3
$\begingroup$

Given the periodic potential Hamiltonian $H=\frac{p^2}{2} - \omega_0^2 \cos(q)$ I would like to show that near the separatrix the period has this behavior: $T(E)\sim |\log(\delta E)|$ with $\delta E=|E-\omega_0^2|$.

More generally given an Hamiltonian system of the form $H=\frac{p^2}{2} + V(q)$ with $V''(q^*)\ne 0$ for a non stable fixed point, I would like to show that near the separatrix we get the same kind of law.

I could prove that $p$ is a solution on the separatrix and found an infinite period. Then I tried doing different development of $E$ to first order and second order but didn't get any result. Do you have any idea on how to do that for the first case and then maybe the general case?

$\endgroup$
4
  • 1
    $\begingroup$ Do you really mean $V'''(q^*) \neq 0$, or do you mean $V''(q^*) \neq 0$? Because the third derivative of $\cos(q)$ is zero at its non-stable fixed points. $\endgroup$ Commented Jan 4, 2021 at 0:31
  • 1
    $\begingroup$ Moreover what’s the meaning of a period near a separatrix. Do you mean the initial condition is near the separatrix? Do you want the time to reach the unstable point? $\endgroup$ Commented Jan 4, 2021 at 4:50
  • $\begingroup$ Yes @MichaelSeifert exactly I meant the second derivative $V''(q*)\ne0$ but I copied a typo. Sorry $\endgroup$ Commented Jan 4, 2021 at 15:50
  • $\begingroup$ Yes @ZeroTheHero. Startin very close to the separatrix, but at different small distance $\delta E$ that is the variable $\endgroup$ Commented Jan 4, 2021 at 15:52

1 Answer 1

2
$\begingroup$

The period is calculated by the integral $$ \sqrt{2}\int_0^{2\pi} \frac{d q}{ \sqrt{\omega_0 ^2 \cos (q)+E}} $$ which can be represented by special functions. After applying a replacement $E\to \delta E+\omega^2_0$, you need to expand this integral around separatrix $\delta E=0$, the leading term is $-2\omega_0^{-1}\ln(\delta E)$. Thus the leading term of period around $\delta E=0$ will be $T\sim-2\omega_0^{-1}\ln(\delta E)$.

$\endgroup$
1
  • $\begingroup$ Alright I was on the right track then, but couldn't manage to do the computation properly to the end. Thanks !! $\endgroup$ Commented Jan 4, 2021 at 16:10

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