Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Consider $T\colon\ell^2\to\ell^2$ an operator such that $Te_k=\lambda_k e_k$ with $\lambda_k\to 0$ as $k \to \infty$ how to prove that it is compact?

share|cite|improve this question
up vote 17 down vote accepted

Let $T_n$ such that $T_n(e_k)=\begin{cases}\lambda_ke_k&\mbox{ if }k\leq n\\\ 0&\mbox{ if }k>n \end{cases}$. Then $T_n$ is finite ranked hence compact and for $v\in\ell^2$, $v=(v_0,v_1,\ldots)$ $$\lVert (T-T_n)v\rVert^2=\sum_{k=0}^{+\infty}|\langle((T-T_n)v)_k\rangle|^2=\sum_{k\geq n+1}|(T-T_n)(v_k)|^2=\sum_{k\geq n+1}|\lambda_k|^2\cdot|v_k|^2\\\leq \sup_{k\geq n+1}|\lambda_k|^2\lVert v\rVert_{\ell^2}^2 $$ so $\lVert T-T_n\rVert\leq \sup_{k\geq n+1}|\lambda_k|$ and we conclude that $T_n\to T$ in norm. A norm-limit of compact operators is compact so $T$ is compact.

Conversely, if $T$ is compact, then you can extract from $\{Te_n\}$ a converging subsequence so you can extract from $\{\lambda_ne_n\}$ a converging subsequence, say $\{\lambda_{n_k}e_{n_k}\}$. Since $\lVert \lambda_{n_{k+1}}e_{n_k{+1}}-\lambda_{n_k}e_{n_k}\rVert^2=|\lambda_{n_{k+1}}|^2+|\lambda_{n_k}|^2\to 0$, we should have $\lambda_{n_k}\to 0$. We thus can prove that for each subsequence $\{\lambda_{n_j}\}$, we can extract that a further subsequence which converges to $0$, hence the whole sequence converges to $0$.

share|cite|improve this answer
Sorry to resurrect an old post, but I was stuck on the same question so had a look on here, and I think I have found an error in your proof. (That said, I'm a third year maths student and your page says that you're doing a PhD - the odds aren't in my favour!) – Smiley Sam Nov 28 '14 at 15:35
At the end, you say $$\lVert \lambda_{n+1}e_{n+1}-\lambda_ne_n\rVert^2=|\lambda_{n+1}|^2+|\lambda_n|^2\to 0$$ but actually you only know that a subsequence of $\{ \lambda_n e_n \}_n$ converges, so surely you only know your statement, not for $n$, but for some $n_j$ forming the convergent subsequence $\{ \lambda_{n_j} e_{n_j} \}_j$? This gives $\lambda_{n_j} \rightarrow_j 0$, which doesn't imply $\lambda_n \rightarrow_n 0$. (As I said, you're the one doing the PhD, so apologies if I am mistaken!) – Smiley Sam Nov 28 '14 at 15:36
@SmileySam There was indeed an inaccuracy, thanks for pointed this out. I've edited. – Davide Giraudo Nov 28 '14 at 18:32
Is it correct now? "You can extract a convergent subsequence" is different to "Every subsequence is convergent$. – Smiley Sam Nov 28 '14 at 18:37
Yes it is; you extract a subsequence from a subsequence. – Davide Giraudo Nov 28 '14 at 21:19

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.