-2

Let $(a_n)$ and $(b_n)$ be two sequences of non-negative real numbers such that for all $n \in \mathbb{N}$, $a_{n+1} \leq a_n + b_n$ and $\sum_{n=1}^{\infty} b_n$ converges. Then, $\lim_{n \to \infty} a_n$ exists and is finite.

the only thing i could think of was the same as someone did in this question, How to prove the sequence $\{b_n\}$ with $0 \leq b_{n+1} \leq b_n + a_n$ converges, where $a_n \geq 0$ converges to $0$. which is just bounding the sequence. I guess he wasn't looking for a proof of the statement so here I am.

anton
  • 9
  • 1
    See https://math.stackexchange.com/q/2199456/42969, https://math.stackexchange.com/q/908654/42969, https://math.stackexchange.com/q/2199243/42969 – Martin R Sep 24 '23 at 11:15
  • 2
    The title is different from the question. The answers to the two versions are different. – geetha290krm Sep 24 '23 at 11:24
  • the title is just so different from the question they didn't even suggest it as a link to my question until you pointed it out. – anton Sep 24 '23 at 11:42
  • The title says $(b_n)$ converges to $0$. The body of the question says $\sum_{n=1}^{\infty} b_n$ converges. These two conditions are different. Please either state that you are asking two separate questions (for the price of one), or choose which one of the two questions you really want answered. – Adam Rubinson Sep 24 '23 at 12:29

1 Answers1

0

Let $(a_n),(b_n)$ be nonnegative sequences such that $a_{n+1}\leq a_n+b_n$, we have

  1. If $b_n\to0$, then $a_n$ may not converge (your link already provided a counterexample)
  2. If $\sum_nb_n$ converges, then $a_n$ converges.

To prove 2, assume $a_n$ diverges, then there exist $\varepsilon>0$ and a sequence $(n_1,n_2,...)$ such that $|a_{n_{k+1}}-a_{n_k}|>\varepsilon$, or equivalently, $$(1)\ 0\leq a_{n_{k+1}}<a_{n_k}-\varepsilon\quad\text{or}\quad(2)\ a_{n_{k+1}}>a_{n_k}+\varepsilon.$$ Note that $(1)$ can happen finitely many times, because otherwise $a_{n_k}$ will become negative eventually.

As for $(2)$, since $\sum_nb_n$ converges it is Cauchy and therefore there exists $N\in\mathbb Z^+$ such that $\sum_{n=p}^qb_n<\varepsilon$ for all $q>p>N$. This implies $$a_{q+1}=a_p+\sum_{n=p}^q(a_{n+1}-a_n)\leq a_p+\sum_{n=p}^qb_n\leq a_p+\varepsilon.$$ We can choose $(p,q+1)$ as $(n_k,n_{k+1})$ for a sufficiently large $k$, which means $(2)$ must also happen finitely many times. But this means $(n_k)$ is finite, contradiction.