MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Let $V \subset H \subset V'$ be a Hilbert triple.

We can define a weak derivative of $u \in L^2(0,T;V)$ as the element $u' \in L^2(0,T;V')$ satisfying $$\int_0^T u(t)\varphi'(t)=-\int_0^T u'(t)\varphi(t)$$ for all $\varphi \in C_c^\infty(0,T)$.

Then we define the space $W = \{u \in L^2(0,T;V) : u' \in L^2(0,T;V')\}$. We know for example that for $u, v \in W$, $$\frac{d}{dt}(u(t),v(t))_H = \langle u'(t), v(t) \rangle + \langle v'(t), u(t) \rangle.\tag{1}$$

Now suppose I change my space of test functions and define a weak derivative as an element $u' \in L^2(0,T;V')$ satisfying $$\int_0^T u(t)\varphi'(t)=-\int_0^T u'(t)\varphi(t)$$ for all $\varphi \in C_c^1(0,T)$.

Define $\tilde W = \{u \in L^2(0,T;V) : u' \in L^2(0,T;V')\}$ where the derivative is now with respect to these new test functions. How is this related to $W$? Do properties like (1) still hold?

I think yes, by uniqueness of weak derivatives. But I wanted to check in case I missed something.

share|cite|improve this question

The derivative you define is the distributional derivative, which can be defined on $C^\infty_c$ or $C^1_c$. But then, you restrict to weak derivatives $u' \in L^2(I;V')$, which means, you allow test functions even from from $H^1(I;V)$, which is the class for which the left hand side of the definition of your derivative makes sense.

Let for a moment be $V=\mathbb R$. Let $u$ be the Heaviside function with jump at zero. Then, its distributional derivative according to your first and second definition is the Dirac functional $\delta(x)$, which is not a weak derivative in your sense. By both of your definitions, you can even compute its distributional derivative $\delta'(x)$ with both your definitions. The next derivative $\delta''(x)$ is only defined on $C^2_c$ and thus would not be caught by your second definition. But as soon as a derivative is defined as a functional on $C^1_c$, it will be the same on $C^\infty_c$, since the latter is a subspace.

share|cite|improve this answer
So the second definition ($C^1_c$) is only bad when we want to discuss second order derivatives. – maximumtag Sep 7 '13 at 13:42

The main property that you would want in your weak derivative is that it defines a closed, unbounded operator on $L^2$. Suppose you have an unbounded operator $A$ (for you it would be the derivative) defined on a dense domain $D(A)$ in $L^2$. As such, $A$ need not be closed (it's graph in $L^2\times L^2$ is not a closed subspace). So the next step is to consider its closed extensions (operators whose graphs are closed subspaces of $L^2\times L^2$ which contain the graph of $A$ itself). Among these, there exists a minimal closed extension $\bar{A}$ with domain $D(\bar{A})\supseteq D(A)$ (its graph is the closure of the graph of $A$).

So, when it comes to the minimal closed extension $\bar{A}$, it doesn't matter which domain you pick for $D(A)$ as long as it produces the same $\bar{A}$. Two unbounded operators $A_0$ and $A_1$ (say the derivative on $C^\infty_c$ or $C^1_c$) define the same minimal closed extension, $\bar{A}_0 = \bar{A}_1$, if for instance the graph of $A_1$ is contained and dense in the graph of $\bar{A}_0$ (as it happens to be the case in your example).

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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