The Dirac Delta Function

The Dirac delta function $δ (x)$ is a generalized function (distribution) that is zero everywhere except at $x = 0$ , where it is "infinite" in such a way that its total integral equals one. Its defining property is the sifting property: $\int_{- \infty}^{\infty} f (x) δ (x - a) d x = f (a)$ . It appears throughout physics and engineering wherever a quantity is concentrated at a single point — point masses, point charges, instantaneous impulses, and as the identity element of convolution. Introduced by Paul Dirac (1927) in quantum mechanics and made rigorous by Laurent Schwartz (1945–1950) via the theory of distributions.

The Dirac delta is not a function in the classical sense: no ordinary function can be zero almost everywhere yet integrate to a nonzero value. Instead it is properly understood as a distribution — a continuous linear functional acting on smooth test functions — or informally as the limit of a sequence of increasingly tall, increasingly narrow "spike" functions of unit area. It matters because it lets us write point sources, impulses, and discrete events in the same integral and differential language used for continuous distributions, dramatically unifying the mathematics of physics. The most surprising fact: although $δ$ cannot be evaluated at a point, it can be differentiated infinitely often (giving $δ^{'}, δ^{''}, \dots$ ), and it is the derivative of the Heaviside step function. This single object turns hard discontinuous problems into clean algebraic manipulations.

The Core Insight

The delta function is defined entirely by what it does inside an integral, not by its pointwise values. To know $δ$ is to know that it "samples" or "sifts out" the value of any well-behaved function at a single point. Everything else — its scaling rules, its derivatives, its Fourier transform — follows from this one sifting property. Stop thinking of $δ (x)$ as a graph and start thinking of it as an operation.

Formal Statement

The Dirac Delta as a Distribution

The Dirac delta $δ$ is the continuous linear functional on the space $D$ of smooth, compactly supported test functions $φ$ defined by

⟨ δ, φ ⟩ = φ (0) .

⟳This is the definition —

δ

assigns to each test function

φ

its value at the origin. Linearity and continuity (in the topology of

D

) are immediate.

Symbols: $φ$ is an infinitely differentiable function vanishing outside a bounded set; $⟨ \cdot, \cdot ⟩$ denotes the action of a distribution on a test function. The shifted version $δ (x - a)$ satisfies $⟨ δ (\cdot - a), φ ⟩ = φ (a)$ .

The informal but ubiquitous integral notation encodes the sifting property:

\int_{- \infty}^{\infty} f (x) δ (x - a) d x = f (a) .

⟳Read the integral as shorthand for the functional action

⟨ δ (\cdot - a), f ⟩

; it equals

f

evaluated at the point where the delta's argument vanishes.

The two "informal axioms" often quoted are:

δ (x) = 0 for x \neq = 0, \int_{- \infty}^{\infty} δ (x) d x = 1.

⟳These follow by choosing test functions:

f \equiv 1

near

0

gives unit area, while support away from

0

contributes nothing to the sifting integral.

Defining Properties of the Delta Function

For suitable functions $f$ and constants $a$ , the delta satisfies:

δ (- x) = δ (x) (even), δ (a x) = \frac{1}{∣ a ∣} δ (x) .

⟳Substitute

u = a x

in the sifting integral; the Jacobian

1/∣ a ∣

and the modulus from reversing limits when

a < 0

produce the factor

1/∣ a ∣

x δ (x) = 0, f (x) δ (x - a) = f (a) δ (x - a) .

⟳Test both sides against an arbitrary

φ

⟨ x δ, φ ⟩ = ⟨ δ, x φ ⟩ = (x φ) ∣_{x = 0} = 0

δ (g (x)) = i \sum \frac{δ ( x - x _{i} )}{∣ g ^{'} ( x _{i} ) ∣}, where g (x_{i}) = 0, g^{'} (x_{i}) \neq = 0.

⟳Near each simple root

x_{i}

, linearize

g (x) \approx g^{'} (x_{i}) (x - x_{i})

and apply the scaling rule [4] to each piece.

The delta function as a limit of unit-area spikes (Gaussian nascent deltas)

Why It Holds / Derivation

The Delta as a Limit of Nascent Delta Functions

We construct $δ$ as a limit of ordinary functions and verify the sifting property.

Step 1 — Choose a nascent delta. Let

δ_{ε} (x) = \frac{1}{ε π} e^{- x^{2} / ε^{2}}

, a Gaussian of width

\sim ε

\int_{- \infty}^{\infty} δ_{ε} (x) d x = \frac{1}{ε π} \int_{- \infty}^{\infty} e^{- x^{2} / ε^{2}} d x = 1 for all ε > 0.

