Lecture 1

Dirac Delta Function

1.1 Definition of Dirac Delta function and its

Representations
Consider the function D (x) defined by
(
1/ for − /2 ≤ x ≤ /2
D (x) = (1.1)
0 for |x| > /2

where  is a small parameter. The plot of the function is shown below:

Figure 1.1: Plot of the function defined in Eq. (1.1)

The integral of the function with respect to x is unity, i.e.,

Z ∞
D (x)dx = 1. (1.2)
−∞

Now, imagine making  smaller. As we decrease , the function gets narrower

and taller, but the integral of the function (i.e., the area under the graph of
the function) remains constant at the value 1. In the limit  → 0, the function
D (x) collapses to a single point, namely x = 0, and gets infinitely tall. So,

1
lim→0 D (x) is not a function at all and the procedure of taking the limit
 → 0 is not justified.

However, we can make the limiting procedure meaningful if we multiply

D (x) by some well-defined function f (x), integrate over x and then take the
limit  → 0. Consider the integral
Z ∞
D (x)f (x)dx
−∞

where f (x) is a well-defined function. If  is sufficiently small, the variation of

f (x) over the effective integration interval [−/2, /2] is negligible and f (x)
remains practically equal to f (0). Therefore,
Z ∞ Z ∞
D (x)f (x)dx ≈ f (0) D (x)dx = f (0). (1.3)
−∞ −∞

The smaller the value of , the better the approximation. In the limit  → 0,
the above equation is exact:
Z ∞
lim D (x)f (x)dx = f (0). (1.4)
→0 −∞

Now, we define the Dirac Delta Function δ(x) as

Z ∞ Z ∞
def
δ(x)f (x)dx = lim D (x)f (x)dx = f (0) (1.5)
−∞ →0 −∞

This equation is valid for any function f (x) defined at the origin. More
generally, δ(x − x0 ) is defined as
Z ∞
δ(x − x0 )f (x)dx = f (x0 ). (1.6)
−∞

Actually, the integral notation

Z ∞
δ(x)f (x)dx
−∞

is not justified because δ(x) is not really a function. Physically, there is

no problem since it becomes impossible to distinguish between D (x) and

2
δ(x) as soon as  becomes negligible compared to all distances involved in
a physical problem. Whenever a mathematical difficulty might arise, all we
need to do is to assume that δ(x) is actually D (x) with  extremely small
but not strictly zero.

Formally, we can express δ(x) as a limit of a sequence of proper functions:

lim D (x) ≡ δ(x). (1.7)

→0

Here D (x), which is a proper function of x, is called the representation of

the delta function. The representation we have discussed so far is the “square
function” given in Eq. (1.1). The representation is not unique. There are
other functions which approach the delta function when appropriate limits
are taken.

Gaussian Representation

Consider the function

1 2 2
D (x) = √ e−x / ( > 0). (1.8)
 π

For each value of the parameter , this function satisfies

Z ∞
D (x)dx = 1. (1.9)
−∞

This is the normalized Gaussian function whose plot is shown in the figure
below.
√
The Gaussian function has a peak at the origin. The peak has a height 1/ π
and a width of order  (exactly how the width is defined doesn’t matter). So
if  is allowed to be very small, the peak becomes very tall and very narrow.
Outside the peak the function becomes extremely small. Thus we have
1 2 2
δ(x) = lim √ e−x / . (1.10)
→0  π

3
 Figure 1.2: Plot of the Gaussian function defined in Eq. (1.8)

Mathematical Notes:
We have the integral Z ∞ √
2
e−x dx = π.
−∞

Now let us consider the following integral

Z ∞
2 2
I= e−b x +ax dx.
−∞

First, write
 a 2 a2
b2 x2 + ax = bx + − 2.
2b 4b
Therefore,
Z ∞
a2 /4b2 2
I = e e−(bx+a/2b) dx
−∞
Z ∞
a2 /4b2 2 a
= e (1/b) e−z dz (z = bx + )
−∞ 2b
√
2 2 π
= ea /4b .
b

Damped Sinusoidal Representation of the Delta Function

Consider another function

1 sin(x/)
D (x) = ( > 0). (1.11)
π x

4
 Figure 1.3: Plot of the function defined in Eq. (1.11)

A plot of the function is shown below. The function D (x) has the value
√
1/ π at x = 0 and it oscillates with decreasing amplitude as |x| increases.
The width of the central maximum is of the order of  and the period of
oscillations with respect to x is 2π.
For any value of  we have
Z ∞
D (x)dx = 1. (1.12)
−∞

