In two-dimensional case one can generalize figures of constant width as figures which can rotate in a convex polygon. Here is one example which can be used to drill triangular holes:

alt text

I would like to know what happens with this generalization in dimension $3$ and maybe higher. Obviously a body of constant width $1$ can rotate arbitrary in a unit cube. More formally, given a body $B$ of constant width $1$ and $A\in SO(3)$ there is $v\in \mathbb R^3$ such that $$A(B)+v\subset\square,$$ where $\square$ is unit cube. On the other hand, except for the cube, I do not see any other examples of convex polyhedron which have nontrivial rotating bodies (i.e. distinct from the inscribed ball).

I hope that the answer is known. (= I hope I should wait for the answer and I do not have to think.)

The question is inspired by this one: "Local minimum from directional derivatives in the space of convex bodies."

  • 10
    $\begingroup$ These bodies are called "rotors" of the corresponding cavity. A full classification of non-trivial (i.e. other than bodies of constant width) rotors is known. A good reference on these is "Geometric Applications of Fourier Series and Spherical Harmonics" by Helmut Groemer. As I recall, only the regular simplex and cross-polytope admit non-trivial rotors in dimensions four and above. In three-dimensions, there are more cases. $\endgroup$ Jan 26, 2012 at 18:10
  • 3
    $\begingroup$ @Yoav, why don't you write it as an answer? $\endgroup$ Jan 26, 2012 at 18:57
  • $\begingroup$ Sorry, I'm still getting used to how things work here on MO. Thanks for the MOtiquette pointer. $\endgroup$ Jan 26, 2012 at 19:30

2 Answers 2


I found the reference I was looking for. The full list of cases under which $K$ is a rotor in a cavity shaped like the polytope $P$ is available on page 27 of the notes titled "The use of spherical harmonics in convex geometry" by Rolf Schneider. They are available under "Course Materials" on his website. As I recall, there is one more non-trivial case in $d=3$ if the cavity is allowed to be unbounded (e.g. a cone), and this case appears in the more complete list in Groemer's book.


This adds nothing to Yoav Kallus' answer, but I was curious to see what these rotors look like (Schneider's notes has no figures). I found grainy photos of rotors for the cube, the regular octahedron, and the regular tetrahedron in a 50-year old paper by Michael Goldberg, "Rotors in Polygons and Polyhedra," Mathematics of Computation, Vol. 14, No. 71 (July, 1960), pp. 229-239:
(They remind me of stones found on a beach!)

Of course one can find much better examples of cube rotors, which as Anton points out, are just constant-width bodies. E.g., this is from the cover of Bryant and Sangwin's 2008 How Round Is Your Circle (Wayback Machine and an alternative link):
    Constant width bodies


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.