2
$\begingroup$

The inequality $$\frac{1}{2}(x^2+y^2)+x^2y-\frac{1}{3}y^3 \lt C$$ produces a somewhat triangular shape near the origin for $C \le \frac 16$, where $C$ is a constant.


Example when $C = 1/10$:

enter image description here


I am trying to find a closed form expression for the $(x,y)$ coordinate of the top of the "triangle" in terms of the constant $C$.

$y = 0$ when $C = 0$ and $y=1$ when $C = \frac 16$.

Background: This is the Henon-Heiles potential function

  • 0
    $(x,y)$ are solutions of a Hamiltonian system which is non integrable. Does it means that even on a constant energy level $V$, you can't find explicit expressions of $x$ and $y$ ?2017-02-16
  • 0
    to compute the top, you may consider the following problem $\max_{x,y\in E} y$ where $E=\{\frac{1}{2}(x^2+y^2)+x^2y-\frac{1}{3}y^3 \lt C\}$ and use duality theory to compute the optimal point $(x^\star,y^\star)$2017-02-16
  • 0
    The Hamiltonian describes the motion of a particle inside this potential well but I'm only interested in a 2D cross section of the well and it's extrema, which is a separate problem from the trajectory of the particle.2017-02-16

1 Answers 1

0

Assuming that the maximum can be found when $x=0$ and using the formulas of the "casus irreducibilis" (https://en.wikipedia.org/wiki/Casus_irreducibilis) I get $$ y=\frac{1}{2}-\cos\left(\frac{1}{3} \arccos\left(1-12C\right)+\frac{\pi}{3}\right) $$ The details:

Setting $x=0$ results in $y^3-\frac{3}{2}y^2+3C=0$. We substitute $y=t+\frac{1}{2}$ and get the "depressed" polynomial $$ t^3-\frac{3}{4}t-\frac{1}{4}+3C=0 $$ We set $p=-\frac{3}{4}$ and $q=-\frac{1}{4}+3C$. It can easily be shown that $$ \left(\frac{q}{2}\right)^2+\left(\frac{p}{3}\right)^3 \leq 0\;\; \mbox{for}\;\;0\leq C\leq \frac{1}{6} $$ so the conditions for using the trigonometric method of the "casus irreducibilis" are fulfilled. The formula is as follows: $$ t_k = 2\sqrt{-\frac{p}{3}}\cos\left(\frac{1}{3}\arccos\left(\frac{3q}{2p}\sqrt{-\frac{3}{p}}\right)-k\frac{2\pi}{3}\right)\;\;\mbox{for}\;\;k=0,1,2. $$ We are lucky, because the coefficients of the depressed polynomial allow a lot of simplifications in this formula.

Finally, we only have to consider our initial substitution $y=t+\frac{1}{2}$ and find the appropriate $k$ for the root we are looking for. It turns out that $k=1$ is the correct choice. Additionally, I have used $\cos(\alpha) = -\cos(\alpha+\pi)$ to replace the $-\frac{2\pi}{3}$ with $\frac{\pi}{3}$, which is just a matter of personal preference.

  • 0
    I can tell this is correct by the shape of the curve. Did you use [Cardano's solution](https://en.wikipedia.org/wiki/Casus_irreducibilis#Cardano.27s_solution) to get this? If so, how did you reduce the long root terms into cos and arccos? I expected the solution to have imaginary/complex numbers. Is there an Euler's formula trick in there? Thanks for your help.2017-02-17
  • 0
    I have added the details to my answer,2017-02-17