2
$\begingroup$

Given two non-colinear real unit vectors $v,w$, I believe the rank of $M=vv^\top + ww^\top$ is 2 and I'd like to prove it. $vv^\top$ and $ww^\top$ are obviously of rank one, $v$ is not in the kernel of $M$ because $vv^\top v=v$ and $\|ww^\top v\|<1$ (because $v$ and $w$ are not colinear), same with $w$. So the rank of $M$ is at least 2 [edit just realised this argument is wrong; the non-colinearity of the images is required]. Also, by subadditivity of the rank, $\text{rank}(vv^\top + ww^\top )\leq \text{rank}(vv^\top) +\text{rank}(ww^\top )$. Do you see any mistakes/do you have another proof?

$\endgroup$
3
$\begingroup$

Choose a basis in which $e_1=v$ and $e_2=w$. Then $vv^\top+ww^\top=E_{11}+E_{22}$, and it has rank two.

$\endgroup$
1
  • $\begingroup$ Hard to do simpler! $\endgroup$
    – anderstood
    Sep 8 '16 at 19:28
1
$\begingroup$

Hint: Take a vector $x$ and write down an expression for $Mx$. Then, regardless of $x$, $Mx$ is a linear combination of...

$\endgroup$
2
  • $\begingroup$ ... its projection on $v$ and its projection on $w$. $\endgroup$
    – anderstood
    Sep 8 '16 at 19:37
  • 1
    $\begingroup$ So every value of $Mx$ is a linear combination of $v$ and $w$, and hence its rank is 2. $\endgroup$ Sep 8 '16 at 19:43
1
$\begingroup$

Yet another proof: note that $$ vv^T + ww^T = \pmatrix{v&w} \pmatrix{v & w}^T $$ and for any matrix $M$, $M$ has the same rank as $MM^T$.

$\endgroup$

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.