5

I found this problem in real analysis and I have no idea how to start, I just need a hint

let $x_n$ be a bounded sequence of real numbers, that satisfies:

1- $\lim_{n\to \infty} x_{n+1} - x_n = 0 $

2- if $ A = \{x_n: \forall n\in \mathbb{N}\} $ then $A'$ is finite

(This defines A as the range of $x_n$ and A' is the set of limit points of A)

prove that $x_n$ is convergent

for $x_n$ to converge it's enough to show that A' has only one element, that if $x \in A' $ and $y \in A'$ then $x = y$ I tried to deduce that from the definition of the set of limit points and no luck there. and since $x_n$ is bounded then it has a convergent subsequence, which converges to a point in A'. I don't know what to do next

Linom
  • 67
  • Welcome to MSE. Your question is phrased as an isolated problem, without any further information or context. This does not match many users' quality standards, so it may attract downvotes, or closed. To prevent that, please [edit] the question. This will help you recognise and resolve the issues. Concretely: please provide context, and include your work and thoughts on the problem. These changes can help in formulating more appropriate answers. – José Carlos Santos Mar 29 '20 at 07:02
  • (2) says that all but finitely many $x_n$ are integers. We can toss out any finite set without affecting convergence, so you may as well solve the problem when $x_n$ is a sequence of integers. Then, use the definition of (1) with $\epsilon = 1/2$... – Jair Taylor Mar 29 '20 at 07:04
  • @JoséCarlosSantos is it better now? I'm sorry I'm new at this – Linom Mar 29 '20 at 07:11
  • @JairTaylor How does it say that "all but finitely many $x_n$ are integers"? – Linom Mar 29 '20 at 07:12
  • The complement of "x_n is integer" is finite => x_n is not integer is finite => finite non-integer values – Gareth Ma Mar 29 '20 at 07:19
  • I'm sorry I think I wasn't clear, the set A is just the range of the sequence, and the set A' is the set of limit points of A. the question doesn't say that A = ${x_n \in \mathbb{N} }$ it says that A = ${x_n:\forall n \in \mathbb{N} }$ – Linom Mar 29 '20 at 07:22
  • 2
    @Linom Yes, it's better now. – José Carlos Santos Mar 29 '20 at 07:25
  • @Linom Ah, I misread. $A'$ is sometimes used to mean the complement not the closure. Also, I read it as ${a_n: a_n \in \mathbb{N}$. – Jair Taylor Mar 29 '20 at 16:11

3 Answers3

2

If $S\subset \Bbb R$ where $S$ has at least 2 members and $S$ has a member strictly between any 2 of its members, then $S$ cannot be finite.

We show that if $u,v$ are any members of $A'$ with $u<v$ then there exists a member of $A'$ in $(u,v)$ and conclude that if $A'$ has more than 1 member then $A'$ is infinite.

Suppose $u,v\in A'$ with $u<v.$ Let $r=(v-u)/3.$ We show that for any $n\in \Bbb N$ there exists $n'>n$ such that $$(\bullet)\quad x_{n'}\in [u+r,u+2r].$$ So $\{n'\in \Bbb N: x_{n'}\in [u+r,u+2r]\}$ is infinite, so $A'$ has a member in $[u+r,u+2r],$ which is a subset of $(u,v)$.... Here is how:

Given $n\in \Bbb N,$ take $n_1\ge n$ such that $|x_{m+1}-x_m|<r$ whenever $m\ge n_1.$

Now take $n_2\ge n_1$ such that $|u-x_{n_2}|<r,$ which is possible because $u\in A'.$

And $v\in A'$ so take $n_3>n_2$ such that $|v-x_{n_3}|<r. $

We now have $n\le n_1\le n_2<n_3$ and $x_{n_2}<u+r<u+2r<x_{n_3}.$

Finally let $n'$ be the $least$ $j>n_2$ such that $x_j\ge u+r.$

Obviously $n'> n$ (as $n'>n_2\ge n$).

The main point is that $x_{n'-1}<u+r$ and $n'-1\ge n_1$ so $$u+r\le x_{n'}=x_{n'-1} +(x_{n'}-x_{n'-1})<u+r+|x_{n'}-x_{n'-1}|<u+r+r.$$ So $x_{n'}\in [u+r,u+2r] $ as required in $(\bullet)$ above.

1

It is correct that you have to show that $A'$ has only one element. But for this you have to use that $A'$ is finite – otherwise the statement becomes wrong, see for example A sequence of real numbers such that $\lim_{n\to+\infty}|x_n-x_{n+1}|=0$ but it is not Cauchy.

