathbb{C})}$ we have that $\operatorname{rank}(XY)\ge \operatorname{rank}X+\operatorname{rank} Y-n$.
Since $A$ is singular, we have $\operatorname{rank}A\le n-1$.
We have the equality $A\cdot \operatorname{adj}A=O_n$.
Hence, from Sylvester's rank inequality $0=\operatorname{rank}(A\cdot \operatorname{adj}A)\ge \operatorname{rank}A + \operatorname{rank}(\operatorname{adj}A)-n$.(1)
Case 1. $\operatorname{rank}A\le n-2=>\operatorname{adj}(A)=O_n$ as in Julien's solution.
Case 2. $\operatorname{rank}A=n-1$.
From (1) we have that $\operatorname{rank}(\operatorname{adj}A)\le 1$, so $\operatorname{rank}(\operatorname{adj}A)\in \{0,1\}$.
If $\operatorname{rank}(\operatorname{adj}A)=0$, then $\operatorname{adj}A=O_n$ and this contradicts $\operatorname{rank} A=n-1$.
Hence, $\operatorname{rank}(\operatorname{adj}A)=1$ in this case.
In both cases $\operatorname{rank}(\operatorname{adj}A)\le 1$.

$\endgroup$

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.