2
$\begingroup$

I have a polynomial

$$ \gamma_k(x) = x^{2k+2} - m x^{2k} - n x^k - 1 $$

where $m$ and $n$ are positive real numbers and $k$ is a positive even integer. How can I prove $\gamma_k(x)$ has distinct roots for any $m$,$n$ and $k$ ?

$\endgroup$
  • 3
    $\begingroup$ Have you tried finding $\gcd(\gamma_k,\gamma_k')$? That would be my first approach, as a root is multiple iff it also is a root of the derivative. Also, i think you should swap $k$ for $2\ell$ so you don't have to go around remembering all the time that $k$ is even. $\endgroup$ – Arthur Dec 14 '17 at 18:04
  • 1
    $\begingroup$ Does "roots" here refer only to real roots, or to complex roots, too? $\endgroup$ – Travis Willse Dec 14 '17 at 18:05
  • $\begingroup$ @Arthur of course I tried to find $gcd(\gamma_k,\gamma_k ')$ but couldn't achieve. If you prove it with your way, could you please share it here? $\endgroup$ – drxy Dec 14 '17 at 18:25
  • $\begingroup$ @Travis both of them. $\endgroup$ – drxy Dec 14 '17 at 18:25
  • 1
    $\begingroup$ That's not "of course" (as evidenced by my comment getting three up votes). It's cool that you did, but you should've said so in your post. $\endgroup$ – Arthur Dec 14 '17 at 18:30
3
$\begingroup$

This claim is false over $\Bbb C$. For example, for $k = 2$ we can generate polynomials of the form $x^6 - m x^4 - n x^2 - 1$ with repeated roots by taking $$\left(x - \frac{1}{\lambda}\right) \left(x + \frac{1}{\lambda}\right)(x^2 + \lambda)^2 = x^6 - \left(\frac{1}{\lambda^2} - 2 \lambda\right) x^4 - \left(\frac{2}{\lambda^2} - \lambda\right) x^2 - 1 .$$ It can be checked that that the coefficients of $x^4, x^2$ are negative as required for $0 < \lambda < \frac{1}{\sqrt[3]{2}}$. For these values, the repeated roots are (nonreal) imaginary: $\pm i \sqrt\lambda$.

(On the other hand, a deleted answer shows that for any even $k$ the polynomial has a single positive real root $\mu$ as a straightforward application of Descartes' Rule of Signs, and since the polynomial is even, a single negative real root, $-\mu$.)

$\endgroup$
  • $\begingroup$ That's it @Travis. Thank you. $\endgroup$ – drxy Dec 14 '17 at 20:03

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.