Let $X$ and $Y$ be norm spaces, and let $T:X\to Y$ be a linear transformation which is continuous under weak convergence.

That is, if $\forall x^\star\in X^\star:x^\star x_n\to x^\star x $ then $\forall y^\star \in Y^\star:y^\star Tx_n \to y^\star Tx$.

Prove that $T$ is continuous under norm convergence.

That is, if $\|x_n\|\to\|x\|$ then $\|Tx_n\|\to\|Tx\|$.

Any hints would be most welcomed.

TIA, Shai


Suppose that $T$ is not continuous under norm convergence. Then there is a sequence $\{x_n\}$ such that $$ \Vert x_n\Vert\le 1,\quad \text { and }\quad\Vert T(x_n)\Vert \ge n^2 \quad\text{ for each }n. $$ (there is a sequence $z_n$ of non-zero vectors in $X$ converging in norm to $0$ for which $\lim\limits_{n\rightarrow\infty} \Vert Tz_n\Vert\ne 0$. Then $\{z_n/\Vert z_n\Vert\}$ is in the unit ball of $X$ and $\{T( z_n/\Vert z_n)\Vert\}$ is unbounded).

Now $\Vert x_n/n \Vert\rightarrow 0$, and thus $y^* (x_n/n)\rightarrow0$ for each $y^*\in Y^*$. Since $T$ is assumed to be continuous under weak convergence, $$y^*T(x_n/n)\rightarrow 0\quad, \text{ for any }y^*\in Y^*. $$

But then, it follows from the Uniform Boundedness Principle$^\dagger$ that $\{T(x_n/n)\}$ is norm bounded.

However, $\Vert T(x_n/n)\Vert\ge n$ for each $n$.

$^\dagger$ We consider $\{Tx_n :n=1,2,\ldots\}$ as a subset of $Y^{**}$. That is we consider $Tx_n$ as a continuous linear functional on $Y^*$ (which is a Banach space). The $Tx_n$ are pointwise bounded, and, thus, norm bounded in $Y^{**}$. But the norms of the $Tx_n$ in $Y^{**}$ are the same as the norms in $Y$.

  • $\begingroup$ Uniform boundness principle requires that the space $X$ is complete, but in the original question this wasn't assumed. $\endgroup$ – Norbert Jan 13 '12 at 19:03
  • $\begingroup$ @Norbert I added some additional detail. Perhaps I was too cavalier with my phrasing. $\endgroup$ – David Mitra Jan 13 '12 at 19:15

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.