How can I prove that if $\sum_{n=1}^\infty a_n$ is Cesàro summable and if $\sum_{n=1}^\infty n |a_n|^2 < \infty$, then $\sum_{n=1}^\infty a_n$ converges?

  • 1
    $\begingroup$ I was working on the same problem. My final goal is to prove the fact $f\in H^{\frac{1}{2}}([0,1])\cap C([0,1])$ implies $S_Nf\rightarrow f$. $\endgroup$ – amathnerd Oct 14 '14 at 14:12
  • $\begingroup$ Yes that's exactly my goal as well! $\endgroup$ – Krishnan Mody Oct 14 '14 at 14:53

@Yiorgos S. Smyrlis I think your counter-example does not hold:

You can easily prove that if $S_n =\sum_{k=1}^n a_k $ :

$S_n$ ~ $\ln(\ln(n))$

Both being terms of diverging series you can say that partial sums are equivalent:

$ \sum_{k=1}^N S_k $ ~ $\sum_{k=1}^N \ln(\ln(k))$ ~ $\int_A^N \ln(\ln(t)) dt$

What we're trying to know is whether $\frac{1}{N}\sum_{k=1}^N S_k$ converges or not ( $\sum a_n$ Cesaro summable):

Let $t=Nu$ :

$B_n = \frac{1}{N}\int_A^N \ln(\ln(t)) dt =\int_{\frac{A}{N}}^1 \ln(\ln(Nu)) du =\int_{\frac{A}{N}}^1 \ln(\ln(N) +\ln(u)) du $

$ B_n = (1-\frac{A}{N})\ln(\ln(N)) + \int_{\frac{A}{N}}^1 \ln(1 +\frac{\ln(u)}{\ln(N)}) du$

$\ln(1+x) = x+ o(x)$

$\implies B_n =(1-\frac{A}{N})\ln(\ln(N)) + \frac{1}{\ln(N)}\int_{\frac{A}{N}}^1\ln(u) du + o(\frac{1}{\ln(N)})$

$\implies B_n=(1-\frac{A}{N})\ln(\ln(N))+K\frac{1}{\ln(N)}+o(\frac{1}{\ln(N)}) $

Finally that proves : $ \frac{1}{N}\sum_{k=1}^N S_k $ ~ $B_n$ ~ $\ln(\ln(N))$

So $\sum a_n$ is not Cesaro summable. I'll come back on the original question in I find something..

Edit Hello, it's been a while and I don't know if you found an answer already but I have something:

Let: $S_n =\sum_{k=1}^n a_k $ ; $C_n =\frac{1}{n}\sum_{k=1}^n S_k $

Then: $C_n =\sum_{k=1}^n (1-\frac{k}{n})a_k= S_n - \frac{1}{n}\sum_{k=1}^n ka_k $ Now we know that $(C_n)$ converges, and we are left with the right hand side to handle.

Let's use the second hypothesis, and the well-known result: $u_n\rightarrow u $ => $\frac{1}{n}\sum_{k=1}^n u_k\rightarrow u$

Let $H_n = \sum_{k=1}^n k|a_k|²$; $K_n = \frac{1}{n}\sum_{k=1}^n H_k$

$K_n = \sum_{k=1}^n (1-\frac{k}{n})k|a_k|² = H_n - \frac{1}{n}\sum_{k=1}^n k²|a_k|² $

Now because of the result I just mentioned, $(K_n)$ and $(H_n)$ have the same limit, hence: $\frac{1}{n}\sum_{k=1}^n k²|a_k|² \rightarrow 0$

$|\frac{1}{n}\sum_{k=1}^n ka_k| \leq \frac{1}{n}\sum_{k=1}^n 1*k|a_k| \leq \frac{1}{n} \sqrt{\sum_{k=1}^n 1²}\sqrt{\sum_{k=1}^n k²|a_k|²}=\sqrt{\frac{1}{n}\sum_{k=1}^n k²|a_k|²}$

So we get: $\frac{1}{n}\sum_{k=1}^n ka_k \rightarrow 0$

And from : $S_n= C_n + \frac{1}{n}\sum_{k=1}^n ka_k $ we get:

$S_n\rightarrow lim(C_n)$ : $(S_n)$ converges, CQFD


Take $$ a_n=\frac{1}{n(\log(n+1))}. $$ Then $a_n\to 0$, and hence it is Cesaro summable. Also $$ \sum_{n=1}^\infty na_n^2=\sum_{n=1}^\infty\frac{1}{n(\log(n+1))^2}<\infty, $$ but $$ \sum_{n=1}^\infty \frac{1}{n(\log(n+1))}=\infty! $$

  • $\begingroup$ $\sum (-1)^n$ is Cesaro summable even though the terms are not going to zero. On the other hand, a divergent positive series will not be Cesaro summable. $\endgroup$ – Aaron Meyerowitz Dec 11 '14 at 4:15
  • 1
    $\begingroup$ The sequence $(a_n)$ is cesaro summable , but the series $\sum a_n$ is not. The difference is you've got to study the sequence ($S_n$) where $S_n = \sum_{k=1}^n a_k$ and see if ($S_n$) is cesaro summable. $\endgroup$ – mvggz Dec 11 '14 at 9:40

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.