⟳Substitute

u = x / ε

and use the Gaussian integral

\int e^{- u^{2}} d u = π

Step 2 — Apply to a continuous test function. For any

f

continuous at

0

\int_{- \infty}^{\infty} f (x) δ_{ε} (x) d x = \frac{1}{π} \int_{- \infty}^{\infty} f (ε u) e^{- u^{2}} d u .

⟳Same substitution

u = x / ε

d x = ε d u

, which cancels the prefactor's

ε

Step 3 — Take the limit. As

ε \to 0

f (ε u) \to f (0)

pointwise, and dominated convergence gives

ε \to 0 lim \int_{- \infty}^{\infty} f (x) δ_{ε} (x) d x = f (0) \cdot \frac{1}{π} \int_{- \infty}^{\infty} e^{- u^{2}} d u = f (0) .

⟳Pull out

f (0)

and recognize the remaining normalized Gaussian integral as

1

Step 4 — Identify the limit. The functional

f \mapsto lim_{ε \to 0} \int f δ_{ε}

equals

f \mapsto f (0)

, which is exactly

δ

in [1]. The same limit results for rectangular, Lorentzian, or

sinc

nascent deltas — so

δ

is independent of the approximating family.

■

Paul Adrien Maurice Dirac | 1902–1984

Dirac introduced the delta function in his 1927 paper and his 1930 monograph The Principles of Quantum Mechanics as a tool to handle continuous spectra and to express orthonormality of position eigenstates, $⟨ x ∣ x^{'} ⟩ = δ (x - x^{'})$ . He called it an "improper function" and was fully aware it was not a function in the classical sense, justifying it by the consistency of the results it produced. Trained as an electrical engineer before turning to physics, Dirac prized formal manipulation that "worked," and his confident, almost cavalier use of $δ$ challenged mathematicians for two decades — culminating in Laurent Schwartz's rigorous theory of distributions, which vindicated Dirac entirely.

Building Intuition

Picture pouring exactly one liter of water onto a flat table. Spread thin, it covers a wide area at low height — that is a broad nascent delta. Now confine it to a smaller and smaller puddle: the height shoots up while the volume stays fixed at one liter. The delta is the idealized limit — infinitely tall, infinitely thin, total "amount" exactly one. The crucial mental shift is that you never care about the height; you only care about how much "stuff" gets deposited when you integrate. When you multiply $δ (x - a)$ by a function $f$ and integrate, the spike acts like a probe that asks, "What is $f$ doing right here at $a$ ?" and reports back $f (a)$ , ignoring $f$ everywhere else. This is why $δ$ is the perfect language for anything point-like: a point charge has all its charge at one location, an idealized hammer-strike delivers all its impulse in one instant.

Worked Examples

Sifting and the Scaling Rule

Given: Evaluate

I = \int_{- \infty}^{\infty} (3 x^{2} + 2) δ (2 x - 4) d x

. Step 1 — Locate the root. The argument

2 x - 4 = 0

x = 2

. Step 2 — Apply the scaling rule [6] with

g (x) = 2 x - 4

g^{'} (x) = 2

, so

δ (2 x - 4) = \frac{1}{∣2∣} δ (x - 2)

. Step 3 — Sift.

I = \frac{1}{2} \int_{- \infty}^{\infty} (3 x^{2} + 2) δ (x - 2) d x = \frac{1}{2} (3 \cdot 4 + 2) .

⟳Factor out

1/2

, then the sifting property replaces

x

2

in the polynomial.

Step 4 — Compute.

\frac{1}{2} (12 + 2) = \frac{14}{2}

I = 7 (dimensionless)

Impulse on a Mass — Physical Units

Given: A mass

m = 0.50 kg

at rest receives a force

F (t) = J δ (t - t_{0})

with impulse magnitude

J = 3.0 N\cdots

. Find the velocity just after the impulse. Step 1 — Newton's law.

m \frac{d v}{d t} = F (t) = J δ (t - t_{0})

. Step 2 — Integrate across the impulse over

[t_{0}^{-}, t_{0}^{+}]

m (v (t_{0}^{+}) - v (t_{0}^{-})) = \int_{t_{0}^{-}}^{t_{0}^{+}} J δ (t - t_{0}) d t = J .

⟳The integral of

δ

over any interval containing its spike is

1

; the smooth velocity term contributes the jump on the left.

Step 3 — Solve. With

v (t_{0}^{-}) = 0

v (t_{0}^{+}) = J / m = (3.0 N\cdots) / (0.50 kg)

v (t_{0}^{+}) = 6.0 m/s

The delta encodes the instantaneous nature of an idealized collision: finite velocity change in zero time.

Common Misconceptions

$δ (x)$ is not a function and $δ (0)$ is not a number. Writing "

δ (0) = \infty

" is a heuristic, not a value you may use in arithmetic. Likewise, products of deltas at the same point are undefined:

δ (x)^{2}

has no meaning in distribution theory, because distributions cannot generally be multiplied. Statements like

\int δ (x)^{2} d x

are ill-defined (and the heuristic "answer"

δ (0)

is precisely why). Always interpret

δ

through its action on a test function, never by evaluating it pointwise.

