Similar question: An open interval as a union of closed intervals

My approach to this question feels ham-fisted. I would appreciate,

  1. Correction if supplied proof is wrong
  2. Discussion to allieviate ham-fistedness
  3. Avoidance of the word 'clearly'

To Prove: For every $x$ in an open interval $I$, there exists a closed interval $\bar{I}_{\frac{\delta}{2}}$ strictly contained in $I$, with $x \in \bar{I}_{\frac{\delta}{2}}$.

(Here we consider $(\mathbb{R},|\cdot|)$ where the $|\cdot|$ is the Euclidean metric.)


An interval $I\subset \mathbb{R}$ is open iff \begin{align*} \exists \delta >0, \delta \in \mathbb{R}, \text{ such that } (x-\delta,x+\delta) \subset I \quad \forall x \in I \end{align*}


Suppose $I$ is open and $I=(a,b)$ with $a<b$, and $a,b \in \mathbb{R}$. Then pick an $x' \in I$. Since $I$ open, $\exists \delta>0$ with $(x'-\delta,x'+\delta)=I_\delta \subset I$.

But then $(x'-\frac{\delta}{2},x'+\frac{\delta}{2})=I_{\frac{\delta}{2}} \subset I_\delta \subset I$, is also an open set.

Claim: $\bar{I}_{\frac{\delta}{2}}=[x'-\frac{\delta}{2},x'+\frac{\delta}{2}]$ is a closed set and strictly contained in $I$.

We prove this by checking that $\bar{I}^c=(-\infty,x'-\frac{\delta}{2})\cup(x'+\frac{\delta}{2},+\infty)$ is an open set.

Suppose $y \in \bar{I}^c$. Then either $y \in (-\infty,x'-\frac{\delta}{2})$ or $y \in(x'+\frac{\delta}{2},+\infty) $. Let $y \in(x'+\frac{\delta}{2},+\infty) $.

Define: $\delta_1< |y-(x'+\frac{\delta}{2})|$.

Then $(x'+\frac{\delta}{2})<y<\delta_1+(x'+\frac{\delta}{2})$ and so $y \in \Big((x'+\frac{\delta}{2}),\delta_1+(x'+\frac{\delta}{2})\Big)\subset \Big((x'+\frac{\delta}{2}),+\infty \Big)$, so $\Big((x'+\frac{\delta}{2}),+\infty \Big)$ is open.

A similar argument holds for $\Big(-\infty,(x'-\frac{\delta}{2})\Big)$. Hence $\bar{I}_{\frac{\delta}{2}}\subset I$ is a closed set. Moreover it is an interval , since the closure of an open interval is again an interval.


You’re working much too hard. Let $I=(a,b)$ be an open interval, and let $x\in I$. Let $a'=\frac12(a+x)$ and $b'=\frac12(x+b)$; then




By definition $[a',b']$ is a closed interval, and it follows from $(1)$ that $x\in[a',b']\subseteq(a,b)=I$.

Some points to note:

  • It’s not necessary to choose a closed interval that has $x$ at its centre. You can do so, but it’s extra work. If you do choose to do so, one way is to let $\epsilon=\min\{x-a,b-x\}$ and let $\delta$ be any positive real number less than $\epsilon$; then $[x-\delta,x+\delta]$ is a closed interval centred at $x$ and contained in $(a,b)$, since $$a\le x-\epsilon<x-\delta<x<x+\delta<x+\epsilon\le b\;.$$

  • It’s not necessary to show that $[a',b']$ is a closed set in the topology on $\Bbb R$: you were simply asked for a closed interval, meaning a subset of $\Bbb R$ of the form $[u,v]$ for some $u,v\in\Bbb R$ with $u\le v$. Of course it happens to be true that every closed interval is a closed set in the usual topology on $\Bbb R$, but that’s a separate issue.

  • $\begingroup$ Thanks, very nice answer. In my proof I wasn't assuming $x'$ in the center of $I$, but I wanted to put two open intervals around $x'$ so that there was 'space' to ensure a closed set 'between' them. However, as you pointed out there is clearly a more elegant approach. :) $\endgroup$ – zornslemmings Nov 4 '16 at 19:45
  • $\begingroup$ @zornslemmings: You’re welcome. (I like that screen name. :-)) $\endgroup$ – Brian M. Scott Nov 4 '16 at 19:51
  • $\begingroup$ One might also just note that $[\frac34a+\frac14b,\frac14a+\frac34b]\subset(a,b)$. +1 for a nice answer. $\endgroup$ – MPW Nov 4 '16 at 19:51

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.