How prove that for any curve $\alpha(s)$ of length $L=1$ in the real plane, there is a semicircle of diameter $2R=1$ that contains it.

enter image description here

Any hints would be appreciated.

  • 1
    $\begingroup$ nice picture. how'd you make it? $\endgroup$ – Stefan Smith Dec 29 '13 at 23:47
  • 3
    $\begingroup$ using GeoGebra. $\endgroup$ – felipeuni Dec 29 '13 at 23:55
  • $\begingroup$ Do you know for sure that this is true? Where does the question come from? $\endgroup$ – Seub Dec 30 '13 at 2:18
  • $\begingroup$ Well, how can you know for sure if this is true without the underlying reason? $\endgroup$ – hot_queen Dec 30 '13 at 2:37
  • $\begingroup$ @Seub foromatematicoperuano.wordpress.com/page/2 $\endgroup$ – felipeuni Dec 30 '13 at 2:44

I gave up, googled and found an extremely elegant agrument (due to A. Meir) here. I guess the hard part was to locate the center of the semicircle.

  • 1
    $\begingroup$ It took me more than 30 minutes to understand the argument in its entirety… How does anyone think of such a construction… Thanks for the link +1 $\endgroup$ – Prism Dec 30 '13 at 4:43

This (real analysis) is not really a subject I'm good at, but I think I have a solution.
Suppose $\alpha$ has length two, so the radius of the circle is $1$. Take the midpoint of $\alpha$ (the point for which the two curves you get when cutting $\alpha$ have the same length) and call it $M$. Now, at both sides of $M$, the curve has length $1$. Now look at the (closed) disk with midpoint $M$ and radius $1$. Because this disk is per definition the set of points with distance at most $1$ from the center, all points not in this disk have a distance to $M$ strictly larger than $1$. Such a point (say $P$) cannot be part of $\alpha$, because then, the distance $|PM|$ would be larger than $1$ and a straight line segment is shorter than any other curve connecting $P$ and $M$ (this follows intuitively from the triangle inequality) the part $PM$ of the curve must be longer than $1$, but it has length $1$, so this is a contradiction. Therefore, we can always draw a disk with radius $1$ that fully covers a curve with length $2$.

The only possible flaws I see are that a straight line segment is the shortest possible connection and that the boundary of the disk (set of points $P$ with $d(P,M)\leq 1$) is not trivially equal to the circle with radius $1$ around $M$.

  • $\begingroup$ Read the question again. You need to cover the curve by a semicircle. $\endgroup$ – hot_queen Dec 30 '13 at 1:03
  • $\begingroup$ @hot_queen Oh, I missed that. Time to go to bed... $\endgroup$ – Ragnar Dec 30 '13 at 1:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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