8
$\begingroup$

Consider a damped pendulum whose equation of motion is given in general by $$m\ddot{x}=-\mu\dot{x}-kx$$ where $\mu,k>0$

Rewrite this equation as

$$\ddot{x}+2\gamma\dot{x}+\omega^2x=0,$$

where $2\gamma = \frac{\mu}{m}$ and $\omega^2 = \frac{k}{m}$.

If $\gamma>\omega$, the roots of this equation are real and distinct. Define $\beta^2=\gamma^2- \omega^2$. Then the roots are $-\gamma \pm\beta$.

I have no issue with this, the solution is of the form $$x(t)=e^{-\gamma t}\left(Ae^{\beta t}+Be^{-\beta t}\right).$$

Given some initial conditions $$x(0)=x_0 \quad \text{and} \quad \dot{x}(0)=v_0 $$ Does the system return to rest in a finite time?

I have no issue here as well. I can rewrite the solution to read as $$x(t)=Ae^{(-\gamma+\beta)t}\left[1+\frac{B}{A}e^{- 2\beta t}\right]$$ where the next time, say $t_1$ such that $x(t_1)=0$ is given by $$t_1 = \frac{1}{2\beta}\ln\left(-\frac{A}{B}\right)$$

Here is where I struggle though. I can picture what happens if $$\frac{A}{B}<-1$$ I can choose some arbitrary number, say -2, so that $\ln(-(-2))=\ln(2)$ and thus $t_1$ is finite. In other words given some initial velocity the pendulum will move from its initial position and without oscillating return to its equilibrium position after time $t_1$.

However, now I am stuck.

If $\dfrac{A}{B}\to -\infty$ then $t_1 \to \infty$ I can make peace with. Am I right in saying that the system is so over damped that this pendulum almost comes to a complete stop and sloooooooooooooowly over an infinitely long period returns to equilibrium? If so, my brain can picture that, and I am happy.

But now the cases I can't make peace with.

  1. If $A=-B$, i.e. $x(0)=0$, so we start at equilibrium, regardless of my initial velocity $\dot{x}(0)=v_0$, it is true that $t_1$ is always zero as $\ln(1)=0$. In my brain, I can picture a pendulum starting at equilibrium position, flying with some crazy initial velocity, and that equation is telling me, yeah.. The next time it's at $0$ is now instantaneously at $0$.

    So what happened? It's impossible that the system is so overdamped that it's almost not damped at all and the pendulum flies infinitesimally fast to the equilibrium and then stops suddenly without shooting past. I kind of almost want to say it's negatively overdamped.. and that can't make sense.. can it?

  2. Similarly, If $-1 <\frac{A}{B} - <0$, then $t_1$ is negative.

  3. Similarly, If $\frac{A}{B} = 0$, or rather as $\frac{A}{B} \to -0$ then $t_1 \to -\infty$ is problematic.

  4. Self-explanatory ... If $\frac{A}{B} > 0$ then we have a ln of a negative number which is also problematic.

So my question is really this. Are there physical relationships to the other 4 cases or is it one of those "the maths is possible but the physics is meaningless" scenarios. i.e. If $\frac{A}{B} > 0$ then that equation cannot physically ever be representing an over damped oscillator... or at least not in this universe. I am really hoping that's the case.

TL;DR - What is happening with over damped pendulums? Or does the equation of motion become meaningless under certain initial conditions?

PS: I am putting this under homework because my brain might be missing something obvious that someone might illuminate. My background is in applied mathematics and fluid mechanics. For the life of me I can't picture this mentally or find literature online that addresses these issues and unluckily I am curious to know things that bother me.

