Question 16 of an Analysis Textbook

Axiom Of Completeness: If $X$ and $Y$ are nonempty subsets of $\mathbb{R}$ having the property that $x\leq y$ for every $x\in X$ and $y\in Y$ then there exists a $c\in\mathbb{R}$ such that $x\leq c\leq y$ for all $x\in X$ and $y\in Y$.

Dedekinds Theorem (or part d) ): If $X$ and $Y$ are sets defined in $c)$, then either $X$ has a maximal element, or $Y$ has a minimal element.

Apologies for the confusion; I did not know there were multiply versions of the theorems above.

I am currently on part e); I have managed to show that the Axiom of Completeness implies Dedekind's theorem from pard d), but I am having trouble showing the Axiom of Completeness from Dedekind's Theorem.

My goal: To show that Dedekinds Theorem $\implies$ Axiom of Completeness

I assumed all the necessary properties that $X, Y$ are non-empty subsets of $\mathbb{R}$ and that $x\leq y$ for all $x\in X$ and $y\in Y$. I noticed in part c) that if $X\cup Y=\mathbb{R}$ then $\mathrm{sup}\ X=\mathrm{inf}\ Y$ which would be a really nice choice for $c\in\mathbb{R}$ to close off that $x\leq c\leq y$ (showing the axiom of completeness) however, I am not sure how to show that $X\cup Y=\mathbb{R}$. It's obvious that $X\cup Y\subset \mathbb{R}$(since they are both subsets), but the other inclusion is tripping me up; perhaps I went down a wrong path, or there is a trick to this. Any hints or directions?

  • 1
    $\begingroup$ I am not 100% sure I understand what you mean here. Can you clearly state which version of the axiom of completeness you are using? Also the first paragraph is a bit hard to digest. Did you already show that AC implies point d and want to show d implies AC? Or vice-versa? $\endgroup$
    – AnCar
    Aug 20, 2022 at 0:08
  • $\begingroup$ Apologies for the confusion, I wrote some extra information in the question just now. $\endgroup$
    – kman
    Aug 20, 2022 at 0:40

1 Answer 1


Ok thanks for the clarification. So let $X$ and $Y$ be nonempty sets satisfying the property that for all $x\in X$ and $y\in Y$, we have that $x\leq y$. We need to show there is some intermediate $c$ assuming that point $\textit{d}$ holds.

From point $\textit{b}$ we have that $\sup{X}\leq\inf{Y}$. Now let us define two auxiliary sets $\hat{X}$ and $\hat{Y}$ by \begin{equation*} \begin{split} \hat{X} & :=\{z\in\mathbb{R}:z\leq\sup{X}\}\\ \hat{Y} & :=\{z\in\mathbb{R}:z\geq\inf{Y}\} \end{split} \end{equation*}

Clearly these are nonempty as $X\subset\hat{X}$ and $Y\subset\hat{Y}$ and we also have that for all $x\in\hat{X}$ and all $y\in\hat{Y}$ it holds that $x\leq y$.

$\textit{Case 1:}$ $\mathbb{R}\setminus(\hat{X}\cup\hat{Y})\neq\varnothing$. Then let $c$ be a real number in the difference. By definition then $x\leq \sup{X}\leq c\leq \inf{Y}\leq y$ for all $x\in X$ and $y\in Y$.

$\textit{Case 2:}$ $\mathbb{R}=\hat{X}\cup\hat{Y}$. Then by point $\textit{d}$, $\hat{X}$ has a maximal element, call it $c$ (analogous treatment for the case when $c$ is the minimum of $\hat{Y}$). Then $c=\sup{\hat{X}}\geq\sup{X}\geq x$ for all $x\in X$. But then by point $\textit{b}$ applied to the sets $\hat{X}$ and $\hat{Y}$, we know that $c=\sup{\hat{X}}\leq \inf{\hat{Y}}\leq\inf{Y}\leq y$ for all $y\in Y$.

For the sake of completeness, a certain version of Dedekind's theorem can be used to prove the existence of suprema and infima assuming that you don't want to assert their existence despite how you've presented the problem.

$\textit{Axiom:}$ Whenever $A$ and $B$ are non-empty sets of reals satisfying $a\leq b$ for all $a\in A$ and $b\in B$ and $\mathbb{R}=A\cup B$, then either $A$ has a maximum or $B$ has a minimum.

We use the above axiom to prove that a non-empty set $A$ that has an upper bound has a supremum.

Indeed, let $U(A):=\{y\in\mathbb{R}:a\leq y\text{ for all }a\in A\}$ and let $V(A):=\mathbb{R}\setminus U(A)$. Clearly $\mathbb{R}=U(A)\cup V(A)$. Now pick arbitrary $y\in U(A)$ and $x\in V(A)$. Since $x\in V(A)$, there must exist some $a_x\in A$ such that $x<a_x$. But $a_x\leq z$ for any $z\in U(A)$ by construction, therefore $x<a_x\leq y$. So we can apply the axiom to conclude that either $V(A)$ has a maximum or that $U(A)$ has a minimum and call this number $t$.

If $t=\min{U(A)}$ then $t$ is by definition the least upper bound of $A$ and therefore the sup of $A$. So assume wlog that $t=\max{V(A)}$. But then there exists a unique $a_t\in A$ such that $a_t>t$. The existence is given by the fact that $t$ is not an upper bound for $A$ and the uniqueness by the fact that if there were a different $a_{t'}\in A$ satisfying $a_{t'}>t$, then $\min \{a_t,a_{t'}\}>t$ would be a bigger element in $V(A)$. But then $a_t$ is clearly the maximum of the set $A$ and therefore the sup of $A$.

  • $\begingroup$ I believe this is actually circular, how can we define $\mathrm{sup}\ X$ and $\mathrm{inf}\ Y$ without using the axiom of completeness? $\endgroup$
    – kman
    Aug 20, 2022 at 3:49
  • $\begingroup$ Ummm, isn't this included in the definition of point d? I mean the sets $X$ and $Y$ that satisfy the setup of point d have a supremum respectively an infimum based on point b. If you mean that you cannot assume sups and infs exist, how am I supposed to understand what assumption d means in this scenario? $\endgroup$
    – AnCar
    Aug 20, 2022 at 4:25

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .