I have equations of the following form:

$$(p_1(y) + p_2(y) x)^2 + p_3(y) = 0$$

Where $x \in \mathbb{R}$, $y = e^x$ and the $p_i$ are polynomials with real coefficients.

Is there any way to say something about the set of roots to this equation?


  • some reasonable bound $B$ such that the zeros are $|x| \le B$
  • some bound on the number of roots

I'm trying to write a program to solve the equation numerically (analytical is clearly hopeless), but not sure how to proceed without some info about the roots.

If I knew the roots to an $n$th derivative of the equation, I could work my way "up" and find the roots of the original equation too, but not sure if that helps.

Some related problems I've looked at:

Any help/ideas would be much appreciated.

  • $\begingroup$ what is the degree of $p_3$? $\endgroup$ – uniquesolution Feb 22 at 21:50
  • $\begingroup$ It might be worthy to check the derivative - when is it eventually positive/negative. Note that $$ \frac {\text d}{\text dx} p(y) = yp'(y) $$ which doesn't have the free term. However, it's harder to solve the derivative for $x$, as you lose the square. $\endgroup$ – eudes Feb 23 at 0:19

Solving for $x$ results in

$$x = \dfrac{-p_1(y)\pm\sqrt{-p_3(y)}}{p_2(y)}=\dfrac{-p_1(\exp(x))\pm\sqrt{-p_3(\exp(x))}}{p_2(\exp(x))}=\Phi(x).$$

You might be able to make some statements about the roots if you are able to check the conditions for the Banach fixed-point theorem.

  • $\begingroup$ How about putting $\log y$ instead of $x$? And asymptotically $$ (-p_1 \pm \sqrt p_3)/p_2 \approx \text{const} \cdot y ^\wedge [\max(\deg p_1, \ 1/2 \ \deg p_3) - \deg p_2], $$ for large $y$ unless $\deg p_1 = \: 1/2 \ \deg p_3$ and leading coefficients of $p_1$ and $p_3$ are appropriate. For small $y$, consider $\log (1/y)$ and polynomials in $1/y$ (degrees will possibly change). $\endgroup$ – eudes Feb 22 at 23:55

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.