$\endgroup$
  • 2
    $\begingroup$ You don't have a restoring force here $\endgroup$ – Aaron Stevens Aug 16 at 12:31
  • 1
    $\begingroup$ I'm a bit concerned by the differential equation you've specified. Note that you have pole in the RHP so your system isn't stable. (oops, I see Aaron beat me to this as I wrote the comment). $\endgroup$ – Alfred Centauri Aug 16 at 12:33
  • $\begingroup$ If you choose initial conditions so that $B = 0$, the mathematical solution will return to zero. But as others have said, this is an unstable system since the "restoring force" from the "stiffness" term (interpreting the equation of motion as a vibrating mechanical system) has the wrong sign. Physically, such a solution is like giving a pencil some initial motion so that it stops, balanced perfectly on its point! Mathematically that is possible, but not in real life. $\endgroup$ – alephzero Aug 16 at 13:16
  • $\begingroup$ @Alfred I have a feeling this ODE behaves like a damped oscillator (for certain values) since the restoring force is indeed in the wrong direction. For those values this equation quacks like a duck and walks like a duck, it is however a goose. Is there any physical system that has (I know this sounds weird) a restoring force that works together in the same direction as the motion? In which case it wouldn't make sense to call it a restoring force but rather a '______' force. $\endgroup$ – Ken Aug 16 at 13:19
  • 2
    $\begingroup$ Your changed your system. FWIW it's usually poor form to edit the question such that it invalidates answers that have already been posted. $\endgroup$ – Aaron Stevens Aug 16 at 14:23
3
$\begingroup$

I think something has gone wrong very early in your analysis. For example:

1: If $A=-B$, i.e. $x(0)=0$, so we start at equilibrium, regardless of my initial velocity $\dot{x}(0)=v_0$, it is true that $t_1$ is always zero as $\ln(1)=0$.

Yep! There is only one root and it is at $t=0$. But then you say,

In my brain, I can picture a pendulum starting at equilibrium position, flying with some crazy initial velocity and that equation is telling me, yeah.. the next time it's at 0 is now instantaneously at 0. So what happened?

And the key is, nothing happened.

It started at $x = 0$ with $\dot x \ne 0.$ That is where it started.

That is also the only place it is ever at zero. Looking really abstract, if you draw out the factor of $B$ rather than the factor of $A$, you have this equation $(A/B) + e^{-2\beta t} = 0$ and it only has at most one root, namely it only has one if $A/B < 0$. It had no other roots it could possibly have. You started it at $x=0$ and it will never be at zero again.

It didn’t “return back to zero after no time”... it never returned back to 0. How could it? The graph is $A~\sinh(\beta t)~e^{-\gamma t}$ and for $t > 0$ that is the product of three positive terms! How are we going to multiply positives to get anything that is not a positive number? (OK so $A$ could be negative but, y'know, “without loss of generality” etc. we can assume that we orient our axes so that either $x(0) > 0$ or if $x=0$ then so that $\dot x(0) > 0.$)

More to the point the single root only exists for positive $t$ if you fire the particle at the $x=0$ position with a certain negative velocity. If you do not fire it hard enough, then the damping just gets rid of this initial momentum. If you do fire it hard enough, it crosses over with some finite remaining velocity and then it sits at $x < 0$ until the infinity of time.

“Aha, but I have good physical intuitions and there must be some boundary between these where I give the particle just enough energy to hit zero but not cross over.” That is a beautiful idea which unfortunately also does not work... because the time when the crossing-over happens happens to be $t=\infty$ at this boundary case; the root “comes in from infinity” as you increase $v_0.$

There is only one root, and it is never the end of the dynamics, except in the trivial case where $x_0 = v_0 = 0.$

