# Stability of Orbits

#### CAF123

Gold Member
1. Homework Statement
A particle of mass m moves under a central force $\mathbf{F}(\mathbf{r}) = -\frac{\mu}{r^2} e^{-kr} \hat{r}$.
The particle undergoes motion in a circle if $h^2 = (a\mu/m)e^{-ka}$. I have shown that if $u(\theta) = 1/r,$ then the orbit eqn for $u(\theta)$ becomes $$u'' + u = \frac{1}{a}e^{-k(1/u - a)}.$$ This is a show that so I know it is correct.
I can't make further progress with the following:
Assume that the motion is slightly disturbed. Use the orbit eqn for $u(\theta)$ to show that in the perturbed orbit, the furthest distance from the centre occurs with a period of $2\pi/w$ where $w = \sqrt{1-ka}$

3. The Attempt at a Solution

Assume the perturbation, p(t), is small. That is p(t=0) << a. Let r = a + p(t) => u = 1/(a+p).
Differentiate twice wrt theta using the chain rule and I get $$u'' = -\frac{\ddot{p}}{a\mu e^{-ka}/m u^2}.$$ Sub into the orbit eqn: $$-\frac{\ddot{p} (a + p)^2 m}{a\mu e^{-ka}} + \frac{1}{a+p} = \frac{1}{a} e^{-k(1/(a+p) - a)}$$

Now Taylor expand the $(a+p)^2$, $(a+p)^{-1}$ and exp terms. Doing this, putting everything together, I get that $$\ddot{p} + \frac{\mu e^{-ka}}{ma^4} p = \frac{e^{1/a}}{ma^2}$$

I was thinking my next step would be to solve this for $p$, find the maximum of this (i.e the furthest distance it is perturbed) and then add this maximum to a. Does it sound good? (Can anyone confirm what I got above before I continue - it looks somewhat strange since when I find the solution, the arguments of sin/cos will contain an exponential). Actually something has had to have gone wrong since there will be a dimension associated with the arg of the trig terms.

MAny thanks.

Last edited:
Related Introductory Physics Homework Help News on Phys.org

#### voko

Differentiate twice wrt theta using the chain rule and I get $$u'' = -\frac{\ddot{p}}{a\mu e^{-ka}/m u^2}.$$
Seeing how you got that would be helpful. I am not saying this is incorrect, though.

Sub into the orbit eqn: $$-\frac{\ddot{p} (a + p)^2 m}{a\mu e^{-ka}} + \frac{1}{a+p} = \frac{1}{a} e^{-k(1/(a+p) - a)}$$
Not so fast. What does $$u'' = -\frac{\ddot{p}}{a\mu e^{-ka}/m u^2}$$ really mean? Is that $$u'' = -\frac{\ddot{p}}{\frac {a\mu e^{-ka}} {m u^2}}$$? If so, then $$-\frac{\ddot{p}}{a\mu e^{-ka}/m u^2} = -\frac{m u^2\ddot{p}}{a\mu e^{-ka}} = -\frac{\ddot{p} m}{a(a + p)^2\mu e^{-ka}}$$

#### CAF123

Gold Member
Seeing how you got that would be helpful. I am not saying this is incorrect, though.
First I computed u'. $$u' = -(a+p)^{-2} \dot{p} \frac{dt}{d\theta} = -\frac{\dot{p}}{(a+p)^2} \frac{1}{\sqrt{\frac{a\mu e^{-ka}}{m}}} u^2,$$ where $$\dot{\theta} = h/r^2 = \sqrt{\frac{a\mu e^{-ka}}{m}} \cdot \frac{1}{r^2}$$.

Then u'' = d/dθ (u') = $$-\frac{1}{\sqrt{\frac{a\mu e^{-ka}}{m}}} \frac{d}{d\theta} \frac{\dot{p}}{(a+p)^2} (a+p)^2 = -\frac{1}{\sqrt{\frac{a\mu e^{-ka}}{m}}} \frac{d}{dt} \dot{p} \frac{dt}{d\theta} = -\frac{\ddot{p}m}{a\mu e^{-ka} u^2}$$

I made a typo in the OP, it should be m/u^2 not /mu^2

Edit: is my method correct?

Last edited:

#### haruspex

Homework Helper
Gold Member
2018 Award
$$e^{-k(1/(a+p) - a)}$$
Something wrong in the exponent. Should it just be -k(1/u - a) = -kp?

