Assume $a_n\geq 0$ is a sequence of positive real numbers which satisfy the following inequality: for each $n,m\in\mathbb{N}$, we have $$(n+m)a_{n+m}\leq na_n+ma_m.$$ I can't show the convergence of this seemingly well-behaved sequence (I am guessing it does converge, and that one should use a monotonic trick?). Any hint is appreciated. I found this in a list of exercises in sequences and convergence, and I believe it should be elementary because so are the rest of the problems in that list. Thanks in advance!

For completeness, my -failed- attempts so far:

  • Using induction, one proves esily that for any $p,n\in\mathbb{N}$ the inequality $$a_{pn}\leq a_n.$$ From here one might conclude convergence of subsequences of the form $\{a_{{p^k}n}\}_k$ for any $p,n\in\mathbb{N}$. But unfortunately, I can't get an argument from here.

  • Another promising-looking inequality is $$a_{n+1}\leq \frac{n}{n+1}a_n+\frac{a_1}{n},$$ but again this does not suffice.

  • Finally, I tried playing with the sequence (for some $r$ fixed) $$A_n=\min\left\{\frac{nx_n+x_1}{n+1},\frac{(n-1)x_{n-1}+x_2}{n+1},\dots,\frac{(n-1)x_{n-r}+x_{r+1}}{n+1}\right\},$$ becase its a trick I have seen elsewhere when solving exercises on convergence of bounded sequences. I don't think this is the way, because I believe this would only work if you had an estimate of $a_{n+1}$ as a "convex" combination of the $r$-tail.

EDIT: It seems like everyone can prove this result one way or another. A friend of mine immediately suggested to apply Fekete's Subadditive Lemma here to the sequence $b_n=n a_n$.


This is a classic exercise, I give below a solution. I do not know if it is the simplest one. You have done a part of the work.

Put $L=\inf\{a_n\}$. Let $\varepsilon>0$.

a) There exists $m\geq 1$ such that $a_m<L+\varepsilon$. You have proven that then, for every $k\geq 1$, one has $a_{mk}\leq a_m\leq L+\varepsilon$.

b) Put $M=\max\{a_0,\cdots a_{m-1}\}$. Then for any $r\in \{0,\cdots,m-1\}$

$$a_{mk+r}\leq \frac{km}{km+r}a_{km}+\frac{r}{km+r}a_r\leq L+\varepsilon+\frac{M}{mk}$$

c) There exist $k_0$ such that $\frac{M}{mk}<\varepsilon$ for $k\geq k_0$.

d) Take $n\geq (k_0+1)m$. We can write $n=mk+r$, with $r\in \{0,\cdots,m-1\}$ and $k\geq k_0$. Hence by the above, $a_n\leq L+2\varepsilon$. As $a_n\geq L$, we are done.

  • $\begingroup$ Thank you very much! Your solution looks really nice, I was even starting to hesitate whether $a_n$ was even converging. $\endgroup$ – Simulacra May 12 '17 at 13:47
  • $\begingroup$ You are welcome. $\endgroup$ – Kelenner May 12 '17 at 13:49

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.