2

How do you prove that $\left|A \times A\right|$ is equal to $\left|A\right|$ for every infinite set?

I'm trying to prove this basic fact of cardinal arithmetic, but I'm getting stuck on the uncountable sets.

I suspect there's an obvious technique that I'm missing for proving statements like this.

I'm okay with taking the axiom of choice and am not trying to prove that this fact is equivalent to the axiom of choice as in this question.


Let $\varepsilon$ refer to the empty set.

Let $\mathbb{N}$ refer to the positive integers.


I'm trying to prove that

$$ \text{$\left|A\right| = \left|A \times A\right|$ if and only if $\left|A\right| \in \{0, 1\}$ or $A$ is infinite.} $$

Here's my attempt to prove it.

Let's consider five different mutually exclusive cases for the size of $A$.

  • $A$ is empty
  • $A$ is a singleton
  • $\left|A\right| \in \mathbb{N}_{\ge2}$
  • $A$ is countably infinite.
  • $A$ is uncountably infinite.

case 1/5: $A$ is empty.

$ \varepsilon \times \varepsilon $ is equal to $\varepsilon$.

case 2/5: $A$ is a singleton.

$ \{\sigma_1\} \times \{\sigma_2\} $ is $\{(\sigma_1, \sigma_2)\}$.

case 3/5: $\left|A\right| \in \mathbb{N}_{\ge2}$

Let $x$ be a real variable.

$$ \text{If $x = xx$, then $x \in \{0, 1\}$ } $$

Therefore, the left and right components of the theorem are both never true.

case 4/5: $A$ is countably infinite.

Since $A$ is countably infinite, there exists a bijective function $f : A \to \mathbb{N}$.

There's a bijective function $g$ from $\mathbb{N} \times \mathbb{N}$ to $\mathbb{N}$, Arguments to $g$ are shown below with a leading dot, for example ·7. * marks a value that has been omitted for expository purposes.

                x
         ·1  ·2  ·3  ·4
     ·1   1   3   6  10
  y  ·2   2   5   9   *
     ·3   4   8   *   *
     ·4   7   *   *   *

We notice that the first row is the triangle numbers, so $g(x, 1) = \frac{x+xx}{2}$.

For the cells below the first row, we note that $g(x, y) = g(y+x-1, 1) + (1 - y)$.

$g$ is a bijection from $\mathbb{N} \times \mathbb{N}$ to $\mathbb{N}$.

Thus $h$, as defined below, is a bijection from $A \times A$ to $A$.

$$ h(a, b) = f^{(-1)}(g(f(a), f(b))) $$

case 5/5: $A$ is uncountably infinite.

This case, I can't prove, but I can prove it for some of the uncountable cardinalities.

Let $\varphi$ be a bijection from $\mathbb{R}$ to $\mathbb{R} \cup \{ -\infty, \infty \}$.

Let $a$ be a bijection from $\mathbb{R}$ to $[0,1]$.

$$ a(t) = \frac{\arctan(t) + \frac{\pi}{2}}{\pi} $$

$$ a(-\infty) = 0 $$ $$ a(+\infty) = 1 $$

Let $j$ be a bijection from $[0,1] \times [0,1]$ to $[0,1]$.

Let $j(x, y)$ be given as follows. Write out $x$ and $y$ in base 2.

Using the following table, convert each digit pair one at a time in $x$ and $y$ to base 4.

          x
       ·0  ·1
 y  ·0  0   1
    ·1  2   3

for example

$$ j(.01\cdots, .11\cdots) = .23\cdots $$

We can use this to define a bijection from $\mathbb{R} \times \mathbb{R}$ to $\mathbb{R}$.

$$ \psi(u, v) = \varphi^{(-1)} \circ a^{(-1)} \circ j(a\circ \varphi(u), a \circ \varphi(v)) $$

Let $S_X$ be the set of functions from $X$ to $\mathbb{R}$, then we can define a bijection $p$ from $S_X \times S_X$ to $S_X$ as follows. Let $\lambda \cdots \mathop. \cdots$ denote an anonymous function.

$$ p(a, b) = \lambda x \mathop. \psi(a(x), b(x)) $$

The same arguments works for the set $D$, where $x \in D$ if $x$ is positive or negative infinity or if $\arctan(x)$ is a dyadic rational. In the case of dyadic rationals, after a prefix of finite length all the digits are zero in both base 2 and base 4. Therefore $D$ is countable.

Therefore, we have bijections between $A$ and $A \times A$ for all cardinalities of the form $\left|\mathbb{R}\right|$, $\left|X \to \mathbb{R}\right|$, and $\left|X \to D\right|$.

However, I don't have any reason to believe that all of the uncountable cardinalities are expressible in this way.

Greg Nisbet
  • 11,657
  • This is not my area of expertise, but this seems to be about the same problem: Show that an infinite set $C$ is equipotent to its cartesian product $C\times C$. – Martin R Jul 03 '20 at 04:40
  • You can use $\emptyset$ (\emptyset) to denote the empty set. – K.defaoite Jul 03 '20 at 04:41
  • My understanding is you will need to use the axiom of choice for this. – Q the Platypus Jul 03 '20 at 04:50
  • @MartinR, Asaf Karagila's answer in the other post does answer my question. I am curious though whether the approach I tried to take (defining bijections on $\mathbb{N}$ and $\mathbb{R}$ and trying extend them to functions targeting those sets) is workable or not. I don't know whether this is enough of a difference to make my question a non-duplicate, however. – Greg Nisbet Jul 03 '20 at 04:54
  • @QthePlatypus I am okay with taking the axiom of choice to cover the cardinals I didn't hit in section 5. I'm also curious whether the argument in section 5 works; I think it does, but I'm not certain. – Greg Nisbet Jul 03 '20 at 04:58
  • 1
    @GregoryNisbet This isn't my area either, but it is Asaf's area, so I absolutely believe him when he says that this problem is equivalent to the axiom of choice. What this means is that there will be no constructive proof of this theorem. So, starting from functions constructed from explicit formulas and modifying/extending them to suit your needs, will not get you to a full solution. – user804886 Jul 03 '20 at 05:05
  • @user804886, I'm sorry I think I misspoke. I am not questioning or doubting Asaf's answer at all. I'm just curious if extending a bijection from $B$ to $B \times B$ to a bijection from $X \to B$ to $(X \to B) \times (X \to B)$ is workable ... I think you need the axiom of choice or possibly even stronger axioms to talk about cardinal exponentiation. – Greg Nisbet Jul 03 '20 at 05:14

0 Answers0