Let C be a compact convex subset of 3-dimensional Euclidean space E(3) whose interior is non-empty and whose diameter is d. What is the largest volume that C can have if every subset of C that is a straight line segment of length d is a subset of the boundary of C (or-equivalently-if no chord of C that contains an interior point of C has length d)? It looks as though this maximum might be close to the volume of a right circular cone whose base radius is d/2 and whose height is ((3^(1/2))/2)*d. I am unable to say more. Perhaps there is no maximum volume-only a least upper bound. Although C is a convex body, it is not what would be called stricly convex. To keep things simple I am avoiding any discussion of higher dimensional versions of this question.
-
$\begingroup$ The equivalent form of the question is to maximize the volume of a convex set of diameter $d$ such that it contains <i>one</i> segment of length $d$ on its surface. (Then you may compress everything a bit to make this segment unique.) $\endgroup$– Ilya BogdanovSep 1, 2013 at 10:28
2 Answers
I would suggest yet another larger set. Take two points $A$ and $B$ in the plane with $|AB|=d$. Let $S$ be the set of all points $X$ in this plane lying in one halfplane defined by $AB$ and satisfying $|XA|,|XB|\leq d$. Then rotate $S$ around the perpendicular bisector of $AB$; you get the convex body containing your cone. Then, after smashing it a bit you get a desired set --- still of larger volume.
It seems you could curve the base of the cone, making it a sector of a sphere of
radius $r > d$ slightly larger than $d$, centered at a point $c$ directly above the
apex of the cone:
(Arrows indicate segments of length $d$ that cannot fit.)
If this is correct, the max volume can be approached but not achieved, as you suggest.
Update. And here is Ilya Bogdanov's larger body, a half-lune:
-
$\begingroup$ Thanks to all of you for your many interesting examples. They all reinforce my feeling there is a least upper bound $\endgroup$ Sep 2, 2013 at 17:57
-
$\begingroup$ Thanks to all of you for your many interesting examples. They all reinforce my feeling that there is a least upper bound on the volume but no actual maximumum volume. But apart from playing around with examples, I can see no way to make further progress on this problem and actually prove something. It is interesting that a solid hemisphere having diameter d is a convex set of this kind which is larger than the right circular cone I mentioned. But one cannot seem to add volume to the hemisphere as one can to the cone. $\endgroup$ Sep 2, 2013 at 18:22