Since the irrationals and the reals have the same cardinality, there must be a bijection between them. Somewhere on this forum I found something like this:

Map all of the numbers of the form $q + n \sqrt2$ where $q \in \mathbb Q$ and $n \in \mathbb N$ into $q + (n+1) \sqrt2$ and map the rest of these numbers into themselves.

The problem is that I can't see how this function could possibly work. For example, we will never get the value of $2$, since $2$ is not an irrational number and adding multiples of $\sqrt 2$ will never make a number rational. Could you please explain to me why this function works?

  • 3
    $\begingroup$ This function is going from $\mathbb{R}$ to the irrationals. $\endgroup$ Feb 5, 2018 at 18:15
  • $\begingroup$ Could we use some other irrational number, like $e$ or $\pi$ instead of $\sqrt 2$ ? $\endgroup$
    – Aemilius
    Feb 5, 2018 at 18:17
  • $\begingroup$ @Aemilius Yes, you could. $\endgroup$ Feb 5, 2018 at 18:17
  • $\begingroup$ Why should you get the value $2$ if it is not an irrational number? $\endgroup$
    – Asaf Karagila
    Feb 5, 2018 at 18:23
  • $\begingroup$ the map maps R to R/Q so there is no $f(x) = 2 \in \mathbb Q$. However $f^{-1}(2 + \sqrt 2) = 2$ as $f(2) = 2 +\sqrt 2$ so there IS a number that is mapped from 2. So the function works. It just goes in the opposite direction than you expected. $\endgroup$
    – fleablood
    Feb 5, 2018 at 19:04

1 Answer 1


(Let $\mathbb I$ be the notation for the irrationals).

This is actually a map from $f:\mathbb R \to \mathbb I$ and not the other way, so the fact that your don't ever get $2$ (or any other rational number) is a feature, not a bug.

Of course this means the there must be an inverse from $f^{-1}:\mathbb I \to \mathbb R$. This wold be. If $x = q + \sqrt 2*n; n\ge 1; n\in \mathbb Z$ then map $x \to q + (n-1)\sqrt 2$; otherwise map all things to themselves. Thus we can get $2$ via $2 + \sqrt 2 \to 2$.

And this seems weird because .... what does the $\sqrt 2$ have to do with anything. And as you probably guessed, and irrational $\omega$ will do as well.

So to demonstrate:

1) $f: \mathbb R \to \mathbb I$.

If $q \in \mathbb Q$ then $q = q + 0*\omega$ so $f(q) = q + \omega \in \mathbb I$.

If $x = q + k\omega; q\in \mathbb Q; k \in \mathbb Z; k \ge 1$ then $f(x) = q + (k+1) \omega\in \mathbb I$.

And if neither of those cases $f(x) = x \in \mathbb I$.

So $f: \mathbb R \to \mathbb I$.

2) $f$ is unto.

If $x \in \mathbb I$ then either $x = q + k\omega$ where $q $ is rational and $k$ is a positive number or .... $x$ doesn't.

If $x$ does then $f(x - \omega) = f(q + (k-1)\omega) = q + k\omega = x$.

And if $x$ doesn't then $f(x) = x$.

So either way there exist a real number that maps to $x$.

So $f$ is onto.

3) $f$ is one-to- one.

Let $f(w) = x \in \mathbb I$. If $x \ne q + k*\omega$ for any rational $q$ and positive integer $z$ then if $w = q + n*\omega$ for some rational $q$ and natural $n$ then $f(w) = q + (n+1)\omega \ne x$ so that's impossible. So $f(w) = w = x$. That's the only one possible value that maps to $x$.

If $x = q + k \omega$ for some rational $q$ and positive integer $k$ and $w$ does not equal any $r + n*\omega$ for any rational $r$ not any rational $n$ then $f(w) = w \ne q+k\omega = x$ so that's impossible. So $w = r + n*\omega$. So $f(w) = r + (n+1)\omega = q + k\omega$.

So $r-q = (k- n-1)\omega$. But if $k-n-1\ne 0$ then the RHS is irrational and that is impossible. So $k - n - 1 = 0$ and $n = k-1$ and $r = q$ and $w = q + (k-1)\omega$ and that is the only one possible value that maps to $x$.

So $f$ is one-to-one.

... and so $f$ is a bijection between $\mathbb R\to \mathbb I$.


Okay, this feels ... wierd. But why?

What's really going on is

$\mathbb I = \mathbb R \setminus \mathbb Q$ and $|\mathbb R| >|\mathbb Q|=$ countable. When we want to make a bijection from a set to itself minus a countable subset of itself the "trick" is to leave the vast majority of the subset alone but to shove the "chain" of countable members to a countable number of exceptions that just follow an infinite but countable chain out of the way. So we "shove" all rational $q$ to $q + \omega$ That obvioulsy works. But were do we shove all the $q + \omega$ to? Well, to $q + 2\omega$ of course! But where do we shove those to? Well, we just shove all the $q + k\omega$ to $q + (k+1)\omega$.

Hopefully seen like that (albeit informal and handwavy) it is clearer why id does (and must) work.


You must log in to answer this question.

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