What I know is that every bounded sequence in $\Bbb R^n$ has a converging sub sequence regardless of the chosen norm, since this is true under the Euclidean norm, and all norms on $\Bbb R^n$ are equivalent. My question is

Is every bounded sequence has a converging sub sequence for any norm-induced metric space?

If not, a counter-example is appreciated. Thanks!

  • 2
    $\begingroup$ Sequentially compact=compact in a metric space, so your first claim is wrong. Do you mean to say complete? $\endgroup$ – Hayden Jan 22 '17 at 15:37
  • $\begingroup$ @Hayden I am new to this stuff and I am a bit confused right now. So any bounded sequence has a converging sub sequence in $\Bbb R^n$ with respect to the euclidean norm? Since all norms on $\Bbb R^n$ are equivalent, then the sequence is still bounded w.r.t. other norms, and the sub sequence is still convergent w.r.t. those other norms? Then we say $\Bbb R^n$ is sequentially compact regardless of the chosen norm? Where am I going wrong...? $\endgroup$ – River Jan 22 '17 at 15:50
  • $\begingroup$ For $\mathbb R^n$ to be sequentially compact, every sequence, not only the bounded ones would need to have a convergent subsequence. That is the definition of sequential compactness. However $\mathbb R^n$ is locally compact, i.e. each point has a compact neighborhood. From this it follows that each bounded sequence has a convergent subsequence. Is that probably what you mean? $\endgroup$ – Tim B. Jan 22 '17 at 15:52
  • $\begingroup$ @LeBtz Thanks a lot. I misunderstood the concept. Sorry. I edited the question. $\endgroup$ – River Jan 22 '17 at 15:56
  • $\begingroup$ As a real vector space $\mathbb{R}^n$ is complete for any norm, so yes any bounded sequence has a convergent sub-sequence. Now with $d(a,b) = |\arctan(b)-\arctan(a)|$, $\mathbb{R}$ is a metric space, but it is not complete because the sequence $u_n = n$ is Cauchy but its limit is $+\infty \not \in \mathbb{R}$ $\endgroup$ – reuns Jan 22 '17 at 16:10

I assume that by finite dimensional norm-induced metric space you mean a vector space in the first place because otherwise the concept of a norm or dimension doesn't make any sense.

If $(V, \|\cdot\|)$ is any $n$-dimensional normed $\mathbb F$-Vector space, where $\mathbb F\in\{\mathbb R,\mathbb C\}$ and $\{b_1,...,b_n\}$ is a basis of $V$, then the unique linear map $\varphi$ that maps $b_i$ to $e_i$ where $\{e_1,...,e_n\}$ is the standard base of $\mathbb F^n$ is an isomorphism. It is easily checked that $|\cdot|$ defined by $|x| := \|\varphi^{-1}(x)\|$ defines a norm on $\mathbb F^n$. For this norm it is true that each bounded sequence has a convergent subsequence. If $(x_n)_{n\in\mathbb N}$ is a bounded sequence in $V$, then $(\varphi(x_n))_{n\in\mathbb N}$ is bounded in $(\mathbb F^n, |\cdot|)$, so it has a convergent subsequence $(\varphi(x_{n_k}))_k$, but then $(x_{n_k})_k$ must converge in $V$ because $\varphi^{-1}$ is continuous.


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.