i want to compute the product of convolution $1 * (\delta' * H)$ where $\delta$ is distribution of Dirac and $H$ is function of Heaviside.

first, we compute $\delta' * H.$ We have by definition that $\delta' * H (x) = \displaystyle\int_{-\infty}^x \delta' (y) dy$ how we can finish?


  • 1
    $\begingroup$ Take a test function $\varphi$ and try to compute $\langle \delta' * H,\varphi\rangle$ instead... $\endgroup$ Apr 27 '13 at 13:04
  • $\begingroup$ ok, so let $\varphi \in \mathcal{D}$ then, $$(\delta ' * H , \varphi) = (\delta', H * \varphi) = (\delta' , \displaystyle\int H(y-x) \varphi(y) dy) = (\delta' , \displaystyle\int_{-\infty}^x \varphi(y) dy)$$ how we can finish ? $\endgroup$
    – lili
    Apr 27 '13 at 13:12
  • 1
    $\begingroup$ @lili: Apply $(S*T)' = S'*T$ for any convolutable distributions $S,T$. $\endgroup$
    – Vobo
    Apr 27 '13 at 13:39

In fact, this can easily be done if you already know some properties of distributions.

The distribution $\delta$ is the identity element of the convolution and it's easy to see that:

$\hspace{3cm}\delta^{(k)}*T=(\delta*T)^{(k)}=T^{(k)}$ for all $k\in\Bbb N$ and $T\in\cal{D}'(\Bbb R)$.

Now using the fact that $H'=\delta$, you get : $$1*(\delta'*H)=1*H'=1*\delta=1$$

If you don't know all this, you take $\varphi\in\cal{D}(\Bbb R)$, and start like you did, except you made an error :

$$\langle \delta'*H,\varphi\rangle=\langle\delta',H*\tilde{\varphi}\rangle$$

where $\tilde{\varphi}:x\mapsto\varphi(-x)$.

For the part where you were stuck, don't forget that $\langle T',\varphi\rangle=-\langle T,\varphi'\rangle$.

  • $\begingroup$ okay, so $$< \delta' * H , \varphi> = - < \delta , H * \tilde{\varphi}'>$$ and for the rest? and i have another question, in general if $T, S \in \mathcal{D}'$ then $T * g \in \mathcal{D}'?$ and if $T \in \mathcal{D}'$ and $f \in \mathcal{C}^{\infty}$ then $T * f \in $? $\endgroup$
    – lili
    Apr 27 '13 at 16:19
  • $\begingroup$ Dear lili, maybe you should read a course before asking questions! $\endgroup$ Apr 27 '13 at 16:34

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.