4
$\begingroup$

I was working on a question which defined a new operation between vectors.

Let $\vec{u}=(u_i)_{i\in\lbrack n\rbrack},\vec{v}=(v_i)_{i\in\lbrack n\rbrack}\in\Bbb R^n$, define $\star:\Bbb R^n\times\Bbb R^n\to\mathcal{M}_n(\Bbb R)$ where $\vec{u}\star\vec{v}=A_{ij}=(u_iv_j)$ as a matrix.

The question is as follows:

If $\vec{u},\vec{v}$ are linearly independent, find the rank of the matrix $\vec{u}\star\vec{v}-\vec{v}\star\vec{u}$.

What I did was try to conjecture the rank with a couple of common l.i. vectors. I chose $e_1, e_2$. This gave me the matrix with entries $1$ in $(1,2)$, $-1$ in $(2,1)$, and $0$ elsewhere. Thus if the question is true, the matrix's rank should be $2$.

To prove this I tried checking what happened when $\vec{u},\vec{v}$ were linearly dependent. The only possibility is that $\vec{v}$ is a scalar multiple of $\vec{u}$, then $\vec{v}=\lambda\vec{u}$ for a real $\lambda$.

Should $\lambda$ be $0$, the matrix's rank would be 1. Else the matrix would be the zero matrix.

Now to prove the result, I think that if I prove that the rank is less than or equal to $2$ then I've got the proof. This is because the other cases are covered in what I just wrote.

Still, I feel that even if it were true, my proof would be too flimsy and not elegant. I'm looking for any hints and help to approach the solution.

$\endgroup$
  • 1
    $\begingroup$ Hint: Can you find an expression for the column vectors of $\vec u\ast\vec v-\vec v\ast\vec u$ and show that they are linearly independent? This would conclude that the rank is $n$. $\endgroup$ – Prasun Biswas Mar 13 '18 at 23:03
  • 1
    $\begingroup$ Note that $u \star v = u v^T$. Hence ${\cal R} (u \star v) = \operatorname{sp} \{ u \}$ if $v \neq 0$. $\endgroup$ – copper.hat Mar 13 '18 at 23:07
6
$\begingroup$

We have $u \star v= uv^T$.

$$uv^T-vu^T = \begin{bmatrix}u &-v \end{bmatrix} \begin{bmatrix} v^T \\ u^T\end{bmatrix}= \begin{bmatrix}u &v \end{bmatrix}\begin{bmatrix} 1 & 0 \\ 0 & -1\end{bmatrix} \begin{bmatrix} v^T \\ u^T\end{bmatrix}$$

Hence the rank is at most $2$.

Let $B$ be the right inverse of $\begin{bmatrix} v^T \\ u^T\end{bmatrix}$ which exists since the rows are linearly independent.

$$B^T(uv^T-vu^T)B=\begin{bmatrix} 1 & 0 \\ 0 & -1\end{bmatrix}$$

Hence the rank is $2$.

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ $\begin{bmatrix} u & v\end{bmatrix} \in \mathbb{R}^{n \times 2}$ right? so the matrix in the middle should in $\mathbb{R}^{2 \times 2}$? $\endgroup$ – Siong Thye Goh Mar 14 '18 at 22:48
  • $\begingroup$ Sorry, brain fart. ;-) $\endgroup$ – egreg Mar 14 '18 at 22:50
  • $\begingroup$ no worries, i make lots of such mistakes too. $\endgroup$ – Siong Thye Goh Mar 14 '18 at 22:57

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.