MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

In the first pages of SGA4 I read

[...] Cependant le seul univers connu est l'ensemble des symboles du type {Ø,{Ø},{Ø,{Ø}}, ... } etc. (tous les éléments de cet univers sont des ensembles finis et cet univers est dénombrable). En particulier, on ne connaît pas d'univers qui contienne un élément de cardinal infini. [...]

(the sole known universe is like {Ø,{Ø},{Ø,{Ø}}, ... }, and we don't know any universe with a infinite cardinal).

Mais, c'est vrai? I wonder if during all these years somebody discovered a universe "bigger" than that exhibited by Grothendieck.

share|cite|improve this question
I bring Grothendieck into this because the quotation I wrote comes from SGA... I'm only asking if, from 1963 to now, someone found out a universe different from {Ø,{Ø},{Ø,{Ø}}, ... }. – Fosco Loregian Jul 31 '10 at 10:02
The question looks very interesting to me. I don't see Ryan Budney's point. – Pierre-Yves Gaillard Jul 31 '10 at 10:24
This question is very vague. What do you mean by "discover"? I think it is likely that you're taking Grothendieck's quotation a bit too literally. Sure there are other universes, for example a model of ZFC. Or the initial topos. Or the effective topos. OR a model of ZFC + measurable cardinals. And there are permutation models. And so on. – Andrej Bauer Jul 31 '10 at 10:26
You should understand Grothendieck as saying "I am not sure infinite sets actually exist". This of course is a matter of opinion, but most mathematicians don't have a problem with the existence of the set of natural numbers. – Andrej Bauer Jul 31 '10 at 10:32
I don't understand where all these querulous comments are coming from. Like PYG, I think the question is perfectly clear. And indeed, it seems to be understood and answered correctly below (modulo some trivial quibbling about the empty set). But also the sentence "I assume you are referring to Grothendieck universes" in the answer seems to suggest that there is some doubt in the matter -- but the OP refers to a specific passage from SGAIV and this pasage is talking about [what are now called] Grothendieck universes. So what's the problem? – Pete L. Clark Jul 31 '10 at 11:45
up vote 19 down vote accepted

The universe that Grothendieck intends to suggest by his notation is known in set theory as HF, the class of hereditarily finite sets, the sets that are finite and have all elements finite and elements-of-elements, and so on (the transitive closure should be finite). The set HF is the same as $V_\omega$ in the Levy hiearchy, and can be built by starting with the emptyset and iteratively computing the power set, collecting everything together that is produced at any finite stage. This is the smallest nonempty transitive set that is closed under power set. It satisfies all the Grothendieck universe axioms, except that it doesn't have any infinite elements, since none appear at any finite stage of this consrtruction.

There is an interesting presentation of this universe by a simple relation on the natural numbers. Namely, define $n\ E\ m$ if the $n^{\rm th}$ bit in the binary expansion of $m$ is $1$. The structure $\langle\mathbb{N},E\rangle$ is isomorphic to $\langle HF,{\in}\rangle$ by the map $\pi(n)=\{\pi(m)\,|\,m\,E\,n\}$, which set-theorists will recognize as the Mostowski collapse of $E$.

Since HF doesn't have any infinite elements, it is a rather impoverished universe for many applications of that concept. And so we naturally seek larger universes. But the difficulty is that we cannot prove they exist. The difficulty is not one of "discovery," but rather just that we can prove that the hypothesis of the existence of a univese containing infinite sets is too strong for us to prove from our usual axioms. The reason is, as has been remarked in some of the other answers and comments, all other Grothendieck universes have the form $H_\kappa$, the hereditarily size less than $\kappa$ sets, for an inaccessible cardinal $\kappa$. So this is just like HF, which is $H_\omega$, but on a higher level, and in this sense, these higher universes are not so mysterious. They are intensely studied in set theory, a part of the research effort in large cardinals.

