Take the 2-minute tour ×
MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

The question is in the title, here is my motivation:

$\require{AMScd}$Let $(\mathcal C,\otimes,I)$ be a monoidal symmetric closed category. Then, the tensor product commutes with colimits, and if $\mathcal C$ has infinite coproducts, the object $N:=\displaystyle\coprod_{n\in\mathbb N}I$ has the following properties :

  • it is a natural numbers object (NNO) in the sense of Lawvere, ie it has natural morphisms $0:I\to N$ and $S:N \to N$ such that any given $I\to X\to X$ uniquely defines a morphism $f:N\to X$ such that the following diagram commutes

$$\begin{CD} I @>0>> N @>S>> N\\ @| @VfVV @VfVV\\ I @>>> X @>>> X \end{CD}$$

  • it is also the free monoid on $I$, ie the initial object in the category of monoids under $I$

If such a coproduct does not exist, a natural numbers object (if it exists) is the free monoid on $I$, but the converse is not necessarily true (or at least, I haven't been able to prove it). To better understand what might happen, It would help to see a workable example of such a category.

Of course for such an example to be interesting, I would like to have a category where an NNO or the free monoid exists.

share|improve this question
Trivial example: the category of finite sets with the cartesian product... –  Zhen Lin Jun 26 '14 at 8:14
Actually, I would like a category where a Natural Number Object could exist. I have edited my question accordingly. –  Cyrille Corpet Jun 26 '14 at 8:23
I don't get your definition of NNO. It looks like incomplete. –  Fernando Muro Jun 26 '14 at 8:59
This definition allows to define sequences in $X$ by recursion: given a first "element" $1\to X$ and a "recursion law" $X\to X$, it defines a sequence $N\to N$. Nicer properties are given by the symmetric closed structure, for instance, addition is given as a map $N\otimes N\to N$ which is defined by adjunction by a map $N\to \operatorname{Hom}(N,N)$, itself defined by the usual recursive definition in $\operatorname{\mathbf{Set}}$. –  Cyrille Corpet Jun 26 '14 at 9:09
Could you add "with NNO" to the title of your question? (A counterexample to the question in your title, but which I think has no NNO anyway, is the category of Banach spaces and continuous linear maps between them.) –  Yemon Choi Jun 26 '14 at 12:33

2 Answers 2

up vote 7 down vote accepted

The effective topos is an example of a (locally!) cartesian closed category with NNO where infinite coproducts do not exist. Indeed, the main feature of the NNO $N$ in the effective topos is that the endomorphisms of $N$ are precisely the computable functions. As such, $N$ cannot be (isomorphic to) $\coprod_{n \in \mathbb{N}} 1$: if that were the case, then there would be uncountably many endomorphisms of $N$. (Of course, $\coprod_{n \in \mathbb{N}} 1$ does not even exist in the effective topos; as you say, if it existed, it would have to be the NNO.)

Here is a closely related construction. Suppose we have a countable model $M$ of Zermelo set theory, or even just Mac Lane set theory. (For instance, by Skolem, we may take $M$ to be a countable elementary substructure of $V_{\omega + \omega}$.) Then the category of sets in $M$ is an elementary topos (so locally cartesian closed) with NNO, but its NNO cannot be $\coprod_{n \in \mathbb{N}} 1$, for the same reason.

share|improve this answer

Another example is a filterquotient topos. E.g. let $\mathcal{U}$ be a nonprincipal ultrafilter on $\mathbb{N}$ and consider the filterquotient of $Set/\mathbb{N}$ by $\mathcal{U}$. This is a world of "nonstandard analysis" which has a NNO (the image of $\mathbb{N}\times\mathbb{N}\to\mathbb{N}$) but not every external infinite sequence of elements is an internal one, so the NNO is not $\coprod_{n\in\mathbb{N}} 1$.

share|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.