Marcos
Posted on Friday, 11 June, 2004 - 07:20 am:

You've have an n×n board with 1's in each cell. You make one of them into a -1. You're only allowed to change the signs of a whole k×k subboard (1 £ n). Your aim is to have 1's in every cell.

My question is, what is the number of initial configurations (i.e. in how many cells can the -1 go) so that you can get a solution.

I'm going to call this number b(n).

We have b(2)=b(3)=0

It's also not very difficult to show that b(5)=1 (where the -1 goes in the central cell: this cell works for all odd n)

This of course automatically implies b(7) ³ 5 but I'm not sure of its exact value.

Does anybody have any ideas?

Marcos

Michael McLoughlin
Posted on Friday, 11 June, 2004 - 07:44 am:

I don't quite understand the question. What do you mean when you say "You're only allowed to change the signs of a whole k x k subboard"?
Kerwin Hui
Posted on Friday, 11 June, 2004 - 08:54 am:

Here are my thoughts so far:

The 5 by 5 case shows the central 3 by 3 in the 7 by 7 are doable (find a 5 by 5 square which contains it in the middle).

Also, we have the following `device':
- + +
+ + +
+ + +


Turning all 3 by 3 gives
+ - -
- - -
- - -


Now turn the three 2 by 2's that doesn't contain the top left corner, we get
+ + +
+ - +
+ + -


So we can move any -1 down the a diagonal to get 2 -1's.

We can also iterate this process again, which gives us: if we have a 4 by 4 square with a -1 at a corner, we can move this -1 to the opposite corner.

So we can do the 37 squares marked by an 'X'
X X X X X
X X X X X X
X X X X X
X X X X X
X X X X X
X X X X X X
X X X X X

(and this implies all 8 by 8 and above are doable). I have no idea how to tackle the remaining 12 squares at this moment.

Kerwin
Kerwin Hui
Posted on Friday, 11 June, 2004 - 09:01 am:

We also have the device:
+ + +
- + +
+ + +

changing the two left hand side 2 by 2 and then the 3 by 3 gives
+ + -
+ - -
+ + -


So the remaining 12 are also doable.

Kerwin
George Barwood
Posted on Friday, 11 June, 2004 - 09:12 am:

Answer in white:


B(n) = (n-4)^2 ( n > = 5 )

Since B(5) = 1, we can manipulate any square than is not in the 2-cell wide border.

For any cell in the border there is an adjacent square that is always flipped at the same time.

Therefore we can never find a solution for a border cell.
George Barwood
Posted on Friday, 11 June, 2004 - 09:17 am:

Oops, ignore the above post, obviously my statement doesn't hold for corners!
Kerwin Hui
Posted on Friday, 11 June, 2004 - 09:44 am:

Just have a little more play with my devices - can show that b(6)=36. (Try it! It is not too difficult.)

Kerwin

Marcos
Posted on Friday, 11 June, 2004 - 01:52 pm:

That's great Kerwin! Thanks!

So we have,

b(2)=b(3)=b(4)=0

b(5)=1

b(n)=n2 for n > 5?

I'm having a bit of trouble showing b(6)=36 although that's probably because I haven't really tried doing it properly.

Marcos

Kerwin Hui
Posted on Friday, 11 June, 2004 - 08:37 pm:

Oops, that is not true... beta(6)=20, not 36. There are 16 impossible position
X X
X X X X
X X
X X
X X X X
X X

(any nontrivial subboard contains an even number of those). The other 20 are possible (the only slightly nontrivial ones are those in the middle of the edge).

Kerwin