On a $2004 \times 2004$ chess table there are 2004 queens such that no two are attacking each other\footnote[1]{two queens attack each other if they lie on the same row, column or direction parallel with on of the main diagonals of the table}. Prove that there exist two queens such that in the rectangle in which the center of the squares on which the queens lie are two opposite corners, has a semiperimeter of 2004.
Problem
Source: Romanian IMO TST 2005 - day 5, problem 1
Tags: geometry, rectangle, modular arithmetic, combinatorics proposed, combinatorics
24.04.2005 17:24
Remember that ISL problem which said that if we have an even number of people around a table and they change places, there will be two who will have the same number oif people between them before and after the rearrangement? Basically, that says that if $n$ is even, and $\sigma\in S_n$, then we can either find $i,j$ s.t. $|i-j|=|\sigma(i)-\sigma(j)|$ or we can find $i,j$ s.t. $|i-j|+|\sigma(i)-\sigma(j)|=n$. We apply it to $n=2004$, together with the observation that the conditions imply $|i-j|\ne|\sigma(i)-\sigma(j)|,\ \forall i\ne j\ (*)$, where $\sigma(i)$ is the row on which the queen standing on the column $i$ is placed ($(*)$ means that no two queens lie on the same line parallel to one of the diagonals).
24.04.2005 20:06
I think we need more detailed references here, as it was obvious that we need to prove fact mentioned above.
24.04.2005 20:14
I can't find the problem on Kalva anymore, because there are lots of problems to look through, and I don't remember the yewr, but I was pretty sure it had been discussed before.
24.04.2005 20:19
So the problem is unsolved...
24.04.2005 20:50
Let's look at that problem I mentioned earlier: For $i\in\overline{0,n-1}$, let $\sigma(i)-i$ denote the distance modulo $n$, in a counter-clockwise direction, between $i$ and $\sigma(i)$. If we assume that all $\sigma(i)-i$ are distinct, it means that they are $0,1,\ldots,n-1\pmod n$. If we add all these quantities, we get something which is not divisible by $n$ (it's $\frac n2\cdot(n-1)\pmod n$), but, on the other hand, $\sum(\sigma(i)-i)=0\pmod n$ because the residues $\{\sigma(i)\}$ are the same as the residues $\{i\}$, so we have a contradiction. This means that there are $i,j$ s.t. $\sigma(i)-i=\sigma(j)-j\pmod n$. This proves the assertion, I believe.