MHB Closed Graph Theorem: Proving T Has Closed Graph

Click For Summary
The discussion centers on proving that the linear mapping T from l_p to l_infinity has a closed graph and that the sequence a defining T belongs to l_q, where (1/p) + (1/q) = 1. It is established that if x converges in l_p and Tx converges in l_infinity, then Tx must equal y, demonstrating that T has a closed graph. The proof for a being in l_q involves constructing a sequence z^(n) in l_p and showing that T is bounded, leading to the conclusion that the series of a_k converges in the l_q space. The participants also discuss the relationship between convergence in norm and coordinatewise convergence across sequence spaces. The conversation emphasizes the importance of these properties in functional analysis.
Cairo
Messages
61
Reaction score
0
Suppose that 1<p<inf and a=(a_k) a complex sequence such that, for all x in l_p, the series (which runs from k=1 to inf) Sigma(a_k x_k) is convergent. Define T:l_p--->s by

Tx=y, where y_j=Sigma(a_k x_k) (where j runs from 1 to j).

I need to prove that

1) T has a closed graph (as a linear mapping from l_p to l_inf)

2) If a is the sequence defining T, then necessarily a in l_q, where (1/p)+(1/q)=1

Could somebody help me out with this please?
 
Physics news on Phys.org
What is $s$ here?
 
Here, s denotes the space of all possible sequences.
 
dray said:
Here, s denotes the space of all possible sequences.
Thanks.

In fact I don't understand the order of the question. Once we have proved that $a\in\ell^q$ (for example by the principle of uniform boundedness, or by contradiction, exhibiting $x\in\ell^p$ such that $\sum_{j=1}^{+\infty}|x_ja_j|=+\infty$), we can see that $T$ is continuous.
 
The question requires us to prove that T has a closed graph (i.e is continuous) first before we go on to show that a in l_q.
 
dray said:
Suppose that $1<p<\infty$ and $a=(a_k)$ a complex sequence such that, for all $x$ in $l_p$, the series (which runs from $k=1$ to $\infty$) $\sum (a_k x_k)$ is convergent. Define $T:l_p\to s$ by

$Tx=y$, where $y_j=\sum (a_k x_k)$ (where $k$ runs from 1 to $j$).

I need to prove that

1) $T$ has a closed graph (as a linear mapping from $l_p$ to $l_\infty$)

2) If $a$ is the sequence defining $T$, then necessarily $a \in l_q$, where $(1/p)+(1/q)=1$.

Could somebody help me out with this please?
For 1), you have to show that if $x^{(n)}\to x$ in $l_p$ and $Tx^{(n)}\to y$ in $l_\infty$, then $Tx=y.$ To see that this condition holds, all you need to know is that, in each of $l_p$ and $l_\infty$, convergence in norm implies coordinatewise convergence. The $j$th coordinate of $Tx^{(n)}$ is $$(Tx^{(n)})_j = \sum_{k=1}^j a_kx^{(n)}_k$$ (where $x^{(n)}_k$ denotes the $k$th coordinate of $x^{(n)}$).

Let $n\to\infty$ on both sides of that displayed equation to see that $y_j = (Tx)_j$. Since that holds for each $j$ it follows that $Tx=y$, as required.

For 2), fix $n$ and define $z^{(n)}\in l_p$ by $$z^{(n)}_k = \begin{cases}\theta_k a_k^{q-1}&\text{if }1\leqslant k\leqslant n, \\0&\text{if } k>n, \end{cases}$$ where $\theta_k$ is a complex number of absolute value 1 chosen so as to make $\theta_k a_k^q$ real and nonnegative. Then $$\|z^{(n)}\|_p = \Bigl(\sum_{k=1}^n |a_k|^{p(q-1)}\Bigr)^{1/p}.$$ The $n$th coordinate of $Tz^{(n)}$ satisfies $$Tz^{(n)}_n = \sum_{k=1}^n a_kz^{(n)}_k = \sum_{k=1}^n |a_k|^q.$$

But it follows from 1) and the closed graph theorem that $T$ is bounded, and therefore $\|Tz^{(n)}\|_\infty \leqslant \|T\|\,\|z^{(n)}\|_p.$ You should be able to deduce that $$\Bigl(\sum_{k=1}^n |a_k|^q\Bigr)^{1/q} \leqslant \|T\|.$$ Finally, let $n\to\infty$ to conclude that $a\in l_q$.
 
How do I prove that for each sequence space, convergence in norm implies coordinatewise convergence?
 
Because for each norm and each sequence $x=\{x_k\}$, we have $|x_k|\leq \lVert x\rVert$.
 
girdav said:
Because for each norm and each sequence $x=\{x_k\}$, we have $|x_k|\leq \lVert x\rVert$.

Thanks girdav, but I'm still looking for a proof that this is the case.
 
  • #10
I don't understand, you want a proof of the last inequality? If it's the $\infty$ norm it's obvious, and for $p$-norms, just use the fact that the series which defines the norm is greater than only one term of this series.
 

Similar threads

  • · Replies 2 ·
Replies
2
Views
2K
  • · Replies 7 ·
Replies
7
Views
3K
  • · Replies 2 ·
Replies
2
Views
2K
  • · Replies 2 ·
Replies
2
Views
2K
Replies
2
Views
2K
Replies
6
Views
2K
Replies
2
Views
2K
  • · Replies 44 ·
2
Replies
44
Views
6K
  • · Replies 2 ·
Replies
2
Views
1K
Replies
2
Views
2K