Suppose that $f$ is continuously differentiable on $\mathbb{R}$ except at $x_1,\cdots,x_m$ where $f$ has jump discontinuities, and that its pointwise derivative $df/dx$ (defined except at $x_1,\cdots,x_m$) is in $L_{loc}^1(\mathbb{R})$. Prove that the distributional derivative of $f$ is given by: $$f' = (df/dx)+\sum_{j=1}^m[f(x_j+)-f(x_j-)]\tau_{x_j}\delta$$

So let $f \in L_{loc}^1(\mathbb{R})$ and wlog, let $x_1 < \cdots < x_m$. I know the following:

$$<f',\phi> = -<f,\phi'> = -\int_\mathbb{R} f\phi' dx = -\int_{-\infty}^{x_1}f\phi' dx - \int_{x_1}^{x_2}f\phi' dx - \cdots -\int_{x_m}^{\infty}f\phi' dx $$

Now using integration by parts for each term, I (think) I get:

$$\int_{-\infty}^{\infty} \frac{df}{dx}\phi(x) dx + \sum_{j=1}^m [f(x_j+)-f(x_j-)]\phi(x_j)$$

It's close, but I don't know how to make it equal to the above expression. Any help would be greatly appreciated!

  • $\begingroup$ I would substract $\sum_{j=1}^m[f(x_j+)-f(x_j-)]\tau_{x_j} H$ to $f$, where $H$ is the Heaviside function. the obtained function $g$ is continuous and piecewise $C^1$, hence $g(x) = g(0) + \int_0^x g'(t)dt$, and the distributional derivative of $\sum_{j=1}^m[f(x_j+)-f(x_j-)]\tau_{x_j} H$ is clearly $\sum_{j=1}^m[f(x_j+)-f(x_j-)]\tau_{x_j} \delta$ $\endgroup$ – reuns May 18 '16 at 20:03

You are basically finished - note that you can write $$ \sum_{j = 1}^{m}[f(x_j^+) - f(x_j^{-})]\phi(x_j) = \langle \sum_{j = 1}^{m}[f(x_j^+) - f(x_j^{-})]\tau_{x_j}\delta, \phi\rangle $$ where $\langle\tau_{x_j}\delta, \phi\rangle = \phi(x_j)$ denotes the Dirac measure at $x_j$. Now combine with $$ \int_{-\infty}^{\infty}\frac{df}{dx}\phi(x)d x = \langle \frac{df}{dx}, \phi\rangle. $$


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.