I was looking at the following definition of the weak derivative:

Let $\Omega$ be a domain (ie an open connected subset of $\mathbb{R}^n$). Suppose $u,v \in L_{1,loc}(\Omega)$ and

\begin{equation} \int_{\Omega}u(x)\partial^{\alpha}\eta(x)dx = (-1)^{|\alpha|}\int_{\Omega}v(x)\eta(x)dx, \forall \eta \in C^{\infty}_c(\Omega) \end{equation} where $\alpha$ is a multiindex. Then $v$ is called the weak partial derivative of $u$ in $\Omega$, and is denoted by $\partial^{\alpha}u$.

Now suppose the classical derivative $\partial^{\alpha}u$ exists and is continuous on some domain $\Omega$. Is the classical derivative the weak derivative on $\Omega$?

I suppose this reduces to the integration by parts formula (ie the Guass-Green Theorem). But I'm having trouble justifying its use since for any particular $\eta \in C^{\infty}_c(\Omega)$ I seem to require a subdomain $\Omega' \subset \Omega$ such that $\text{supp} \eta \subset \Omega'$ and $\partial\Omega'$ is $C^{|\alpha|}$. Is there a standard construction, or do I need to have stronger restrictions on the domain $\Omega?$

Edit: The $\partial^{\alpha}u$ exists and is continuous on all of $\Omega$, not all of $\mathbb{R}^n$ as initially stated.


If $\eta$ has compact support, then let $b$ be a smooth function that is one on $S=\operatorname{supp} \eta$, and is zero on the complement of some compact set that contains $S$ in its interior. Then we have $u(x) = u(x) b(x)$ for all $x \in S$, that is, $ \operatorname{supp} (u \cdot b) \subset S$ ($\cdot$ means pointwise multiplication). (Note that we have $\operatorname{supp} D^\alpha \eta \subset S$ as well.)

Then, for appropriate $\alpha$, we have $\int_\Omega u D^\alpha \eta = \int_\Omega (u b)D^\alpha \eta = (-1)^{|\alpha|} \int_\Omega D^\alpha (u \cdot b) \eta= \int_\Omega (D^\alpha u)\eta$.

Hence the classical derivative is the weak derivative.

  • $\begingroup$ I think I understand. There's no need to integrate over only the domain when we've extended our functions by zero to all of R^n. To apply integration by parts we need only choose a large enough ball. $\endgroup$ – Dogomien Dec 20 '13 at 18:21
  • $\begingroup$ Well, sort of. We show that for any particular $\eta$, the integral is equivalent to one where both functions have compact support, and hence the equality follows. $\endgroup$ – copper.hat Dec 20 '13 at 18:23
  • $\begingroup$ Sorry I'm a bit confused. How does one apply the "integration by parts" step? $\endgroup$ – Dogomien Dec 20 '13 at 18:48
  • $\begingroup$ I don't have a straightforward answer for that. Maybe math.stackexchange.com/a/171925/27978 might be of help? $\endgroup$ – copper.hat Dec 20 '13 at 19:48

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.