In this MO answer, I mention a number of weaker universe concepts that we can prove exist, and which I believe serve most of the uses of the universe concept in category theory, if one wanted to care more about such set theoretic issues.

share|cite|improve this answer

Let me rephrase part of what Joel David Hamkins and Anon already said, but without mentioning inaccessible cardinals:

A Grothendieck universe strictly bigger than the one in the question would be a model of ZFC. (More precisely, it would become a model once we interpret the membership symbol of ZFC as actual membership.) So the existence of such a Grothendieck universe would imply the consistency of ZFC. G"oedel's second incompleteness theorem implies that ZFC cannot prove the consistency of ZFC. Therefore ZFC cannot prove the existence of a Grothendieck universe.

share|cite|improve this answer
Clear and concise, I understood what was the "problem". Thanks. – Fosco Loregian Jul 31 '10 at 22:17

I assume you are referring to Grothendieck universes.

The existence of a bigger Grothendieck universe is equivalent to the existence of an inaccessible cardinal, which cannot be proved from ZFC, because it implies the consistency of ZFC.

There is a smaller example of a Grothendieck universe: the empty set. This is the only other Grothendieck universe that can be proven to exist in ZFC.

share|cite|improve this answer
A universe is nonempty by definition. – Pierre-Yves Gaillard Jul 31 '10 at 11:00
Thank you. I think it should be noticed that in SGA4 an universe is defined nonempty, but in the sequent Appendix (redige' par N. Bourbaki, pages 185--...) the empty set is accepted as a universe... It seems like a problem, isn't it?! – Fosco Loregian Jul 31 '10 at 12:32
Dear tetrapharmakon: I just looked at the article you took the quotation from. I'd have been surprised if Grothendieck and Verdier had overlooked the empty set. I don't know why Grothendieck and Verdier excluded the empty set, nor why Bourbaki included it. – Pierre-Yves Gaillard Jul 31 '10 at 12:50
Thanks for your answer. I don't know why Grothendieck and Verdier excluded the empty set, nor why Bourbaki included it. Excuse me but I can't understand what is your point: what "philosophical" position shall I adopt about the empty set? Must I follow Grothendieck&Verdier or Bourbaki? – Fosco Loregian Jul 31 '10 at 17:56
Dear tetrapharmakon: Is it better to include the non-emptiness condition into the definition of a universe, or not to include it? I don't have the slightest idea. I don't think it's an important question. – Pierre-Yves Gaillard Jul 31 '10 at 19:27

Just to make perfectly sure: Grothendieck is absolutely not questioning the existence of infinite sets in this quotation. (He had, and has, some eccentricities, but not in this direction!)

Remember that "universe" is a technical term for a certain type of set, essentially one which has maximally nice closure properties. The universe he is talking about corresponds to the cardinal $\aleph_0$, a countably infinite set whose elements may themselves be identified with the finite cardinals (as is a standard operating procedure since von Neumann, although those who don't think that much about infinite sets can and often do safely forget this point). He is not discussing universes as models of formal set theory or anything like that, so the idea that "internally" in this countably infinite universe, infinite sets do not exist, is not at all what he is getting at. Rather, since he has written down an example of an infinite set, we can conclude (from this passage alone, notwithstanding the rest of his work) that he believes in and is comfortable working with infinite sets.

share|cite|improve this answer

There exist plenty of other universe. Recall that most of the times one proves the (relative) consistency of some axiom independent from ZF, one actually builds a model that satisfy that axiom. So for instance, the method of forcing invented by Cohen enables to list a infinite number of (elementary) different universes. Of course the universe you mentioned is much more tangible and intuitive then the one built with forcing, but if you accept AC, then they have the same dignity.

share|cite|improve this answer
Once again, I believe you are confusing the notion of universe studied in set theory with the (strictly narrower) notion of universe Grothendieck is discussing here, since called Grothendieck universes. See and – Pete L. Clark Jul 31 '10 at 13:13

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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