$\sup_{r \in A} r$ I am stuck at proof of $\cup_{r\in A}^\infty(-r,r)=(-\sup_{r \in A} r,\sup_{r \in A} r)$

First question is how we know that $ \sup$ of given set exists?

Let's take $\{x\in\mathbb{Q}:x\geq 0, \ x^{2}<2\}$ this set does not have a supremum.

Is this equality true for all sets $A$?

For proof.

Let $x \in \cup_{r\in A}^\infty(-r,r)$ that means there exists at least one $(-r,r)$ s.t.$x\in(-r,r)$ and if we know that $\sup$ exists then it will belong to $(-\sup r,\sup r)$

Now let's $x\in(-\sup r,\sup r)$ have difficulties for proving this part and proof of first part is right?

This can be helpful.

  • $\begingroup$ Your example does not have a supremum in $\mathbb Q$, but it does have one in $\mathbb R$. In fact every bounded subset of $\mathbb R$ has a supremum. $\endgroup$
    – Alex G.
    Nov 3, 2021 at 13:38
  • $\begingroup$ Please, change $supr_{r \in A}$ with $\sup_{r \in A} r $ $\endgroup$
    – VIVID
    Nov 3, 2021 at 13:39
  • $\begingroup$ What is the definition of $A$? $\endgroup$
    – user987907
    Nov 3, 2021 at 13:41
  • 1
    $\begingroup$ hint: if $x\in(-\sup A,\sup A)$ then there exists $\epsilon > 0$ such that $-\sup A + \epsilon < x < \sup A - \epsilon$ $\endgroup$
    – podiki
    Nov 3, 2021 at 13:59
  • 1
    $\begingroup$ @podiki the sup equals infinity case need a little different treatment. $\endgroup$ Nov 3, 2021 at 14:22

1 Answer 1


From the formulation it seems that we implicitly have the assumption that $A\subseteq\mathbb R$. (After all, this is probably motivated by the discussion in the comment to this question. The discussion then continued in chat.)

Also the notation in the question is rather non-standard, but looking at the comments in chat it is clear that we're trying to show that $$\bigcup_{r\in A}(-r,r)=(-\sup A,\sup A).$$

First let's have a look at the case $A\subseteq(0,\infty)$ and $A\ne\emptyset$. This is the case needed in the linked question - but we can check later what happens if we allow negative values or if we consider the empty set. (Although such cases are much less relevant.)

Let us denote $R=\sup A$. We know that either $A$ is bounded (and $R\in\mathbb R$) or $A$ is unbounded (and $R=+\infty$).

In either case, we have $(-r,r)\subseteq(-R,R)$ for each $r\in A$, and thus $$\bigcup_{r\in A}(-r,r) \subseteq (-R,R).$$

The relevant part of the proof is the opposite inclusion.

If $R$ is finite. Let us suppose that $R$ is finite. Let us take arbitrary $x\in(-R,R)$, i.e., we have $|x|<R$. Since $R=\sup A$, there exists $r\in A$ such that $|x|<r<R$ and $$x\in(-r,r).$$ So we see that for every $x\in(-R,R)$ we also have $$x\in\bigcup_{r\in A}(-r,r).$$

If $R=+\infty$. In this case $(-R,R)=(-\infty,\infty)=\mathbb R$. For any $x\in\mathbb R$ there exists an $r\in A$ such that $|x|<r$ and $x\in(-r,r)$. Consequently, we get $$x\in\bigcup_{r\in A}(-r,r).$$

In both cases we have shown $$(-R,R)\subseteq \bigcup_{r\in A}(-r,r)$$ and we have the desired inequality.

What if $A$ is empty? For $A=\emptyset$ we have $\sup A=\sup\emptyset=-\infty.$ (If we want to define $\sup\emptyset$ in the context of sets of real numbers, this is the only reasonable value. The other option is to leave this supremum undefined.) So we have $(-\sup A,\sup A)=(\infty,-\infty)=\emptyset=\bigcup_{r\in\emptyset}(-r,r)$.

What if $A$ contains some $r\le0$? Whenever $r\le 0$, we have $(-r,r)=\emptyset$. So adding such $r$'s does not change the union. So we have1 $$\bigcup_{r\in A}(-r,r)=\bigcup_{r\in A\cap(0,\infty)} (-r,r).$$ How taking the intersection influences the supremum? We either have $A\cap(0,\infty)\ne\emptyset$, an in this case $\sup A=\sup(A\cap(0,\infty))$. And in the case that $A\cap(0,\infty)=\emptyset$ we get that2 $R=\sup A\le 0$ and thus $(-R,R)=\emptyset$.

So if we were able to solve the cases $A\subseteq(0,\infty)$ and $A=\emptyset$, we get that the same is true for any $A\subseteq\mathbb R$.

1Just notice that the only difference is that on the LHS we have added some empty sets. I.e., we have $$\bigcup_{r\in A}(-r,r)=\left(\bigcup_{r\in A\cap(-\infty,0]} (-r,r)\right) \cup \left(\bigcup_{r\in A\cap(0,\infty)} (-r,r)\right) = \emptyset \cup \bigcup_{r\in A\cap(0,\infty)} (-r,r) = \bigcup_{r\in A\cap(0,\infty)} (-r,r).$$

2If $A\subseteq (-\infty,0]$ then $0$ is an upper bound for the set $A$. So for the least upper bound we have $\sup A \le 0$. (This is a general property of supremum. The assumption that $a\le x$ for each $a\in A$ implies that $\sup A\le x$.)

  • $\begingroup$ Thanks for answer @MartinSleziak.Can you please explain how you get $\bigcup_{r\in A}(-r,r)=\bigcup_{r\in A\cap(0,\infty)} (-r,r)$ and for case $A \cap (0,\infty)=\emptyset$ how you get that $R=supA \leq 0$ $\endgroup$
    – unit 1991
    Nov 7, 2021 at 13:39
  • $\begingroup$ @unit1991 I have expanded on these two points a bit more in the footnotes at the end of the post. $\endgroup$ Nov 7, 2021 at 13:49
  • $\begingroup$ Thank you very much and one last maybe silly question.$\cup_{r\in A}^\infty(-r,r)=(-\sup_{r \in A} r,\sup_{r \in A} r)$ if we take union up to $n$ not infinity then equality is true? $\endgroup$
    – unit 1991
    Nov 7, 2021 at 14:04
  • $\begingroup$ @unit1991 In my opinion, writing something like $\bigcup_{r\in A}^\infty (-r,r)$ or $\bigcup_{r\in A}^n (-r,r)$ is a non-stardand notation with rather unclear meaning. I would suggest that we could continue the discussion in this chatroom. If you have time to come to chat, feel free to ping me there. $\endgroup$ Nov 7, 2021 at 14:34

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.