If $A' = \{ a_1, \ldots, a_k \}$ then you can choose a $\epsilon > 0$ such that all the $k$ intervals $(a_i - 2\epsilon, a_i + 2\epsilon)$ are pairwise disjoint.

Show that all but finitely many $x_n$ are in one of the $k$ intervals $(a_i - \epsilon, a_i + \epsilon)$.

Finally use the condition $\lim_{n\to \infty} x_{n+1} - x_n = 0$ to show that for sufficiently large $n$, all $x_n$ are in the same interval $(a_i - \epsilon, a_i + \epsilon)$. Which means that this $a_i$ is the only limit point of the sequence $(x_n)$.

Martin R
  • 113,040
  • Thanks, but I'm still a little bit lost, How do I show that? I could choose $\epsilon$ to be the same as the one that defines the neighborhood then there exist some N such that $$n>N \implies |x_{n+1} -x_n| < \epsilon$$

    How do I proceed?

    – Linom Mar 29 '20 at 08:05
  • @Linom: If $x_n \in (a_i - \epsilon, a_i + \epsilon)$ and $|x_{n+1} -x_n| < \epsilon$ then $x_{n+1}$ cannot lie in a different $(a_j - \epsilon, a_j + \epsilon)$, because those intervals are all at least $2 \epsilon$ apart. – Martin R Mar 29 '20 at 09:00
1

This is building on Martin R.'s answer, which I think misses a crucial element to show that at some point all the sequence elements must lie "near" the same limit point.

Since $A'$ is finite, let's assume the contrary of what we want to show, so we assume that $|A'| \ge 2$. Then we can choose $a_1,a_2 \in A', a_1 < a_2$ such that no real number in the open interval $(a_1,a_2)$ belongs to $A'$.

Set $\epsilon_0=\frac{a_2-a_1}3$.

Because $a_1,a_2$ are limit points of $A$, there must be two infinite sets of indices $I_1, I_2$ such that $\forall i \in I_1: |a_1-x_i| < \epsilon_0$ and $\forall i \in I_2: |a_2-x_i| < \epsilon_0$.

Let now be $a$ any real number in $[a_1+\epsilon_0, a_2-\epsilon_0]$ and $\epsilon>0$. We can find an $N_\epsilon$ such that $\forall n \ge N_\epsilon:|x_{n+1}-x_n| < \epsilon$ (exists because of condition 1).

Because $I_1$ is infinite, we can find an $i_1 \in I_1$ with $i_1 \ge N_\epsilon$. Because $I_2$ is infinite, we can find $i_2 \in I_2$ with $i_2 > i_1$.

So we have

$$x_{i_1} < a_1+\epsilon_0 \le a \le a_2-\epsilon_0 < x_{i_2}.$$

Now it follows that at least one of $x_{i_1}, x_{i_1+1},\ldots,x_{i_2}$ must lie in the interval $(a-\epsilon,a+\epsilon)$. If either $x_{i_1}$ or $x_{i_2}$ do, we are done. Otherwise $x_{i_1}$ is to the left of that interval and $x_{i_2}$ is to the right of it.

Since $i_1 \ge N_\epsilon$, we know that the "step-length" $|x_{n+1}-x_n|$ is less than $\epsilon$ so the sequence cannot "step over" the interval $(a-\epsilon,a+\epsilon)$ of length $2\epsilon$, even if that interval is open.

So what have we shown? For an arbitrary point $a \in [a_1+\epsilon_0, a_2-\epsilon_0]$ and an arbitrary $\epsilon > 0$ we know that some $x_n \in (a-\epsilon, a+\epsilon)$. But that means that $a$ is a limit point of the sequence!

So contrary to our choice of $a_1,a_2$, the interval $(a_1,a_2)$ contained limit points. That means with finite $A'$, we must have $|A'|=1$.

Ingix
  • 14,494
  • How does that follow? "Now it follows that at least one of $x_{i_1}, x_{i_1+1},\ldots,x_{i_2}$ must lie in the interval $(a-\epsilon,a+\epsilon)$. If either $x_{i_1}$ or $x_{i_2}$ do, we are done. Otherwise $x_{i_1}$ is to the left of that interval and $x_{i_2}$ is to the right of it." – Linom Mar 29 '20 at 13:27
  • We have (just above the quoted part) $x_{i_1} < a < x_{i_2}$. So if $x_{i_1}$ is not in the interval $(a-\epsilon, a+\epsilon)$, it must be to the left of it (otherwise we would have $x_{i_1} \ge a+\epsilon > a$ in contradiction to $x_{i_1} < a$ from above. The same argument works for $x_{i_2}$. – Ingix Mar 29 '20 at 20:45