2

The following ODEs came up in a physics problem: $$m\ddot r-mr\dot \theta^2+F=0$$ $$r\ddot \theta+2\dot r \dot \theta = 0 $$ I have been told that if the initial conditions are: $$F=mr(0) \dot \theta(0)^2$$ $$\dot r(0)=0$$ (We are given $r(0)$ and $\dot \theta(0)$ and suppose $\theta(0)=0$.)

Then the only solution is: $$r=r(0)$$ $$\dot \theta = \dot \theta(0)$$ which corresponds to uniform circular motion, but I haven't seen why this is true.

So my question is: Why must the above solution be the only one when the conditions are met?

Roee
  • 43

2 Answers2

2

$$ r\ddot \theta+2\dot r \dot \theta = 0 \implies 2\frac{\dot r}{r} + \frac{\ddot\theta}{\dot\theta} = 0 \implies \ln(r^2\dot\theta) = C \implies r^2\dot\theta = C \tag{1} $$ By the condition $F=mr(0) \dot \theta(0)^2$ and the equation $$ \ddot r-r\dot \theta^2+\frac{F}{m}=0 \tag{2} $$ at $t= 0$ you can see that $\ddot r(0) = 0$. Substituting $(1)$ in $(2)$ to eliminate $\dot \theta$ yields $$ \ddot r -C^2\frac{1}{r^3}+\frac{F}{m}=0 \tag{3} $$ Now taking all derivatives of $(3)$ at $t = 0$, knowing that $\ddot r(0) = \dot r(0) = 0$, should show recursively that $\frac{d^n}{dt^n}r(0) = 0, n\ge 1$, so the Taylor expansion of $r(t)$, (if it has one,) will have only one nonvanishing term, namely the constant one $r(0) \ne 0$, (it is essential for this method that this term to be nonzero).

This result and the same idea can be use to find $\frac{d^n}{dt^n}\theta(0) = 0, n \ge 2$ from the equation ($1$) rewritten as $$ \dot \theta = \frac{C}{r^2} $$

Uniqueness

Set $\dot r = u$ and $\dot \theta = v$, then \begin{align} \dot r &= u \\ \dot \theta &= v \\ \dot u &= rv^2-\frac{F}{m} \\ \dot v &= -2\frac{u}{r} v \\ \end{align} The function $$ f:\mathbb R^4\backslash\{(0,\theta,u,v)\} \to \mathbb R^4, \qquad (r,\theta,u,v) \mapsto \left( u,v,rv^2-\frac{F}{m},-2\frac{u}{r} v \right) $$ is locally Lipschitz in every neighborhood with $r \ne 0$ because it is differentiable. So the uniqueness of the solution follows, as the "mathcounterexamples.ne" noted, from the Picard-Lindelöf Theorem.

Physor
  • 4,682
  • You do not need derivatives of (3), just insert the construction of $F$ to get $\ddot r=C^2(\frac1{r^3}-\frac1{r_0^3})$, together with $\dot r(0)=0$ this means that $r=r_0=const.$ is the solution at the local equilibrium for $r$. – Lutz Lehmann Oct 22 '20 at 12:36
  • If $r=r_0$, then $\ddot r=0$. Together with the IC $\dot r(0)=0$ this specifies one unique solution. On the other hand, $r(t)=r_0=r(0)$ is a solution. – Lutz Lehmann Oct 22 '20 at 13:17
  • @Physor Thank you for adding this solution to your answer! So if I understand it correctly, we have to assume that $r$ is analytic, and then we can conclude that the solution is unique? I would also like to see a solution that does not require $r$ to be analytic if anyone knows about such a solution... – Roee Oct 22 '20 at 14:33
  • @user800827 : The constant $F$ is constructed in such a way that it fits the constant solution for $r$. For the system that gives of course a circular solution with constant angular velocity. – Lutz Lehmann Oct 22 '20 at 15:00
  • @Physor Is it necessary that the neighborhood is compact? When It tried to look it up I ran into this link – Roee Oct 23 '20 at 14:37
  • You're right again, It is not necessary for the neighorhood to be compact, open is ok, but the Lipschitz continuity is local any way. – Physor Oct 23 '20 at 14:59
  • However I'd like to note that when I tried to find the general solution of the problem I found $$ \frac{dr}{\sqrt{C''-\frac{r^2}{C}-2\frac{F}{m}}r} = \pm dt $$ which you'd not like to integrate.

    For my solution above I didn't like invoking the Taylor series, but I couldn't find other ways in the time being

    – Physor Oct 23 '20 at 15:04
  • 1
    @Physor Thank you very much! It has been a tremendous help. – Roee Oct 23 '20 at 15:16
1

You can use at least two different approaches to prove the uniqueness of the solution.

Use Picard–Lindelöf theorem by proving that the map defining the ODE is Lipschitz continuous.

Solve directly the ODE, starting by noticing that $$r\ddot \theta+2\dot r \dot \theta = 0 $$ implies that $$r^2 \dot\theta = r^2(0) \dot\theta(0)$$ is constant. And then replacing $ \dot\theta$ in the first equation.

  • Thank you very much!! I wasn't aware that this ODE could be solved directly, could you please add the solution to your answer, as I do not have the proper background to solve it. – Roee Oct 22 '20 at 09:54