You guys give up too easily. Why is the equation cubic? What is the meaning of the root a = 9.00544849? Try to figure it out before you look.
» Click to show Spoiler - click again to hide... «
The equation is cubic because there are three answers to the question which match the equation as derived. Both whimbrel and anthonytc22 got an equivalent cubic, from different derivations. This equation reflects the requirement that the total length of the bent pole is still 100 feet, but does not quite require that the bent pole actually touch the corner of the box.
So, given that hint, what does the diagram look like for the third case?
» Click to show Spoiler - click again to hide... «
If you cut the pole at 9.005... feet, and let if bend over to the left, and then sight along the pole, you will see it is aimed straight at the corner of the box. This matches the equations, but does not satisfy the written problem, which wants the bent flagpole to actually touch the corner of the box.
Now, when I solved it, I got a 6th degree polynomial. All 6 roots have a physical interpretation which sort-of meet the problem. It turns out the reason I got more roots is I considered the two small triangles which each share a side with the square, and wrote that the sum of their hypotenuses plus the upright pole must be 10 (after dividing everything by 10), and the two small triangles must be right triangles. I didn't use any similar triangles, law of cosines, or trig functions. As a result, I got more solutions.
Here are my definitions:
Divide by 10 to get rid of all those useless zeros. Call the apex of the bent flagpole point P = (0,1+b), and call the tip touching the ground point Q = (1+e,0). Call the lower left corner of the box O = (0,0), and the upper right corner I = (1,1).
For the third root, b is negative because point P is below the upper left corner (0,1).
Also, e is negative, because the tip of the pole is to the left of the lower right corner (1,0). The meaning of negative b and e is clearly defined by using coordinates instead of talking about lengths.
Here is one derivation of an equation that requires P, I and Q to be on a line:
» Click to show Spoiler - click again to hide... «
In retrospect, I could ask for an equation that states P, I, and Q are co-linear. I would write P = I + x*(Q-I) for some x. When x = 0, P = I; when x = 1, P = Q; otherwise, P is elsewhere on the line. This translates to two equations so I can eliminate x.
(0,1+b) = (1,1) + x*((1+e,0)-(1,1))
(0,1+b) = (1,1) + x*(e, -1)
(0,1+b) = (1,1) + (x*e, -x)
(0,1+b) = (1 + x*e, 1 - x)
0 = 1 + x*e, and 1 + b = 1 - x
-x = 1/e, and -x = b
b = 1/e.
Here is the derivation I worked out originally:
» Click to show Spoiler - click again to hide... «
Actually, I wrote
(b+1)^2 + (e+1)^2 = (c+d)^2, expanded, and subtracted b^2+1 = c^2 and e^2+1 = d^2. This simplifes to b+e = cd. I squared this (which allows c or d to be negative without requiring they both be negative) and substituted for c^2 and d^2. This gives a perfect square, (be - 1)^2 = 0, or be = 1, b = 1/e.
Interesting. That means if you look at the X and Y intercepts of any line through (1,1), so that Y intercept is b+1 (f(0) = b+1) and X intercept is e (f(e) = 0), then b = 1/e.
Here is another derivation:
» Click to show Spoiler - click again to hide... «
A general line is y = m x + (b+1). When x = 0, y = (b+1). y intercept is b+1.
Now add the requirement that the line goes through (1,1). 1 = m + (b+1). m = -b. So a general line through (1,1) with y intercept (b+1) is y = -b x + (b+1).
Solve for x. x = (-1/b)(y - (b+1)) = -y/b + (b+1)/b = -y/b + (1 + 1/b). When y = 0, x is (1 + 1/b).
So an equation which requires P, Q, and I to be colinear is just (b e) = 1.
Next, I want to require that the pole has length 10.
» Click to show Spoiler - click again to hide... «
The length of the pole extending from O to P is (1+b). The length from P to I is c, and the length from I to Q is d. So, the equation is
10 = 1 + b + c + d.
I also have that the two small triangles are right triangles,
b^2 + 1 = c^2, and e^2 + 1 = d^2,
So I get
b + sqrt(b^2+1) + sqrt(1/b^2+1) = 9.
Now this equation is a perfectly good definition of b, and it has three roots. Two are shared with the cubic found before, and the third is different.
The third root of the cubic arising from similar triangles is also a root of
b - sqrt(b^2+1) + sqrt(1/b^2+1) = 9.
Allowing either c or d to be negative yields a 6th degree polynomial:
» Click to show Spoiler - click again to hide... «
The 6th degree polynomial turns out equivalent to:
b ± sqrt(b^2+1) ± sqrt(1/b^2+1) = 9.
The equation with plus or minus square roots is just as solvable with numerical methods as the 6th degree polynomial, and the bisection method yields 6 roots.
To get the polynomial square c = 9 - (b+d). This permits c to be negative. Substitute for c^2 and d^2 (but not d). b^2 cancels. Move d(2b - 18) to the right hand side and square both sides. This permits d to be negative as well. Substitute (1/b^2 + 1) for d^2. Multiply through by b^2. Don't expand (avoid useless work). The result is
(1 + 81b^2 - 18b^3)^2 - b^2(b^2 + 1)(2b - 18)^2 = 0.
This equation obviously expands to a 6th degree polynomial. As written, it solves just fine numerically and gives the same roots as the simpler equation with plus or minus square roots.
Images of 6 solutions
» Click to show Spoiler - click again to hide... «

The first root satisfies b + sqrt(b^2+1) - sqrt(1/b^2+1) = 9, but the pole doesn't even reach point I, let alone Q. This root comes from allowing d to be negative while requiring
1 + b + c + d = 10.

The second root is one of the expected solutions.

The third root is the other expected solution.

The fourth root satisfies b - sqrt(b^2+1) + sqrt(1/b^2+1) = 9. The pole does not reach the ground. It comes from allowing c to be negative while requiring 1 + b + c + d = 10.

The fifth root also satisfies b - sqrt(b^2+1) + sqrt(1/b^2+1) = 9. The pole is bent to the left at point P and touches the ground at point Q. It does not touch point I but point I is on the line determined by P and Q. In this case allowing c to be negative makes perfect sense, since it reflects the fact that the pole does not actually go over to point I and back.

The last root satisfies b + sqrt(b^2+1) + sqrt(1/b^2+1) = 9, with b negative. The pole bends at point P to the right and extends to I. Then it reverses and extends back through P to Q.
--
Archimerged is founder of the
Renewable Energy Design Wikicity. What are you waiting for? Contributors needed!