Definition

Let ΩRn\Omega\subseteq\mathbb R^n be open and let ff be locally integrable, meaning on every compact subset of Ω\Omega. A locally integrable function gg is the weak partial derivative jf\partial_j f if

Ωfjφdx=Ωgφdx\int_\Omega f\,\partial_j\varphi\,dx =-\int_\Omega g\,\varphi\,dx

for every φCc(Ω)\varphi\in C_c^\infty(\Omega). Equivalently, the of the regular distribution defined by ff is the regular distribution defined by gg. A weak derivative, when it exists, is unique up to equality .

Agreement with classical derivatives

If fC1(Ω)f\in C^1(\Omega), shows that its classical is also its weak derivative. More generally, a locally function on an interval has a weak derivative equal almost everywhere to its ordinary derivative. Weak differentiation therefore extends rather than replaces classical differentiation.

Existence and failure

The absolute-value function on R\mathbb R has weak derivative sgn(x)\operatorname{sgn}(x), even though it is not differentiable at 00. The Heaviside function has no weak derivative represented by a locally integrable function: its distributional derivative is δ0\delta_0. Thus every locally integrable function has a distributional derivative, but not every such derivative is weak in the function-valued sense.

Sobolev-space role

A function belongs to the W1,p(Ω)W^{1,p}(\Omega) when it and all its first weak derivatives belong to Lp(Ω)L^p(\Omega); higher-order Sobolev spaces use weak derivatives indexed by multi-indices. This makes differentiability compatible with norm completion and variational methods Evans, §5.2.

References
  1. Lawrence C. Evans, Partial Differential Equations, 2nd ed., American Mathematical Society, 2010. DOI record. Relevant: §5.2 on weak derivatives and Sobolev spaces.
  2. Robert A. Adams and John J. F. Fournier, Sobolev Spaces, 2nd ed., Academic Press, 2003. Publisher record. Relevant: Chapter 3 on weak derivatives.