1
$\begingroup$

If $f(t)$ an $g(t)$ are piecewise continuous functions on $[\ 0, \infty)$ then the convolution integral of $f(t)$ and $g(t)$ is,

$$(f*g)(t) = \int_{0}^{t}f(t-\tau)g(\tau) \text{d} \tau.$$

The text then gives a 'fact':

$\mathcal{L} \{f*g \} =F(s)G(s),$ where $\mathcal{L} \{ f(t) \} = F(s)$.

I tried to show this, but I'm not sure if it's correct. First

$$ \mathcal{L} \{f*g \} = \int_{0}^{\infty} e^{-st} \left [\ \int_{0}^{t}f(t-\tau)g(\tau) \text{d} \tau \right]\ \text{d}t. $$

I'm weary of swapping integral signs because don't exactly know when it is valid, but

$$ \mathcal{L} \{f*g \} = \int_{0}^{\infty} g(\tau) \left [\ \int_{\tau}^{\infty}f(t-\tau) e^{-st} \text{d} t \right]\ \text{d} \tau. $$ Using the substitution $u = t - \tau,$

$$ \mathcal{L} \{f*g \} = \int_{0}^{\infty} g(\tau) e^{-s \tau} \left [\ \int_{0}^{\infty}f(u) e^{-su} \text{d} u \right]\ \text{d} \tau = \left [\ \int_{0}^{\infty}f(u) e^{-su} \text{d} u \right]\ \left [\ \int_{0}^{\infty} g(\tau) e^{-s \tau} \text{d} \tau \right] $$

Does this make any sense?

Edit: $e^{-st}$ replaced with $e^{-su}$ and corrections to errors

$\endgroup$
4
  • $\begingroup$ What if I made both limits $\infty$ at the beginning? $\endgroup$
    – user110503
    Jun 25, 2015 at 15:19
  • 1
    $\begingroup$ integrating $\tau $ from $0$ to $t$ and then integating $t$ from $0$ to $\infty $ is the same as integrating $t$ from $\tau $ to $\infty $ and then $\tau $ from $0$ to $\infty $ $\endgroup$ Jun 25, 2015 at 15:22
  • $\begingroup$ Ok I found the limits and they match the ones you have just given. So in the substitution $u = t - \tau$, the lower limit becomes $0$. Thanks. I feel bad because I should know this. $\endgroup$
    – user110503
    Jun 25, 2015 at 15:25
  • 1
    $\begingroup$ But now you do know it and that's what matters. $\endgroup$ Jun 25, 2015 at 15:33

1 Answer 1

2
$\begingroup$

As you know we have $$\mathcal{L} \{f*g \} = \int_{0}^{\infty} e^{-st} \left [\ \int_{0}^{t}f(t-\tau)g(\tau) \text{d} \tau \right]\ \text{d}t.$$ Note that the exponential has the following property $$ e^{-st} = e^{-s(t - \tau )} e^{-s\tau} $$ We may exploit this by sneaking the exponential into the $\tau$ integral with Fubini's to obtain $$\int_0^\infty \left [\int_0^t e^{-s ( t- \tau)} f( t - \tau) e^{-s \tau} g ( \tau) d \tau \right ] dt $$ Now let's change variables, say $u = t-\tau $ and $v = \tau$. What happens to the integral under this change of variables?

$\endgroup$
6
  • $\begingroup$ Thanks, but did you read my attempt? The problem was my limits were incorrect. With your substitution, $g(\tau)$ becomes $g(u - t)$ and the integral upper and lower limit become $t$ and $0$ respectively, but I'm not sure how that would help. $\endgroup$
    – user110503
    Jun 25, 2015 at 15:35
  • $\begingroup$ Your change of integrals was incorrect, I thought I'd show you how to justify the swap with the above. In addition to that, your change of variables wasn't quite right (since you forgot the limits). I thought I'd rewrite the change using $u$ and $v$ since that's what you've most likely seen. You're glancing over the Jacobian of the transformation and what not... Hence why I rewrote it the way I did. Also, your comment applies to your direct change of variables as well if you notice... you may want to have another look. $\endgroup$
    – Jeb
    Jun 25, 2015 at 15:51
  • $\begingroup$ I'm still a bit confused. I thought the problem was that when I swapped the integrals I didn't have the right limits. With the right limits and the substitution it looks like it comes out correctly. The reason I kept the lower limit as $0$ in the substitution originally is because $f(t) = 0$ for $t < 0$, so I thought it made no difference. $\endgroup$
    – user110503
    Jun 25, 2015 at 16:15
  • $\begingroup$ It's much more than that, consider when you wrote $$ \int_0^\infty f(u) e^{-st} du = e^{-st} \int_0^\infty f(u) du \neq \mathcal{L}\{f\} [s]$$, it's still a function of $t$! That's not at all what you wanted to show. Have a look at what I've written, and perform the suggested change of variables according to the definition. $\endgroup$
    – Jeb
    Jun 25, 2015 at 16:24
  • $\begingroup$ I'm not sure when I wrote that. I thought that I had $$\int_{0}^{\infty} f(u)e^{-s(t + \tau)} du = e^{-s \tau } \int_{0}^{\infty} f(u)e^{-st} du $$ $\endgroup$
    – user110503
    Jun 25, 2015 at 16:29

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.