$\endgroup$
  • $\begingroup$ Ok I think this is exactly what I was looking for. Let me make sure I understand correctly. 1) This system is over damped, so I fire this object with some velocity. 2) The velocity is either a) great enough to overcome the damping but now the object is stuck in the void as it can't return from x<0. b) not great enough and the damped system laughs at the failed attempt to even try escape, the object was never leaving. 3) There is a sweetspot region that is impossible to hit thanks to our friend infinity. $\endgroup$ – Ken Aug 16 at 16:18
  • $\begingroup$ Then there is the boring case I wasn't interested in, in which we start at some point away from the equilibrium position i.e. x(0) not 0 and watch this object come to a halt as in the critically damped case. $\endgroup$ – Ken Aug 16 at 16:20
  • $\begingroup$ Yep, those are the three options! I am not sure that it matters that it is “over”-damped, either... if it is underdamped the same basic story holds except it happens to cross 0 infinitely many times; the energy is still leaving the system slower as there is less energy to go around. If it is critically damped then we have something like $e^{-\gamma t}(A + B t)$ and that line $A + B t$ is similarly not able to converge on 0 the way you are thinking might happen. $\endgroup$ – CR Drost Aug 16 at 16:27
  • 1
    $\begingroup$ @Ken (If it helps, “overdamped” does not mean that the damping is more effective; it actually means that the damping is getting in its own way of helping to remove the energy from the system. So like as damping gets larger and larger, the 1/e decay time for the amplitude of the overdamped system is also getting larger and larger; as it goes to infinity $x(t) = x_0$ becomes a solution of the system for any $x_0$ and there is no possibility of restoration back to $x=0$ even approximately, as any attempt to move the thing results in an infinite damping force.) $\endgroup$ – CR Drost Aug 16 at 16:36
  • $\begingroup$ Thank you! It drives me nuts if I can't picture what's going on. I believe that It's just as important to understand the physics behind the problems and their solutions. While I feel a lot of people are only interested in getting the right answer, I am more interested in making sure I understand fully why that answer is indeed right. Today you helped me do that =) $\endgroup$ – Ken Aug 16 at 16:41
5
$\begingroup$

Let us solve the differential equation to see what we get. (spoiler: it is not what you originally posted).

You have $$\ddot{x}+2\gamma\dot x+\omega^2x=0$$

The characteristic equation of the DE is,

$$\lambda^2+2\gamma\lambda+\omega^2=0$$

And solving for $\lambda$ you get:

$$\lambda_{1,2}=\frac{-2\gamma\pm \sqrt{4\gamma^2-4\omega^2}}{2}=-\gamma\pm\sqrt{\gamma^2-\omega^2}$$

And the usual way to express this is,

$$\lambda_{1,2}=-\gamma\pm i\sqrt{\omega^2-\gamma^2}$$

The physical parameters here are $\omega$, the natural frequency of the system (the frequency of oscillation of the undampened system) and $\gamma$, which is related to the damping factor $\zeta:=\gamma/\omega$. The value of $\zeta$ tells you whether the system is underdamped ($\zeta<1$), critically damped ($\zeta=1$) or overdamped ($\zeta>1$). You should be able to find these concepts in more details in a System Modelling book.

Basically, the only kind of system for which you will observe oscillatory motion is $\zeta<1$, because there you have an imaginary component on the roots, and you know that translates to sines and cosines.

You can see the prototype response for these different kind of systems in the picture I attached.

Regarding your question, you can see that the general solution of the equation is:

  • For $\gamma<\omega$

$$e^{-\gamma t}\left[A\text{ }\mathrm{sin}\left(t\sqrt{\omega^2-\gamma^2}\right)+B\text{ }\mathrm{cos}\left(t\sqrt{\omega^2-\gamma^2}\right)\right]$$

  • For $\gamma=\omega$

$$A\text{ }e^{-\gamma t}+B\text{ }te^{-\gamma t}$$

  • And for $\gamma>\omega$

$$A\text{ }\exp\left[-\gamma{\left(1+\sqrt{1-\frac{\omega^2}{\gamma^2}}\right)}t\right]+B\text{ }\exp\left[-\gamma{\left(1-\sqrt{1-\frac{\omega^2}{\gamma^2}}\right)t}\right]$$

Where I rearranged thing a little in the last equation to make it clearer that the exponents are negative. I graphed this with $\omega =10, A=B=\frac{1}{2}$ for the three cases, with $\zeta=\frac{1}{2},1,2$, and I get the following:

Damping

The red is underdamped, the blue is critically damped and the green is overdamped. They all decay exponentially to zero.

