0

$(X, ||• ||) $ be a $d$ dimensional normed linear space over $K$.

$\beta = \{e_1,e_2,e_3,...,e_d\} \text {be a basis of } X $.

Given any, $x\in X$ has a unique representation of the form

$x= x_1 e_1 + x_2 e_2 +... +x_n e_n (x_j \in K,\forall j\in \mathbb{N}_d ) $

Then, $(x_1, x_2, x_3,..., x_d) $ is defined to be the coordinate of $x$ with respect to $\beta$.

Question : $$\text{ Given any sequence} (x^{(n)}) _{n\in \mathbb{N}}\text{ and } x \text{ in } X $$

$(x^{(n)})$$ \text{ converges to } x \text { iff it converges Co-ordinatewise. }$

My attempt :

$(x^{(n)}) _{n\in \mathbb{N}} \text{ is sequence in } X $

$\text{ where the n-th term }$

$x^{(n)}=(x_{1}^{(n)},x_{2}^{(n)},..., x_{d}^{(n)}) $

And, $x= (x_1, x_2,..., x_d) $

And, $(x_{j}^{(n)})$ converges to $x_j$

$|x_{j}^{(n)} - x_j | < \frac{\epsilon}{\sum_{n=1}^{d}{||e_{j}||}} $ $(\forall n\ge N_{j} \text { and } j \in \mathbb {N}_d) $

To show, $(x^{(n) })_{n \in \mathbb{N}}$ converges to $x$ in $(X, ||•||) $ \begin{align} ||x^{(n) }- x || &=||\sum_{n=1}^{d}{(x_{j}^{(n)} - x_j) e_j ||} \\ &\le \sum_{n=1}^{d}{|(x_{j}^{(n)} - x_j)| \text{ } {|| e_j||}}\\ &< \epsilon \text { }(\forall n \ge max{\mathbb\{N_j\}} \end{align}

Hence, $(x^{(n) })_{n \in \mathbb{N}}$ converges to $x$ in $(X, ||•||) $

Is my proof of first part is correct?

How can I prove the reverse implication?

Thanks.

SoG
  • 13,327
  • 1
    Your proof of the 1st half is correct. – DanielWainfleet Nov 23 '21 at 16:44
  • Thank you sir. For the converse part can we show it for any general norm other than assuming euclidean norm? In fact any finite dimensional linear space admits an inner product that is a 2-norm . – SoG Nov 23 '21 at 16:53

1 Answers1

1

Let $\lim_{n\to\infty}\|x(n)-x\|=0$ where $x(n)=\sum_{j=1}^de_jx_{n,j}$ and $x=\sum_{j=1}^de_jx_j.$ For simplicity of notation, let $z(n)=x(n)-x=\sum_{j=1}^de_j(x_{n,j}-x_j)=\sum_{j=1}^de_jz_{n,j}.$ Then $\lim_{n\to\infty}\|z(n)\|=0$ and we wish to show that $\lim_{n\to\infty}z_{n,j}=0$ for each $j\le d.$

Suppose by contradiction that $z(n)$ does not converge co-ordinate-wise to $0.$ Take $r>0$ such that tne set $A=\{n\in\Bbb N:\max_{j\le d}|z_{n,j}|>r\}$ is infinite. For each $n\in A$ take some $j(n)\le d$ such that $|x_{n,j(n)}|=\max_{j\le d} |x_{n,j}|.$ Now $d$ is finite so we may take some fixed $j'\le d$ such that the set $B=\{n\in A: j(n)=j'\}$ is infinite.

Consider the sequence $S_B=(\,(|z_{n,j'}|^{-1}\cdot z_{n,1},...,|z_{n,j'}|^{-1}\cdot z_{n,d})\,)_{n\in B}=((w_{n,1},...,w_{n,d}))_{n\in B}$ of members of $\Bbb R^d$ or $\Bbb C^d.$

We have $\forall n\in B\,\forall j\le d\,(|w_{n,j}|\le 1).$ Therefore there exists an infinite $C\subseteq B$ such that $\lim_{n\to\infty ;n\in C}w_{n,j}=w_j$ exists for each $j\le d$. This is a finite-dimensional generalization of the fact that a bounded sequence in $\Bbb R$ or $\Bbb C$ has a convergent subsequence.

We have $|z_{n,j'}|>r$ for $n\in C$ so we have $$(\bullet)\quad 0=r^{-1}\lim_{n\to\infty ;n\in C}\|z_n\|=$$ $$=\lim_{n\to\infty ;n\in C}r^{-1}\|z_n|\ge$$ $$\ge \lim_{n\to\infty ;n\in C}|w_{n,j'}|^{-1}\|z_n\|=$$ $$=\lim_{n\to\infty ;n\in C}\|\sum_{j=1}^de_jw_{n,j}\|.$$ And the sequence $T=(\sum_{j=1}^de_jw_{n,j})_{n\in C}$ converges co-ordinate-wise to $w=\sum_{j=1}^de_jw_j$. So by the first part (proved in your Q), $T$ converges in norm to $w$. And by $(\bullet)$ we must have $w=0.$

BUT $w_{n,j'}=|z_{n,j'}|^{-1}z_{n,j'}$ so we have $|w_{j'}|=\lim_{m\to\infty ;n\in C}|w_{n,j'}|=1.$ Therefore $$0=w=\sum_{j=1}^dw_je_j$$ but at least one co-ordinate of $w$ (namely $w_{j'}$) is non-zero, and this contradicts the linear independence of $\{e_1,...,e_d\}.$

Remark: The linear independence of $\{e_1,...,e_d\}$ must be used eventually. Here it is used only at the finish.

  • Thank you sir. That's the proof I am looking for. It's a really great explanation. – SoG Nov 23 '21 at 19:01
  • Note also that if $X$ is a finite-dimensional vector subspace of a normed vector space $Y$ then $X$ is closed in $Y$. Because if a convergent sequence $x(n)$ in $X$ converges to $x\in Y$ then co-ordinate-wise convergence of $x(n)$ implies $ x=\sum x_je_j\in X.$ – DanielWainfleet Nov 25 '21 at 06:20