Thus, the limit of the function as  → 0 has all the properties of the delta
function: it becomes infinitely large at x = 0, and infinitely rapid oscillations
as |x| increases means that the entire contribution to an integral containing
this function comes from an infinitesimal neighborhood of x = 0. We can
therefore write
1 sin(x/)
δ(x) = lim (1.13)
→0 π x

5
Other representation of the Delta Function

We can also show that

1 −|x|/
δ(x) = lim e (1.14)
→0 2
1 
δ(x) = lim (1.15)
→0 π x + 2
2

 sin2 (x/)
δ(x) = lim . (1.16)
→0 π x2

1.2 Properties of the Delta Function

It is important to note that, because of its singular character, the δ-function
cannot be the end result of a calculation, and has meaning only so long as a
subsequent integral over its argument is carried out. With this understanding
we can write down some relations between delta functions:

1. The delta function is an even function, i.e.,

δ(−x) = δ(x). (1.17)

2.
xδ(x) = 0. (1.18)

3.
1
δ(ax) = δ(x). (1.19)
|a|
Proof:
Consider the integral
Z ∞
I= δ(ax)f (x)dx.
−∞

Since the delta function is an even function it doesn’t matter if we

replace a by |a| in the argument. Thus
Z ∞
I= δ(|a|x)f (x)dx .
−∞

6
 Making the change of variable

y = |a|x

we have Z ∞
1 1
I= δ(y)f (y/|a|)dy = f (0) ,
|a| −∞ |a|
or, Z ∞ Z ∞
1
δ(ax)f (x)dx = δ(x)f (x)dx ,
−∞ |a| −∞

i.e.,
1
δ(ax) = δ(x)
|a|

4. More generally, we have

X δ(x − xi )
δ(φ(x)) =  ∂φ  (1.20)
 
i ∂x xi

where the sum runs over the xi ’s which are the simple roots of φ(x).
Proof:
Let x1 , x2 , · · · , xN be the simple roots of φ(x) (figure below):

Figure 1.4: Simple roots of φ(x).

In the neighborhood of any of the simple roots xi , we can write

φ(x) = (x − xi )ψ(x)

where ψ(xi ) 6= 0. We have


∂φ(x) 
ψ(xi ) = .
∂x x=xi

7
 Now, consider the integral
Z ∞ N Z
X xi +
δ(φ(x))f (x)dx = δ [(x − xi )ψ(xi )] f (x)dx
−∞ i=1 xi −
N Z xi +
X 1
= δ(x − xi )f (x)dx
i=1
|ψ(xi )| xi −
N Z ∞
X 1
= ∂φ
δ(x − xi )f (x)dx
i=1
| ∂x |x=xi −∞
N
X 1
= f (xi )
| ∂φ |
i=1 ∂x x=xi

The above result is obtained if we write

N
X δ(x − xi )
δ(φ(x)) =  ∂φ  . Proved. (1.21)
 
i=1 ∂x x=xi

5. A frequently used example of the above result is

1 1
δ(x2 − a2 ) = δ(x − a) + δ(x + a), (a > 0). (1.22)
2a 2a
Here
φ(x) = x2 − a2 = (x − a)(x + a).

The two simple roots of φ(x) are at x = a and x = −a. Now

 
 ∂φ 
 
 ∂x  = |2x|x=a = 2a
x=a

and  
 ∂φ 
 
 ∂x  = |2x|x=−a = 2a.
x=−a

Therefore, the above result follows.

6.
f (x)δ(x − a) = f (a)δ(x − a).

7. Z
δ(x − y)δ(y − a)dy = δ(x − a).

8
Note 1:
We have the identity
xδ(x) = 0.

The converse is also true and it can be shown that the equation

xu(x) = 0

has the general solution

u(x) = cδ(x).

Note 2:
We will now prove an identity which is particularly useful in quantum me-
chanics: Z ∞ Z ∞
1 dx
lim+ f (x)dx = P f (x) ∓ iπf (0), (1.23)
→0 −∞ x ± i −∞ x
or, in short  
1 1
lim+ =P ∓ iπδ(x), (1.24)
→0 x ± i x
where it is understood that the second of these two equations have meaning
only within an integral. The symbol P means the the principal part of an
integral where the integrand has a simple pole. The principal part is defined
as Z B Z −η Z B
dx dx
P f (x) = lim+ + f (x). (1.25)
−A x η→0 −A η x
Proof:
We can write
1 x ∓ i x i
= 2 2
= 2 2
∓ 2 . (1.26)
x ± i x + x + x + a2
Now, we have
1 
lim+ = δ(x).
→0 π x + 2
2

Therefore,

lim+ (∓)i = ∓iπδ(x). (1.27)
→0 x2 + a2

9
Next, consider the first term on the right and side of Eq. (1.26). We multiply
this term by a function f (x) which is regular at the origin and then integrate
over x. We get