Why No Ordinary Function Can Be the Delta

Suppose $δ$ were an ordinary (Lebesgue-integrable) function equal to $0$ for all $x \neq = 0$ . Then it differs from the zero function only at the single point $x = 0$ , a set of measure zero. By the standard theory of Lebesgue integration, any function that is zero almost everywhere has integral zero:

\int_{- \infty}^{\infty} δ (x) d x = 0 \neq = 1.

⟳Modifying a function on a measure-zero set never changes its Lebesgue integral; hence the unit-area requirement is impossible for a genuine function.

This contradiction with [3] proves

δ

cannot live among ordinary functions — it requires the larger framework of distributions.

Connections

The delta function ties together many areas. It is the derivative of the Heaviside Step Function: $H^{'} (x) = δ (x)$ in the distributional sense, since integrating $δ$ produces the unit jump. It is the identity element of Convolution: $(f * δ) (x) = f (x)$ , making it the natural starting point for Green's Functions, which are responses to a delta-source. Its Fourier Transform is the constant $1$ (and conversely the transform of $1$ is $2 π δ$ ), revealing that an infinitely localized signal contains all frequencies equally. The rigorous home of $δ$ is the Theory of Distributions, and in Measure Theory it corresponds exactly to the Dirac measure $δ_{a} (A) = 1$ if $a \in A$ , else $0$ .

Real-World Applications

Quantum mechanics — Position and momentum eigenstates obey $⟨ x ∣ x^{'} ⟩ = δ (x - x^{'})$ and $⟨ p ∣ p^{'} ⟩ = δ (p - p^{'})$ , expressing continuous-spectrum orthonormality.
Electromagnetism — A point charge $q$ at $r_{0}$ has charge density $ρ (r) = q δ^{3} (r - r_{0})$ , feeding directly into Poisson's equation $\nabla^{2} ϕ = - ρ / ε_{0}$ .
Signal processing — The unit impulse $δ (t)$ defines a system's impulse response $h (t)$ ; the output for any input is the convolution $x * h$ , the foundation of LTI system theory.
Structural mechanics — A point load on a beam is modeled as $w (x) = P δ (x - a)$ in the Euler–Bernoulli equation, yielding discontinuities in shear.
Probability — A discrete or mixed random variable has a probability density containing deltas, e.g. $p (x) = \frac{1}{2} δ (x) + \frac{1}{2} g (x)$ for a point mass plus a continuous part.
Crystallography — An ideal infinite lattice has density $\sum_{n} δ (r - R_{n})$ (a Dirac comb), whose Fourier transform produces the reciprocal lattice and sharp diffraction peaks.
Sampling theory — The Dirac comb $\sum_{n} δ (t - n T)$ models ideal sampling, underpinning the Nyquist–Shannon theorem.

Historical Development

1822 — Joseph Fourier writes the Fourier integral theorem in a form implicitly containing the integral representation of the delta.
1828 — George Green introduces "Green's functions," responses to point sources, foreshadowing delta methods in potential theory.
1880s — Gustav Kirchhoff and Oliver Heaviside use impulse-like quantities informally in physics and electrical engineering.
1894 — Heaviside's operational calculus manipulates step and impulse functions, drawing sharp criticism from rigor-minded mathematicians.
1927 — Paul Dirac introduces $δ (x)$ explicitly in his quantum mechanics papers, listing its key properties.
1930 — Dirac's The Principles of Quantum Mechanics popularizes the "improper function" $δ$ worldwide.
1936 — Sergei Sobolev develops generalized functions while studying hyperbolic PDEs, a key precursor to distributions.
1945–1950 — Laurent Schwartz formalizes the theory of distributions, rigorously defining $δ$ as a continuous linear functional; he earns the Fields Medal (1950) partly for this work.
1955 — Israel Gelfand and Georgi Shilov publish Generalized Functions, systematizing the subject for physics and applied mathematics.

Subtleties & Limits

While the linear theory of distributions is complete, multiplying distributions remains fundamentally problematic: products like $δ \cdot δ$ or $δ \cdot H$ are not defined within Schwartz's framework, a serious obstacle in nonlinear PDEs, quantum field theory (where products of field operators produce divergences requiring renormalization), and general relativity (defining distributional curvature for singular spacetimes). Colombeau algebras and related theories of "generalized functions" attempt to allow consistent multiplication, but they sacrifice some natural properties and have not achieved universal adoption. Deciding the "right" framework for nonlinear operations on singular objects — and giving rigorous meaning to formal manipulations physicists perform daily — remains an active area at the intersection of analysis and mathematical physics.