2

I'm getting hung up on a proof that I remember being fairly easy... Showing that the power set of $\mathbb N$ is uncountable. Supposing it's countable, say $A=\{A_1,...\}$, we choose a set $B$ composed of elements $b_i$, where $b_i\notin A$. How to we guarantee that $B$ isn't the naturals itself, an element of its own power set? Thanks!

Eugene Zhang
  • 17,100
kfriend
  • 889
  • 5
    The usual proof is different, let $f\colon \mathbb{N}\to \mathcal{P}(\mathbb{N})$ be any map. Let $A = { n \in \mathbb{N} : n \notin f(n)}$. Then $A$ is not in the image of $f$, so $f$ is not surjective. – Daniel Fischer Sep 08 '15 at 18:52
  • You'll always get an element of the power set of $\mathbb N$, and you could get $\mathbb N$ itself. The point is that the element you get will not be in the sequence. ${}\qquad{}$ – Michael Hardy Sep 08 '15 at 19:12
  • 1
    @DanielFischer : When you say the usual proof is different, is the difference you have in mind only that it doesn't start with the assumption that it's surjective? ${}\qquad{}$ – Michael Hardy Sep 08 '15 at 19:13
  • @MichaelHardy No, I mean I'm too sleepy to see that it's in fact the same. – Daniel Fischer Sep 08 '15 at 19:15
  • 1
    @DanielFischer : In some contexts it's important to distinguish between the version that is by contradiction and the one that's not done that way. But in such cases, I'd be explicit about that being the difference. ${}\qquad{}$ – Michael Hardy Sep 08 '15 at 19:18

3 Answers3

3

It does not matter if $B$ is empty, $\Bbb N$, or anything. The point is that $B$ is not one of the $A_i$'s.

Just to clarify, $B$ should be defined as $\{i\in\Bbb N\mid i\notin A_i\}$, rather than involving $b_i$'s.

Asaf Karagila
  • 405,794
1

It is better proved in this way.

Enumerate all natural number as $a_1,a_2,\cdots,a_n\cdots$. Suppose $A=\{a_{n_i}|\:i\in\Bbb{N}\}$ be a subset of $a_n$.

Now let $$ r=c_1c_2\cdots c_n\cdots\quad\text{where }c_n=\begin{cases}1, \quad\text{if }a_n\in A\\0,\quad\text{if }a_n\notin A\end{cases} $$ Then it is clear that each $A$ corresponds to a $r$. We prove that $r$ can not be enumerated.

If not, let $r_k=c_{1k}c_{2k}\cdots c_{nk}\cdots$ be an enumeration, where $k\in\Bbb{N},\:c_{nk}\in\{0,1\}$. Let $r'=d_{1}\cdots d_{n}\cdots, \:d_{n}\ne c_{nn}$. Then $\forall k\:r'\ne r_k\:$ for $d_k\ne r_{kk}$, which means $r'$ is not in the enumeration.

The contradiction means $r$ can not be enumerated and must be uncountable, and so is all the number of $A$, or power set of $\Bbb{N}$.

Eugene Zhang
  • 17,100
  • 1
    This seems to really obfuscate the heart of the proof; it takes us from $\mathcal P(\mathbb N)$ to $\mathbb R$ and back for no particular reason (which is actually problematic when we note that binary representations aren't unique and which doesn't really accomplish anything). – Milo Brandt Sep 09 '15 at 00:01
0

First we prove Cantor's theorem $f:A\rightarrow \text{pow}(A)$ is not surjective.\ The approach is show that there exists a set $T\in \text{pow}(A)$ that is not in the range of $f(A)$. In other words, set T in the codomain is not mapped from the domain. For each $a \in A$ its mapping to subset $f(a)$ could either contain $a$ as a set member or doesn't. For example $f:1\rightarrow \{2,9\}$ doesn't but $f:2\rightarrow \{1,2,9\}$. Thus to construct such a T we simply let it be equal $A_f$ which we define to be $\{a\in A_f | a\notin f(a)\}$. Once again, this is well defined subset of A and hence a member of pow(A)

Why? Consider pow({1,2,3}). Say if 1 is not a member of f(1) and 2 is not a member of f(2), surely (1,2) is neither a member of f(1) or f(2).Thus it is obviously not equal to f(1) or f(2). If 3 is a member of f(3), we simply don't include it in T because there is a chance that it equals f(3) given that it contains 3. For example f(3) could be {1,2,3},{1,3},{3} etc. This implies $\forall a \in A, f(a)\neq A_f $ which is exactly what we want.

More precisely, consider the case if $\exists a \in A_f$ where $f(a)=A_f$ implies $a \in f(a)$ which is a contradiction to the definition of $A_f$. Now consider if $\exists a \notin A_f$ where $f(a)=A_f$. However that is impossible as $a \in f(a)$. Therefore $f(a)$ contains $a$ which is not a member of $A_f$. Therefore proposing $f(a)=A_f$ leads to a contradiction. Now let $A$ be $\mathbb{N}$ and we are done