Z ∞
xf (x)
lim+ 2 2
dx
→0 −∞ x + a
 Z −η Z η Z ∞ 
xf (x) xf (x) xf (x)
= lim+ lim+ 2 2
dx + lim+ 2 2
dx + lim+ dx
→0 η→0 −∞ x + a η→0 −η x + a η→0 η x2 + a2
(1.28)

Note that we take the limit over η first and the we take the limit over η. Now
consider the second integral of the above equation.
Z η
xf (x) 1 2 2 x=η

lim+ 2 2
dx = f (0) lim ln(x +  ) x=−η
= 0.
η→0 −η x + a η→0+ 2

If we now reverse the order of the evaluation of limits in Eq. (1.28), the
 → 0+ limit causes no difficulties in the other two integrals. We thus have
Z ∞ Z −η Z ∞ 
xdx xdx
lim+ 2 2
f (x) = lim+ lim+ + f (x)
→0 −∞ x + a η→0 →0 −∞ η x + a2
2
Z −η Z ∞ 
dx
= lim+ + f (x)
η→0 −∞ η x
Z ∞
1
= P f (x)dx
−∞ x

This establishes the identity.

1.3 Derivative of the Delta Function

One may define the derivative δ 0 (x) of the delta function. When  is small,
the derivative of D (x) has two peaks close to the origin, one peak being
positive and the other negative as shown in the figure below.
As  → 0, each of the peaks become very narrow and very tall, and each
of the two peaks approach very close to the origin. Now, an integration by

10
 Figure 1.5: Plot of D (x) and D0 (x).

parts gives
Z ∞ Z ∞
D0 (x)f (x)dx = [D (x)f (x)]∞
−∞ − D (x)f 0 (x)dx . (1.29)
−∞ −∞

Because D (x) tends to zero as x → ±∞, the first term on the right hand
side vanishes unless f (x) explodes violently at infinity. So by letting  → 0,
we arrive at the definition of δ 0 (x):
Z ∞ Z ∞
0
δ (x)f (x)dx = − δ(x)f 0 (x)dx = −f 0 (0). (1.30)
−∞ −∞

From this we immediately get

xδ 0 (x) = −δ(x). (1.31)

Conversely, it can be shown that the general solution of the equation

xu(x) = δ(x)

can be written as
u(x) = −δ 0 (x) + cδ(x)

11
where the second term arises from the homogeneous equation

xδ(x) = 0.

From the definition (1.30) it also follows that

δ 0 (−x) = −δ 0 (x). (1.32)

The nth order derivative of δ(x) can be defined in the same way. We find
Z ∞
δ (n) (x)f (x)dx = (−1)n f (n) (0). (1.33)
−∞

We can also prove the following properties of the derivatives of the delta
function:

δ (m) (x) = (−1)m δ (m) (−x)

xm+1 δ (m) (x) = 0
xδ (m) (x) = −(m − 1)δ (m−1) (x).

1.4 Integration of the Delta Function

Consider the indefinite integral
Z x
θ (x) = D (y)dy. (1.34)
−∞

A graph of θ (x) versus x is shown below.

As  → 0, the step in the function θ (x) gets progressively steeper, until,
finally, the function changes abruptly from 0 to 1 at x = 0. Therefore,
taking the limit  → 0 in Eq. (1.34) we have
Z x
θ(x) = δ(y)dy (1.35)
−∞

where (
1 for x > 0
θ(x) = (1.36)
0 for x < 0.
If we differentiate Eq. (1.35) with respect to x, we get
dθ(x)
= δ(x). (1.37)
dx
12
 Figure 1.6: The θ-function as an integral of the delta function

1.5 Three dimensional delta function

The three-dimensional delta function δ(~r ) is defined as
def
δ(~r ) ≡ δ(x)δ(y)δ(z). (1.38)

In other words, δ(~r ) is zero if any of the coordinates x, y and z is not equal
to zero and δ(~r ) tends to infinity at the origin, i.e., when x = 0, y = 0 and
z = 0, such that Z
δ(~r )d3 r = 1 (1.39)
volume
if the volume of integration contains the origin. We also have
Z
δ(~r )f (~r )d3 r = f (0) (1.40)

where again the volume of integration contains the origin.

Note:
• δ(~r − ~r 0 ) = δ(x − x0 )δ(y − y 0 )δ(z − z 0 )

• V δ(~r − ~r 0 )d3 r = 1
R

where the volume of integration includes the point ~r 0 . Otherwise, the

integral is zero.

• V δ(~r − ~r 0 )f (~r )d3 r = f (~r 0 )

R

if V includes the point ~r 0 .

