I'm currently working on some problem regarding Dirichlet forms and, as the title states, I'm trying to figure out if $$\sum\limits_{k=1}^{\infty} (a_{k+1} - a_k)^2 <\infty \Rightarrow \frac{1}{n^2} \sum\limits_{k=1}^n a_k^2 \to 0$$ holds. I've already tried it with some examples (mostly 'mainstream' sequences) and it seems to always work. However, I don't seem to be able to proof it and thus I tend to think the statement is false...

My initial idea for the proof was this:

We have $a_n = a_1 + \sum\limits_{k=1}^{n-1} a_{k+1} - a_k$. Using $(a+b)^2 \leq 2(a^2+b^2)$ and the Cauchy-Schwarz inequality we get

$$a_n^2 = \left( a_1 + \sum\limits_{k=1}^{n-1} (a_{k+1} - a_k) \right)^2 \leq 2 a_1^2 + 2 \left( \sum\limits_{k=1}^{n-1} (a_{k+1}-a_k) \right)^2$$ $$\leq 2a_1^2 + 2 (n-1) \cdot \sum\limits_{k=1}^n (a_{k+1}-a_k)^2 \leq 2a^2_1 + C \cdot (n-1)$$

But this only implies that $\frac{1}{n^2} \sum\limits_{k=1}^n a_k^2$ is bounded. Any hints/ideas or a counterexample would be much appreciated.


Here is a sledgehammer solution: Using the Hardy's inequality, we get

$$ \sum_{n=1}^{\infty} \frac{a_{n}^{2}}{n^{2}} < \infty. $$

Now you can apply the Kronecker's lemma to obtain the desired conclusion.

This at least shows that your guess is correct. I believe that there is a much simpler solution, and I will update my solution when I find it.

  • $\begingroup$ Thanks for the timely answer. It indeed feels like one of those things were one thinks "there has to be an easy way" :) $\endgroup$ – GenericNickname Oct 24 '14 at 16:47

Here is a solution using Cauchy's inequality and Stolz–Cesàro theorem.

By Stolz–Cesàro theorem, we need to prove

$$\lim\dfrac{1}{n^2}\sum_{k=1}^n a_k^2 = \lim \dfrac{a^2_{n+1}}{2n+1} =0$$ provided the second limit exists.

Let $b_k = a_{k+1}- a_k$ with $b_0 = a_1$, we have $\sum_k b_k^2 < \infty$ and $a_{n+1} = \sum_{k=0}^{n}b_k$.

\begin{align} a_{n+1}^2 = (\sum_{k=0}^{n}b_k)^2 \leq 2(\sum_{k=0}^{N}b_k)^2 + 2(\sum_{k=N+1}^{n}b_k)^2 \leq 2(\sum_{k=0}^{N}b_k)^2 + 2(n-N)(\sum_{k=N+1}^{n}b_k^2) \end{align}

Thus we have \begin{align} \limsup_n\dfrac{a_{n+1}^2}{n} &\leq \limsup_n \dfrac{2(\sum_{k=0}^{N}b_k)^2 + 2(n-N)(\sum_{k=N+1}^{n}b_k^2)}{n}\\ &\leq 2(\sum_{k=N+1}^{\infty}b_k^2) \end{align}

Since it's true for all $N$ and $\sum_{N}^\infty b_k^2 \to 0$ as $N\to \infty$, we have the conclusion.

  • $\begingroup$ Indeed we have a simple solution! (+1) $\endgroup$ – Sangchul Lee Oct 24 '14 at 19:18

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.