P in a Square


Given a square ABCD and a point P in the plane (either picture above is correct), find PD and AB in the problems below.

Problem 1: (PA, PB, PC) = (4, 3, sqrt(10) )
Problem 2: (PA, PB, PC) = (7, 3, 9)

For each problem, there is only one answer for PD. There are two answers for AB, corresponding to the point P being inside or outside of the square. Finding PD is easier than finding AB.

Source: 1. Internet newsgroup rec.puzzles, 2. Original
[Taken from "Ken's Puzzle of the Day" February 24, 1994]

Solution:
Wilfred Theunissen and Evert Offereins provided solutions:
Solution to Problem 1:

AB = 5 and PD = sqrt(17) with P inside the square,
AB = 1 and PD = sqrt(17) with P outside the square.

Problem 2:

PD = 11 and AB = sqrt((539-98*sqrt(17))/106) +
sqrt((8171-98*sqrt(17))/106) with P inside the square.
PD = 11 and AB = sqrt((4655-98*sqrt(17))/106) -
sqrt((415-98*sqrt(17))/106)) with P outside the square.
[These are the correct values (though I have a little simpler form for the last numbers.) -KD]
From Ken Burres:
I have a method for finding PD (I have not really put the time into finding AB).

Drawing perpendicular lines through point p (parallel to the sides) you receive 4 rectangles, with PB, PC, PA, and PD as the diagonals. I labelled the shorter segments on the sides (made from the perpendicular lines above) x, y, z, and w so that:
x^2 + y^2 = PB^2 (x^2 + y^2 = 9),
y^2 + w^2 = PC^2 (y^2 + w^2 = 10),
x^2 + z^2 = PA^2 (x^2 + z^2 = 16).
By manipulating these equations (or solving them with a matrix) one will find that:
z^2 + w^2 = 17.
Therefore PD^2 = 17, PD = sqrt{17}.

With the other situation, z^2 + w^2 = 121, therefore PD = 11.


Wilfred solved the generic problem for AB and sent this solution:
At last, I solved this problem:

Let AP=a, BP=b ,CP=c and DP=d. AB=x.

It is easy to see that a^2 + c^2 = b^2 + d^2, but I don't need this
result in my proof.

Let angle(ABP)= Beta, so angle(CBP)=90-Beta.

Looking at Triangle ABP and the cosine-rule:

(1)     a^2 = b^2 + x^2 -2*b*x*cos(Beta)

Looking at Triangle CBP and the cosine-rule:

(2)     c^2 = b^2 + x^2 -2*b*x*cos(90-Beta)

Because cos(90-Beta) = sin(Beta):

(3)     c^2 = b^2 + x^2 -2*b*x*sin(Beta)

Rewrite (1) and (3):

(4)     2*b*x*cos(Beta) = b^2 - a^2 + x^2
(5)     2*b*x*sin(Beta) = b^2 - c^2 + x^2

And thus:

(6)     4*b^2*x^2*cos^2(Beta) = (b^2 - a^2 + x^2)^2
(7)     4*b^2*x^2*sin^2(Beta) = (b^2 - c^2 + x^2)^2
        -------------------------------------------   +
(8)     4*b^2*x^2 = (b^2 - a^2 + x^2)^2 + (b^2 - c^2 + x^2)^2

(Because cos^2 + sin^2 = 1 for each possible angle)

So formula (8) is a quadratic formula in x^2
Rewriting this formula results in:

x =
0.5*sqrt(2*a^2+2*c^2+sqrt((2*a^2+2*c^2)^2-8*(b^2-a^2)^2-8*(b^2-c^2)^2))
(If P is inside the square)

AND

x =
0.5*sqrt(2*a^2+2*c^2-sqrt((2*a^2+2*c^2)^2-8*(b^2-a^2)^2-8*(b^2-c^2)^2))
(If P is outside the square)

This results in (for Problem2):

x = 0.5*sqrt(260 + 28*sqrt(17)) if P is inside the square and
x = 0.5*sqrt(260 - 28*sqrt(17)) if P is outside the square.

These forms are indeed a lot simpler.

My (Ken's) Solutions:

Use P inside the square (both solutions will be found this way.) Divide the square into four rectangles, with internal corners at P. Let the side-lengths of the rectangles be w, x, y, z, with w+x = y+z = AB.
Pythagorus shows us:
w^2 + y^2 = AP^2
w^2 + z^2 = BP^2
x^2 + z^2 = CP^2
x^2 + y^2 = DP^2 = AP^2 + CP^2 - BP^2
So DP = sqrt(AP^2 + CP^2 - BP^2)

To solve for AB, let x=AB, let point P be at (Px, Py), A be at (0,0), B be at (0,x), and let a=AP, b=BP, c=CP, d=DP. Then:
(1) Px^2 + Py^2 = a^2
(2) Px^2 + (Py-x)^2 = b^2
(3) (Px-x)^2 + (Py-x)^2 = c^2
(4:2-1) Py = [x^2 - (b^2 - a^2)]/2x
(5:3-2) Px = [x^2 - (c^2 - b^2)]/2x
(6:4,5->1) 2x^4 - 2x^2(a^2+c^2) + (b^2-a^2)^2+(c^2-b^2)^2 = 0
Quadratic Formula:
x^2 = [ (a^2 + c^2) +/- sqrt( (2bd)^2 - (a^2 - c^2)^2 ) ] / 2

For Problem 1, DP = sqrt(17), AB = 1 or 5
For Problem 2, DP = 11, AB = sqrt(65 +/- 7sqrt(17))


Extra Credit: Find AB, AP, BP, CP, DP to all be integers. If you do, let me know!

Update 2 Sept. 2008: Bojan Basic writes:

I just want to inform you that this is a very hard problem. In fact, it is well-known and still unsolved. You may check chapter D19 of Guy's book "Unsolved Problems in Number Theory" for much informations about it. If you don't have that book, its status can be seen here: http://mathworld.wolfram.com/RationalDistanceProblem.html.


kend@rosemail.rose.hp.com