Does the system return to rest in a finite time?

No, it does not (if, by returning to rest you mean that the system returns to the equilibrium point with $v=0$). As you can see in the three equations for the solutions, the amplitude of $x$ decreases exponentially. This means that you asymptotically approach the rest state ($x=p=0$), but never reach it.


Edit: I will add, per the request of the OP, another graph. In this case with $-A=B=\frac{1}{2}$, $\omega=10$ and $\zeta=2$.

A=-B

We can think of this system as one where the oscillator starts at the equilibrium position but with a high velocity (we push the mass very hard and reach desired velocity exactly at the equilibrium point of the oscillator and we start recording the position at that instant ($t=0$)).

$\endgroup$
  • $\begingroup$ My question specifically states over damped pendulum. The roots of that equation are not wrong, they are the only roots that satisfy an over damped pendulum as per my specific inquiry. While I appreciate your reply it is off topic as you have given all three cases and avoided the question at hand, which is, the physical interpretation behind the cases that can't be plotted. $\endgroup$ – Ken Aug 16 at 14:49
  • $\begingroup$ Well, you have edited your question so much that it is hard to post a proper answer. Still, you can see in my answer all the possible solutions for the system you post. I think part of the problem is you are getting confused because of the roots because you post a "general solution" that can also involve complex numbers... That's why in basic DE courses you see the solutions broken down in three cases, like I posted here. $\endgroup$ – S V Aug 16 at 14:53
  • $\begingroup$ And in the original question the roots were clearly wrong. As you can see here all the roots are negative. The only way they could be positive is if $\gamma$ was negative... I have yet to see a physical system where the damping actually gives energy to the thing being damped. $\endgroup$ – S V Aug 16 at 14:54
  • $\begingroup$ And also note that for a given system $A/B$ is a constant, and is determined entirely by the initial (or boundary) conditions imposed. $\endgroup$ – S V Aug 16 at 14:55
  • $\begingroup$ This I am highly aware of, consider your values for $\omega=10$, however, let $A=-0.5$ and $B=0.5$ which corresponds to an initial condition of x(0)=0. Now let $\zeta=2$... plot that solution.. How does that solution make physical sense. This is the essence of my question. (spoiler: it is what I originally posted). $\endgroup$ – Ken Aug 16 at 15:18
3
$\begingroup$

1) the system will never get to $0$ in a finite time. Therefore, it makes sense that the only time to can find where $x=0$ is at $t=0$. It isn't saying that the object starts moving but then instantly snaps back to $0$. Your solution is showing you that you are only at equilibrium at $t=0$, which is true.

For all other cases you have picked initial conditions such that the object never reaches the origin. Or in other words, you would need to have been at the origin initially in the past (hence the negative time).


Keep in mind that this is an over damped system, so the last time the object will "reach equilibrium" will be due to an essentially exponential decay where the origin is only arrived at as $t\to\infty$. For the cases where $t_1>0$ you have just set the initial conditions such that the system overshoots equilibrium before this exponential decay occurs.

$\endgroup$
3
$\begingroup$

(Answer written for v2 of question with given stable system ODE)

First, for the case of real, distinct roots, I believe there are just two cases to consider given a non-zero initial position and $t\ge 0$:

  • The initial velocity is such that the system asymptotically approaches $x=0$

  • The initial velocity is such that the system goes through $x=0$ for some $t>0$ and then asymptotically approaches $x=0$

To see this, stipulate that the general solution is

$$x(t) = Ae^{-\alpha t}+Be^{-(\alpha + \beta)t}$$

where the decay constants $\alpha,\,\beta$ are positive real numbers. The initial conditions are then

$$x(0) = A + B$$

$$\dot x(0) = -\alpha x(0) - \beta B$$

Now, solve for $t$ such that $x(t) = 0$:

$$t = -\frac{1}{\beta}\ln\left(-\frac{A}{B}\right)$$

Clearly, this equation can only be satisfied for finite $t\gt0$ when $B\lt -A$

For $-A \lt B \lt 0$, there are solutions for finite $t\lt 0$, and for $B\ge 0$, there are no solutions for finite $t$.

