|
5. & 6.
A Problem Seminar, D. J. Newman
26. Given a finite collection of closed squares of total area 3, prove that they can be arranged to cover the unit square.
The construction we give will do this covering without "tilting" the squares and then it becomes a very pretty result because the 3 is best possible. Namely, 3 squares each of area $1-\varepsilon$ cannot cover: they cannot even cover the 4 vertices, since each covers at most one!
Sometimes the "dumbest" approach to a problem is the one that leads to success, and this happens to be the case here. Just arrange the squares in decreasing (non-increasing) order and line them up in rows until the length of each row first exceeds 1. We obtain, thereby, rows of length $>1$ and minimum heights which we may call $h_1, h_2, h_3, \ldots$ and finally a row of length $\leqslant 1$ with a height of perhaps 0 . We need only prove that $h_1+h_2+h_3+\cdots \geqslant 1$ for then the juxtaposition of these rows will cover the whole unit square. Now the maximum heights of these row's are bounded by $1, h_1, h_2$, etc., and so the areas of said rows are bounded by $1 \cdot\left(1+h_1\right), h_1\left(1+h_2\right), h_2\left(1+h_3\right), \ldots$ which are then bounded by $1+h_1, h_1+h_2, h_2+h_3, \ldots$. Comparing areas then gives
\[
1+2 h_1+2 h_2+\cdots \geqslant \text { Total Area } \geqslant 3
\]
or $h_1+h_2+\cdots \geqslant 1$, as desired.
27. Given a finite collection of squares of total area $\frac12$, show that they can he arranged so as to fit in a unit square (with no overlaps).
Since two squares of side $\frac{1}{2}+\varepsilon$ cannot be so fit into the unit square, this is also a best possible result.
We proceed as in Problem 26, arranging the squares in decreasing order and then making rows of them this time, however, we stop each row just before its length exceeds 1. So, calling the maximum heights $H_1, H_2, \ldots$, we see that this time we want the inequality $H_1+H_2+\cdots \leqslant 1$ (so that the "pile up" of all the rows will then fit into the unit square).
To obtain this inequality, we call the lengths of the rows $L_1, L_2, \ldots$, respectively, and note that each $L_k+H_{k+1}>1$. Hence, $L_k+H_{k+1}-H_k>1-H_k \geqslant 1-H_1$, and so $H_{k+1}\left(L_k+H_{k+1}-H_k\right)>\left(1-H_1\right) H_{k+1}$. If we sum these inequalities we note that the left-hand sides add up to an area bounded by the full sum of the individual squares less the first one; i.e., to $\frac{1}{2}-H_1^2$. So we have $\frac{1}{2}-H_1^2 \geqslant\left(1 -H_1\right)\left(H_2+H_3+\cdots\right)$ and this, in turn, gives $\frac{1}{2}+H_1-2 H_1^2 \geqslant\left(1-H_1\right)\left(H_1+H_2+H_3+\cdots\right)$. The proof is completed by noting that $\frac{1}{2}+H_1-2 H_1^2=1-H_1-\frac{1}{2}\left(1-2 H_1\right)^2 \leqslant 1-H_1 ;$ which gives our required inequality, namely, $1-H_1 \geqslant\left(1-H_1\right)\left(H_1+H_2+H_3+\cdots\right)\implies1\geqslant H_1+H_2+H_3+\cdots$.
|
|