§ Derivative of step is dirac delta
I learnt of the "distributional derivative" today from my friend, Mahathi.
Recording this here.
§ The theory of distributions
As far as I understand, in the theory of distributions, A distribution
is simply a linear functional F:(R→R)→R.
The two distributions we will focus on today are:
- The step distribution, step(f)≡∫0∞f(x)dx
- The dirac delta distribution, δ(f)=f(0).
- Notationally, we write D(f) where D is a distribution, f is a function as ∫D(x)f(x)dx.
- We can regard any function as a distribution, by sending f to F(g)≡∫f(x)g(x)dx.
- But it also lets us cook up "functions" like the dirac delta which cannot actually exist as a function. So we move to the wider world of distributions
§ Derivative of a distribution
Recall that notationally, we wrote D(f) as ∫D(x)f(x)dx. We now want
a good definition of the derivative of the distribution, D′. How? well,
use chain rule!
∫0∞UdV=[UV]∣0∞−∫0∞VdU∫0∞f(x)D′(x)dx=[f(x)D(x)]∣0∞−∫0∞Df′(x)
Here, we assume that f(∞)=0, ahd that f is differentiable at 0.
This is because we only allow ourselves to feed into these distribtions
certain classes of functions (test functions), which are "nice". The test
functions f (a) decay at infinity, and (b) are smooth.
The derivation is:
∫0∞UdV=∫0∞f(x)δ(x)=f(0)[UV]∣0∞−∫0∞VdU=[f(x)step(x)]∣0∞−∫0∞step(x)f′(x)=[f(∞)step(∞)−f(0)step(0)]−step(f′)=[0−0]−(∫0∞f′(x)dx)=0−(f(∞)−f(0))=0−(0−f(0))=f(0)
- Thus, the derivative of the step distribution is the dirac delta distribution.