Suppose $X, Y$ are compact sets in $\mathbb{R}^2$ and $F$ is an ambient isotopy carrying $X$ onto $Y$.

Is there an ambient isotopy $F'$ agreeing with $F$ on $X$ and which is constant in a neighborhood of $\infty$?

Basically, to move one compact set to another, you don't need to radically deform the whole space. I was trying to prove it using the Annulus Theorem, and reduced it to this question:

If $J$ is a Jordan curve and $D$ is an open disc containing the trace of $J$ under some isotopy $G$ on $J$, then $G$ can be extended to $\overline{D}$ such that it's fixed on $\partial D$.

In other words, you'll have a Jordan curve sliding around inside a fixed, larger circle and want to extend it to the larger disc in a well-behaved way.

Does anyone know a reference or proof of this? If you can extend it in the annular region, then that's sufficient; you can concoct the extension inside $J$ from that. I've been stuck on it for a few days, now; I know how to prove it when $F$ is smooth, but I don't know how to use that to get the topological version.


  • 1
    $\begingroup$ Is your ambient isotopy a smooth family $[0,1] \times \mathbb{R}^n \to \mathbb{R}^n$? If so, the answer is yes. Look for example at Hirsch's proof of the isotopy extension theorem in his textbook. He gives strong control over the support of the isotopy. $\endgroup$ – Ryan Budney Mar 12 at 2:52
  • 1
    $\begingroup$ Mine won't be smooth, it will just be some isotopy of a Jordan curve in the plane, that I want to extend to an annulus such that it's fixed on the outer boundary component. If it happens in a disc instead of annulus, that's also sufficient; I can just replace what happens inside the Jordan Curve with the original $F$. I think I'll restate the question like that, since it seems more natural. $\endgroup$ – John Samples Mar 12 at 3:05
  • $\begingroup$ There is also a fiber bundle version: We have the closed plane annulus $A$ and a product $A \times I \subset \mathbb{R}^3$ with slices parallel to the $xy$-plane, where the outer cylindrical shell is just the unit circle times $I$ but the inner cylindrical shell just has a continuous family of Jordan curves. Is there an isotopy to a 'standard' annular prism that's slice? $\endgroup$ – John Samples Mar 16 at 3:33
  • $\begingroup$ For the two-dimensional version you can use complex analysis: surely a Riemann mapping for the region bounded by a Jordan curve depends on the curve continuously in whatever sense you need. $\endgroup$ – Tom Goodwillie Mar 17 at 23:06

While this might have been known earlier, one way to derive this result is to apply Corollary 1.2 in

Edwards, R. D.; Kirby, R. C., Deformations of spaces of imbeddings, Ann. Math. (2) 93, 63-88 (1971). ZBL0214.50303.

enter image description here

To apply this result (I will do it in $R^n$), start with an isotopy $F(\cdot ,t), t\in [0,1]$ carrying a compact $C=X\subset R^n$ to $Y$ (the second space will play no role); in their notation, $h_t(x)=F(x,t)$, in particular, $h_0=id_X$. Let $K$ be the image $F(X\times [0,1])$ and let $M$ be a closed $n$-ball $B(0,R)$ whose interior contains $K$. According to Corollary 1.2 above, $F$ extends to an isotopy $H: M\times [0,1]\to M$. It is then easy to extend $H$ to an isotopy of $R^n$ which is the identity outside the ball $B(0,2R)$.

  • $\begingroup$ I'll have to look at this paper, but seems promising! Is $M$ allowed to have boundary? Does their 'isotopy' mean $H$ is onto? If it's constantly a homeo on $M$ then yes we'll have a nice, circular boundary and can explicitly slide the points outside $M$ along the action of $H$ on $\partial(M)$ radially, and then use your gradient trick (on the MSE thread) to get it eventually constant. I've now also been told that it's in researchgate.net/publication/… by Oversteegen but I looked in here already; will have to look again x_x $\endgroup$ – John Samples Mar 18 at 1:14
  • $\begingroup$ Ok, I don't understand the proof yet but this Edwards-Kirby paper contains several theorems which imply the above result. I will post another answer in this thread in a while (weeks?) extracting/translating a direct proof of these results in the plane, which I assume is much easier. I already planned to look at the Oversteegen papers soon, so I will hopefully have some interesting things to say. $\endgroup$ – John Samples Mar 18 at 4:25
  • $\begingroup$ Alright, I've gone through it and everything checks out! Especially, $M$ can have a boundary and they're using "isotopy" in the usual sense for that result. $\endgroup$ – John Samples Apr 14 at 20:40
  • 1
    $\begingroup$ Oh yeah, I should add for future readers that you'll actually want $B(0,R)$ to contain the trace of some neighborhood $U$ of $C$. It always exists, but to directly apply the result it should be mentioned. $\endgroup$ – John Samples Apr 14 at 21:17

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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