6
$\begingroup$

Attempt : Suppose $A$ has a non-zero eigenvalue $\lambda$. Then corresponding to it's non-zero eigen vector $X$, we have $AX=\lambda X \Rightarrow A^2X=\lambda^2 X\Rightarrow 0=\lambda^2 X$. Which is a contradiction. Hence $\lambda=0$ with algebraic multiplicity $5$.

Looking at all the Jordan normal forms $J$ of $A$, I found one $J$ which gives $J^2=0$ whose $\text {Rank} J=2$. (There is other Jordan normal form with rank$=1$)

Here is that J := $$J = \begin{pmatrix} 0 & 1 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 1 & 0 \\ 0 & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 \\ \end{pmatrix} $$

I found that Jordan normal forms with ranks $3,4$ don't satisfy $J^2=0$. So the least upper bound for the rank is $2$. Is my attempt correct? Thanks.

$\endgroup$
  • 1
    $\begingroup$ The minimal polynomial of the matrix divides $x^2$. This gives you an upper bound on the size of the Jordan blocks. $\endgroup$ – amd May 5 '17 at 7:25
  • $\begingroup$ In your title, use "maximum" instead of "least upper bound": this should be reserved to quantities that vary in $\mathbb{R}$, not in $\mathbb{N}$. $\endgroup$ – Jean Marie May 5 '17 at 7:30
  • $\begingroup$ @amd True. I know this i.e. Largest size of Jordan block of $\lambda_i =$ degree of factor $(x-\lambda_i)$ in minimal polynomial. So by this way I have two Jordan forms, one with rank $1$, and second with rank $2$. Awesome +1. $\endgroup$ – Error 404 May 5 '17 at 8:09
  • $\begingroup$ @JeanMarie I see. I am changing it to "maximum". I think for closed and bounded set, maximum=supremum=least upper bound. Here in this case, we have at the max 6 different ranks viz., $\{0,1,2,3,4,5\}$. So we should be able to use both words maximum or supremum. $\endgroup$ – Error 404 May 5 '17 at 8:13
  • $\begingroup$ Yes of course, but in general, in mathematics, we try to work at "minimum conceptual energy": it's better not to use an involved concept when not necessary. $\endgroup$ – Jean Marie May 5 '17 at 8:17
8
$\begingroup$

Here is a proof not relying on Jordan normal form.

You have exhibited a $5 \times 5$ matrix that has rank $2$.

Let us work by contradiction for the cases $n \geq 3$.

Let us assume that there exists an $A$ with

  • (1) rank$(A)\geq 3$ and

  • (2) $ \ A^2=0.$

By definition of rank function: (1) means that dim$(Im(A)) \geq 3$;

thus, by rank-nullity theorem : dim$(Ker(A)) \leq 5-3=2.$

Let us take any $X \in Im(A)$, it is thus of the form $X=AY$ for a certain $Y$.

Using (2) : $AX=A^2Y=0$ ; thus $X \in Ker(A)$.

As a conclusion: $Im(A) \subset Ker(A)$, which is impossible considering the dimensionalities $ \geq 3$ and $ \leq 2$ obtained above.

$\endgroup$
  • $\begingroup$ This is really cool!! Wow. +1. $\endgroup$ – Error 404 May 5 '17 at 8:23
  • 2
    $\begingroup$ It works also nicely without contradiction: $A^2=0$ $\Rightarrow$ $\text{im}\,A\subset\ker A$ $\Rightarrow$ rank$\,A\le\frac{n}{2}$. $\endgroup$ – A.Γ. May 5 '17 at 13:11
  • $\begingroup$ @ A.Γ. you are right, thanks for the comment! $\endgroup$ – Jean Marie May 5 '17 at 13:16

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.