3
$\begingroup$

Let $(u_n)_{n\geq 1}$ be a sequence of real numbers such that $\displaystyle \forall n\geq 1, \;\;\;\;0\leq u_n\leq \frac {1}{n^2}\sum_{k=1}^nu_k$.

Prove that $\sum u_k$ converges.

Let $S_n=\sum_{k=1}^nu_k$ and $A_n=\sum_{k=1}^n \frac 1{k^2}$.

It suffices to prove that $S_n$ converges.

Summing the inequalities $0\leq u_n\leq \frac {1}{n^2}\sum_{k=1}^nu_k$ from $n=1$ to $n=N$ yields $S_N\leq \sum_{k=1}^N \frac{S_k}{k^2}$

Summation by part gives $\displaystyle \sum_{k=1}^N \frac{S_k}{k^2}= S_NA_N-\sum_{k=1}^{N-1}u_{k+1}A_k\leq S_NA_N$

Hence the estimate $S_N\leq \sum_{k=1}^N \frac{S_k}{k^2}\leq S_NA_N$

As a consequence, $S_N$ converges if and only if the series $\displaystyle \sum_{k}\frac{S_k}{k^2}$ is convergent.

This is quite interesting, but it doesn't answer the problem...

$\endgroup$
  • $\begingroup$ Is it true that if you will prove it for upper bound , you will prove it for all interval $\endgroup$ – openspace Apr 27 '16 at 17:38
2
$\begingroup$

You can prove by induction that : $0 \leqslant u_n \leqslant \frac{2}{n(n+1)}u_1$.

Base case for $n=1$, that is okay

Inductive step Assume it is true for all $k < n$. Then :

$u_n \leqslant \frac{u_n}{n^2} + \frac{1}{n^2}\sum \limits_{k=1}^{n-1} u_k$, so $(1 - \frac{1}{n^2})u_n \leqslant \frac{1}{n^2}\sum \limits_{k=1}^{n-1} \frac{2}{k(k+1)}u_1 = \frac{1}{n^2}\sum \limits_{k=1}^{n-1} \Big( \frac{2}{k} - \frac{2}{k+1} \Big)u_1$.

Hence $\frac{n^2 - 1}{n^2} u_n \leqslant \Big( 2 - \frac{2}{n} \Big)u_1 = \frac{2(n-1)}{n} u_1$. We end up with $u_n \leqslant \frac{2}{n(n+1)}u_1$.

Finally you deduce that $\sum \limits_{k=1}^n u_k \leqslant 2 u_1$ (and all the $u_k$ are non-negative), so $\sum u_k$ converges.

$\endgroup$
  • $\begingroup$ Hmm, what led you to believe that $0 \leqslant u_n \leqslant \frac{2}{n(n+1)}u_1$ ? $\endgroup$ – Gabriel Romon Apr 27 '16 at 17:41
  • $\begingroup$ The only possibility for it to not work was if the $u_k$ were too big, and trying to find an upper bound for $u_1$, $u_2$, $u_3$ and $u_4$ using only the inequality, I ended up noticing this pattern $\endgroup$ – charmd Apr 27 '16 at 17:42

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.