Since it is evident by inspection that $x(t)\rightarrow 0$ as $t\rightarrow\infty$, the two cases I mention above follow.

Note: the $B=-A$ (zero initial position) case is the case that is 'on the boundary' between the solutions with $x=0$ for $t\gt 0$ and $x=0$ for $t\lt 0$.

Here are plots of four sets of initial conditions for the general solution $x(t) = Ae^{-t}+Be^{-2t}$ that illustrate the cases above:

enter image description here

$\endgroup$
2
$\begingroup$

I have a feeling this ODE behaves like a damped oscillator (for certain values)

(sigh) Link to the version of this question I wrote this answer to

There are a family of initial conditions that yield an $x(t)$ that goes to zero as $t\rightarrow\infty$:

$$\dot x(0) = -(3 + \sqrt{19})\cdot x(0)$$

This corresponds to $A=x(0),\, B=0$

But the essential problem with this system (in the context of your lengthy analysis of it) is that it isn't stable which is to say that the slightest perturbation away from these initial conditions will yield an $x(t)$ that goes to $\pm\infty$ as $t\rightarrow\infty$

Here's a graph of a $A=1, B=0$ solution as well as a $A/B=-2$ solution that you mention early in your post:

enter image description here

This is really all there is to it. There's a family of solutions that look suspiciously like a (decaying) first order system (no damped oscillation evident), and then the rest - all of which 'run away' from $x=0$. I just don't see any evidence damped oscillator behavior in any of the solutions.


Re your edited question: IIRC, for an overdamped (stable) 2nd order system with non-zero initial conditions, the trajectory for $t>0$ either asymptotically approaches $x=0$ or crosses $x=0$ once before asymptotically approaching $x=0$. Isn't that all there is to it?

$\endgroup$
  • $\begingroup$ thank you, I did push this through mathematica as well to see the solution space. I made the question applicable to an ODE that definitely is for a damped oscillator. The solution form is however identical, the point of my question was perhaps obscured by the ode that I used as the example. That family of solutions is precisely what I was interested in. If you didn't know the original ODE, those solutions behave exactly like a damped oscillator but fail for everything else. But if the ode is a damped oscillator those solutions still exist. $\endgroup$ – Ken Aug 16 at 14:19
  • $\begingroup$ "with non-zero initial conditions" <<--- This is what I am interested in. That ODE for an over damped oscillator pick the initial condition $x(0)=0$ for the velocity I am not really fussed. As soon as you pick the equilibrium position for the initial condition on an over damped pendulum the time solution is always 0 regardless of the initial velocity. But why? What is the physical reason that starting at the equilibrium results in this solution. I can see the solution, I can plot the solution, but I don't understand the physical reason for that solution. $\endgroup$ – Ken Aug 16 at 15:28
  • $\begingroup$ "As soon as you pick the equilibrium position for the initial condition on an over damped pendulum the time solution is always 0 regardless of the initial velocity" - Of course the time solution for $x=0$ is $t=0$ - that's the initial condition. Are you expecting that should be two values of $t$ that give $x=0$? If so, why? $\endgroup$ – Alfred Centauri Aug 16 at 15:42
  • $\begingroup$ Yes I am, maybe I am expecting 2 times for the wrong reason? Let $x(0)=0$ we start at equilibrium yes? Now $\dot{x}(0)>0$ would mean that we start at equilibrium but with some positive velocity, say, I dunno. 500km/s. Surely we would move away from the initial starting position.. but then we have to return again, and since it is over damped it would have to be in a finite time. $\endgroup$ – Ken Aug 16 at 15:47
  • 1
    $\begingroup$ @Ken, keep in mind that responses due to the $x(0)=0,\,v(0)\ne0$ initial condition(s) are proportional to the impulse response of the system. From what I've learned in the past, the impulse response either (1) crosses zero infinitely many times (underdamped case) or (2) asymptotically returns to zero (critically damped / overdamped case). See, for example, this plot $\endgroup$ – Alfred Centauri Aug 16 at 16:04
1
$\begingroup$

In my work, I am dealing a lot with cases where A=-B in a somewhat related case of aroma release (It's not an overdamped pendulum but instead two coupled first-order differential equations that can be written as a single second-order differential equation like yours).

