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

Is this true? For any hyperfinite $n$ that isn't finite, there is a hyperfinite set $A$ such that $\mathbb R \subset A$ and $|A|\le n$ (that's the crucial part, of course)? Intuitively it seems right, but I haven't found a reference and I am not very good at NSA.

share|cite|improve this question
The buzzword here is "enlargement". A nonstandard extension is an enlargement if each standard $A$ is a subset of a hyperfinite subset of ${}^\ast A$. – Kevin O'Bryant Nov 8 '12 at 17:52
Kevin, do you really mean every $A$? That seems impossible without some restriction on the sizes of $A$ to be considered. For example, if $A$ has cardinality larger than $\mathbb{N}^\ast$, then it cannot be covered with any hyperfinite set. – Joel David Hamkins Nov 8 '12 at 18:53
I do mean every $A$, but I was just giving a definition, not a proof of existence! Technically, it goes like this: you fix a superstructure $V$ that contains all of the sets of your interest, say $V$ has cardinality $\kappa$. Then you take a nonstandard extension of $V$ that is $\gamma$-saturated, for some $\gamma>\kappa$, and that extension is an enlargement. This avoids your concern because $\gamma$-saturation forces ${}^\ast {\mathbb N}$ to be sufficiently enormous. – Kevin O'Bryant Nov 9 '12 at 2:57
I see. By fixing the superstructure, you have in effect limited the cardinality of the $A$ that arise, as I suspected. (By the way, I notice that you always but the asterisk to the left, whereas I have also elsewhere seen it on the right---which is best or more correct?) – Joel David Hamkins Nov 9 '12 at 4:16
I learned to put the asterisk on the left, but it shouldn't go so far to the left as I put it. At least, that's how I see it in articles concerning NSA. It's certainly easier to TeX it to the right, though. – Kevin O'Bryant Nov 9 '12 at 4:58
up vote 4 down vote accepted

This depends on how strong your axioms for non-standard analysis are. Certainly any sufficiently saturated model will have an $A$ of the sort you ask about. [Proof: The collection of formulas consisting of "$x$ is a set", "$|x|\leq n$", and "$r\in x$" for all standard reals $r$ is finitely satisfiable. So by saturation there is an $A$ satisfying the whole collection.) But there are various weaker assumptions than $(2^{\aleph_0})^+$-saturation that are sometimes used in NSA and that might not suffice to provide the $A$ that you want.

share|cite|improve this answer
Andreas, do you know a context in which having less saturation is better/important/useful? I agree that in principle one could have less saturation, but I´m wondering if you really meant the "sometimes used in NSA". – Ramiro de la Vega Nov 8 '12 at 15:18
Ramiro, one answer to your question could be that there are advantages to building $\mathbb{R}^\ast$ as an ultrapower by an ultrafilter on $\mathbb{N}$, and these are not supersaturated. – Joel David Hamkins Nov 8 '12 at 16:04
My recollection (perhaps faulty after all these years) is that some of the early work on nonstandard analysis used "richness" conditions that varied from one paper to another. I believe that some of these were essentially saturation weakened to allow only parameters from the standard model, so that I couldn't use the parameter $n$ in my answer. – Andreas Blass Nov 8 '12 at 17:51

Andreas has pointed out that in any sufficiently saturated nonstandard model $\mathbb{R}^\ast$, the answer is yes.

Meanwhile, let me point out that if, as is commonly done, one builds one's hyperreals $\mathbb{R}^\ast$as the ultrapower of $\mathbb{R}$ by an ultrafilter on $\mathbb{N}$, then the answer is no. Indeed, in such a nonstandard $\mathbb{R}^\ast$, there is no hyperfinite cover of the standard reals at all. To see this, suppose that $x$ is a nonstandard hyperfinite set of reals. In the ultrapower, $x$ is represented by a function $f$ from $\mathbb{N}$ to the finite sets of reals, so that $f(k)$ is a finite set of reals. By the Los theorem, the reals $r$ with $r^\ast\in x$ must all have $r\in f(k)$ for almost all $k$. But this is a countable set, so $x$ contains at most countably many reals.

share|cite|improve this answer

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.