Construction of Lyapunov function given system of differential equations

57 Views Asked by At

Consider the system following system of differential equations \begin{equation} \dot{x} = y - x^{3} \\ \dot{y} = -2x -y^{3}(1-y^{2}) \end{equation} Construct a (strict) Lyapunov function of the form $\alpha x^{2} + \beta y^{2}$, where $\alpha$ and $\beta$ are real valued constants, in order to show that the origin is asymptotically stable.

I've made the following attempt at this problem.

Let $L(x, y) = \alpha x^{2} + \beta y^{2}$. Then clearly $L(x, y) > L(0, 0) = 0$, regardless of the values for $\alpha$ and $\beta$ (of course assuming they are non-zero). We continue by computing the derivative $\dot{L}$, \begin{align*} & \dot{L} = \frac{\partial L}{\partial x}\cdot \dot{x} + \frac{\partial L}{\partial y} \cdot \dot{y} \\ & = 2\alpha x(y-x^{3}) + 2\beta y(-2x-y^{3}(1-y^{2})) \\ & = 2\alpha xy - 2\alpha x^{4} - 4\beta xy - 2\beta y^{4} + 2\beta y^{6} \\ & = (2\alpha - 4\beta)xy + 2\beta(-y^{4} + y^{6}) - 2\alpha x^{4} \end{align*}

We want this derivative to be strictly negative, as such I either wanted to let $2\alpha - 4\beta = 0$, or perhaps choose a region around the origin for which the derivative is strictly negative and proceed to find possible values for $\alpha$ and $\beta$.

I've made an attempt at the former, $2\alpha - 4\beta = 0$ implies $2\alpha = 4\beta$. Inserting this into the derivative gives us the following: $\dot{L} = 2\beta(-y^{4} + y^{6}) - 8\alpha x^{4}$.

I'm unsure of the continuation of my attempt from this point, clearly $y^{6} > y^{4}$ except when $y \in [0, 1]$. As such I considered $y$ outside this region, which would mean we'd want a negative value for $\beta$. But this would make $\alpha$ positive and we wouldn't have certainty over the sign of the derivative.

Would anyone be able to help me with this problem, or have any tips on how to continue?

1

There are 1 best solutions below

2
On BEST ANSWER

The question is only asking about showing asymptotic stability of the origin and not global asymptotic stability. And your analysis for $\alpha=2\beta>0$ is sufficient for proving this, i.e. $L(x,y)>0\ \forall\ x,y\neq0$ and $\dot{L}(x,y)<0\ \forall\ x,y\neq0$ and $-1<y<1$.

This can also be verified by looking at the stream plot: Stream plot nonlinear ODE

It can be noted that the basin of attraction of the origin is not equal to the values of $(x,y)$ for which $\dot{L}(x,y)\leq0$. Instead, a lower bound of the border of the basin of attraction is given by the level set of the Lyapunov function, i.e. $L(x,y)=c$, such that for all points in that set satisfy $\dot{L}(x,y)\leq0$.