It's pretty straightforward to say that there is an infinite number of different sizes of infinity, but then I thought, "What size of infinity is that?"

My thoughts are that the number of unique cardinalities is equivalent to the number of real numbers, based on the fact that the cardinalities can always be ordered by increasing size. I don't really know how to prove this, though. It's mostly based on intuition, which isn't very reliable when talking about uncountably infinite sets.

I originally asked a somewhat related question at a different (and not math-oriented) forum, and the users there told me that it is not possible to talk about the number of cardinalities without talking about the set of all sets, which forms a paradox. If a set were to contain all of the different sizes of infinity, it would have to contain its own power set, which isn't possible.

However, I'm not completely convinced that it is not possible to talk about a set of all of the cardinalities. Sure, a cardinality represents a size of infinity, but I think that it should be possible to have a set of the cardinalities without having the set actually contain the various infinities. Would this avoid the above paradox?

So, is it possible to measure the number of different sizes of infinity, and what would that size be?


We don't really talk about "infinities", instead we talk about "cardinalities". Cardinality of the a set is the mathematical way of saying how large it is. Of course infinity could easily just mean $\infty$ which is a formal symbol representing a point larger than all real numbers (but the notion can be transferred to other contexts as well). This is not the same sort of infinity as infinite cardinalities. Infinite cardinalities are a whole other beast, and they are related to set theory (as we measure the size of sets, not the length of an interval).

Cantor's theorem tells us that given a set there is always a set whose cardinality is larger. In particular given a set, its power set has a strictly larger cardinality. This means that there is no maximal size of infinity.

But this is not enough, right? There is no maximal natural numbers either, but there is only a "small amount" of those. As the many paradoxes tell us, the collection of all sets is not a set. It is a proper class, which is a fancy (and correct) way of saying that it is a collection which is too big to be a set, but we can still decide whether or not something is in that collection.

In a similar fashion we can show that the collection of all cardinalities is not a set either. If $X$ is a set of sets, $\bigcup X=\{y\mid\exists x\in X. y\in x\}$ is also a set, and its cardinality is not smaller than that of any $x\in X$. By Cantor's theorem we have that the power set of $\bigcup X$ has an even larger cardinality.

What the above paragraph show is that given a set of cardinals, we can always find a cardinal which is not only not in that set, but also larger than all of those in that set. Therefore the collection of possible cardinalities is not a set.

  • $\begingroup$ Ok, Asaf, here are some test questions to make this formal: Given a set $X$, are there $X$-many different cardinalities? If not, are there $Y$-many cardinalities for some $Y$ that surjects onto $X$? $\endgroup$ Jan 21 '13 at 17:46
  • 2
    $\begingroup$ @user1729: Actually, there is a Gimel function too. But you can use $\daleth$... :-) $\endgroup$
    – Asaf Karagila
    Aug 20 '14 at 10:48
  • 2
    $\begingroup$ @Andrey: Your comment is very unclear. What does it mean "a set of numbers"? Can you find a natural number which is not in the set of all the natural numbers? Can you find a real number which is not in the set of all the real numbers? What you can do however, is show that for every finite set of natural/rational/real/complex numbers, there is a number which does not lie in that set. Therefore, we can conclude, the set of all natural/rational/real/complex numbers is not finite. $\endgroup$
    – Asaf Karagila
    Apr 7 '15 at 21:54
  • 2
    $\begingroup$ @Andrey: No, it very much does not work with infinite sets. The sum of all the natural numbers is far from a natural number (and it's not $-\frac1{12}$ either!). Sets to classes are like finite subsets of $\Bbb N$ to infinite subsets of $\Bbb N$. $\endgroup$
    – Asaf Karagila
    Apr 9 '15 at 22:31
  • 2
    $\begingroup$ @Andrey: Yes, something like that. Classes are collections of sets which do not have cardinality (because, at least in $\sf ZFC$, having a cardinality means being in bijection with a cardinal which is a set). $\endgroup$
    – Asaf Karagila
    Apr 9 '15 at 22:45

If $A$ is a set, then the power set $P(A)$ is a set of bigger cardinality. If $\{A_i\}_{i\in I}$ is a family of sets, then $P(\bigcup_{i\in I}A_i)$ is a set of bigger cardinality than any of the $A_i$. This allows us to define an infinite set $F(a)$ for each ordinal $a$ such that $a<b$ implies that $F(a)$ has smaller cardinality than $F(b)$. To do so, let

  • $F(\emptyset)=\mathbb N$,
  • $F(a)=P(F(b))$ if $a=b+1$ is the successor of $b$,
  • $F(a)=P(\bigcup_{b<a} F(b))$ if $a$ is a limit ordinal

Now if a set $S$ were able to enumerate all infinite cardinalities, this would give us an injective map from the proper class of all ordinals into this set, which is absurd.


As the other answers have pointed out (at least within the framework of ZFC), the answer is "proper class many." Let me just point out that this isn't the end of the story.

We can ask whether two proper classes have the same size, by asking about the existence of a definable bijection, etc. In this sense the class of ordinals - or equivalently of cardinalities of well-orderable sets (=initial ordinals) - is actually small: every other proper class surjects onto it! Given a proper class $C$, consider the map $rk: C\rightarrow ON$ sending a set $x\in C$ to the unique $\alpha$ such that $x\in V_{\alpha+1}-V_\alpha$ - that is, its (von Neumann) rank. This is well-defined, and maps $C$ to a cofinal subclass $S$ of $ON$. Now we can "collapse" $S$ onto $ON$ by sending $\alpha\in S$ to $ot(\{\beta\in S: \beta<\alpha\})$; this is surjective. Composing these two maps gives a surjection from $C$ to $ON$.

Meanwhile we can construct models in which there is a proper class $C$ with no surjection from $ON$ (let alone an injection into $ON$): see Joel David Hamkins' answer to https://mathoverflow.net/questions/110799/does-zfc-prove-the-universe-is-linearly-orderable.

Finally, we could ask: can there be a proper class $C$ such that $ON$ does not inject into $C$? This is a subtler question, and class injections are weird things; but I believe the answer is yes via class forcing (I vaguely recall seeing this a long time ago, and it being relatively simple, but I can't remember the details).

The distinction between injection and surjection above might make you think, "But wait a minute! Don't we have the axiom of choice to simplify things?" Indeed we do, but the axiom of choice treats only sets, and at the level of classes "injects into" and "is surjected onto" are still distinct, a priori. We can have an "axiom of choice for classes" (called global choice), if we enlarge our language a bit to talk about classes, and this axiom implies that all proper classes have the same size.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.