I want to find a visual proof of the following fact:

For any convex figure in the plane there is a sequence of Steiner's symmetrizations that makes it arbitrary close to a circular disc.

All proofs I know require some integral estimates. I would prefer a more visual proof (even if it is more involved).

  • $\begingroup$ Do you consider Blaschke selection theorem visual? $\endgroup$ – Ivan Izmestiev Aug 1 '18 at 15:07
  • $\begingroup$ @IvanIzmestiev, no I want a direct proof. (I guess you want to say that optimal shape exists and it must be round disc and therefore everything converges to the disc --- no I do not want it.) $\endgroup$ – Anton Petrunin Aug 1 '18 at 15:16

(1) Let $$ C = {\rm conv}\ B\bigcup \{x\}$$ where $B$ is a closed ball of center $o$ and $x$ is not in $B$.

Consider arc $A\subset B$, $$ \partial B \bigcap {\rm Int}\ C $$

Note that by any Steiner symmetrization $S_L$, $A$ is still in the interior ${\rm Int}\ S_L( C)$.

If $x_1$ is close to $x$ s.t. $\overrightarrow{ox}$ and $\overrightarrow{ox_1}$ are different, then $S_L(C)$ where $L=( \overrightarrow{ox_1})^\perp$ has arc $A_1$ which contains $A$. After a finite process, $\partial B-A_n$ is a one point set. Hence through a some Steiner symmetrization, $A_{n+1}=\partial B$. Hence the resulting convex and compact set contains a ball $B'$ whose radius is strictly larger than that of $B$.

(2) Fix a point $o\in \mathbb{E}^2$ which is the origin. Here assume that all hyperplane pass through $o$.

Assume that $C$ is a convex and compact subset. If $x$ is interior point in $C$, then define $v=\overrightarrow{ox}$ so that $o$ is an interior point of $S_L(C)$ for $v^\perp=L$.

In further, if $B$ is a largest closed ball in $C$ of center $o$, then there is an open arc of non-zero length in $ \partial B \bigcap {\rm Int}\ C$.

Hence we have $B'$ in $(1)$. So repeat the process.

If there is no such arc in $\partial B \bigcap {\rm Int}\ C$, then $B=C$.


Just for illustration, here is an animation of one step of a Steiner symmetrization applied to a rasterized version of a polygon.


          First: a histogram of vertical lengths. Second: lengths lifted to be symmetric around horizontal.

  • 1
    $\begingroup$ Thanks for the visualization. I think it would be a nice addition to the Wikipedia article on Steiner symmetrization :) $\endgroup$ – Shamisen Aug 6 '18 at 15:37
  • 1
    $\begingroup$ @Shamisen: Yes, I was thinking of that myself. Requires a block of time to master exactly how to incorporate a new image into a Wikipedia article. On my ToDo list... $\endgroup$ – Joseph O'Rourke Aug 6 '18 at 16:04

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.