A version of the Closed Graph Theorem states that if $T: X\rightarrow Y$ is a linear operator between Banach spaces then $T$ is bounded iff the graph of $T$ is closed in $X\times Y$.

To check if the graph is closed we suppose that a sequence $x_n \rightarrow x\in X$ and that $Tx_n \rightarrow y\in Y$ and then have to show that $x\in X$ and $Tx = y$.

My question is, what if $x_n\rightarrow x\in X$, but $Tx_n$ doesn't converge. Then the graph of $T$ is not necessarily closed, but $T$ is necessarily unbounded. Doesn't this contradict the theorem?

I'm sure I'm misunderstanding something simple - any help would be greatly appreciated!

  • $\begingroup$ If $x_n\to x$ but $Tx_n$ does not converge, then $T$ is discontinuous, correct? $\endgroup$ – Ilya Jul 28 '14 at 15:05
  • $\begingroup$ If $x_n \to x$ and $T x_n$ does not converge, then $T$ is discontinuous, hence its graph is not closed, hence you can find a sequence $(\tilde{x}_n)$ such that $\tilde{x}_n \to \tilde{x}$ and $T\tilde{x}_n \to \tilde{y} \neq T\tilde{x}$. The point of the criterion is that you take a point $(x,y)$ in the closure of the graph, and then need to deduce that the point is in the graph. $\endgroup$ – Daniel Fischer Jul 28 '14 at 15:06
  • $\begingroup$ Thanks for the responses - but why is there necessarily a convergent sequence $(x_n)$ such that $(Tx_n)$ is also convergent? Couldn't it be the case that $(Tx_n)$ does not converge for any convergent $(x_n)$? $\endgroup$ – Sookias Jul 28 '14 at 15:15
  • $\begingroup$ To show that a set $S$ in a Banach space $X$ is closed: Assume $\{ s_{n}\} \subseteq S$ converges to $x\in X$ and show that $x \in S$. Now translate that to your problem of showing that the graph $G(T)\subseteq X\times Y$ is closed in $X\times Y$: Assume $(x_{n},Tx_{n})$ converges in $X\times Y$ to $(a,b)$, and show that $b=Ta$. In either case you start by assuming a sequence in the set converges. When dealing with a sequence in $X \times Y$, the sequence converges iff both coordinates converge. $\endgroup$ – DisintegratingByParts Jul 28 '14 at 16:10

If there is a sequence $x_n \to x$ such that $Tx_n$ does not converge, then $T$ is unbounded, so the theorem says the graph of $T$ is not closed. And then there is some other sequence $y_n$ such that $y_n \to y$ and $Ty_n \to z$ for some $y$ and $z$ but $z \ne Ty$. There is no contradiction.

  • $\begingroup$ Many thanks @Robert Israel, but could you please justify the existence of your other sequence $(y_n)$? I thought that $T$ being unbounded simply means that there is some sequence $y_n \rightarrow y$ such that $Ty_n$ does not converge to $Ty$. As opposed to $Ty_n$ converges but not does not converge to $Ty$. $\endgroup$ – Sookias Jul 28 '14 at 15:21
  • $\begingroup$ That's where "the graph is not closed" comes in. If $(y,z)$ is some point in the closure of the graph that is not in the graph, ... $\endgroup$ – Robert Israel Jul 29 '14 at 2:10

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.