4
$\begingroup$

Suppose that $P\subset\Bbb R^n, n\ge 3$ is a (full-dimensional) convex polytope with a bipartite edge-graph $G=(V_1\cup V_2,E)$ (for example, a zonotope). Suppose further that there are concentric spheres $S_1,S_2\subset\Bbb R^n$ so that all vertices of $V_i$ are on $S_i$.

Question: if $|V_1|=|V_2|$ do we necessarily have $S_1=S_2$ (that is, $P$ is inscribed)?

Note that there are counterexamples if $n=2$:

$\endgroup$
6
  • $\begingroup$ Do you have a counter example for n>=3 and unequal vertex counts? $\endgroup$ Commented Jun 6, 2022 at 16:56
  • $\begingroup$ For $n=3$ there is for example the rhombic dodecahedron. Prisms over this one provide example for $n>3$. $\endgroup$
    – M. Winter
    Commented Jun 6, 2022 at 18:47
  • $\begingroup$ At least for the 3-dimensional polytope, the converse is true: if $P$ is inscribed, then $|V_1|=|V_2|$. $\endgroup$ Commented Jun 7, 2022 at 13:10
  • $\begingroup$ . ...And for higher-dimensional polytopes this also holds. $\endgroup$ Commented Jun 7, 2022 at 13:50
  • $\begingroup$ @Ilya Yes, this follows for example from Steinitz' criterion for non-inscribability: if $P$ is inscribed, then its edge-graph has no independent set containing more than half the vertices. But the $V_i$ are independent sets. $\endgroup$
    – M. Winter
    Commented Jun 7, 2022 at 14:56

1 Answer 1

3
$\begingroup$

After some simplifications, I arrived to the following general fact (it can be generalized even further).

Paint the vertices in black and white according to the bipartite structure. We prove that, if all black vertices lie on a sphere, then one of the white vertices is (non-strictly) outside the corresponding ball, provided that there are at least as many black vertices as white ones.

Suppose the contrary. Let $W$ and $B$ be the convex hulls of white and black vertices, respectively. By bipartiteness, each edge of $B$ is crossed by $W$ (and vice versa). Moreover, there exist edges of $W$ lying on the boundary of $P$.

Take any point $a$ strictly inside $P$ and project all white vertices from $a$ to the sphere to obtain green points; let $G$ be the convex hull of those. Consider the convex hull $S$ of $B\cup G$. Then $S$ contains no black-black edge but contains some green-green edges (those obtained from white-white edges of $W$ on the boundary of $P$). Moreover, all green and black points are vertices of $S$ (as they are conspheric).

All this is now impossible by the following version of Steinitz’s non-inscribability criterion.

Claim. Suppose that $S$ is a (full-dimensional) convex polytope in $\mathbb R^n$ whose vertices are painted in black and green, such that no edge has two black endpoints. Suppose further that there are at least as many black vertices as green ones, and, if their numbers are equal, then there is at least one edge with two green endpoints. Then $S$ is not inscribed in a sphere.

For completeness, here is a sketch of the proof. Assume otherwise. Performing a projective transform of the sphere, we may assume that $S$ contains the center of the sphere in its interior. Now the tangents to the sphere at the vertices of $S$ (a.k.a. their polars w.r.t. the sphere) form the convex polytope $T$ dual to $S$, and $T$ is circumscribed. The facets of $T$ inherit the colors of the corresponding vertices of $S$. No two black facets have a common face of codimension 2, but some green facets do share such face if the numbers of black and green facets are equal.

Now, each face $f$ of codimension 2 is adjacent to two facets; in each of those, consider the solid angle based on $f$ with apex at the tangency point of the facet with the sphere. The two constructed solid angles are congruent. Using the properties mentioned above, we get that the sum of solid angles in green facets is strictly larger than that in black ones. However, the sum of solid angles in any facet is the same (it equals the area of the $(n -2)$-dimensional sphere). This contradicts the condition on the number of black and green facets.

$\endgroup$
4
  • $\begingroup$ There’s still an implicit $n>2$ assumption here, right? $\endgroup$ Commented Jun 9, 2022 at 10:00
  • $\begingroup$ @SamHopkins Right! Otherwise there will be no green-green edge. (The version of Steinitz’s criterion is still vacuously true...) $\endgroup$ Commented Jun 9, 2022 at 10:04
  • $\begingroup$ @IlyaBogdanov Thank you very much Ilya! This sounds like a very nice argument. I will need to read it carefully and I will accept the answer once I reached a good understanding :) Can you give me a taste of the most general interesting statement that this argument applies to? I ask because you said "it can be generalized even further". $\endgroup$
    – M. Winter
    Commented Jun 9, 2022 at 11:40
  • $\begingroup$ I don’t think it is a large generalization, but you can find even a vertex strictly outside the ball, unless all white vertices are on the same sphere. The argument works literally. Also, surely, bipartiteness can be weakened to the same condition as in Steinitz’s. $\endgroup$ Commented Jun 9, 2022 at 11:44

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .