Fresnel Integrals

David Sirajuddin
Itcanbeshown.com
May 29, 2008
The following integral
_

0
cos x
2
dx (1)
can be evaluated by way of complex calculus (see An Engineers Guide to Complex Integra-
tion). At rst, the problem can prove dicult due to the odd choice of complex contour one
must use in order to obtain a solution. This integral is known as a Fresnel integral and arises
in the eld of optics in the description of near eld Fresnel diraction. While the function
is transcendental when evaluating the integral over denite limits, a solution can be found
when the bounds are treated as semi-innite. In fact, the convergence of the real integral
over semi-innite bounds is suggested when looking at a trace of the function (Figure 1).
Figure 1 - A plot is shown of the Fresnel cosine function. The frequency increases with x.
As evident from above, the frequency of the function increases with x, until the wavelength
of the function tends to zero as x . The function oscillates above and below the x-axis,
suggesting that it is possible in the limit for large x, that sucient contributions from the
area sweeped out by the function will be cancelled out by its negative and positive portions,
leading to a nite result. This is precisely the case, and this nite value of the integral is
found via complex integration.
1
Sirajuddin, David Fresnel Integrals
The real-valued function f(x) = cos x
2
is transposed into the complex domain as a
complex exponential f(z) = exp(iz
2
). The complex function f is identied to hold no
singularities; however, this only suggests that the complex closed contour integral of this
function about any domain is zero, not that all path integrals making up the closed contour
are themselves zero. Thus, a solution could still be obtained in this manner. In order
to choose a proper contour, begin by examining the behavior of the complex exponential
function in the complex plane, this is qualitatively shown in the gure below:
Figure 2 - The behavior of the function e
iz
2
is examined. In (a) the function is described along a positively
oriented closed circular contour. The regions where the function decays and increases are labelled. In (b), a
proposed quarter circle contour is shown that evades the increasing regions of the function. And, in (c) an
eighth circle is given, which is a contour that allows for the integration to yield a proper, nite value.
It is seen by inspection that, when examined along a positively oriented closed circular con-
tour, the complex exponential function exp(iz
2
) increases and decays in dierent quadrants.
The integral of the function in a region of growth is not capable of admitting a nite result
since the limits of integration (and hence the radius of any circular path) must eventually be
extended to innity in accordance with the original problem. However, it would seem that
it is still possible to integrate, so long as the increasing regions of the function are avoided.
Thus, a rst choice of contour could be the quarter-circle illustrated in Figure 2(b). However,
when choosing this contour, applying residue theorem, and taking limits, the solution to the
problem does not yield a value, but rather an identity:
_

0
cos x
2
dx =
_

0
sin x
2
dx (2)
This is certainly nice to know, but a value of both these functions integrated over the
stipulated limits is still yet to be found. In order to integrate the function over the desired
bounds so as to obtain a nite result, the 1/8th circle, shown in Figure 5(c), is used.
Applying residue theorem to the system provides the following statement
2
Sirajuddin, David Fresnel Integrals
_
C
e
iz
2
dz = 0
where C is used to denote the closed contour shown in Figure 2(c). Notice how since there
are no residues contained within the contour, the contour integral of the function is equal
to zero. That is to say, residue theorem is reduced to Cauchys integral theorem in this
instance. The integral can further be broken up into a summation of path integrals, where
the sum of the paths is equivalent to the contour C.
_
C
1
e
iz
2
dz
. .
I
+
_
C
2
e
iz
2
dz
. .
II

_
C
3
e
iz
2
dz
. .
III
= 0 (3)
The integrals have been labelled as I, II, and III for convenient referencing, and each path
C
1
, C
2
, and C
3
are as shown in Figure 5(c). Each integral is evaluated below.
Integral I
The only work that needs to be done on this integral is to parameterize it along the real
line, and take limits. Thus,
_
C
1
e
iz
2
dz =
_
R
0
e
ix
2
dx
since f(z) = f(x, y), and y = 0 for all points on the real line. Letting R , Integral I
becomes
_

0
e
ix
2
dx. (4)
where it is noted that the real part of this integral can be taken to make this equation of the
same form as the original integral in the problem statement (Eqn. (4)).
Integral II
Since integral II has a complex-valued, nonconstant path, it would be convenient to prove
this integral tends to zero. This integral does indeed turn out to be identical to zero, by
way of Jordans lemma. This is shown by initially factoring the function in the integrand in
the following way
_
C
2
e
iz
2
dz =
_
C
2
e
i(z
2
z)
e
iz
dz =
_
C
2
g(z)e
iz
dz.
The above equation is now in the form of the statement in Jordans lemma, where g(z) =
e
iz(z1)
. Since the function g is entire, that is, it is analytic for all points in the complex
plane, it meets the requirements for Jordans lemma. Thus, it can be said that
_
C
2
e
iz
2
dz = 0 (5)
3
Sirajuddin, David Fresnel Integrals
Integral III
For the third integral, the following parametrization is introduced
z(t) = e
i(/4)t
=
1 + i

2
t (0 t t
0
)
it then follows that
dz =
1 + i

2
dt
for an appropriate value of t
0
, such that dist|0, z(t
0
)| = R. Inputting these substitutions into
integral III in Eqn. (3)
_
C
3
e
iz
2
dz =
1 + i

2
_
t
0
0
exp
_
i
_
e
i

4
t
_
2
_
dt
=
1 + i

2
_
t
0
0
exp
_
i
_
e
i

4
t
2
_
2
_
dt
=
1 + i

2
_
t
0
0
exp(i
2
t
2
)dt
=
1 + i

2
_
t
0
0
e
t
2
dt
Furthermore, if t
0
is allowed to extend to innity then
_
C
3
e
iz
2
dz =
1 + i

2
_

0
e
t
2
dt
The parametrization has rendered the original integral into a Gaussian integral. The above
integral can be evaluated in a number of ways. Such methods include Feynmanns so-called
parametric integration, or by integrating the square of the integral in polar coordinates, or
by using the gamma function (t). It is identied that the integral above is equivalent to
(3/2) =

/2. Thus, the solution to the above equation can be written as

2
_

0
e
t
2
dt =
_
1 + i

2
_

2
=
1
2
_

2
+
i
2
_

2
(6)
Inserting Eqns. (4), (5), and (6) into (3), and solving for the real integral reveals
_

0
e
ix
2
dx =
1
2
_

2
+
i
2
_

2
Taking the real part of the above equation, and combining this result with the identity shown
in Eqn. (2), the solution to integral (1) is found to be
_

0
cos x
2
dx =
_

0
sin x
2
dx =
1
2
_

2
(7)
4