8
$\begingroup$

Let $P$ be a simple convex polytope in $\mathbb{R}^d$ (that is, any vertex belongs to exactly $d+1$ facets). Given the collection of outer normals to facets of $P$, combinatorics of $P$ may be different, of course. But is this information enough to reconstruct number of faces of $P$ of all dimensions? If yes, what is specific procedure to do this? I am looking at what happened when we move facets parallel and pass through `singularity', i.e. not-simple polytope, and on first glance it looks that $f$-vector preserves, but I am always not sure with such things. And this is not good proof anyway, I would prefer more direct argument.

If the statement is however false, I wonder for which specific collections it is still true.

$\endgroup$
  • 2
    $\begingroup$ I presume the statement you wish to investigate is this: Every simple polytope with the same set of facet normals has the same $f$-vector. $\endgroup$ – Joseph O'Rourke Oct 17 '15 at 19:03
  • $\begingroup$ @Joseph Yes, exactly. Sorry if it was not clear. $\endgroup$ – Fedor Petrov Oct 17 '15 at 19:26
8
$\begingroup$

Consider a simplex $S$ in five dimensions, and one more face normal $\pi$ in general position with respect to $S$ (no hyperplane perpendicular to $\pi$ passes through more than one vertex of $S$). Intersect $S$ with a halfspace $H$ perpendicular to $\pi$, and translate $H$ continuously towards $S$, starting from a position where it contains all of $S$.

Then when the intersection with $H$ first cuts off one vertex of $S$, the one cut-off vertex is replaced by five new vertices (on the edges of $S$ connecting the cut-off vertex to the others) so the total number of vertices becomes $6-1+5=10$. When the intersection with $H$ cuts off two vertices of $S$, they are replaced by eight new vertices (again, one for each edge of $S$ that connects one of the two cut-off vertices to the four remaining ones), so the total number of vertices becomes $6-2+8=12$. And when the intersection with $H$ cuts off three vertices of $S$, they are replaced by nine new vertices, so the total number of vertices becomes again $6-3+9=12$. All of these are simple $5$-polytopes.

So no, the $f$-vector is not preserved. Not even the number of vertices is preserved.

$\endgroup$
  • $\begingroup$ David, did you use $5$ dimensions because it is not so clear in dimension $4$? As Fedor observed, the statement is true in $\mathbb{R}^3$, and you have given an $\mathbb{R}^5$ counterexample. That leaves $\mathbb{R}^4$...? $\endgroup$ – Joseph O'Rourke Oct 18 '15 at 21:04
  • 1
    $\begingroup$ It is also false in $\mathbb{R}^4$ but the counterexamples are a little more complicated. E.g. consider a hypercube and a face normal parallel to one of its long diagonals. Cutting off one vertex leaves $16-1+4=19$ vertices. Cutting off five vertices leaves $16-5+12=23$ vertices. $\endgroup$ – David Eppstein Oct 18 '15 at 23:03
5
$\begingroup$

Not an answer; just an illustration:


          fvector
          $f$-vector left & right: $(V,E,F)=(6,9,5)$. Middle polyhedron is not simple.


$\endgroup$
  • 4
    $\begingroup$ Well, thank you, but warning: for dimension 3 the answer is positive just by Euler formula (and number of facets is enough, not even their directions), so it may be somehow misleading. $\endgroup$ – Fedor Petrov Oct 17 '15 at 16:07

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.