# Cover a disk with thin rectangles

Let $D$ be a disc of diameter 20, and suppose you are given 19 rectangles, each of which is $1 \times 20$. Can $D$ be covered completely by the rectangles? Note that the rectangles can be arranged "any which-way" for the covering.

Note: this was a popular poser in my grad student days; some of us came up with an answer, but had a hard time nailing down the proof. I'd like to hear from anyone with a good approach, or even a good reference, to this question.

Added note: the rectangles are not to be cut up. And the question may be posed differently in terms of 19 width 1 "strips", each infinitely long (i.e. copies of $[0,1] \times R$). In this form the question becomes: Can we cover the diameter-20 disk using 19 of these width 1 strips? [If one could cover via strips, then each strip could be trimmed to a 1X20 rectangle without uncovering any covered points in the disk.]

-
Are the rectangles permitted to overlap each other? – Michael Joyce Nov 5 '12 at 23:24
The sum of areas of rectangles are $19\times 20=380$ and the area of the disc is $10^2\times \pi=100\times\pi\approx 314$. So there are enough rectangles to cover it!! How to cover? – Sigur Nov 5 '12 at 23:26
For a disc of diameter $n \geq 2$, suppose you have $n-1$ rectangles of dimensions $1 \times n$. Covering the disc when $n = 2$ is clearly impossible, but heuristically likely as $n \rightarrow \infty$, since the area of the disc is $\pi n^2 /4$ and the area of the rectangles is $n^2 - n$, which means for large $n$, (rectangle area) - (disc area) tends to infinity, and we'll have more leeway in how to place the rectangles. This makes me wonder: what is the minimal $n$ for which such a covering is possible? (And will it be possible for all greater integers?) – Benjamin Dickman Nov 5 '12 at 23:54
Yes the rectangles are allowed to overlap each other. And certainly there's enough area to do it, provided we use rectangles which overlap near the center of the disk. In any case, the question is: can it be done, and why/why not. – coffeemath Nov 6 '12 at 0:45
Well, the area of a shape is the measure of the set of points that intersect it. But these thin rectangles are more like lines than like points. Perhaps the analogous measure on the set of lines which intersect the disk (something that is studied in integral geometry) is the relevant quantity here. – Rahul Nov 6 '12 at 1:05

I remember that problem from university! (But I didn't solve it, somebody gave me a big clue)

The answer is that it is impossible, you can't cover the circle of diameter 20 using only 19 strips $1\times 20$.

To see why, suppose you have such cover and imagine a sphere of diameter 20 cut in half by our circle. Find the orthogonal projection of each strip on the sphere, it is a ring, and it is easy to compute it's area $2\pi R x$, where $R$ is the radius of the sphere and $x$ the width of the strip.

But between all the rings, they cover the whole sphere so the total area must be at least $4\pi R^2$, so $$N \times 2 \pi R x \ge 4 \pi R^2$$ where $N$ is the number of strips. Letting $N=19, R=10$ and $x=1$ we obtain a contradiction.

-
+1, very nice solution! – joriki Nov 7 '12 at 22:21
Wow! When our little group "solved" this in grad school, we stayed in the plane to do it, and used the idea that in covering the circumference some strips had to get near the boundary of the disk, and that made their areas too small. We got a bit lost in the details, so I much appreciate your answer. +1. – coffeemath Nov 8 '12 at 17:41
If the disk is partitioned into 20 width one parts of the form $[x_i,x_i+1]$ for $x_i=-10..,9$, the projections on the sphere cover it exactly. And since the surface areas of the slices only depend on their width, this shows you need 20 to cover the surface area, but only have 19. [The surface area of a slice on the sphere being the same as that on a cylinder...] So even without a calculation I should have noticed this, if only I'd thought about the 3-d aspect! – coffeemath Nov 9 '12 at 12:46