Distributions and Generalized Functions

In many physical applications, we encounter objects like the Dirac delta function that are not “functions” in the classical sense. Distribution theory provides a rigorous mathematical framework for these entities.

The Problem with Classical Functions

A classical function $f: \mathbb{R} \to \mathbb{R}$ has a value at every point. However, an idealized point mass or a sudden impulse has a “value” of infinity at a point and zero elsewhere, yet its integral is finite.

Distributions as Linear Functionals

A distribution $T$ is a continuous linear functional on the space of test functions $D$ (smooth functions with compact support). We denote the action of a distribution $T$ on a test function $\phi$ as $\langle T, \phi \rangle$.

The Dirac Delta Distribution

The Dirac delta $\delta$ is defined by its effect under an integral (or pairing): $\langle \delta, \phi \rangle = \phi(0)$ For any test function $\phi$.

import numpy as np

def gaussian_approximation(x, epsilon):
  """The Dirac delta can be viewed as the limit of a zero-centered Gaussian."""
  return (1.0 / (epsilon * np.sqrt(np.pi))) * np.exp(-(x/epsilon)**2)

x_vals = np.linspace(-1, 1, 100)
# As epsilon -> 0, the peak gets higher and narrower
peak_01 = gaussian_approximation(0, 0.1)
peak_001 = gaussian_approximation(0, 0.01)

print(f"Peak with epsilon=0.1:  {peak_01:.2f}")
print(f"Peak with epsilon=0.01: {peak_001:.2f}")
print("The area under these curves is always 1.")

Weak Derivatives

Distribution theory allows us to define the derivative of any locally integrable function. If $f$ is a distribution, its weak derivative $f'$ is defined by: $\langle f', \phi \rangle = -\langle f, \phi' \rangle$ This formula is inspired by the integration by parts formula where boundary terms vanish for test functions.

Sobolev Spaces

Sobolev spaces are spaces of functions whose weak derivatives exist and belong to $L^p$ spaces. They are essential in the study of partial differential equations.