7

(I have found examples 1,2 that answer my original questions, so the question here is refined)

This question has been asked many times in this site, but all examples I see are maps between some complicated spaces. So here I'm asking for some examples for good spaces. In fact, it turns out that every continuous function from a path connected space to $\mathbb R$ is a quotient map

Note that the closed map lemma cannot be generalised, for example $(0,1)\to [0,1]$ is not closed. So there might exist continuous surjection from locally compact space to Hausdorff space but not a quotient map, and in the sequel we focus on non-compact manifolds.

Let's first find surjective smooth map between manifolds that is not a quotient map. Such a map must have critical points since every submersion is open(The only obstruction for a smooth surjection not a submersion is that it doesn't have constant rank).

Example 1: For example, $f:(0,2)\to \mathbb S^1,$ where $f(t)=e^{2\pi i\phi(t)}$ where $\phi$ is a smooth bump function that $\phi(0)=0$, increasing in $(0,1)$ and takes value $1$ in $(1,2)$, is not a quotient map, as $(1/2,2)$ is saturated open but its image is not.

If we consider a map onto its image, we also have following example:

Example 2: Consider figure eight $g:(-1,1)\to\mathbb C,g(t)=(1-e^{2\pi i t})\text{sgn} t$. then this map has constant rank, but is not a quotient map onto its image: $(-1,-1/2]$ is closed but its image is not.

By example 2, there is also a smooth map $\mathbb R^n\to \mathbb R^m$ for every $m\geq 2$ of constant rank, viewed as a surjection onto its image, that is not a quotient map. Thus, the result cannot be generalised into arbitary dimensions.

So far, all of these examples have "non-trivial" images. So here I'm asking for nicer maps with very simple image,

(1)Do we have a smooth surjective map $f:\mathbb R^m\to \mathbb R^n$ for every $n>1$ that is not a quotient map.

(2)Do we have a continuous surjective map $f:\mathbb R^m\to \mathbb R^n$ for every $n>1$ that is not a quotient map? For example, is there a spacefilling curve not a quotient map? (Note that we do not always have smooth surjection $\mathbb R^m\to\mathbb R^n$ if $m<n$, but continuous surjection always exists, so (1) doesn't imply (2)).

Eric
  • 1,194
  • 1
    What are your thoughts? What have you tried? Where did you get stuck? Please show us some work! – Ben Steffan Jul 27 '24 at 10:06
  • 1
    There is a smooth surjective map $(0,1)\to S^1$ which is not a quotient map. – Moishe Kohan Jul 27 '24 at 20:59
  • @Moishe Kohan, could you explain what is this map? Thanks – Eric Jul 28 '24 at 18:01
  • 2
    After your question is reopened. – Moishe Kohan Jul 28 '24 at 18:12
  • Are you seriously saying that $[1/2,2)$ is an open subset and that $f(t)$ is smooth in the 1st example? – Moishe Kohan Jul 29 '24 at 13:22
  • That is a mistake, edited – Eric Jul 29 '24 at 13:24
  • 1
    Better but still wrong: Your map in the first example is only continuous, but is not differentiable. Also your 2nd question is unclear. Are you asking if a space filling curve can be a quotient map? (It can be.) Or you are asking for an example of a space-filling curve which is not a quotient map? – Moishe Kohan Jul 29 '24 at 13:28
  • I am asking for an example of space filling curve not a quotient map in the question (2) But I'm still thinking why can it be a quotient map. – Eric Jul 29 '24 at 13:50
  • It is good now. A space-filling curve $\mathbb R\to \mathbb R^n$ can be a proper map, hence, a quotient map. As for an example I had in mind in my first comment, it is $\exp(it^2)$, $t\in (-2\pi, 2\pi)$. – Moishe Kohan Jul 29 '24 at 13:58
  • How do you see the space-filing curve is proper? As for the example I think you mean $e^{(it^2)/2\pi}$ :) – Eric Jul 29 '24 at 14:15

1 Answers1

4

Here is a construction of surjective continuous maps $f: \mathbb R^m\to \mathbb R^n$ which are not quotient maps, for all $n\ge 2, m\ge 1$. Start with a homeomorphism $h_0: (-\infty, 0]\to [-1,1)$; of course, $h(0)=-1$. Set $J:= (-1,1)\times \{0_{n-1}\}$, an open interval in $\mathbb R^n$ and define $A:= \mathbb R^n\setminus J$. Set $$ p:= (-1,0,...,0)\in \mathbb R^n. $$

The main part will be to find a continuous surjection $f_0: [0,\infty)\to A$ such that $f_0(0)=p$. Once it is found, I will take $f_1: \mathbb R\to \mathbb R^n$ as the map which equals $h_0$ on $(-\infty, 0]$ and equals $f_0$ on $[0,\infty)$. Here I regard $[-1,1)$ as an interval in the 1st coordinate axis in $\mathbb R^n$. Such a map will be clearly not a quotient map since $(-\infty, 0]\cup f_0^{-1}(p)$ is a closed saturated subset of $\mathbb R$ whose image under $f$ is not closed (it is the interval $[-1,1)$). To get a map $f: \mathbb R^m\to \mathbb R^n$ I will then use the composition $$ f_1\circ \pi, $$ where $$ \pi: \mathbb R^m\to \mathbb R $$ is the projection to the first coordinate line.

The existence of $f_0$ is a special case of a more general claim:

Theorem. Suppose that $X$ is a nonempty metrizable space which is an increasing union of countably many compacts $K_i, i\in \mathbb N$, each of which is a Peano continuum (connected and locally connected). Then for every $x_0\in X$ there exists a continuous surjective map $$ g: [0,\infty)\to X, g(0)=x_0. $$ Proof. Without loss of generality, we may assume that $x_0\in K_1$. Then, since $K_1$ is a Peano continuum, there exists a continuous surjective map $g_1=[0,1]\to K_1, g_1(0)=x_0$. I will build surjective maps $g_i: [i-1,i]\to K_i$ inductively, so that $g_{i}(i-1)=g_{i-1}(i-1)$. Then take $g: [0,\infty)\to X$ which equals $g_i$ on $[i-1, i], i\in \mathbb N$. qed

To apply this theorem in the case when $X=A$ and $x_0=p$ as above, observe that $A$ is the increasing union of connected and locally connected compacts $$ K_i=\{x\in \mathbb R^n: 1\le x_1^2 + i x_2^2+...+ i x_n^2\le i^2\}. $$

As for a smooth (actually, real-analytic) example of a surjection $(-1,1)\to S^1$ I would take $$ f(t)= e^{2\pi it^2}. $$

For the existence of dimension-raising proper maps $\mathbb R^m\to \mathbb R^n$, see here.

Moishe Kohan
  • 111,854