Checking the stability of equilibrium point

176 Views Asked by At

Is equilibrium point $x = 0$ stable for $\ddot{x}+\dot{x}=-x\sin x$?

It is a case of a bigger problem, where we need to check the stability of critical points $x \in [-\frac \pi 2, \frac \pi 2]$ for $\ddot{x}+\dot{x} = x(a-1)(\sin x-a)$. I've done with all other situations, but can not do anything with this case. Obviously, we need to say $\dot{x}=y$, $\dot{y} =-y-x\sin x$. Linear approximation doesn't help here. So, it seems we need to find Lyapunov function $V$ in order to use Lyapunov theorem or Chetaev theorem. I've tried some, but I didn't manage to find any. I think that we shoud in some way modify $V = \frac{y^2}2 - x\cos x + \sin x$, because $\dot{V} = -y^2$.

2

There are 2 best solutions below

2
On

If $\mathbf{a = 0}$

The nonlinear system

$$ \left[\matrix{\dot{x} \cr \dot{y}}\right] = \left[\matrix{y \cr - x \sin(x) - y}\right] $$

has multiple equilibrium points at $x = \pi n$, where $n \in \mathbb{Z}$.

Perform a linearization on the nonlinear system and obtain

$$ \left[\matrix{\dot{x} \cr \dot{y}}\right] = \left[\matrix{0 & 1 \cr - x \cos(x) - \sin(x) & -1}\right] \left[\matrix{x \cr y}\right] .$$

If the system is linearized at $x = 0$, then we get

$$ \left[\matrix{\dot{x} \cr \dot{y}}\right] = \left[\matrix{0 & 1 \cr 0 & -1}\right] \left[\matrix{x \cr y}\right] $$

$$ \mathbf{\dot{x}} = \mathbf{A} \mathbf{x} $$

and the eigenvalues of $\mathbf{A}$ are obtained by solving for the roots of the characteristic equation

$$ |\lambda \mathbf{I} - \mathbf{A}| = 0 $$

$$ \lambda_{1} = -1, \quad \lambda_{2} = 0 .$$

Since one eigenvalue is at the origin, the response will increase with time for any $\dot{x}(0) \neq 0$, and we conclude that the equilibrium $(x,\dot{x}) = (0,0)$ is unstable. Thus, it is not surprising that your Lyapunov function doesn't work when it is applied to an unstable point.

On the other hand, if the system is linearized at $x = -\pi$, then we get

$$ \left[\matrix{\dot{x} \cr \dot{y}}\right] = \left[\matrix{0 & 1 \cr -\pi & -1}\right] \left[\matrix{x \cr y}\right] $$

and the eigenvalues are $\lambda_{1,2} = - 0.5 \pm 1.70047i$. Since the eigenvalues have negative real parts, we conclude that the point $(x,\dot{x}) = (-\pi,0)$ is a stable equilibrium (as shown in the phase portrait by Hans).

If $\mathbf{a > 1}$

Moving on to the general problem, the system can be written as

$$ \left[\matrix{\dot{x} \cr \dot{y}}\right] = \left[\matrix{y \cr - f(a,x) - y}\right] $$

where

$$ f(a,x) = x \left(a - 1\right) \left(a - \sin(x)\right) ,$$

we can perform analysis and find that it satisfies $f(0) = 0$ and $x f(a,x) > 0$ for all $x \neq 0$ if $a > 1$ is selected. Although multiple equilibrium points exist in the trigonometric function, only one real equilibrium $(x,y) = (0,0)$ exists.

Consider the Lyapunov function candidate

$$ V(x, y) = \frac{1}{2} \left[\matrix{x & y}\right] \left[\matrix{1 & 1 \cr 1 & 2}\right] \left[\matrix{x \cr y}\right] + 2 \int_{0}^{x} f(\chi) \; d\chi > 0 \quad \text{for} \quad x \neq 0, y \neq 0 .$$

The derivative of $V$ along the trajectories of the system is given by

$$ \dot{V}(x,y) = - x f(a,x) - y^{2} $$

and it satisfies

$$ \dot{V}(x,y) < 0 \quad \text{for} \quad x \neq 0, y \neq 0 .$$

Since there is only one real equilibrium $(x,y) = (0,0)$ for $a > 1$, then it can be concluded that the origin is asymptotically stable.

The streamplot for $a = 1.8$ is shown below.

enter image description here

0
On

It's unstable, as can be seen from the phase portrait for the system $(\dot x,\dot y)=(y,-y-x \sin x)$:

The red curve is the $x$-nullcline $y=0$, and the orange curve is the $y$-nullcline $y=-x \sin x$. In the blue region, the solution curves must go downwards and to the left, and they must cross the orange part of the boundary going straight left, so if a solution starts inside that region (which it can do arbitrarily close to the origin), it can only exit on the left side, and over there it's no longer close to the origin.