The way that I look at is as a sum of two exponetial terms:

$$x(t) = A e^{-c_1t} + B e^{-c_2t} $$

with $c_2 > c_1 > 0$ relating to two decaying terms

  • One term is a more slowly decaying term $A e^{-c_1t}$.
  • The other term $B e^{-c_2t}$ is a faster decaying term (or I actually see it as an increasing term when B is negative).

    This second term makes the end result, $x(t)$, a curve that is moving towards this slowly decaying term (the difference with it is decreasing exponentially).


Image example:

image example for different situations

So you can see the four different situations depicted by the four black curves. In all cases they are moving towards this red curve which is that slowly decaying term $A e^{-c_1t}$. The curves have an asymptote in $x=0$ but differ in the point when they cross the point x=0, with the four cases:

$$\begin{array}{rl}t>0 &\quad\text{for} \quad B/A<-1\\ t=0 &\quad\text{for} \quad B/A=-1\\ t<0 &\quad\text{for} \quad-1<B/A<0\\ \text{never} &\quad\text{for} \quad 0 \leq B/A \\ \end{array}$$


This last case $B/A > 0$ relates to a pendulum that initially has a fast speed, $\dot{x}$, toward $x=0$. This speed get's slowed down quickly due to the large damping force $-\mu\dot{x}$. This damping of the speed relates to the faster $B e^{-c_2t}$ (this term is in the same way a slowing down term when $B/A < 0$ but then the speed is away from $x=0$) . When you reach the slower red curve relating to the $A e^{-c_1t}$ term then the damping force almost equals the spring force and $m\ddot{x}=-\mu\dot{x}-kx \approx 0$.

$\endgroup$
  • $\begingroup$ That's so interesting, I love how DE's and nature combine =) $\endgroup$ – Ken Aug 18 at 15:56
0
$\begingroup$

The equation of motion of an harmonic oscillator can be written as:

$ \ddot{x}+2\xi\omega_n\dot{x}+\omega_n^2x=0 $

with initial conditions $x(0)=x_0$ and $\dot{x}(0)=\dot{x}_0$, where $\omega_n$ is the natural frequency of the system and $\xi$ is called the damping ratio, and can be calculated as

$ \xi=\frac{\mu}{2m\omega_n} $

The system is said to be overdamped if $\xi>1$ and, for this case, the solution of the differential equation is:

$ x(t)=C_1e^{\left(-\xi+\sqrt{\xi^2-1}\right)\omega_nt}+C_2e^{\left(-\xi-\sqrt{\xi^2-1}\right)\omega_nt} $

The constants $C_{1,2}$ are obtained from the initial conditions:

$ C_{1,2}=\frac{\pm x_0\omega_n\left( \xi \pm \sqrt{\xi^2-1} \right) \pm \dot{x}_0}{2\omega_n\sqrt{\xi^2-1}} $

We notice from the response of this system: The motion is aperiodic, regardless of the initial condition. The motion diminishes exponentially with time.

When $\xi$ is appreciably greater than unity, one of the two decaying exponentials decreases much faster than the other, so the faster-decaying exponential term may be neglected. The solution can be approximately written as:

$ x(t)\approx C_2e^{\left(-\xi-\sqrt{\xi^2-1}\right)\omega_nt} $

Now, the decaying exponential will never be exactly zero, but you can get an estimate from zero. Say, $\left| x(t) \right|<\epsilon$, so

$ C_2e^{\left(-\xi-\sqrt{\xi^2-1}\right)\omega_nt} \approx \epsilon $

and you can solve for $t$.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.