Fourier Integral Representation of the Dirac Delta Function Chris Clark December 31, 2009

The Problem
1 It is often claimed in the physics literature that 2 eikx dk is equal to the Dirac delta function, but this relation is not strictly true because the integral is not convergent. To see why this integral cannot converge, 1 just consider the real part, which is 2 cos(kx) dk by Eulers formula. The integral is oscillatory, so it never converges to an exact value. However, it is still possible to dene a correspondence based on a double integral. Let the function g(x) correspond to the integral f (x, k) dk i b b

PV
a

f (x, k) dx dk =
a

g(x) dx

for all nite choices of a, b. Here, P V denotes the principal value of the integral. Notice that the order of integration has been switched. Switching the order of integration back is not justied and will break the . equality. If we write this correspondence as f (x, k) dk = g(x), then we can show that

. eikx dk = 2(x)

The Correspondence
Theorem 1.
ikx e

. dx = 2(x)
b b

Proof. We must show PV

a

eikx dx dk =
a

2(x) dx

for all nite values of a, b.

b 0 b b

PV
a

eikx dx dk = P V
a

eikx dx dk + P V
0 a

eikx dx dk

Performing the substitution k k in the rst integral,

0 b b

= PV
a

ei(k)x dx d(k) + P V
0 b a b a b

eikx dx dk
0 a

= PV
0

eikx dx dk + P V

eikx dx dk

= 2P V
0 a

Re[eikx ] dx dk

Using Eulers formula,

eia

= cos(a) + i sin(a),
b

= 2P V
0 a N

cos(kx) dx dk
b

= 2 lim lim = 2 lim lim

0 N N

cos(kx) dx dk
a

0 N

sin(bk) sin(ak) k k

dk

Since the lower limit of integration never reaches zero, we do not have to worry about the division by zero.
N

= 2 lim lim

0 N

sin(bk) dk k

sin(ak) dk k

Now we do the substitution u = bk in the rst integral and u = ak in the second.

bN

= 2 lim lim

0 N

sin(u) du u

aN a

sin(u) du u

If either of the variables a and b are negative, the limits become 0 to , which negates the value of the integral since the integrand is symmetric about zero. Therefore,
N

= 2 lim lim

0 N

sgn(b)

sin(u) du sgn(a) u

sin(u) du u

The integral here is called the sine integral where the upper limit of integration is the parameter to the sine integral function. = 2(sgn(b) sgn(a))Si() = (sgn(b) sgn(a)) since Si() = /2. Now, we notice that
b a 2(x)

dx = (sgn(b) sgn(a)), which completes the proof.

The Fourier Transform

As in illustration of the usefulness of this correspondence, we will (non-rigorously) derive the expression for the Fourier transform by assuming that the correspondence behaves like an equality when double integrals are present, and assuming that the order of integration is interchangeable. Starting with the sifting property of the Dirac delta function,

(x) =

(x )(x x ) dx 1 2

(t)

eik(xx ) dk

dx

1 = 2

1 2

(t)eikx dx eikx dk (k)eikx dk

1 = 2 where 1 (k) = 2

(x )eikx dx

is the Fourier transform of (x).