#### voko

How did you go from

$$u' = -(a+p)^{-2} \dot{p} \frac{dt}{d\theta} = -\frac{\dot{p}}{(a+p)^2} \frac{1}{\sqrt{\frac{a\mu e^{-ka}}{m}}} u^2,$$

to

$$u'' = -\frac{1}{\sqrt{\frac{a\mu e^{-ka}}{m}}} \frac{d}{d\theta} \frac{\dot{p}}{(a+p)^2} (a+p)^2$$

given that

$$u^2 = \frac 1 {(a + p)^2}$$

?

#### CAF123

Gold Member
How did you go from

$$u' = -(a+p)^{-2} \dot{p} \frac{dt}{d\theta} = -\frac{\dot{p}}{(a+p)^2} \frac{1}{\sqrt{\frac{a\mu e^{-ka}}{m}}} \mathbf{u^2},$$
For some reason, I had a few initial problems writing the above out in latex, so a typo has crept in. u^2 should be on the denominator.

to

$$u'' = -\frac{1}{\sqrt{\frac{a\mu e^{-ka}}{m}}} \frac{d}{d\theta} \frac{\dot{p}}{(a+p)^2} (a+p)^2$$

given that

$$u^2 = \frac 1 {(a + p)^2}$$

?

#### CAF123

Gold Member
Something wrong in the exponent. Should it just be -k(1/u - a) = -kp?
Indeed it should.

#### voko

I think this $\frac{\mu e^{-ka}}{ma^4} p$ should be $\frac{\mu e^{-ka}}{ma^3} p$. Then the units seem OK. What equation do you get when you fix the exponent and linearize?

#### CAF123

Gold Member
I think this $\frac{\mu e^{-ka}}{ma^4} p$ should be $\frac{\mu e^{-ka}}{ma^3} p$. Then the units seem OK. What equation do you get when you fix the exponent and linearize?
That is exactly the case. The final equation I attain after Taylor expanding is $$\ddot{p} + \frac{\mu e^{-ka}}{ma^3} p = \frac{k\mu e^{-ka}}{ma^2}p,$$ which I can subsequently write as $$\ddot{p} + p(\frac{\mu e^{-ka}}{ma^2} \left(\frac{1}{a} - k\right) = 0$$ Here I identify the harmonic solution, with $$w^2 = (\frac{\mu e^{-ka}}{ma^2} \left(\frac{1}{a} - k\right)$$

This can further be simplified using $h^2$ given in the problem statement. Hence $w^2 = \frac{h^2}{a^4} (1-ka)$ which is incorrect by a factor of $h/a^2$.

Last edited:

#### voko

$\exp -kp$, linearized, should have a constant term. Which is actually quite bizarre, because that means p = 0 is an equilibrium point for the perturbed system.

#### CAF123

Gold Member
$\exp -kp$, linearized, should have a constant term. Which is actually quite bizarre, because that means p = 0 is an equilibrium point for the perturbed system.
When I linearised $e^{-kp}$ I got $1 - kp$. Is this not correct?

#### voko

Yes, this is correct. But that means that the equation is $\ddot{p} + \lambda p = c$, which means that $p = c/\lambda$ is the equilibrium point. Not $p = 0$ as expected.

#### CAF123

Gold Member
Yes, this is correct. But that means that the equation is $\ddot{p} + \lambda p = c$, which means that $p = c/\lambda$ is the equilibrium point. Not $p = 0$ as expected.
I brought everything together on the LHS to get the familiar harmonic equation and then identified $w^2$. Is my final eqn correct that I posted in my last two posts? When I rearrange I get 0 on the RHS and so p = 0 would correspond to the equibilrium soln, no?

#### voko

You got the equation in the form $\ddot{p} + \lambda p = 0$. How is that possible if the original RHS had a constant term?

#### CAF123

Gold Member
The eqn I got before rearranging was:
$$\ddot{p} + \frac{\mu e^{-ka}}{ma^3} p = \frac{k\mu e^{-ka}}{ma^2}p,$$ which I can subsequently write as $$\ddot{p} + p(\frac{\mu e^{-ka}}{ma^2} \left(\frac{1}{a} - k\right) = 0$$
You got the equation in the form $\ddot{p} + \lambda p = 0$. How is that possible if the original RHS had a constant term?
The RHS depends on p, so when I bring it over I have the form $\ddot{p} + \lambda p = 0$. (Unless ofcourse I made an error). I am just off to a tutorial now.

#### voko

In #11 you said that the RHS should have a constant term. In #15, however, there is no such term.

