3

Let $(x_n)$ be a sequence in a Hilbert space $H$ which weakly converges to $x$. If $\|x_n\| \rightarrow \|x\|$ also, show that $x_n$ converges strongly to $x$.

So, this statement seems to be true. I was wondering how to show it.

I tried with that:

Since $x_n \overset{w}{\rightarrow} x$ weakly, that means that $|\langle x_n - x, y\rangle| < \epsilon$ for every $y$. Since its true for every $y$, I can pick $y = x_n - x$ , which proves the strong convergence is implied by only the weak. Why is that wrong? And how to fix it?

amWhy
  • 210,739
chuckyy
  • 31
  • 1
    You cannot pick $y=x_n-x$ as $y$ may not depend on $n$. Instead, expand $|x_n-x|^2$ and take the limit. – A.Γ. Feb 06 '18 at 07:30
  • Why y shall not depend on n? – chuckyy Feb 06 '18 at 07:38
  • By definition of weak convergence: for every $y$, but every partricular $y$, it cannot be a sequence of different $y$'s. Otherwise you get strong convergence. For example, $x_n\to 0$ weakly, then $\langle x_n,y\rangle\to 0$. Take $y=x_n$, and you get strong convergence. Nonsense. – A.Γ. Feb 06 '18 at 07:40
  • As no one has mentioned this, this is known as the Kadec-Klee property of norms. There is also a stronger property property of norms, known as the Kadec property, where the norm and weak topologies agree on the sphere (i.e. the two subspace topologies on the unit sphere, derived from the norm and weak topologies, are the same). – Theo Bendit Nov 10 '21 at 01:50

1 Answers1

7

\begin{align*} \|x_{n}-x\|^{2}&=\left<x_{n}-x,x_{n}-x\right>\\ &=\left<x_{n},x_{n}\right>-\left<x_{n},x\right>-\left<x,x_{n}\right>+\left<x,x\right>\\ &=\|x_{n}\|^{2}+\|x\|^{2}-\left<x_{n},x\right>-\left<x,x_{n}\right>\\ &\rightarrow\|x\|^{2}+\|x\|^{2}-\left<x,x\right>-\left<x,x\right>\\ &=0. \end{align*}

user284331
  • 56,315