15
$\begingroup$

A quite elegant and classic exercise of Calculus (in infinite series) is the following:

If the non-negative sequence $\{a_n\}$ is decreasing and $\sum_{n=1}^\infty a_n<\infty$, then $na_n\to 0$.

To show this observe that, if $\,\sum_{n\ge n_0}a_n<\varepsilon/2$, then for for $n\ge 2n_0+1$, $$\frac{\varepsilon}{2}>a_{\lfloor n/2\rfloor}+\cdots+a_n\ge \frac{1}{2}na_n\ge 0.$$

This, in a sense, is related to the fact that $\sum\frac{1}{n}=\infty$. To go one step further, since $\sum\frac{1}{n\log n}=\infty$, can we obtain, with the same assumptions on $\{a_n\}$, that $\,n\log n\, a_n\to 0$?

This conjecture holds with the additional assumption that $b_n=na_n$ is also decreasing. To see this, let $n_0\in\mathbb N$, such that $\sum_{n\ge n_0}a_n<\varepsilon$. Then for $n\ge n_0^2$, we have $$ \varepsilon>\sum_{\sqrt{n}\le k\le n}a_n\ge \sum_{\ell=1}^{\lfloor\log_2 \sqrt{n}\rfloor}\sum_{k=\lfloor2^{\ell-1}\log_2\sqrt{n}\rfloor+1}^{ \lfloor2^\ell\log_2\sqrt{n}\rfloor}a_k\ge \sum_{\ell=1}^{\lfloor\log_2 \sqrt{n}\rfloor} (2^{\ell-1}\log_2\sqrt{n}-1)a_{\lfloor2^\ell\log_2\sqrt{n}\rfloor} \\ \ge \sum_{\ell=1}^{\lfloor\log_2 \sqrt{n}\rfloor} \Big(\frac{1}{2}{\lfloor2^\ell\log_2\sqrt{n}\rfloor}-1\Big) a_{{\lfloor2^\ell\log_2\sqrt{n}\rfloor}}\ge \sum_{\ell=1}^{\lfloor\log_2 \sqrt{n}\rfloor}\Big(\frac{1}{2}n-1\Big)a_n\ge (\log_2 n-1)\Big(\frac{1}{2}n-1\Big)a_n. $$ Hence, $(\log_2 n-1)\big(\frac{1}{2}n-1\big)a_n\to 0$, which implies that $\,\,n\log n\,a_n\to 0$.

One step further: If $a_n>0$, $\sum_{n=1^\infty}a_n<0$ $a_n$, and the sequences $na_n$ and $n\log n a_n$ are decreasing, then $$ n\log n \log\log n\,a_n\to 0. $$

$\endgroup$
9
$\begingroup$

A counterexample is given by $$ a_n=\frac{1}{k^2 e^{k^2}} $$ when $e^{(k-1)^2}\leq n< e^{k^2}$. Then the number of terms equal to $1/(k^2 e^{k^2})$ is bounded above by $e^{k^2}$, so $\sum_n a_n\leq \sum_k 1/k^2<\infty$, but when $n$ is slightly less than $e^{k^2}$, $$ n\log n\,a_n\approx1, $$ so $n\log n\, a_n\not\to 0$.

$\endgroup$
2
$\begingroup$

Actually, there is nothing specific with $\log n$: the best we can say is that $n\cdot a_n\to 0$.

Let $\left(R_n\right)_{n\geqslant 1}$ be a sequence of non-negative numbers with goes to infinity. Then there exists a non-decreasing sequence of non-negative numbers $\left(a_n\right)_{n\geqslant 1}$such that $\sum_{n=1}^{+\infty}a_n$ is finite but $\left(nR_na_n\right)_{n\geqslant 1}$ does not converge to $0$.

Indeed, let $\left(N_k\right)_{k\geqslant 1}$ be an increasing sequence of integers such that $R_{N_{k+1}} \geqslant k^2$ for all $k\geqslant 1$. If $N_k\lt n\leqslant N_{k+1}$, let $$a_n:=\frac 1{k^2N_{k+1}}.$$ Then $$ \sum_{k=1}^{+\infty}\sum_{n=N_k+1}^{N_{k+1}}a_n=\sum_{k=1}^{+\infty}\sum_{n=N_k+1}^{N_{k+1}}\frac 1{k^2N_{k+1}}= \sum_{k=1}^{+\infty}\frac{N_{k+1}-N_k}{k^2N_{k+1}}\leqslant \sum_{k=1}^{+\infty}\frac{1}{k^2}<+\infty $$ and for all $k\geqslant 1$, $$ N_{k+1}R_{N_{k+1}}a_{N_{k+1}}=\frac 1{k^2}R_{N_{k+1}}\geqslant 1. $$

If we want a strictly decreasing sequence, add to $a_n$ a term $b_n$ which decreases to $0$ and such that $\sum_{n\geqslant 1}nR_nb_n$ is finite.

$\endgroup$

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.