3

I tried to prove two directions.

The ($\Leftarrow$) direction is not so hard, it's just in $\mathbb R$, since $\mathbb R$ is complete, Heine-Borel Theorem holds. However, I have a hard time proving ($\Rightarrow$) direction. Which is using Heine-Borel Theorem we can prove the Completeness Axiom.

There is a similar problem posted on this website with solutions. However, it's hard for me to understand either since I have no background on topology...Least Upper Bound Axiom for Heine-Borel Theorem

Any help? Thanks~

Xiao
  • 9,566
Matata
  • 2,088

2 Answers2

4

We show if Heine-Borel theorem holds on $\mathbb{R}$, then $\mathbb{R}$ has the least upper bound property.

Let $F$ be a bounded subset of $\mathbb{R}$, we want to show that $F$ has a least upper bound.

Note that the clousure $\overline F$ is closed and bounded.

If $\overline F$ has a maximal element, we can show this element is actually the least upper bound for $F$, using the fact that the maximal element is in $\overline F$.

Else, by Heine-Borel theorem the open cover of $\overline F\subset\cup_{x\in \overline F}(-\infty, x)$ has a finite sub cover, that is $$F\subset (-\infty, x_1) \cup (-\infty, x_2)\cup \cdots \cup (-\infty, x_n)$$ where each $x_i$ are in side the clousure $\overline F$

Call the largest $x_i$, "$k$", then $k$ is the least upper bound of $F$. You can try to show this using the fact that $k \in \overline F$, so every open ball of radius $\epsilon$ around $k$ contains some element of $F$.

Xiao
  • 9,566
  • $G={(-\infty,x): x\in \overline F}$ is NOT a cover of $\overline F$ when $F$ is bounded and not empty..., It would be if $\overline F$ had no largest member, but then $G$ would have no finite sub-cover..... Therefore $\max \overline F$ exists and must be $\sup F.$ – DanielWainfleet Jul 19 '17 at 05:55
  • Yes, I edited my answer. Thank you – Xiao Jul 19 '17 at 18:05
  • But, again, ${(-\infty,x):x\in\overline{F}}$ is not necessarily a cover of $\overline{F}$. Think of $\overline{F}$ as $[0,1]$. Then, no finite collection of $(-\infty,x)$, with $x\in[0,1]$ is a cover of $[0,1]$. – Vassilis Markos Jul 20 '17 at 08:47
  • @ΒασίληςΜάρκος It breaks down into two cases: 1. $\overline F$ has a max, then this max is the sup of $F$; 2. if $\overline F$ does not have a max, then ${(-\infty,x): x\in \overline F}$ would be an open cover, since there is no max, for each element $y\in \overline F$, there exists another element $\overline F \ni x>y$ and therefore $y\in (-\infty, x)$. – Xiao Jul 20 '17 at 16:18
  • Could you give me an example of a closed and bounded subset of $\mathbb{R}$ that does not include it's maximum? I' ve tried to find one but I feel stuck right now. – Vassilis Markos Jul 20 '17 at 16:28
  • @ΒασίληςΜάρκος Since $\mathbb{R}$ is complete, every closed and bounded subset will include its maximum. If we don't know the space is complete ($\Leftrightarrow$ least upper bound property), closed and bounded set may not include its maximum. For example $(0,1)$ with the subspace topology, $(0,1)$ is closed and bounded but does not have any max since $(0,1)$ is not complete. – Xiao Jul 20 '17 at 18:08
  • I think that the question refers to $\mathbb{R}$ and, hence, $\overline{F}$ falls to the first case. But then, the proof you provide does not make any use of Heine-Borel's theorem, so, either there is a mistake, or it is not yet clear to me, where is H-B Theorem being used. Also, to claim that a closed set contains it's sup - if it ever exists, since we have not accepted Completeness Principle - doesn't it recquire Completeness? – Vassilis Markos Jul 20 '17 at 18:17
  • @ΒασίληςΜάρκος To say "$\overline F$ falls to the first case", you need to know that $\mathbb{R}$ is complete. And when I said $\overline F$ does or does not contain its max, I meant either $\overline F$ has a maximal element or it doesn't, no completeness principle is used. When it has a maximal element, we are done since this element is the l.u.b. for $F$, if it does not have a maximal element, then we use H-B theorem to conclude that $F$ has a l.u.b. – Xiao Jul 21 '17 at 02:07
  • @ΒασίληςΜάρκος Also if we look at a subset $A\subset \mathbb{R}$, if H-B property holds on $A$, then $A$ should be complete (has the l.u.b property). Again if we let a bounded set $F\subset A$ and look at $\overline F$, there is no way to know if "$\overline F$ falls to the first case". – Xiao Jul 21 '17 at 02:38
  • I think I get your thought, now. Thanks! – Vassilis Markos Jul 21 '17 at 08:10
2

Let $A\subseteq\mathbb{R}$ be a non-empty bounded set - actually it is needed to be bounded from above only. Then, the set $B=\overline{A}$, which is the closure of $A$, is a closed and bounded set which, by Heine-Borel theorem means that $B$ is compact and, hence, due to compactness' definition, every open cover of $B$ has a finite sub-cover. We now need to find a "proper" open cover of $B$.

