Question: Let $\{a_n\}$ and $\{b_n\}$ be sequences such that:

$$b_{n+1} = a_n + a_{n+1}$$

for every positive integer $n$, and suppose that $\{b_n\}$ is convergent. Prove that $\lim_{n\to\infty} \left(\frac{a_n}{n}\right)=0$.

Things I know:

  • Since $\{b_n\}$ is convergent with limit $L$ then : $$ \forall\epsilon>0, \exists N\in\mathbb{N}:\forall n\ge N, |b_n-L|<\epsilon $$
  • $b_{n+1}=a_n+a_{n+1} \implies a_n = b_{n+1}-a_{n+1}$
  • $\lim_{n\to\infty} \left(\frac{a_n}{n}\right)=0 \implies \lim_{n\to\infty} \left(\frac{ b_{n+1}-a_{n+1}}{n}\right) = 0$

I'm not necessarily sure how to go about proving this using the above information. It's clear that I have to show that the difference between $\{b_n\}$ and $\{a_n\}$ converges to a point, and this might have something to do with knowing that a convergent sequences is always Cauchy.

Any help or direction would be appreciated.

Thank you

Edit: I would like to use an approach that involves the following topics: Convergent Sequences, Cauchy Sequences, Supremums and Infimums, Limits, Epsilon Characterizations, Bounded Sequences, Bolzano-Weierstrass Theorem.


First we do the case that $(b_n)$ converges to $0$, the general case will follow from this. Define $(c_n)$ by $c_1=a_1$ and $c_n=b_n$ for $n \geq 2$. Then I claim we have $$a_n=(-1)^n\sum_{k=1}^n (-1)^kc_k,$$ for all $n$. This can be seen by induction, it clearly holds for $n=1$ and if it holds for some $n \geq 1$ then $$a_{n+1}=c_{n+1}-a_n=c_{n+1}-(-1)^n\sum_{k=1}^n (-1)^kc_k=\sum_{k=1}^{n+1}(-1)^{n+1+k}c_k=(-1)^{n+1}\sum_{k=1}^{n+1} (-1)^kc_k,$$ so it holds for $n+1$. Now that we have this, we need to show $\frac{a_n}{n}$ goes to zero, which is equivalent to showing $|\frac{a_n}{n}|$ goes to zero. Notice the following $$|\frac{a_n}{n}| \leq \frac{1}{n}\sum_{k=1}^n |c_k|,$$ so we see that $|\frac{a_n}{n}| $ is bounded above by the average of the first $n$ terms of a sequence converging to $0$ since $c_k \rightarrow 0$. The average of the first $n$ elements of a sequence that converges to $0$, converges to $0$ (see the lemma beneith), that is we have $$\lim_{n \rightarrow \infty} \frac{1}{n} \sum_{k=1}^n |c_k| = 0,$$ which establishes the result for the case that $(b_n)$ converges to $0$. If $(b_n)$ does not converge to zero but to some $L \in \mathbb{R}$, then the sequence $(b'_n)$ defined by $b'_n=b_n - L$ does converge to $0$ and if we define $a'_n=a_n-\frac{L}{2}$, we have $$a'_{n+1}+a'_{n}=b'_{n+1}$$ so by the preceding argument we have that $\lim \frac{a'_n}{n}=0$. But from this we have $$\lim \frac{a_n}{n}=\lim (\frac{a'_n}{n}+\frac{L/2}{n})=0,$$ which completes the proof.

EDIT: Now for the lemma:

Let $(s_n)$ be a sequence of real numbers converging to $0$. Then the sequence $(t_n)$ defined by $$t_n=\frac{1}{n}\sum_{k=1}^n s_k$$ converges to $0$ as well.

Proof Let $\epsilon>0$. There is some $N_1 \in \mathbb{N}$ such that if $n>N_1$ then $|s_n|<\frac{\epsilon}{2}$. We can now find $N_2>N_1$ such that if $n>N_2$, then $$|\frac{\sum_{k=1}^{N_1} s_k}{n}|<\frac{\epsilon}{2}.$$ Now by the triangle inequality, if $n>N_2$ we also have $$|t_n|<\frac{\sum_{k=1}^{N_1} |s_k| + \sum_{k=N_1+1}^n |s_k|}{n} < \frac{\epsilon}{2}+\frac{n-N_1}{n} \frac{\epsilon}{2}<\epsilon.$$

  • $\begingroup$ Firstly, thank you for the detailed answer. Although I do understand your method and approach, the book does not touch series or summations until after this chapter, and so I don't know how I would arrive at your claim for the claim of $a_n$ in the form of a summation.Is there anyway to go about this only using things related to convergent sequences, limsups liminfs, and their properties? I would also appreciate a proof in regards to the average converging to 0. $\endgroup$ – Hushus46 Oct 13 '17 at 0:18
  • $\begingroup$ Sure, shall I edit it into my answer? $\endgroup$ – M. Van Oct 13 '17 at 7:48
  • $\begingroup$ That would be fine, thank you very much again! $\endgroup$ – Hushus46 Oct 13 '17 at 7:49
  • $\begingroup$ And to answer your questions: I arrived at the conclusion about $(a_n)$ when I started realizing that if I'd treat the $(b_n)$ as a "given" sequence and $(a_n)$ as a sequence inductively constructed by $a_{n+1}=b_{n+1}-a_n$, I could get a closed form for $(a_n)$ in terms of $(b_n)$ and $a_1$. (write out some terms and you can see it happening, $a_2=b_2-a_1$, $a_3=b_3-a_2=b_3-b_2+a_1$, etc). Moreover, I claim I only use things related to sequences and no theory about series! Maybe it is confusing since I use sigma notation but I could use other notation as well. $\endgroup$ – M. Van Oct 13 '17 at 8:49

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.