Dirac delta function

Notes for FYS3140

Susanne Viefers,
Dept of Physics, University of Oslo

March 23, 2017

Abstract
This is a summary of the lecture on the Dirac delta function – a very useful mathematical tool
that appears many places in physics.

1 Introduction
The pedestrian ”definition” of the delta function δ(x − a) frequently encountered is that it is infinite
at x = a and zero everywhere else, in such a way that its area is normalized,
Z ∞
δ(x − a) dx = 1 (1)
−∞

and with the defining property

Z ∞
δ(x − a) f (x) dx = f (a). (2)
−∞

In physics context, one may think of such an object as representing an idealized version of some
sharp pulse of very short duration. However, the above is not a proper definition of a function, and
the operations (1) and (2) are not well-defined as they stand. The purpose of these notes is to show
how to define delta functions in a more proper way, and how to make sense of such calculus-lookalike
operations. The upshot will be that expressions like (1) and (2) are understood as short-hand
notation for a limit that is perfectly well-defined, and because of that, the results are valid.

2 Delta sequences
There exist sequences of strongly peaked functions {φn (x)} (n = 1, 2, 3...) such that
Z ∞
lim φn (x − a) dx = 1 (3)
n→∞ −∞

and Z ∞
lim φn (x − a) f (x) dx = f (a). (4)
n→∞ −∞

The limits of φn themselves do not exist, they are ”meaningful” only inside integrals. As a concrete
example consider 
0, |x| ≥ 1/n
φn (x) = (5)
n/2, |x| < 1/n

1
 and a = 0 for simplicity. It is easy to see that φn are normalized for all n,
Z ∞
n 1/n
Z
n2
φn (x) dx = dx = = 1. (6)
−∞ 2 −1/n 2n

Next, consider
Z ∞ Z 1/n
n
φn (x)f (x) dx = f (x)dx. (7)
−∞ 2 −1/n

According to the mean value theorem for integrals 1 , there must be some y, −1/n ≤ y ≤ 1/n, such
that
n 1/n
Z  
n 1 1
f (x)dx = − (− ) f (y) = f (y). (8)
2 −1/n 2 n n
Obviously, y → 0 in the limit n → ∞, so f (y) → f (0), and thus
Z ∞
lim φn (x) f (x) dx = f (0). (9)
n→∞ −∞

So this sequence obeys the relations (3) and (4), i.e. acts as a delta function in the limit n → ∞.
Some other possible delta sequences are
n 1
φn (x) = (10)
π 1 + n2 x2
n 2 2
φn (x) = √ e−n x (11)
π
1 sin2 nx
φn (x) = . (12)
nπ x2

3 Heaviside step function

The Heaviside step function is another useful object in many contexts in physics, and is defined as

0, t < a
H(t − a) = (13)
1, t > a.
It can be related to delta sequences in the following way. Consider
Z x
ψn (x) = φn (x0 )dx0 (14)
−∞

which means that dψn /dx = φn . Now, consider again the delta sequence (5). We see that for
x < −1/n Z x
φn (x0 )dx0 = 0 (15)
−∞

while for x > 1/n, Z x

φn (x0 )dx0 = 1. (16)
−∞
In other words,
lim ψn (x) = H(x). (17)
n→∞
Although we used a concrete example here, this can be shown to be true in general for valid delta
sequences. Since we noted previously that dψn /dx = φn , this is the properly defined version of the
often-encountered statement that the derivative of the step function is the delta function,

H 0 (x) = δ(x). (18)

1
Rb
Let f (x) be continuous on [a, b]. Then there exists c ∈ [a, b] such that a
f (x)dx = f (c) (b − a).

2
4 Delta calculus
By ”delta calculus” one means doing calculus-like operations involving delta functions (Like (1) and
(2) ), but with the implicit understanding that this is just a shortcut method (shorthand notation)
for well-defined expressions involving limits of delta sequences – as discussed above. Because of this,
doing calculus with delta functions does produce valid results. This also includes integration by
parts. Since there exist delta sequences of differentiable functions, e.g. (11), we perform integration
by parts on the integral
Z ∞ Z ∞
dφn (x) ∞ df (x)
f (x)dx = [φn (x) f (x)]−∞ − φn (x) dx. (19)
−∞ dx −∞ dx

The surface term vanishes, since for the functions we deal with, φn will suppress it at ±∞. In the
limit n → ∞, the right-hand side is then like (9), only with f (x) replaced by df /dx. Thus,
Z ∞
dφn (x)
lim f (x)dx = −f 0 (0). (20)
n→∞ −∞ dx

This is what is implicitly meant by the well-known, more compact expression

Z ∞
δ 0 (x) f (x)dx = −f 0 (0), (21)
−∞

where one ”pretends” to do integration by parts on a delta function. By a shift of variables we get
the more general result Z ∞
δ 0 (x − a) f (x)dx = −f 0 (a). (22)
−∞

This further generalizes to higher derivatives (see also weekly problems)

Z ∞
δ (k) (x − a) f (x)dx = (−1)k f (k) (a). (23)
−∞