5
$\begingroup$

Suppose $\mathbb{R}^e=A \cup B$ in which $A \cap B=\varnothing$ and there exist real numbers $a_0$ and $b_0$ such that $a_0 \in A$ and $b_0 \in B$. My question is, can we construct $a \in A$ and $b \in B$ such that there exists a natural number $n$ such that $|a-b|<n$?

The restrictions is that I don't want to use the law of the excluded middle and the axiom of countable choice. My motivation is that $\mathbb{R}^e$, classical real numbers described by Troelstra and van Dalen in the book constructivism in mathematics, is not Archimedean. It is just weakly Archimedean that is:

$$ \forall x\in\mathbb{R}^e ~ \neg\neg\exists n\in \mathbb{N} ~ x<n. $$

What I already know about the problem is that, we can assume, without loss of generality, given any rational number $r$, $b_0+r \in B$. This is because either $b_0+r \in A$ or $b_0+r \in B$. If $b_0+r \in A$ we can solve the problem. So we can assume that $b_0+r \in B$.

$\endgroup$
9
  • 2
    $\begingroup$ The last paragraph has some very fishy logic. How do you get constructively from $\forall x \in [b_0 - n, b_0 + n]. x \in A \lor x \in B$ to $(\exists x \in [b_0 - n, b_0 + n] . x \in A) \lor [b_0 - n, b_0 + n] \subseteq B$? If that's not what you're arguing, please write down a bit more formally what it is that you're arguing. $\endgroup$ Commented Oct 22 at 19:25
  • 2
    $\begingroup$ Another thing that seems stange is that $A$ and $B$ are disjoint, but you say they contain respectively $C$ and $D$, which need not be disjoint. $\endgroup$ Commented Oct 22 at 19:36
  • 1
    $\begingroup$ @AndrejBauer Aside from the fishy logic, I suppose something can be said regarding your second comment: $C$ and $D$ having a common element is exactly the positive answer to the question at hand, and them being disjoint is exactly the negative answer. So, it seems that there's a real error in the reasoning (whatever it is) resulting in considering $C\subseteq A$ and $D\subseteq B$ "without loss of generality", as otherwise, one would have found the solution: $A$ and $B$ are disjoint, so $C$ and $D$ would be so; hence negative answer. $\endgroup$ Commented Oct 22 at 20:58
  • $\begingroup$ @AndrejBauer Yes you are right. For example, we can assume, without loss of generality, given any rational number $r$, $b_0+r \in B$. This is because either $b_0+r \in A$ or $b_0+r \in B$. If $b_0+r \in A$ we can solve the problem. So we can assume that $b_0+r \in B$. But this argument is not enough to conclude that we can assume $b_0+\mathbb{Q} \subseteq B$ as you pointed out in your comment. I edited the question. Thank you very much. $\endgroup$ Commented Oct 24 at 17:41
  • 1
    $\begingroup$ Which definition of $\mathbb{R}$ are you using in your main question? $\endgroup$ Commented Oct 24 at 18:41

1 Answer 1

1
$\begingroup$

We can assume, without loss of generality, that $-1 \in A$. Consider $|b_0|$. Either $|b_0| \in B$ (case 1) or $|b_0| \in A$ (case 2).

Case 1:

Put $$ X=\{x \in A: x < |b_0|\}. $$ According to the book Constructivism in mathematics by Troelstra and van Dalen, every inhabited subset of $\mathbb{R}^e$ which is weakly bounded above, has a supremum. So, since $-1 \in A$, $X$ is inhabited and since it is bounded above, it has a supremum. Put $s=sup(X)$. Either $s \in A$ or $s \in B$. If $s \in B$, since $s$ is the supremum we have $$ \exists x \in X (s-1<x). $$ Since $s$ is the supremum, $x \le s$. So we have $s-1<x \le s$. Therefore $|x-s|<1$ and we are done. If $s \in A$, consider $s+1$. Either $s+1 \in A$ or $s+1 \in B$. If $s+1 \in B$, since $|s-(s+1)|<2$ we are done. If $s+1 \in A$, we must have $s+1 \ge |b_0|$. For seeing that, suppose $s+1 < |b_0|$. So $s+1 \in X$. Therefore, since $s$ is the supremum, $s+1 \le s$ which is a contradiction. So we have $s+1 \ge |b_0|$. Since $s$ is the supremum and $|b_0|$ is an upper bound, we have $s \le |b_0|$. So, we have $s \le |b_0| \le s+1$. So $|s-|b_0|| \le 1$ and we are done.

Case 2:

Since $|b_0| \in A$, $b_0 \le 0$. This is because if $b_0 > 0$, then $|b_0|=b_0$. So we have $b_0 \in A$ which is a contradiction because by assumption, $A \cap B= \varnothing$. Put $$ Y=\{x \in A: x > b_0\}. $$ Either $1 \in A$ or $1 \in B$. If $1 \in B$, since $-1 \in A$ and $|-1-1| < 3$ we are done. So, we can assume, $1 \in A$. Therefore $1 \in Y$. So, $Y$ is inhabited. It is also bounded below. So, it has an infimum. The rest of the argument is similar to the case 1.

$\endgroup$

You must log in to answer this question.

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