Imaginary triangles

Algebra Level 5

Consider the set of cubic polynomials, x 3 + ( 1 p ) x 2 + p x + 1 , p Z x^3 + (1-p)x^2 + px + 1, p \in \Z . Then consider the set of triangles formed by the roots of these polynomials in the complex plane. Submit the value of p p which maximizes the area of such triangles.


The answer is 5.

This section requires Javascript.
You are seeing this because something didn't load right. We suggest you, (a) try refreshing the page, (b) enabling javascript if it is disabled on your browser and, finally, (c) loading the non-javascript version of this page . We're sorry about the hassle.

3 solutions

Chris Lewis
Aug 4, 2020

As per @Mark Hennings' post, the easiest way to solve this is numerically. I got curious, though: can we find a formula for the area of the triangle in terms of the parameter p p ?

Define f p ( x ) = x 3 + ( 1 p ) x 2 + p x + 1 f_p(x)=x^3+(1-p)x^2+px+1 .


The first thing to establish is, when do we get a triangle? That is, when does f p f_p have complex roots?

The discriminant of f p f_p turns out to be ( p 2 p + 1 ) ( p 2 p 31 ) \left(p^2-p+1 \right) \left(p^2-p-31 \right) . We get complex roots when this is negative, ie when 1 2 5 5 2 < p < 1 2 + 5 5 2 \frac12-\frac{5\sqrt5}{2}<p<\frac12+\frac{5\sqrt5}{2}

We'll assume p p is in this range for the rest of the discussion.


Since the coefficients of p p are real, the two complex roots of f p f_p are conjugate. Let the roots of f p f_p be a , u + i v , u i v a,u+iv,u-iv , where a , u , v a,u,v are real, and v > 0 v>0 .

We're interested in the area of the triangle formed by these three points in the complex plane. By symmetry, this triangle is isosceles; its area is just ( a u ) v |(a-u)v|

As absolute functions aren't always differentiable, it's easier to maximise the square of this. Define T = ( a u ) 2 v 2 T=(a-u)^2 v^2


Since we know the roots, we can write f p ( x ) = x 3 + ( 1 p ) x 2 + p x + 1 = ( x a ) ( x u i v ) ( x u + i v ) = ( x a ) ( x 2 2 u x + u 2 + v 2 ) f_p(x)=x^3+(1-p)x^2+px+1=(x-a)(x-u-iv)(x-u+iv)=(x-a) \left(x^2-2ux+u^2+v^2 \right)

Comparing coefficients, we find 1 p = a 2 u , p = 2 a u + u 2 + v 2 , 1 = a ( u 2 + v 2 ) 1-p=-a-2u,\;\;\; p=2au+u^2+v^2,\;\;\; 1=a(u^2+v^2)

With a bit of manipulation, we get a = p 1 2 u a=p-1-2u and v 2 = 3 u 2 + 2 ( 1 p ) u + p v^2=3u^2+2(1-p)u+p , so that

T = ( 3 u 2 + 2 ( 1 p ) u + p ) ( p 1 3 u ) 2 T=(3u^2+2(1-p)u+p)(p-1-3u)^2


By definition, f p ( u + i v ) = 0 f_p(u+iv)=0 .

Plugging in, and considering the real part, we find u 3 3 u v 2 + ( 1 p ) ( u 2 v 2 ) + p u + 1 = 0 u^3-3uv^2 + (1-p) \left(u^2-v^2 \right) + pu + 1 = 0

But we already have an expression for v 2 v^2 ; substituting this in, and tidying up, we find 8 u 3 + 8 ( 1 p ) u 2 + ( p 2 p + 1 ) ( 2 u 1 ) = 0 8 u^3 + 8(1-p)u^2 + \left(p^2-p+1 \right) (2u-1) = 0


All that's left is to eliminate u u , and we'll have a formula for T T in terms of p p .

Except...that involves working out the solution of a cubic and then manipulating it in an already fairly tricky expression. This might be possible, and things might cancel out neatly, but I decided on a different approach.

