MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

In 1970 or so, Klee asked if a convex body in $\mathbb R^n$ ($n\ge 3$) whose maximal sections by hyperplanes in all directions have the same volume must be a ball. The counterexample in $\mathbb R^4$ is trivial and can be described as follows:

Let $f:[-1,1]\to\mathbb R$ be continuous, strictly concave and satisfy $f(-1)=f(1)=0$. For every such function, let $Q_f$ be the body of revolution given by $y^2+z^2+t^2\le f(x)^2$. Then $Q_f$ and $Q_g$ have the same maximal sections in every direction if (and, actually, only if) $f$ and $g$ are equimeasurable, i.e., $|\{f>t\}|=|\{g>t\}|$ for all $t>0$. (Of course, there are plenty of concave functions equimeasurable with $\sqrt{1-x^2}$).

With some extra work, one can construct something like this in $\mathbb R^n$ when $n$ is even though I do not know any similarly nice geometric description of such bodies for $n\ge 6$.

What I (and my co-authors) are currently stuck with is the case of odd $n$ (say, the usual space $n=3$). In view of such simple example in $\mathbb R^4$, I suspect that we are just having a mental block. Can anybody help us out?

share|cite|improve this question
what about $n=2$? – Will Jagy Jul 15 '11 at 2:14
Planar domains of constant width have been well-known for a long time (normally one thinks of them as having constant projections but all their projections are realized as sections as well). – fedja Jul 15 '11 at 2:42
That's right. I knew about curves of constant width, I did not connect that properly with your question. – Will Jagy Jul 15 '11 at 2:46
Not really :(. The bodies of constant width (=constant maximal 1-dimensional sections) are not hard to construct because basically you just have to relate pairs of points. When you raise the section dimension, you have to deal with integral transforms. I suspect that this simple R^4 construction went unnoticed because the customary way is to represent convex bodies by their radial or support functions even for the bodies of revolution and the equimeasurability condition is a total mess in such terms. So, it seems like we need a fresh look with some little twist... – fedja Jul 15 '11 at 3:29

To those who are still interested: we've finally made it but it's so ugly that a nice alternative approach will be always welcome :) We are still stuck with Bonnesen's question about the possibility to recover a convex body from the volumes of its maximal sections and projections in odd dimensions, so some help will be appreciated. The even-dimensional case can be found here. I feel a bit like a student asking for help with his homework, of course, but why not? We all get stuck now and then :). This should really be a comment but it's too long to fit the number of characters restriction.

share|cite|improve this answer
More affirmation of Klee's prowess as a problem-poser! May I ask: Is the essence of your proof in odd dimensions concentrated (in the intellectual sense) in $\mathbb{R}^3$? – Joseph O'Rourke Jan 4 '12 at 2:24
It doesn't matter for us whether it is 3,5,7, or any other odd number: there is no simplification or complication when going from one odd dimension $d>1$ to another, if that's what you meant. However, as I said, the odd dimension case is harder than the even one due to the square root popping up everywhere and the non-locality of the Radon transform. – fedja Jan 4 '12 at 12:23

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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