1
$\begingroup$

I posted this question thinking that the response would be two or three answers that say "Counterexamples to this are found in every textbook—for example this one and this one and this one." But the result is that after two days there are 13 views and nobody has said anything except in a private email with a comment that probably misses the mark.

Two chords of a circle of unit radius have equal lengths if their corresponding arcs have equal lengths.

Suppose a polytope is the convex hull of finitely many points on the unit $(n-1)$-sphere in $\mathbb R^n.$ Each of its $(n-1)$-dimensional facets casts a shadow on the sphere, the light-source being at the center. Suppose a proposition says two such facets have equal $(n-1)$-dimensional volumes if their shadows have equal volumes. I suspect that that is false in general, even if both facets have equally many vertices. (Unlike with chords of circles, their shapes can differ—for example not all triangles are equilateral.)

Is a proof or a counterexample known?

$\endgroup$
7
  • $\begingroup$ Counterexample, I think: Construct two tetrahedra with vertices on $S^2$ as follows: both have a vertex at the south pole. The first has three vertices equally spaced in a horizontal circle on $S^2$ just above the equator. The second has three vertices arranged in a slim triangle containing 0, lifted above the equator. The shadow of the top face of each tetrahedron is nearly the entire upper hemisphere. If the heights of the two horizontal circles are chosen correctly, these areas will be equal (by continuity: when the horizontal circle approaches the equator, the shadow is the hemisphere.) $\endgroup$ Commented Jul 9, 2021 at 20:16
  • $\begingroup$ @GevaYashfe : What do you mean by "0"? $\endgroup$ Commented Jul 9, 2021 at 20:42
  • $\begingroup$ Sorry, I unsuccessfully tried to cram the idea into one comment. $0$ meant the origin. More explicitly, I mean that this triangle also lies on a horizontal circle on $S^2$ just above the origin, say at height $\epsilon$ (not the same height as the triangle of the first tetrahedron). The convex hull of its vertices contains the point $(0,0,\epsilon)$. In other words, it is a "slim triangle containing the origin" which has been lifted a little above the equator. $\endgroup$ Commented Jul 9, 2021 at 20:49
  • $\begingroup$ So the first one has a vertex at $90^\circ$ south and three other vertices: one at $1^\circ$ north $0^\circ$ longitude, one at $1^\circ$ north $120^\circ$ west, and one at $1^\circ$ north $120^\circ$ east. And the shadow of one face is nearly the whole northern hemisphere. But I don't understand where the other tetrahedron is. Its vertices should be on $S^2. \qquad$ $\endgroup$ Commented Jul 9, 2021 at 20:56
  • 2
    $\begingroup$ The other tetrahedron has a vertex at $90^\circ$ south and three other vertices, all are at $\epsilon$ north and they have the following longitudes: one at $1^\circ$ east, one at $1^\circ$ west, the third at $180^\circ$. If $\epsilon$ is very small, the shadow is nearly the entire hemisphere (because as it approaches $0$, the shadow approaches the upper hemisphere in the Hausdorff metric). If $\epsilon$ is chosen correctly, the area of this shadow equals the area of the corresponding shadow of the first tetrahedron. $\endgroup$ Commented Jul 9, 2021 at 21:01

1 Answer 1

2
$\begingroup$

The following is a detailed description of the counterexample given in the comments. The construction is the same, but with slightly more notation and detail for clarity. Also, the construction there gave two triangles with different areas but equal-area shadows; this answer constructs one polytope of which both of these triangles are faces (this is inessential, but it's unclear to me whether the original question asked about two facets of one polytope or of two different ones).

  1. Let $T$ be any triangle inscribed in a unit circle in the $xy$-plane. By $``$the lift of $T$ to height $\varepsilon$ in $S^2$$"$ I mean the triangle $$\varepsilon \hat{z} + \sqrt{1-\varepsilon^2}T,$$ which is similar to $T$, inscribed in $S^2$, and lies in the horizontal plane $\{z=\varepsilon\}$. Denote this lift by $L_\varepsilon(T)$.

  2. If $T$ is inscribed in the unit circle in the $xy$ plane and contains the origin in the interior of its convex hull, the shadow of $L_\varepsilon(T)$ on $S^2$ approaches the entire upper hemisphere as $\varepsilon$ approaches $0$ from above, and it approaches the north hemisphere as $\varepsilon \to 1$ from below (these limits of shadows can be taken with respect to the Hausdorff metric. We care only about the areas of the shadows and not about the shadows themselves).

  3. Let $T_1$ be an equilateral triangle inscribed in the unit circle in the $xy$-plane and let $T_2$ be a very narrow triangle inscribed in the same circle and containing the origin in its interior, such that $T_2$ has small area $s$ (we will find a bound on $s$ later, but any $s<0.01$ will certainly do). Fix $\varepsilon_1 = 0.1$. By part (2), we can choose $\varepsilon_2>0$ such that the shadows of $L_{\varepsilon_1}(T_1)$ and of $L_{\varepsilon_2}(T_2)$ have equal area on $S^2$. It is clear that the areas of these lifts are different (the area of $L_{\varepsilon_2}(T_2)$ is much smaller, as it is bounded above by $s$, which we may modify to be smaller than $A(L_{\varepsilon_1}(T_1))$ if necessary).

  4. The triangles $L_{\varepsilon_1}(T_1)$ and $L_{-\varepsilon_2}(T_2)$ are inscribed in $S^2$, have shadows of equal area on $S^2$, and have different areas. The planes they lie in do not intersect. Therefore the convex hull of their six vertices is a convex polytope inscribed in $S^2$ which has them as faces. This is a counterexample.

$\endgroup$
2
  • $\begingroup$ My suspicion remains that nearly everything is a counterexample to this . . . . . $\qquad$ $\endgroup$ Commented Jul 12, 2021 at 23:06
  • $\begingroup$ @MichaelHardy I'm certain you are right. $\endgroup$ Commented Jul 13, 2021 at 9:17

You must log in to answer this question.

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