2

I have a quick question about the Poincare-Bendixson Theorem in 2D. The way I understand the theorem is that, loosely, if we have some region $R$ that is bounded, connected, and closed, and contains no fixed points, then if any trajectory that enters $R$ remains in $R$ as $t\rightarrow\infty$ then $R$ must contain a limit cycle.

My question is about the applicability of the theorem; I have recently done a problem which had an unstable fixed point at the origin, a stable limit cycle at $r=\sqrt{2}$, and an unstable limit cycle at $r=\sqrt{7}$

Using the P-B Theorem in an annulus between $r=1$ and $r=2$ I can easily show the existence of the first limit cycle, since $\dot{r}$ changes sign at either boundary and therefore any trajectories which enter must remain within the annulus. So far so good.

However, when I came to analyse the second limit cycle, because it is unstable I found that no trajectories entered the region, but there was a net flow out of the annulus from both sides.

My question is two fold:

1) Is the P-B Theorem not applicable in this case? I wondered if it was a technicality that because no trajectories entered, there were no trajectories to leave HAVING ALREADY entered, but I feel like the answer to this is probably an emphatic no. It may also be that my understanding of the Theorem isn't rigorous or expansive enough, in which case I would love to know what I'm missing.

2) I deduced that there must be a limit cycle in the region anyways, since $\dot{r}$ changed signs in a similar way to the previous case, just in the "opposite order", indicating an unstable limit cycle.

The problem is that in the problem itself $\dot{r}$ was easily factorised so I knew where the limit cycles were anyways, which sort of made me think this was a case of "obvious in hindsight".

enter image description here

arcturus7
  • 183
  • Can you post the original problem so we can see what you're working with? – DMcMor May 15 '17 at 21:43
  • Sure thing - I'll dig it out now! – arcturus7 May 15 '17 at 21:43
  • Speaking about 1) : you can reverse the flow (change of time $t \mapsto -t$), check the vector field on the boundary after that and Poincaré-Bendixson becomes applicable again. But limit cycle is a limit cycle no matter in which direction time is changing, so the original system has this limit cycle too. – Evgeny May 16 '17 at 05:30
  • I had a suspicion this might be the case - so really in truth the expression of the Theorem should be that any flows that enter the system and remain in the system for t going to plus or minus infinity, rather than infinity, so that you catch the alpha limit sets as well. Thank you for your help! – arcturus7 May 16 '17 at 05:46
  • Well, many texts have a phrase that every statement that is true for attractors has a "symmetric" statement that is true for repellors that is obtained by reversing the time in the system (and the same for $\alpha$- and $\omega$-limit sets). Some have this phrase explicit, but often it goes implicit :) – Evgeny May 16 '17 at 06:23
  • It was definitely implicit in my lecture notes in that case... haha. Thank you for your help :) – arcturus7 May 16 '17 at 07:01

0 Answers0