nLab Dirac distribution




A Dirac distribution or Dirac δ\delta-distribution δ(p)\delta(p) is the distribution that is given by evaluating a function at a point pp.

It is closely related to Dirac measures, in the language of measure theory.



The distributional derivative of the Heaviside distribution Θ𝒟()\Theta \in \mathcal{D}'(\mathbb{R}) is the delta distribution δ𝒟()\delta \in \mathcal{D}'(\mathbb{R}):

Θ=δ. \partial \Theta = \delta \,.

For bC c ()b \in C^\infty_c(\mathbb{R}) any bump function we compute:

Θ(x)b(x)dx =Θ(x)b(x)dx = 0 b(x)dx =(b(x)| xb(0)) =b(0) =δ(x)b(x)dx. \begin{aligned} \int \partial\Theta(x) b(x) \, d x & = - \int \Theta(x) \partial b(x)\, dx \\ & = - \int_0^\infty \partial b(x) d x \\ & = - \left( b(x)\vert_{x \to \infty} - b(0) \right) \\ & = b(0) \\ & = \int \delta(x) b(x) \, dx \,. \end{aligned}

Fourier transform


(Fourier transform of the delta-distribution)

The Fourier transform (this def.) of the delta distribution, via this example, is the constant function on 1:

δ^(k) =x nδ(x)e 2πikxdx =1 \begin{aligned} \widehat {\delta}(k) & = \underset{x \in \mathbb{R}^n}{\int} \delta(x) e^{- 2\pi i k x} \, d x \\ & = 1 \end{aligned}

This implies by the Fourier inversion theorem (this prop.) that the delta distribution itself has equivalently, in generalized function-notation, the expression

(1)δ(x) =δ^^(x) = k ne 2πikxdk \begin{aligned} \delta(x) & = \widehat{\widehat{\delta}}(-x) \\ & = \int_{k \in \mathbb{R}^n} e^{2 \pi i k \cdot x} \, d k \end{aligned}

in the sense that for every function with rapidly decreasing partial derivatives f𝒮( n)f \in \mathcal{S}(\mathbb{R}^n) we have

f(x) =y nf(y)δ(yx)dvol(y) =y nk nf(y)e 2πik(yx)dvol(k)dvol(y) =k ne 2πikxy nf(y)e 2πikydvol(y)=(1) nf^(k)dvol(k) =f^^(x) \begin{aligned} f(x) & = \underset{y \in \mathbb{R}^n}{\int} f(y) \delta(y-x) \, dvol(y) \\ & = \underset{y \in \mathbb{R}^n}{\int} \underset{k \in \mathbb{R}^n}{\int} f(y) e^{2 \pi i k \cdot (y-x)} \, dvol(k)\, dvol(y) \\ & = \underset{k \in \mathbb{R}^n}{\int} e^{- 2 \pi i k \cdot x} \underset{= (-1)^n\widehat{f}(-k)}{ \underbrace{ \underset{y \in \mathbb{R}^n}{\int} f(y) e^{2 \pi i k \cdot y} \, dvol(y) } } \, dvol(k) \\ & = \widehat{\widehat{f}}(-x) \end{aligned}

which is just the statement of the Fourier inversion theorem for smooth functions (this prop.).

Relation to point-supported distributions

It is clear that:

The delta distribution is a compactly supported distribution, and in fact a point-supported distribution.


Every point-supported distribution uu with supp(u)={p}supp(u) = \{p\} is a finite sum of multiplies of derivatives of the delta distribution at that point:

u=α n|α|kc α αδ(p) u = \underset{ {\alpha \in \mathbb{N}^n} \atop { {\vert \alpha\vert} \leq k } }{\sum} c^\alpha \partial_\alpha \delta(p)

where {c α} α\{c^\alpha \in \mathbb{R}\}_\alpha, and for kk \in \mathbb{N} the order of uu.

(e.g. Hörmander 90, theorem 2.3.4)


  • Lars Hörmander, The analysis of linear partial differential operators, vol. I, Springer 1983, 1990

Last revised on October 26, 2019 at 14:25:22. See the history of this page for a list of all contributions to it.