To avoid any topological reference - well, at least as much as we can - we will first show that: $$\mbox{Heine-Borel Theorem}\Rightarrow\mbox{Cantor's Completeness Principle}$$ Let us, at first, remember Cantor's Completeness Principle:

Let $I=[a,b]$ be an interval in $\mathbb{R}$ and let $a_n,b_n\in I$, $n\in\mathbb{N}$, such that:

  1. $a_n<b_n$ and $I_n=[a_n,b_n]$,

  2. $b_n-a_n\to0$,

  3. $I_n\subseteq I_{n-1}\subseteq\dots\subseteq[a,b]$ for every $n\in\mathbb{N}$.

Then, $$\bigcap_{n=1}^\infty I_n\neq\varnothing$$ and, especially, there exist a $c\in I$ such that: $$\bigcap_{n=1}^\infty I_n=\{c\}$$

So, let $I=[a,b]$ and $I_n=[a_n,b_n]$ as supposed above. We will show, given the truth of Heine-Borel Theorem, that $\bigcap_{n=1}^\infty I_n=\{c\}$.

At first, it is evident that, if $\bigcap_{n=1}^\infty I_n\neq\varnothing$ then it cannot contain more than one elements, since, let $c,d\in \bigcap_{m=1}^\infty I_m\subseteq I_n$ for every $n\in\mathbb{N}$. Then, $\{c,d\}\subseteq I_n=[a_n,b_n]$ for every $n\in\mathbb{N}$, so: $$|c-d|<b_n-a_n\to0\Rightarrow|c-d|=0\Rightarrow c-d=0\Rightarrow c=d$$ So, it remains to prove that $\bigcap_{n=1}^\infty I_n\neq\varnothing$.

For this purpose, let us suppose that $\bigcap_{n=1}^\infty=\varnothing$, and let $$V_n=\mathbb{R}\setminus I_n$$ Now, note that: $$\bigcup_{n=1}^\infty V_n=\bigcup_{n=1}^\infty\mathbb{R}\setminus I_n=\mathbb{R}\setminus\bigcap_{n=1}^\infty I_n=\mathbb{R}\setminus\varnothing=\mathbb{R}$$ So, $\mathcal{V}=\{V_n:n\mathbb{N}\}$ is an open cover of $\mathbb{R}$ and, as a result, of $[a,b]$. Since $[a,b]$ is close and bounded, from Heine-Borel Theorem, we have that $[a,b]$ is compact and, as a result, there exist a finite sub-cover of $\mathcal{V}$, let: $$\mathcal{U}=\{V_{n_k}:k=1,2,\dots,m\}$$ So, we have: $$[a,b]=\bigcup_{k=1}^mV_{n_k}=V_{n_m}$$ since $V_{n}$ is an increasing sequence of sets - as long as $I_n$ is decreasing. But $$V_{n_m}=(-\infty,a_{n_m})\cup(b_{n_m},+\infty)$$ and, this is crearly not a cover of $[a,b]$, since, for instance $\frac{a_{n_m}+b_{n_m}}{2}\not\in V_{n_m}$, which leads to a contradiction.

So, we have proved that: $$\mbox{Heine-Borel Theorem}\Rightarrow\mbox{Cantor's Completeness Principle}$$ Now, we weill prove that: $$\mbox{Cantor's Completeness Principle}\Rightarrow\mbox{Completeness Principle}$$ For this purpose, let us consider a non-empty set $A\subseteq\mathbb{R}$ which is bounded from above. Then, since $A$ is non-empty, there exists an $a\in A$ and, due to boundedness, there exists a $x\in\mathbb{R}$ such that: $$y\leq x\mbox{, for every }y\in A$$ If, $x=a$ or $x\in A$, we have nothing to prove, since $x$ is the wanted lesser upper bound. Let us then suppose that $a<x$. Then, if there exists an element of $A$, let $y_1$, such that: $$y_1>\frac{a+x}{2}$$ we set $$a_1=\frac{a+x}{2}\mbox{ and }b_1=x$$ In any other case, we set: $$a_1=a\mbox{ and }b_1=\frac{a+x}{2}$$ Then, if there exist a $y_2\in A$ such that $$y_2>\frac{a_1+b_1}{2}$$ we set $$a_2=\frac{a_1+b_1}{2}\mbox{ and }b_2=b_1$$ In any other case, we set: $$a_2=a_1\mbox{ and }b_2=\frac{a_1+b_1}{2}$$ In this way we can define, by induction, a sequence of closed intervals $I_n=[a_n,b_n]$, such that:

  1. $b_n-a_n=\frac{x-a}{2^{n+1}}\to0$,
  2. $I_n\subseteq I_{n-1}\subseteq\dots\subseteq[a,x]$

Now, from Cantor's Completeness Principle, there exists a $c\in[a,b]$: $$\bigcap_{n=1}^\infty I_n=\{c\}$$ Now, it is evident by $I_n$'s construction that for every $n\in\mathbb{N}$, $b_n$ is a bound of $A$ and $I_n\cap A\neq\varnothing$, so, $c$ is trully $A$'s lesser upper bound since: $$a_n\nearrow c\mbox{ and }b_n\searrow c$$