Consider the equation 8 u 3 + 8 ( 1 p ) u 2 + ( p 2 p + 1 ) ( 2 u 1 ) = 0 8 u^3 + 8(1-p)u^2 + \left(p^2-p+1 \right) (2u-1) = 0 again. It's a cubic in u u , but it's also a quadratic in p p ! This is much easier to deal with.

Things get messy here, but after solving this quadratic, substituting in to the expression for T T and tidying up, we find the faintly ridiculous T = ( 8 u 6 24 u 5 + 10 u 4 4 u 3 + 18 u 2 4 u 1 ) ± 3 + 24 u 44 u 2 40 u 3 + 84 u 4 + 80 u 5 + 16 u 6 2 ( 2 u 1 ) 2 T=\frac{-\left(8 u^6 - 24 u^5 + 10 u^4 - 4 u^3 + 18 u^2 - 4 u - 1\right) \pm \sqrt{-3 + 24 u - 44 u^2 - 40 u^3 + 84 u^4 + 80 u^5 + 16 u^6}}{2(2u-1)^2}

The ± \pm in there should be taken as negative when u < 1 2 u<\frac12 and positive when u > 1 2 u>\frac12 . When u = 1 2 u=\frac12 we get a special case; we find p = 3 2 p=\frac32 and that T = 7 4 T=\frac74 .


We now have a formula for T T in terms of u u , and p p in terms of u u .

Unfortunately, the formula I have for T T is still pretty horrific; maximising it involves solving an equation numerically anyway.

We find the overall maximum area is 2.73606 2.73606 at u = 1.87103 u=1.87103 and p = 4.55206 p=4.55206 .

Amazing effort, Chris. Dogged determination.

Fletcher Mattox - 10 months, 1 week ago

Excellent work, and like Fletcher said, great determination.

Matthew Feig - 10 months ago

Log in to reply

Thanks - I've corrected the sign error (I'd be amazed if that was the only one!)

Chris Lewis - 10 months ago
Mark Hennings
Aug 4, 2020

If we define F p ( X ) = X 3 + ( 1 p ) X 2 + p X + 1 F_p(X) = X^3 + (1-p)X^2 + pX + 1 , then we note that F p ( 0 ) = 1 > 0 F_p(0) = 1 > 0 , F p ( 1 ) = 3 > 0 F_p(1) = 3 > 0 , while F p ( 2 ) = 13 2 p F_p(2) = 13 - 2p and F p ( 1 2 ) = 1 8 ( 11 + 2 p ) F_p(\tfrac12) = \tfrac18(11 + 2p) . Thus F p ( 2 ) < 0 F_p(2) < 0 if p 7 p \ge 7 , and F p ( 1 2 ) < 0 F_p(\tfrac12) < 0 if p 6 p \le -6 . Thus we deduce that the three roots of the cubic F p ( X ) F_p(X) are all real if p 7 p \ge 7 or p 6 p \le -6 , and hence form a triangle of area 0 0 .

The area of a triangle with complex vertices a , b , c a,b,c is (to within a sign) equal to i 4 1 1 1 a b c a b c \frac{i}{4}\left| \begin{array}{ccc} 1 & 1 & 1 \\ a & b & c \\ a^\star & b^\star & c^\star \end{array}\right| At this point it is simplest to find the roots of F p ( X ) F_p(X) numerically, and calculate the areas of the triangles formed by the roots, in the finite number of cases p = 5 , 4 , 3 , 2 , 1 , 0 , 1 , 2 , 3 , 4 , 5 , 6 p=-5,-4,-3,-2,-1,0,1,2,3,4,5,6 , to discover that the area is maximized at 2.64932 2.64932 when p = 5 p = 5 .

Matthew Feig
Aug 9, 2020

To add to what the others have already said, here is a Desmos visualization of the triangles in the complex plane as p p varies from 8 -8 to 8 8 . The roots and the area are calculated numerically from Cardano's version of the cubic formula.

Over integer values of p p , the area attains a maximum of approximately 2.64931954905 2.64931954905 when p = 5 \boxed{p = 5} .

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...