13
A useful formula

Consider the integral

Z ∞ Z L
ikx
e dx = lim eikx dx
−∞ L→∞ −L
1
eikL − e−ikL


= lim
L→∞ ik
2 eikL − e−ikL
 
= lim
L→∞ k 2i
2
= lim sin kL
L→∞ k
sin kL
= 2π lim
L→∞ πk
= 2πδ(k),

where we have used

sin(x/)
lim = δ(x). (1.41)
→0 πx
Thus, we have the very important formula
Z ∞
eikx dx = 2πδ(k) (1.42)
−∞

In Eq. (1.42) if we integrate with respect to k, we would have δ(x) on the

right hand side,
Z ∞
eikx dk = 2πδ(x) (1.43)
−∞

Also note that in Eq. (1.42) we integrate over the full domain of x from −∞
to ∞. Making a change of variable x → −x does not change the value of the
integral. Hence we also have
Z ∞
e−ikx dx = 2πδ(k) (1.44)
−∞

Similarly, in Eq. (1.43), making the change of variable k → −k, doesn’t

change the value of the integral. So we could also write
Z ∞
e−ikx dk = 2πδ(x) (1.45)
−∞

14
Thus, in summary Z ∞
e±ikx dx = 2πδ(k), (1.46)
−∞
and Z ∞
e±ikx dk = 2πδ(x), (1.47)
−∞
In three dimensions we have
Z
~
e±ik.~r d3 r = (2π)3 δ(~k ). (1.48)
all space
Z
~ ~ 0 ).~
e±i(k−k r 3
d r = (2π)3 δ(~k − ~k 0 ). (1.49)
all space
Z
~ 0
e±ik.(~r−~r ) d3 k = (2π)3 δ(~r − ~r 0 ). (1.50)
all space

1.6 Fourier transform

We can always express a function f (x) in the form
Z ∞
f (x) = eikx f˜(k)dk (1.51)
−∞

where f˜(k) is a function of k, called the Fourier transform of f (x). From Eq.
(1.51) we can write
Z ∞ Z ∞ Z ∞
−ik0 x 0
e f (x)dx = ei(k−k )x
f˜(k)dkdx
−∞ −∞ −∞
Z ∞
= 2π δ(k − k 0 )f˜(k)dk
−∞
= 2π f (k ) ˜ 0

Thus, Z ∞
˜ 1
f (k) = e−ikx f (x)dx. (1.52)
2π ∞
The functions f (x) and f˜(k) are Fourier transform of each other. We can
write Eqs. (1.51) and (1.52) in a more symmetrical fashion as follows
Z ∞
1
f (x) = √ eikx f˜(k)dk (1.53)
2π −∞
Z ∞
˜ 1
f (k) = √ e−ikx f (x)dx. (1.54)
2π ∞

15
 In three dimensions we can write
Z
1 ~
f (~r ) = eik.~r f˜(~k )d3 k. (1.55)
(2π)3/2
~ 0 .~
Multiplying this equation by e−ik r
and integrating over ~r, we have
Z Z Z
~0 1 ~ ~0
e−ik .~r f (~r )d3 r = 3/2
d r d3 k ei(k−k ).~r f˜(~k )
3
all space (2π)
Z
1
= d3 k (2π)3 δ(~k − ~k 0 )f˜(~k )
(2π)3/2
= (2π)3/2 f˜(~k 0 )

Therefore, Z
1 ~
f˜(~k ) = e−ik.~r f (~r )d3 r. (1.56)
(2π)3/2
In summary,

Z
1 ~
f (~r ) = 3/2
eik.~r f˜(~k )d3 k, (1.57)
(2π)
Z
1 ~
f˜(~k ) = e−ik.~r f (~r )d3 r. (1.58)
(2π)3/2

16
1.6.1 Parseval Identity
We will now prove the important identity
Z Z  2
|f (~r )| d r = f˜(~k ) d3 k.
2 3
(1.59)
 

Proof:
Z Z
2 3
|f (~r )| d r = f (~r )f ∗ (~r )d3 r
Z Z Z
1 i~k.~
r ˜~ 1 ~0
= dr 3
3/2
e f (k )d k 3
3/2
e−ik .~r f˜∗ (~k 0 )d3 k 0
(2π) (2π)
Z Z
1 ~ ~0
= 3
d3 kd3 k 0 f˜(~k )f˜∗ (~k 0 ) d3 r ei(k−k ).~r
(2π)
| {z }
(2π)3 δ(~k−~k 0 )
Z
= d3 kd3 k 0 f˜(~k )f˜∗ (~k 0 )δ(~k − ~k 0 )
Z
= d3 k f˜(~k )f˜∗ (~k )
Z  2
= d3 k f˜(~k ) . (1.60)
 

The proof is now complete.

17