7
$\begingroup$

I got stuck in my calculations trying to solve the following problem:

Given the ODE $$\dot{x} = -\alpha x + a\sum_{n=0}^\infty\delta(t-n\tau)$$ where $\alpha \gt 0$, define $$x_k = x(k\tau +0 )$$ and find $x_k$ as a function of $x_{k-1}$. What is the value of the following limit $$\lim_{k\rightarrow \infty}x_k?$$

My attempt

First of all we define the following function

$$f(t) = a\sum_{n=0}^\infty\delta(t-n\tau)$$ then the ODE becomes pretty much standard $$\dot{x}(t) = -\alpha x(t) + f(t)$$

for which the general solution is

$$x(t) = e^{-\alpha t} x(0)+\int_0^t e^{-\alpha(t-t')}f(t')\,dt'\\ x(t) = e^{-\alpha t} x(0)+\int_0^t e^{-\alpha(t-t')}a\sum_{n=0}^\infty\delta(t'-n\tau)\,dt'.$$

Now we interchange the sum with the integral (being a physicist I impudently change them without checking uniform convergence!) and get

$$x(t)= e^{-\alpha t} x(0)+\sum_{n=0}^\infty a\int_0^t e^{-\alpha(t-t')}\delta(t'-n\tau)\,dt'$$

By the definition of $x_k$

$$\begin{align}x_k = x(k\tau+0) &= e^{-\alpha k\tau} x(0)+\sum_{n=0}^\infty a\int_0^{k\tau} e^{-\alpha(k\tau-t')}\delta(t'-n\tau)\,dt'\\&=\underbrace{e^{-\alpha k\tau}x(0)}_{\text{first term}}+\underbrace{\sum_{n=0}^kae^{-\alpha(k\tau-n\tau)}}_{\text{second term}}\end{align}$$

We can easily calculate $x_{k-1}$ from the value of $x_k$

$$ x_{k-1} = \underbrace{e^{-\alpha(k-1)\tau}x(0)}_{\text{first term}}+\underbrace{a\sum_{n=0}^{k-1}e^{-\alpha(k\tau-n\tau)}}_{\text{second term}}.$$

It is clear that is we want to make a relation between $x_k$ and $x_{k-1}$, the first terms of both can be written as $$e^{-\alpha k\tau}x(0) = e^{-\alpha\tau}e^{-\alpha(k-1)\tau}x(0)$$ so I'm bound to say that, at list for the first terms $$x_k = e^{-\alpha\tau}x_{k-1}\tag2$$ The real problem arises when we try to make adjustments to $(2)$ to make the second terms equal, mainly from definition $(2)$ we get

$$x_k = e^{-\alpha\tau}e^{-\alpha(k-1)\tau}x(0) + \color{red}{e^{-\alpha\tau}a\sum_{n=0}^{k-1}e^{-\alpha(k-n)\tau}}$$

My questions now are

Question 1: How can I adjust that second term to get $x_k$ as a function of $x_{k-1}$?

Question 2: Is there an easier way to solve this problem?

$\endgroup$
  • $\begingroup$ When you plug in what $f(t')$ is, shouldn't it be $\delta(t'-n\tau)?$ $\endgroup$ – Adrian Keister Aug 22 '18 at 15:50
  • $\begingroup$ @AdrianKeister Sure should! I'll correct it thanks! $\endgroup$ – Davide Morgante Aug 22 '18 at 15:51
  • 1
    $\begingroup$ Several typos: first, the prefactor of $x(0)$ in $x(t)$ is $e^{-at}$, not $e^{at}$; second, you forgot to replace $t$ by $k\tau$ in the second terms of $x_k$, thus missing the simplifications that follow. I believe that if you modify your try with respect to these two points, everything should run smoothly afterwards... $\endgroup$ – Did Aug 22 '18 at 15:57
  • 1
    $\begingroup$ I think your (2) is suspect. There has to be an additive relationship between $x_k$ and $x_{k-1},$ because the $x_{k-1}$ exponential sum doesn't go as far as the $x_k$ one does.I would write $$x_k=e^{\alpha\tau}e^{\alpha(k-1)\tau}x(0)+a\sum_{n=0}^{k-1}e^{-\alpha(t-n\tau)}+ae^{-\alpha(t-k\tau)}.$$ $\endgroup$ – Adrian Keister Aug 22 '18 at 15:58
  • 1
    $\begingroup$ "Is that not the case?" No. As a way to memorize this, simply note that $y'=by$ is solved by $y(t)=y(0)e^{bt}$, for every $b$, thus, in the absence of $f$, your equation would read $x'=-\alpha x$ and be solved by $x(t)=x(0)e^{-\alpha t}$. $\endgroup$ – Did Aug 22 '18 at 16:02
4
$\begingroup$

So let's start from $$x_k(\tau) = x(k\tau)= e^{-\alpha k\tau}x(0)+a\sum_{n=0}^ke^{-\alpha\tau(k-n)}.$$ Then $$x_{k-1}(\tau)=e^{-\alpha (k-1)\tau}x(0)+a\sum_{n=0}^{k-1}e^{-\alpha\tau((k-1)-n)}.$$ In comparing the two, we want to get $x_k$ to look like $x_{k-1}.$ So we write \begin{align*}x_k(\tau) &= e^{-\alpha\tau}e^{-\alpha (k-1)\tau}x(0)+a\sum_{n=0}^{k-1}e^{-\alpha\tau(k-n)}+a \\ &=e^{-\alpha\tau}e^{-\alpha (k-1)\tau}x(0)+ae^{-\alpha\tau}\sum_{n=0}^{k-1}e^{-\alpha\tau((k-1)-n)}+a \\ &=e^{-\alpha\tau}x_{k-1}(\tau)+a.\end{align*}

To compute the limit, I would figure exactly what $x_{0}(\tau)$ is (looks like $x_0(\tau)=x(0)+a$), then write $x_{k}(\tau)$ in terms of that. That's something you could take the limit of.

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ Great! Simple and ingenious. I was very close to it! Thank you very much for the help, I've greatly appreciated $\endgroup$ – Davide Morgante Aug 22 '18 at 16:37
  • 1
    $\begingroup$ @DavideMorgante: Yes, I'd agree: you certainly did the heavy lifting here. $\endgroup$ – Adrian Keister Aug 22 '18 at 16:37

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.