1
$\begingroup$

Let $X,Y$ be compact, connected, simply-connected, and separable, metric spaces each with at-least $2$-points, and let $f,g:X\rightarrow Y$ be continuous functions. Does there always exist a homeomorphism $\Phi:X\times Y \rightarrow X\times Y$ such that $$ g(x) =\pi_Y\circ \Phi(x,f(x)) $$ for all $x \in X$, where $\pi_Y$ is the canonical projection of $X$ onto $Y$?

Heuristically (but not exactly): "Can we always perturb the graph of a continuous function so that it becomes the graph of another function?"

$\endgroup$

1 Answer 1

1
$\begingroup$

No. Take $I=[0,1]$, $Q=[0,1]\times[0,1]$ and $J=\{0\}\times[1,2]$. Let $X=I$ and $Y=Q\cup J$. Let $f:X\to Y$ be the constant function $f(x)=(0,\frac32)$ and let $g$ be the constant function $g(x)=(\frac12,\frac12)$. Then any homeomorphism $\Phi$ of $X\times Y$ will preserve the square $I\times J$, so $\pi_Y(\Phi(x,f(x)))\in J$ and therefore cannot be equal to $g(x)$.

Added a bit later: The answer is negative even if you additionally assume that the two spaces are homogeneous. Take $X=Y=S^2$. Let $f:X\to Y$ be the identity map of $S^2$ and let $g:X\to Y$ be a constant map taking the value $c\in S^2$. Then the requirement $g(x)=\pi_Y(\Phi(x,f(x)))$ becomes $\pi_Y(\Phi(x,x))=c$. In other words, $\Phi(x,x)=(h(x),c)$ where $h:S^2\to S^2$ is a homeomorphism. But $(S^2\times S^2)\setminus\Delta$ (here $\Delta$ is the diagonal) and $(S^2\times S^2)\setminus (S^2\times\{c\})$ are not homeomorphic, so there is no such $\Phi$.

$\endgroup$
5
  • $\begingroup$ What if $X=Y=\mathbb{R}^n$? Would things workout better? $\endgroup$
    – ABIM
    Aug 18, 2020 at 11:12
  • 1
    $\begingroup$ Yes, then you can take $\Phi(x,y)=(x,y-f(x)+g(x))$. $\endgroup$
    – Dejan Govc
    Aug 18, 2020 at 15:59
  • 1
    $\begingroup$ You could also e.g. replace $\mathbb R^n$ by a Lie group in which case $\Phi(x,y)=(x,yf(x)^{-1}g(x))$ would do the job. $\endgroup$
    – Dejan Govc
    Aug 18, 2020 at 16:08
  • $\begingroup$ Do we really need continuity of both $f$ and $g$ in this case.. I guess not? $\endgroup$
    – ABIM
    Aug 19, 2020 at 14:49
  • 1
    $\begingroup$ @James_T: Well, mainly to ensure $\Phi$ is continuous. But if the discontinuities of $f$ and $g$ "cancel out", this still works, I guess. $\endgroup$
    – Dejan Govc
    Aug 19, 2020 at 16:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.