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

Suppose $K$ is a centrally symmetric, strictly convex body in $\mathbb{R}^2$. Let denote the curvature and the support function of $\partial K$, boundary of $K$, respectively with $\kappa$ and $s$. If $m\le\frac{\kappa}{s^3}(K)\le M$ for some positive numbers $m$ and $M$, does it mean there are ellipsoids $E_1$ and $E_2$ such that $E_1\subseteq K\subseteq E_2$ and $$\frac{\kappa}{s^3}({E_1})=M,~~~ \frac{\kappa}{s^3}(E_2)=m ~~~~? $$

share|cite|improve this question
Could you say more about what $s$ is? I think of a "support function of $X$" as being a function that is 1 for points in $X$ and 0 otherwise. But apparently not in this case. – Carl Feynman Apr 9 '11 at 16:38
@Carl Feynman The support function $h_A:\mathbb{R}^n\to\mathbb{R}$ of a non-empty closed convex set $A$ in $\mathbb{R}^n$ is given by :$ h_A(x)=\sup\{ x\cdot a: a\in A\},$ – MINI Apr 9 '11 at 17:06
I think it should be $M$ for the smaller ellipsoid and $m$ for the larger one. – Sergei Ivanov Apr 9 '11 at 18:31
@Sergei Ivanov edited. – MINI Apr 9 '11 at 18:33
Sorry, I was confused by the word "ellipsoid" (rather than "ellipse") and thought that it was in $\mathbb R^3$. – Sergei Ivanov Apr 10 '11 at 22:09
up vote 4 down vote accepted

Yes this is true. Let me handle the inner ellipse, the outer one is similar.

For brevity, denote $\kappa/s^3$ by $a$. It is easy to see that $$ a = \frac{\dot\gamma\wedge\ddot\gamma}{(\gamma\wedge\dot\gamma)^3} $$ where $t\mapsto \gamma(t)$ is any counter-clockwise parametrization of the boundary curve. For an ellipse, this is a constant inverse proportional to the square of the area. So we have to prove that the maximum-area ellipse contained in $K$ has $a\le M$.

So let $E$ be the maximum-area ellipse contained in $K$. Since the problem is centro-affine invariant, we may assume that $E$ is the unit circle. Then we have to prove that $M\ge 1$.

Consider the points where $\partial K$ touches $\partial E$. One easily sees that the intervals between these points on the circle are no greater than $\pi/2$, otherwise $E$ could be made larger. Choose coordinates so that one of the touch points is $(1,0)$, then there is another touch point of the form $(\cos\theta,\sin\theta)$ where $0<\theta\le\pi/2$. Observe that the arc of $\partial K$ between these touch points is contained in the square $[0,1]^2$.

Parametrize $\partial K$ by spanned area, i.e. by a curve $t\mapsto\gamma(t)$ such that $\gamma\wedge\dot\gamma=1$. Then $\gamma\wedge\ddot\gamma=0$, hence $\ddot\gamma(t)=-a(t)\gamma(t)$ for all $t$ where $a(t)$ is the centro-affine curvature. Let $\gamma(t)=(x(t),y(t))$, then $\ddot x(t)=-a(t)x(t)$ and $\ddot y(t)=-a(t)y(t)$. Suppose that $M=\sup a(t)<1$. Since $x(0)=1$, $\dot x(0)=0$, $y(0)=0$ and $\dot y(0)=1$, a standard comparison theorem for equations of the form $\ddot x=-ax$ implies that $x(t)>\cos t$ and $y(t)>\sin t$ for all $t\in(0,\pi/2]$. Therefore $x(t)^2+y(t)^2>1$ for all $t\in(0,\pi/2]$ and $y(\pi/2)>1$. This means that $\gamma$ leaves the square $[0,1]^2$ before is has a chance to touch the circle again, a contradiction.

share|cite|improve this answer
@Sergei Ivanov Superbe! – MINI Apr 11 '11 at 20:46

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.