1
$\begingroup$

Let $T\in\mathcal L(V)$ and $T^{m-1}v\neq 0$ but $T^mv=0$ for some positive integer $m$ and some $v\in V$. Show that $v,Tv,\ldots,T^{m-1}v$ are linearly independent.

I had written a proof but Im not sure if it is correct. And in the case it would be correct I dont know how to write it better and clearly. So I have two questions:

  1. It is the proof below correct?

  2. If so, how I can write it better using the same ideas?


The attempted proof:

1) If $T^{m-1} v$ would be linearly dependent of $T^{m-2} v$ then exists some $\lambda\neq 0$ such that

$$T^{m-1}v=\lambda T^{m-2}v\implies T^mv=\lambda T(T^{m-2}v)=\lambda T^{m-1}v=0\implies \lambda=0$$

Then $T^{m-2}v$ is linearly independent of $T^{m-1}v$.

2) Now observe that

$$\lambda_1v+\lambda_2Tv+\lambda_3T^2v=0\implies T^{m-2}(\lambda_1v+\lambda_2Tv+\lambda_3T^2v)=\lambda_1T^{m-2}v+\lambda_2T^{m-1}v=0$$

so $\lambda_1,\lambda_2=0$ as we had shown previously, so the original equation reduces to

$$\lambda_3T^2v=0\implies \lambda_3=0$$

thus $v,Tv,T^2v$ are linearly independent.

3) Repeating recursively the analysis in 2) for longer lists of vectors of the form $v,Tv,\ldots,T^kv$ for $k< m$ we can show that the list $v,Tv,\ldots,T^{m-1}v$ is linearly independent.

$\endgroup$
4
  • $\begingroup$ This is a related post $\endgroup$ – Bijesh K.S Jul 11 '17 at 19:15
  • $\begingroup$ The if and only if in the first line looks problematic as $T$ isn't invertible. $\endgroup$ – DMath Jul 11 '17 at 19:19
  • $\begingroup$ @DMath thank you... I overlook it completely. $\endgroup$ – Masacroso Jul 11 '17 at 19:20
  • 1
    $\begingroup$ I'm afraid that the last part is the bulk of the proof and you just cut it off by saying “repeating the analysis”. Note that for $\{v,Tv,T^2v,T^3v\}$ you need to go down to $T^{m-3}$, which is not as easy as the initial step for $\{T^{m-2}v,T^{m-1}v\}$. $\endgroup$ – egreg Jul 11 '17 at 20:33
3
$\begingroup$

Hint

$$a_0v+a_1Tv+\ldots+a_{m-1}T^{m-1}v=0 \quad(1)$$

multiply by $T^{m-1}$, then

$$a_0T^{m-1}v=0\to a_0=0$$

so, from $(1)$,

$$a_1Tv+\ldots+a_{m-1}T^{m-1}v=0$$

multiply by $T^{m-2}$, then

$$a_1T^{m-1}v=0\to a_1=0$$

can you finish?

$\endgroup$
2
  • $\begingroup$ ok but, it is my proof correct? $\endgroup$ – Masacroso Jul 11 '17 at 19:17
  • 1
    $\begingroup$ @Masacroso: it seems correct to me. You just need induction to finish it. $\endgroup$ – Arnaldo Jul 11 '17 at 19:19
1
$\begingroup$

See below for a simple proof.

Suppose that they are not linear independent; that is for some choice of $\alpha_{k}$,

$T^{m-1}v = \sum_{k=0}^{m-2}\alpha_{k}T^{k}v$.

Then $0 =T^{m}v = T\cdot T^{m-1}v = \sum_{k=0}^{m-2}\alpha_{k}T^{k+1}v$,

so $Tv,...,T^{m-1}v$ are linearly dependent.

But that means that $T^{m-1}v = \sum_{k=1}^{m-2}\beta_{k}T^{k}v$.

Inductively repeat the argument to eventually conclude that for some $\gamma \neq 0$, $T^{m-1}v = \gamma T^{m-2} v$, but then we would have to say that $0 = T T^{m-1} v = \gamma T^{m-1} v \neq 0$, a contradiction.

$\endgroup$
1
$\begingroup$

More simply put: if you had a linear combination $c_0 v + c_1 T v + \ldots c_{m-1} T^{m-1} v = 0$ with $c_j$ not all zero, let $c_i$ be the first nonzero coefficient. Then $$T^i v = - \sum_{j=i+1}^{m-1} (c_j/c_i) T^j v$$ Applying $T^{m-i-1}$ to both sides, $$T^{m-1} v = - \sum_{j=i+1}^{m-1} (c_j/c_i) T^{m+j-i-1} v = 0$$ But all terms on the right are $0$, so $T^{m-1} v = 0$, contradiction!

$\endgroup$

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.