Can one construct an embedding of $\mathbb{S}^1$ into $\mathbb{R}^3$ so that every orthogonal projection onto a two dimensional plane is a unit disc?

It is easy to construct an embedding of $\mathbb{S}^1$ into $\mathbb{R}^3$ so that one orthogonal projection is a unit disc: take a Peano-type curve $\gamma=(\gamma_1,\gamma_2):[0,1]\to\mathbb{R}^2$ that fills the unit disc and define $\Gamma(t)=(\gamma_1(t),\gamma_2(t),t):[0,1]\to\mathbb{R}^3$. Clearly $\Gamma$ is one-to-one so it is an embedding and its projection onto the first two coordinates fills the disc. It remains now to "glue" the ends to make it an embedding of $\mathbb{S}^1$.



Assume it is possible, that is, there is an embedding $f\colon\mathbb{S}^1\to \mathbb R^3$ such that any projection of $f(\mathbb{S}^1)$ is a unit disc.

By this answer, the convex hull of the image $f(\mathbb{S}^1)$ is a unit ball. Further note that every extreme point lies in the image; that is, $f(\mathbb{S}^1)\supset\mathbb{S}^2$ --- a contradiction.

P.S. There are embeddings $f\colon\mathbb{S}^1\to \mathbb R^3$ such that any projection of $f(\mathbb{S}^1)$ contains a unit disc. Moreover one can assume that the convex hull $W=\mathop{\rm Conv}f(\mathbb{S}^1)$ is arbitrary close to the unit ball. (Typically, the set of extreme points of $W$ is a Cantor set.) To construct $f$, modify a space filling curve by making it injective, but still intersecting all the lines pass thru the unit ball.

  • 2
    $\begingroup$ Add the link to this question mathoverflow.net/questions/39127/… $\endgroup$ – Taras Banakh Dec 4 '18 at 8:08
  • $\begingroup$ @TarasBanakh Thank you for the link. It greatly complements Anton's answer. $\endgroup$ – Piotr Hajlasz Dec 4 '18 at 12:59
  • $\begingroup$ @PiotrHajlasz You are welcome. $\endgroup$ – Taras Banakh Dec 4 '18 at 13:25
  • 1
    $\begingroup$ This answer is (was) completely impossible to understand if one does not click the link "is a unit ball", I spent 5 minutes puzzled before clicking... I rephrased to emphasize that this is the main point, not just "note that". $\endgroup$ – YCor Dec 4 '18 at 21:14
  • 1
    $\begingroup$ What if we only ask for every projection to contain a disk? i.e. have nonempty interior? $\endgroup$ – Wojowu Dec 4 '18 at 21:29

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.