MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

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

This is a follow up of an interesting recent question on the topic. The answer given there by fedia shows that the matter is rich and complicated, and I can't resist to submit here a further question.

Q1. What is the closure of the integer valued polynomials in $C^0(\mathbb{R},\mathbb{R})$ the uniform convergence on bounded sets?

For instance, a nontrivial element in the closure is, by the Euler infinite product

$$\frac {\sin(\pi x)} \pi x = \lim_{n \to \infty}\, \binom {x-1} n \binom {x+n} n $$

but can one reach all integers-to-integers continuous functions on $\mathbb{R}$?

(After a quick Google search, among the large literature on integer valued polynomials, I was only able to find this book concerning approximation by IVP, that however treats approximation of functions on the p-adic integers rather than on $\mathbb{R}$; I'm not p-adic enough to understand if there are implications to the real case).

[edit] A further question raised by SJR

Q2. What is the closure of the polynomials with integer coefficients in $C^0(\mathbb{R},\mathbb{R})$ the uniform convergence on bounded sets?

At a first glance I'd have said $\mathbb{Z}[x]$ is closed in $C^0(\mathbb{R}, \mathbb{R})$ for some elementary reason, but now I don't see it. It is certainly a countable, metrizable commutative group; so it is closed iff it is discrete. It has an additional ring structure and it's a monoid by composition, both structure being compatible with the topology.

share|cite|improve this question
The $p$-adic situation is very different since $\mathbb{Z}$ is dense in $\mathbb{Z}_p$. – Felipe Voloch Dec 30 '10 at 1:25
Pietro--I believe it's known that Z[x] is closed in C^0(R,R); see the reference I give in my comment to your answer. – paul Monsky Dec 31 '10 at 2:49
up vote 4 down vote accepted

I'd say that the closure of the integer-valued polynomials (IVP) by uniform convergence on bounded sets is the whole set $C^0\big((\mathbb{Z},\mathbb{R}),(\mathbb{Z},\mathbb{R})\big)$ of continuous functions mapping integers to integers.

Let $r\le s\in\ \mathbb{N}$. Then $\big|\binom x s \big| < 1 $ holds for all $x \in ]0,r[$. Consider the IVP $f_n(x):=\frac 1 2 (x + x^n)$, that maps $]-1,1[$ into itself, and converges to $\frac x 2 $ uniformly on compact sets of $]-1,1[$. So th $k$-fold iterate $f_n^k:=f_n\circ\dots\circ f_n$ converges to $\frac x {2^k}$, and for all $p\in\mathbb{Z}$ the sequence of IVP $pf^k _n\big(\binom x s \big)$ converges to $\frac p {2^k} \binom x s $ as $n\to \infty$ on $]0,r[$. By density of the dyadic rationals, and by linearity, we get at least any polynomial multiple of $\binom x r$ as a uniform limit of IVP on compact sets of $]0,r[$, and of course, any polynomial taking integer values on $x=0,1\dots, r$ as well. By density, we also get any function $f\in C^0(]0,r[)$ that takes integer values on the integer points; in particular the restriction to $]-1,r[$ of any $f\in C^0\big((\mathbb{Z},\mathbb{R}),(\mathbb{Z},\mathbb{R})\big)$. By translation invariance, and since $r$ is arbitrary, this is true for any interval in place of $]0,r[$, and we conclude by a diagonal argument.

share|cite|improve this answer
I see, but let me say that you should think more before posting a question, and avoid these schyzoid dialogues with yourself. – Pietro Majer Dec 30 '10 at 8:40
No, this is very nice - Thanks for being two. What about the closure of $\mathbb{Z}[x]$? – SJR Dec 30 '10 at 9:16
Or maybe Z[x] is closed? – SJR Dec 30 '10 at 11:03
Pietro, SJR--I believe the answer is yes and that a proof is to be found in Le Baron O. Ferguson's book: "Approximation by polynomials with integral coefficients", Math Surveys and monographs vol. 17. He also had a survey article in the Monthly on the topic; perhaps there's a proof there as well. – paul Monsky Dec 31 '10 at 2:14
@Paul: Hi Paul! I wrote a survey of Ferguson's results which you can find at the very bottom of this page (it's in French, sorry). The main results are that if $K$ is a compact subset of $R$ then $Z[X]$ is discrete in $C^0(K,R)$ if the capacity of $K$ is $\geq 1$ and otherwise there is a finite set $J(K) \subset K$ such that a function $f$ is a limit of elements of $Z[X]$ if and only if there exists some $P(X) \in Z[X]$ such that $f(a)=P(a)$ for all $a \in J(K)$. For example if $K=[0,1]$ then $J(K)=\{0,1\}$. – Laurent Berger Dec 31 '10 at 10:25

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.