The surreals are described as a game where finitely-many numbers are generated each day. In the limit, things like the non-dyadic rationals, the reals, and even hyperreals like $\omega$ and $\epsilon$ can be defined.

However, the way I've usually seen this defined is something like

on day $\omega$ we can define $\omega = \{\ 1, 2, 3, \ldots\ |\ \}$

and then they go on to talk about what numbers are generated on day $\omega + 1$, etc.

My question is, how is this valid? What axiom allows us to talk about "what happens after infinite days"? Shouldn't the number of days be countable?

  • 1
    $\begingroup$ en.wikipedia.org/wiki/… $\endgroup$ – saulspatz May 1 '20 at 11:31
  • $\begingroup$ This is a number of questions in one, and it's hard to know which one(s) you're asking to be addressed. Do you care about "What literal axiom(s) of set theory allow us to form the Surreal $\omega$?", "What is the intuition behind the order implied by the use of the word 'after' when discussing infinite birthdays of surreal numbers?", "If you add one element to a countably infinite set, doesn't it become uncountable?", "How are surreal birthdays defined [given a particular definition of the Surreals I found at place X]?", etc. $\endgroup$ – Mark S. May 1 '20 at 12:41
  • $\begingroup$ I'm not sure how you got those last two questions from what I wrote here. $\endgroup$ – BlueRaja - Danny Pflughoeft May 1 '20 at 13:01
  • $\begingroup$ "Shouldn't the number of days be countable?" made it sound like you thought a birthday of $\omega+1$ forced something to be uncountable. And you seemed to be asking in part "how is birthday $\omega+1$ valid?" One answer to that is basically "by definition of birthday". But that epends on your choice of definitions for surreal and birthday, but for each definition of surreal there is a most natural definition of birthday. As an aside, tag me with @Mark if you want me to be notified of any future comments. $\endgroup$ – Mark S. May 1 '20 at 13:28

What axiom allows us to talk about "what happens after infinite days"?

The axiom schema of replacement "asserts that the image of any set under any definable mapping is also a set."

I really like Vsauce's How To Count Past Infinity (see Axiom of Replacement at 15:05) which shows how "we can keep our climb going forever."

Shouldn't the number of days be countable?

I'm not eniterly sure what you're referring to here. However, $\omega$ is a "relatively small" countable ordinal. For more information on large countable ordinals, check out Wikipedia's page on large countable ordinals or the excellent 3 part series by John Baez.

  • $\begingroup$ I don't see how "large" countable ordinals are relevant in a discussion of ordinals like $\omega+1$, but I agree that at least for most surreals the axiom schema of replacement would be needed. $\endgroup$ – Mark S. May 12 '20 at 0:52
  • $\begingroup$ @MarkS. I just thought it might be of interest to the OP. $\endgroup$ – user820789 May 12 '20 at 1:08

Intuition/informal argument:

how is this valid?

Let's start by reviewing the finite days.

On day $0$, we take no numbers, put them in the left and right sets, and have $0:=\{ \,\mid\,\}$.

On day $1$, we take that number, put it in the left or right set ($\{ 0\mid0\}$ would break the inequality numbers are supposed to satisfy), and get $1:=\{ 0\mid\,\}$ and $-1:=\{ \,\mid0\}$.

On day $2$, we take the numbers we have so far, and put some of them in a pair sets in all legal ways that obey the inequality numbers are supposed to satisfy (so $\{ 1\mid-1\}$ is not allowed), and get new numbers like $\frac{1}{2}:=\{ 0\mid1\} =\{ -1,0\mid1\}$ , and things equal to old numbers like $\{ -1\mid1\} =0$.

In general, on day $n$ for finite $n$, we take all the numbers so far, and put them in sets in all legal ways.

Now, just call $F$ the set of all the numbers we could get on some finite day. Then we can take all the numbers in $F$, and put them in sets in legal ways, getting new numbers like $\{ 0,1,2,\ldots\mid\,\}$ and $\left\{ 0\mid1,\frac{1}{2},\frac{1}{4},\dots\right\}$ .

We can then do the "next day" step again, to build new numbers like $\left\{ \{ 0,1,2,\ldots\mid\,\} \mid\,\right\}$ .

For historical reasons dating back to Cantor, it's conventional to use the Greek letter $\omega$ to denote both the step we did with $F$ ("day $\omega$") and the surreal number $\{ 0,1,2,\ldots\mid\,\}$ .

Axiomatic/Formal arguments

What axiom allows us to talk about "what happens after infinite days"?

This depends very precisely on what you meant. Throughout, I'll assume you want to deal with standard ZF. Different books have slightly different statements of axioms, but I'll sketch the main points.

Building an Infinite Set

If you meant something like "how can we build a set like $\{ 0,1,2,\ldots\}$?", then the main axiom is the axiom of infinity, which basically declares the existence of a superset of $\{\emptyset,\{\emptyset\},\{\emptyset,\{\emptyset\}\},\{\emptyset,\{\emptyset\},\{\emptyset,\{\emptyset\}\}\}\ldots\}$. Then, as Wikipedia sketches, we can then use things like (restricted) comprehension to get down to exactly that set. Under von Neumann's convention, the elements of that set would be labeled $0,1,2,\dots$, so that we would have a set $\{0,1,2,\ldots\}$. This set is the ordinal (not the surreal number) known as $\omega$.

The Day After

If you meant something like "how can we have a set to represent 'day $\omega+1$'?", then we can build on the previous construction. Once we have $\omega$, then we can use pairing to form $\{\omega,\omega\}=\{\omega\}$. Then we can use it again to form $\{\omega,\{\omega\}\}$. Finally, we can use union to make (at least a set containing) $\bigcup\{\omega,\{\omega\}\}=\{\omega,0,1,2,\ldots\}$. Traditionally, and generalized by ordinal addition, this set is called the ordinal (not the surreal number) $\omega+1$.

Building Surreals

But to actually and truly build a surreal like $\{0,1,2,\ldots\mid\,\}$, we need something that lets us write down that left set of surreals. We want to write something like $\{n^\text{th}\text{ nonnegative surreal with empty right set}\mid n\in\mathbb{N}\}$. We can use the ordinal $\omega$ in place of $\mathbb N$, and write down a complicated logical formula for $n^\text{th}\text{ nonnegative surreal with empty right set}$. But to actually form this set, we need something special, like a way to legally replace each $n$ in the ordinal $\omega$ (a set that exists by the axiom of infinity) with the corresponding surreal. This almost certainly requires replacement (and if it doesn't for some technical reason, I'm certain it'd require it by the time you get to $\omega+\omega$ for set theory reasons).


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.