Linear Stability Analysis question/clarification

127 Views Asked by At

I've been stuck on this problem for the past few days, and was wondering if someone could clarify part ii) of the following question. In order to understand the question, I will also include my solution for part i).

The question is as follows:

Part i) For the dynamical system below, derive a linear stability analysis:

$$\qquad \qquad \qquad \qquad \qquad \frac{dx}{dt}=f(x) \qquad \qquad \qquad \qquad \qquad \qquad (1)$$

Additional information: The time-independent variable $\bar{x}$ denotes a fixed point for the dynamical system and thus $f(\bar{x})=0$.

Part ii) Repeat the derivation in part i) and continue the Taylor expansion to 4th order, for the case of $f'(\bar{x})=f''(\bar{x})=f'''(\bar{x})=0$. What is the stability condition of $f$ in this case?

Hint: look for real solutions in the ODE for the small perturbation function, $\epsilon(t)$, and consider the sign of both the initial condition, $\epsilon(0)$, and the fourth derivative, $f^{(4)}(\bar{x})$.


My solution to part i):

If we consider a small perturbation $\epsilon(t)$. Then let: $$\qquad \qquad \qquad \qquad \qquad x(t)=\bar{x}+\epsilon(t) \qquad \qquad \qquad \qquad \qquad \qquad (2)$$

Taking the derivative of equation $(2)$, we find:

$$\qquad \qquad \qquad \qquad \qquad \frac{dx}{dt}=\frac{d\epsilon}{dt} \qquad \text{(since $\bar{x}$ is time independent)} \qquad (3)$$

Thus substituting equation $(2)$ & $(3)$ into equation $(1)$, we have:

$$\frac{dx}{dt}=f(\bar{x}+\epsilon(t))$$ $$\quad \Rightarrow \frac{dx}{dt}=\frac{d\epsilon}{dt}= f(\bar{x}+\epsilon(t))$$ $$\qquad\qquad\qquad\qquad\quad\quad\Rightarrow \frac{d\epsilon}{dt}=f(\bar{x}+\epsilon(t)) \qquad\qquad\qquad\qquad\qquad (4) $$

Now taking the Taylor series expansion of RHS of equation $(4)$ about $x=\bar{x}$ we have:

$$f(x)=f(\bar{x}+\epsilon(t))=f(\bar{x})+f'(\bar{x})((\epsilon+\bar{x})-\bar{x})+\frac{f''(\bar{x})((\epsilon+\bar{x})-\bar{x})^2}{2!}+\frac{f'''(\bar{x})((\epsilon+\bar{x})-\bar{x})^3}{3!}+... \qquad \qquad (5)$$ $$\Rightarrow f(x)=f(\bar{x})+f'(\bar{x})(\epsilon)+O(\epsilon^2)$$ $$\qquad\qquad\qquad\qquad\qquad \Rightarrow f(x)=f'(\bar{x})(\epsilon) \qquad \text{(since $f(\bar{x})=0$ and $O(\epsilon^2)$ becomes negligible)}$$ $$\Rightarrow\frac{d\epsilon}{dt}=f'(\bar{x})(\epsilon)$$

Thus we are left with a separable equation, solving for $\epsilon(t)$ as follows:

$$\frac{d\epsilon}{dt}=f'(\bar{x})(\epsilon)$$ $$\Rightarrow\frac{d\epsilon}{\epsilon}=f'(\bar{x})dt$$ $$\Rightarrow log(\epsilon)=f'(\bar{x})t+c$$ $$\Rightarrow \epsilon(t)=e^{f'(\bar{x})t+c}$$ $$\qquad\qquad\qquad\qquad\qquad\Rightarrow \epsilon(t)=\epsilon_0e^{f'(\bar{x})t}\qquad\qquad\qquad\qquad\qquad (6)$$

Thus from this result, we can find the stability of our fixed points using the following properties:

  • If $f'(\bar{x})<0$ then $\epsilon(t)$ decays to zero, and thus we approach our fixed point $$\Rightarrow \text{Our fixed point $f(\bar{x})$ is stable} $$

  • If $f'(\bar{x})>0$ then $\epsilon(t)$ grows exponentially, and thus we move away from our fixed point $$\Rightarrow \text{Our fixed point $f(\bar{x})$ is unstable} $$

  • If $f'(\bar{x})=0$ linear stability analysis cannot determine the stability of the fixed point, requiring other techniques (i.e. graphical approach)

My solution to part ii) so far:

Now if we expand the Taylor expansion to fourth order as in equation $(5)$ then we have:

$$f(x)=f(\bar{x})+f'(\bar{x})(\epsilon)+\frac{f''(\bar{x})(\epsilon^2)}{2!}+\frac{f'''(\bar{x})(\epsilon^3)}{3!}+\frac{f^{(4)}(\bar{x})(\epsilon^4)}{4!}+O(\epsilon^5)$$ $$\Rightarrow \frac{d\epsilon}{dt}=f(\bar{x})+f'(\bar{x})(\epsilon)+\frac{f''(\bar{x})(\epsilon^2)}{2!}+\frac{f'''(\bar{x})(\epsilon^3)}{3!}+\frac{f^{(4)}(\bar{x})(\epsilon^4)}{4!}+O(\epsilon^5)$$ $$\qquad\qquad\Rightarrow \frac{d\epsilon}{dt}=\frac{f^{(4)}(\bar{x})(\epsilon^4)}{4!} \qquad\qquad \text{(since $f(\bar{x})=f'(\bar{x})=f''(\bar{x})=f'''(\bar{x})=0$ and $O(\epsilon^5)$ becomes negligible)} $$

Thus we are left with a nonlinear equation:

$$\qquad\qquad\qquad\qquad\qquad\frac{d\epsilon}{dt}-\frac{f^{(4)}(\bar{x})(\epsilon^4)}{24}=0\qquad\qquad\qquad\qquad\qquad (7)$$

This is where I'm stuck on how to solve this.


My question/clarification for part ii)

I was wondering if someone could point me in the right direction as to how to solve the above nonlinear equation. Do I use my solution for $\epsilon(t)$ in equation $(6)$ somewhere? Or am I allowed to substitute $\frac{dx}{dt}$ instead of $\frac{d\epsilon}{dt}$, making equation $(7)$ a linear ODE?

Additionally, in the hint they say use the real solutions of the ODE for the perturbation function $\epsilon(t)$. Do they mean the solution I found for $\epsilon(t)$ in equation $(6)$?

Finally, could someone explain why constants of integration can be ignored or subsumed into $\epsilon_0$ such as in equation $(6)$? If we aren't specified an initial condition for $\epsilon(t)$, do we just define the initial condition as $\epsilon(t)=\epsilon_0$?

I'd appreciate any clarifications/suggestions on where to go from here.

Thanks