Can the polynomial $$y^2(1-x)^4+x^2(1-y)^4=xy(1-x)^2(1-y)^2(n^2+2)$$ be solved over the roots of unity for an integer $n\ge 3$? Known solutions are $$n=1:\quad x=\zeta_5\quad y=\zeta_5^2$$ $$n=2:\quad x=\zeta_8\quad y=\zeta_8^3$$

As the problem is posed, $x$ and $y$ do not have to necessarily be roots of unity of the same degree. But as far as I know, that may be a necessary condition for solvability.

This is a reformulation of an earlier question Rational Trig Solutions for $n\ge 3$ I was advised to formulate the question as it is presented here because there are apparently strong results on what polynomials can be solved over the roots of unity. There are a few papers on such solutions that pop up on a naif google search, but all have been too dense for me to understand.


Performing a brute force search for solutions of the form

$$x=\zeta_j \quad y=\zeta_j^k$$

reveals that real solutions for $n$ abound, but integer real solutions for $n$ are not likely for solutions of this form.

In addition to your solutions, I found these solutions $$n = 1: \quad x=\zeta_5 \quad y=\zeta_5^3$$ $$n = 2: \quad x=\zeta_8 \quad y=\zeta_8^5$$ $$n = 9.0009: \quad x=\zeta_{33} \quad y=\zeta_{33}^{11} \quad or \quad y=\zeta_{33}^{22}$$ $$n = 10.007: \quad x=\zeta_{34} \quad y=\zeta_{34}^{13} \quad or \quad y=\zeta_{34}^{21}$$

Here is the simple MatLab/Octave script I used to do the brute force search:

for k = [2:40]
  x = exp(1i*2*pi/k);
  for j = [2:k]
    y = x^j;
    n = sqrt(((y*(1-x)^2-x*(1-y)^2)^2)/(y*(1-x)^2*x*(1-y)^2))

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.