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

Let me make the question more precise:

Let A be the set of all points in R^n such that each of its coordinates is rational, and let B=R^n-A. My question is, is there a smooth map f:R^n --> R^{n-1} such that Im(f|_A) \cap Im(f|_B)=\emptyset?

This might be a solved question, but I could not find any reference for it. I would guess no such smooth map exists. Here is some of my thinkings: if Im(f|_A) contains a regular value, say a is such a value, then f^{-1}(a) is a 1-dimensional submanifold of R^n, hence contains some point in B; if Im(f|_A) has no regular value, then by density of A in R^n, Im(f) consists of singular values. By Sard's theorem, Im(f) is a Lebesgue measure 0 (connected) subset of R^{n-1}. As a simple example, if n=1, then in this case, f must be constant. I don't know how to continue in general.

share|cite|improve this question

Let us show that for an arbitrary countable subset $Q=\{x_1,x_2,\dots\}\subset \mathbb R^m$, there is a smooth map $f:\mathbb R^m\to \mathbb R^2$ such that $f(Q)$ is disjoint from $f(\mathbb R^m\backslash Q)$. In the construction, $f$ will be a limit of a sequence of maps $f_n$. The maps $f_n$ will be constructed recursively from $f_{n-1}$ in such a way that $f_n$ is smooth and the $C^n$-norm of $f_n-f_{n-1}$ is arbitrarily small.

Define $f_0(z)=0$. Assume $f_{n-1}$ is constructed. Take a smooth $h_n:\mathbb R^m\to\mathbb R^m$ which maps a neighborhood of $x_n$ to $x_n$ and does not move any $x_i$ for $i < n$. Then $f_n$ is a slight deformation of $f_{n-1}\circ h_n$ in a neighborhood of $x_n$ so that the image of $x_n$ will have a unique preimage. The latter is possible if the image of $f_{n-1}$ is a finite graph and we can construct $f_n$ so that this property survives. So the image of $f_0$ is a point, the image of $f_1$ is an interval, the image of $f_n$ is a tree with $n$ endpoints $f_n(x_1),\dots, f_n(x_n)$.

Clearly the $C^n$-norm of $f_n-f_{n-1}$ can be made arbitrarily small. Therefore one can construct the sequence $f_n$ so that it converges to a smooth $f$. One has to be a bit careful in choosing this sequence so that there are no “overlaps” in the limit, so the limit $f$ has the needed property.

share|cite|improve this answer
Anton, I think you are right from n=1 (by cardinality) and n=2 (by connectedness), but how to prove your statement in general? Thank you. Even that is true, this still does not tell me Im(f|_A) \cap Im(f|_B) = \emptyset. – user16706 Jul 25 '11 at 3:01
Enxin, the statement is a simple cardinality argument. Pick any point $x$ in $\mathbb{R}^n$; then there are only countably many $r$ for which the sphere of radius $r$ and center $x$ intersects $A$ in a nonempty set. So pick any $r$ not in that countable set. – Todd Trimble Jul 25 '11 at 5:21
The $n$ indexing the steps of the induction and the $n$ that is the dimension of the ambient space would like to be different letters. – Andreas Blass Jul 25 '11 at 8:05
@Andreas, now it is fixed. – Anton Petrunin Jul 25 '11 at 8:15
My answer was wrong, but now it is corrected. – Anton Petrunin Jul 25 '11 at 8:16

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.