#### CAF123

Gold Member
In #11 you said that the RHS should have a constant term. In #15, however, there is no such term.
I had the form $$-\frac{\ddot{p} (a+p)^2m}{a\mu e^{-ka}} + \frac{1}{a+p} = \frac{1}{a} e^{-kp}$$

Now linearise: $$-\frac{m}{a\mu e^{-ka}} \ddot{p} \left[a^2(1 + 2\frac{p}{a}) \right] + \frac{1}{a} (1-\frac{p}{a}) = \frac{1}{a}e^{-kp}$$

When I bring that single 1/a term on the LHS over to the RHS, I get $\frac{1}{a}(e^{-kp} - 1)$. Now expand the exp and the 1's cancel, leaving only a term in p.

Edit: I neglected the term containing $\ddot{p}p$

Last edited:

#### voko

I can't see any mistake in your derivation (assuming the original equations for u and h are correct).

Unless I am mistaken, $\frac {\mu e^{-ka}} {m a^3} = \Omega^2$, where $\Omega$ is the angular velocity of the circular motion. So $\omega = \Omega \sqrt {1 - ka}$. Can't this be what you are really asked about?

#### CAF123

Gold Member
I can't see any mistake in your derivation (assuming the original equations for u and h are correct).

Unless I am mistaken, $\frac {\mu e^{-ka}} {m a^3} = \Omega^2$, where $\Omega$ is the angular velocity of the circular motion. So $\omega = \Omega \sqrt {1 - ka}$. Can't this be what you are really asked about?
Yes, the question as stated in the OP is the one I need to answer. I see that $h/a^2 = \dot{\theta}$ so $\omega = \dot{\theta} \sqrt{1 - ka}$

I don't see how the dimensions of $\sqrt{1-ka}$ are correct anyway. I believe k has dimensions of 1/length. So $\sqrt{1-ka}$ would be dimensionless. $\dot{\theta} \sqrt{1-ka}$ has units of 1/s, as required. The question is a mistake maybe?

#### voko

They may have asked a bogus question to have you defend your "wrong" answer, they do that sometimes.

They might also have meant the "relative" period of the perturbed motion (in units such that the orbital period = 1), but then the answer is still off by a factor of $2 \pi$.

#### CAF123

Gold Member
They may have asked a bogus question to have you defend your "wrong" answer, they do that sometimes.

They might also have meant the "relative" period of the perturbed motion (in units such that the orbital period = 1), but then the answer is still off by a factor of $2 \pi$.
I do see they mention '...show that in the perturbed orbit, the furthest distance from the centre occurs with...'. Did I ever make use of this 'furthest distance'?

Do you agree with my comments about the dimensionality of the solutions?

#### voko

I can't see how that could change anything. $r = a + p = a + \sin (\omega t + \alpha )$, so the period between two maxima is still $2 \pi / \omega$.

Your dimensional analysis is quite correct, too.

#### CAF123

Gold Member
I spoke to some other people about this and they get the required $\sqrt{1-ka}$. Their argument was along the lines of say, for example, you have sin(2θ). The period of this is 2pi/2,
where the 2 is dimensionless.

This seems to make sense mathematically, but physically since $T = \frac{2\pi}{\omega}$, I don't really understand why it can't have units. (Unless of course there is intrinsic units associated with the '2', but that would contradict the statement of the 2 being dimensionless.

#### voko

This is along the lines of the second sentence in #20. If the orbital period is taken to be the unit of time (just like we do with the "year"), then you could measure the features of the perturbed motion in this unit. When I thought about that yesterday, though, I think I was not getting the $2 \pi$ factor. Please give it a try and see where that gets you.

#### CAF123

Gold Member
This is along the lines of the second sentence in #20. If the orbital period is taken to be the unit of time (just like we do with the "year"), then you could measure the features of the perturbed motion in this unit. When I thought about that yesterday, though, I think I was not getting the $2 \pi$ factor. Please give it a try and see where that gets you.
What's the $2 \pi$ factor? Even if I were to assume that defintion, (i.e it is okay to have null dimensionality associated with the answer), it doesn't explain where the $\dot{\theta}$ comes from.

"Stability of Orbits"

### Physics Forums Values

We Value Quality
• Topics based on mainstream science
• Proper English grammar and spelling
We Value Civility
• Positive and compassionate attitudes
• Patience while debating
We Value Productivity
• Disciplined to remain on-topic
• Recognition of own weaknesses
• Solo and co-op problem solving