I'm having some trouble with the following problem from Using the Borsuk-Ulam Theorem.

Let $P = \mathrm{conv}\{e_1, -e_1, \dots, e_d, -e_d\}$ be a crosspolytope. Show that if a subset $F \subset \{e_1, -e_1, \dots, e_d, -e_d\}$ forms the vertex set of a proper face of the crosspolytope then there is no $i \in \{0, 1, \dots , d\}$ with both $e_i, -e_i \in F$.

The definition of face that I am working with is as follows. A face of a convex polytope $P$ is either $P$ itself or an intersection $P \cap h$ where $h$ is a hyperplane that does not dissect $P$.

As a hint the book suggests considering $h = \{x: \left<a_F, x\right> = 1\}$ for $a_F = \Sigma_{v\in F}v$ as a defining hyperplane.

So let's proceed by way of contradiction assuming that $F$ contains an $e_i$ and $-e_i$ for some $i$. For a hyperplane $h$ as defined above I'm guessing I want to show that $P$ intersects both of the open halfspaces $h^+ = \{x: \left<a_F, x\right> > 1\}$ and $h^- = \{x: \left<a_F, x\right> < 1\}$. Since $e_i - e_i = \vec{0}$ and $a_F$ is just a sum of standard basis vectors there will be a zero in the $i$th component of $a_F$. This means $<a_F, \pm e_i> = 0$ so $\pm e_i \in h^-$.

This is the point in the proof where I get stuck. My question is twofold:

  1. How do we know that $h$ as defined above is a defining hyperplane? Why was it chosen?

  2. How should I proceed at trying to find an element of $P$ in $h^+$? An element in $h^+$ is not immediately obvious to me.


If $F$ contains both $e_i$ and $-e_i$ it contains their midpoint $0$, since each face is a convex set. But $0$ is an interior point of $P$, so the only face of $P$ which contains $0$ is $P$ itself.

To see this using the definition in your book, the hyperplanes through $0$ have the equation $a_1x_1+\cdots+a_dx_d=0$. If $a_j\ne0$ then $e_j$ and $-e_j$ are on opposite sides of the hyperplane, so it does dissect $P$ non-trivially.

  • $\begingroup$ Why are we allowed to look at the hyperplanes through 0? How do we know there is not another hyperplane through $F$ where $e_j$ and $-e_j$ are not on opposite sides? $\endgroup$ – EgoKilla May 9 '17 at 3:57
  • 1
    $\begingroup$ The point is that an interior point of a polytope is not in any of its proper faces. You are working from an eccentric definition of a face, that talks about hyperplanes "dissecting" the polytope. The usual definition of a proper face is that it is the intersection of the polytope with a supporting hyperplane. No supporting hyperplane can pass through the interior point of a hyperplane. $\endgroup$ – Lord Shark the Unknown May 9 '17 at 6:31

you also could see that by mere induction. For d=2 you have a square. Its sides obviously do not go through the center (origin). Any higher cross-polytope then is a bipyramid of the former one. Thus again no boundary facet will contain the origin.

--- rk


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.