Elliptic limit cycle

302 Views Asked by At

The following equations,

\begin{equation} \begin{split} \dot{F} &= -C + \frac{F}{\sqrt{F^2 + C^2}}\left(\alpha -\left(\frac{F}{a}\right)^2 - \left(\frac{C}{b}\right)^2\right),\\ \dot{C} &= F + \frac{C}{\sqrt{F^2+ C^2}}\left(\alpha -\left(\frac{F}{a}\right)^2 - \left(\frac{C}{b}\right)^2\right), \label{eq:LiCy} \end{split} \end{equation}

illustrates a elliptic limit cycle.

$F = F(t)\\ C = C(t) \\ a \rightarrow semi-major axis \\ b \rightarrow semi-minor axis \\ \alpha \, \text{is a scaling constant, and is normally represented as 1 in the equation of an ellipse}$

We would like to represent the limit cycle through the functions $\dot{r}(t)$ and $\dot{\theta}(t)$. Where the solution is found by solving the following equations,

$\dot{r}(t)\, \text{from} \\ x\dot{x} + y\dot{y}$

and

$\dot{\theta}(t) \, \text{from}\\ x\dot{y} - y \dot{x}$

The example we tried to follow, gave the following polar coordinates for the circle,

$x(t) = r(t)\cos(\theta(t))\quad \text{and} \quad y(t)=r(t)\sin(\theta(t))$ where $r(t) = \sqrt{x^2 + y^2}$ and we found $\dot{r}= \alpha -r^2$ and $\dot{\theta}(t)=1$.

Therefore we tried to implement new polar coordinates for an ellipse, where $x(t) = a\cdot r(t)\cos(\theta(t))\quad \text{and} \quad y(t)=b\cdot r(t)\sin(\theta(t))$

which did not work out as we had hoped. We are wondering where it went wrong? And how we could illustrate this elliptic limit cycle by using the given equations?

1

There are 1 best solutions below

0
On

In polar coordinates $r \cos(\theta) = F$, $r \sin(\theta) = C$, the equations become $$ \eqalign{\dot{\theta} &= 1\cr \dot{r} &= 1 + \dfrac{(a^2-1) b^2 \cos^2(\theta) - a^2}{a^2 b^2} r^2\cr}$$ Modulo a change of origin, we can take $\theta = t$. The equation for $r$ then has an integrating factor $$ \mu(t) = \exp\left(-{\frac { \left( {a}^{2}{b}^{2}-2\,{a}^{2}-{b}^{2} \right) t}{2{a} ^{2}{b}^{2}}}-{\frac { \left( a^2-1 \right) \sin \left( 2\,t \right) }{4{a}^{2}}} \right)$$ and we find that $ \mu(t) r(t) - \int \mu(t) \; dt$ is constant. Thus (since $\mu(0)=1$)

$$ \mu(t) r(t) - \int_0^t \mu(s)\; ds = r(0) $$

In order to have a periodic solution with period $2\pi$, what we need is $$ (\mu(2\pi) -1) r(0) = \int_0^{2\pi} \mu(s)\; ds $$

If $r^*$ is the value of $r(0)$ that makes this true, we have $\mu(2\pi) (r(2\pi) - r^*) = r(0) - r^*$. Thus the periodic solution will be unstable if $\mu(2\pi) < 1$, i.e. $a^2 b^2 - 2 a^2 - b^2 > 0$, and stable if $a^2 b^2 - 2 a^2 - b^2 < 0$.