Math 122B: Complex Variables

Types of Real Integrals

I. Integrals of the form P.V.
_

p(x)
q(x)
dx where p(x) and q(x) are polynomials and
q(x) has no zeros (for < x < )
We will consider the complex function f(z) =
p(z)
q(z)
and evaluate its integral along the fol-
lowing contour :
which consists of the line L
R
(R < x < R) and of

C
R
, the upper-half circle centered at the
origin.
Let R be large enough that all poles of f(z) in the upper-half plane (z
1
, ..., z
n
) are located
inside the closed contour . Then, using residue theory,
_

f(z) dz = 2i
n

k=1
Res
z=z
k
f(z).
The value of the integral is found by splitting the integral over into two parts. In the limit
as R , the integral over L
R
will give the desired real integral, and the integral over

C
R
will go to zero. Following this plan, let be the constant 2i

n
k=1
Res
z=z
k
f(z), so the above
equality becomes
_
L
R
f(z) dz +
_
b
C
R
f(z) dz = .
Then, parameterizing L
R
by z = x (R < x < R), we have
_
R
R
f(x) dx +
_
b
C
R
f(z) dz = .
Finally, we take the limit of both sides as R :
_

f(x) dx + lim
R
_
b
C
R
f(z) dz = .
All that remains is to show that
_
b
C
R
f(z) dz 0. This is done by estimating the integral as
follows:

_
b
C
R
f(z) dz

C
R
| max
z
b
C
R
|f(z)| Rmax
|z|=R
|f(z)|
Since f(z) =
p(z)
q(z)
, we estimate |f(z)| by using the triangle inequality to bound |p(z)|, but in
order to bound
1
|q(z)|
, remember that we need |q(z)| larger than something! As an example,
let p(z) = z
2
+ 1 and let q(z) = 2z
6
+ 3. Using the triangle inequality, |p(z)| = |z
2
+ 1|
|z|
2
+ 1, which equals R
2
+ 1 on the circle |z| = R. For the denominator, use the inequality
|2z
6
+3| |2|z|
6
3|, which equals 2R
6
3 on the circle |z| = R (if R
6
>
3
2
). This implies that
1
|2z
6
+ 3|

1
2R
6
3
on the circle. If our nal bound for the integral is something that goes to
zero as R goes to innity that is,

_
b
C
R
f(z) dz

F(R) 0 as R
then, the limit of the integral must be 0. (In the example above, F(R) =
R
2
+ 1
2R
6
3
.) Finally,
using this we have found the value of the real integral:
_

f(x) dx = .
II. Integrals of the form P.V.
_

f(x)sin(x) dx (or P.V.

_

f(x)cos(x) dx) where f(x)

has no singularities on the real axis.
As before, we will integrate a complex function, but we wont simply replace x with z. In-
stead, we consider the complex function
g(z) = f(z) e
iz
= f(z) cos(z) + if(z) sin(z).
Using the same contour (with R large enough that all poles of f(z) in the upper-half plane
(z
1
, ..., z
k
) are inside ),
_

f(z)e
iz
dz = where = 2i
n

k=1
Res
z=z
k
g(z).
Breaking up the integral over into two parts, we have
_
R
R
f(x) cos(x) dx + i
_
R
R
f(x) sin(x) dx +
_
b
C
R
f(z)e
iz
dz = .
Again, we only need to take the limit as R and show that
_
b
C
R
f(z)e
iz
dz 0. Then,
matching up the real and imaginary parts of both sides,
_

f(x) cos(x) dx = Re and

_

f(x) sin(x) dx = Im .
To show that the integral over the half-circle disappears in the limit, we often need Jordans
inequality:
_
2
0
e
Rsin()
d <

R
(for R > 0). If f is a polynomial with large enough power in
the denominator (so with enough decay at innity), we can use the easier method of bounding
the integral, as in II but if not, we require Jordans inequality. For example, consider
_
b
C
R
ze
iz
z
2
+1
dz. Parameterizing the upper-half circle, we have

_
b
C
R
ze
iz
z
2
+ 1
dz

_
2
0
(Re
i
)e
iRe
i
R
2
e
2i
+ 1
iRe
i
d

R
2
R
2
1
_
2
0
e
Rsin
d
R
R
2
1
.
In the last step, we used Jordans inequality... without this extra decay from the integral, we
cant prove that the right-hand side goes to zero!
III. Integrals of the form
_
2
0
F(cos , sin ) d, where F(
z+z
1
2
,
zz
1
2i
) has no poles on
the unit circle.
Notes: The integral may be taken over any interval of length 2 since the function is pe-
riodic. Also, you may see integrals where, for example, goes from 0 to by symmetry,
these often equal one-half the value of the integral from 0 to 2!
The integral in this case is really a parameterized version of a contour integral on the unit
circle. Parameterizing the unit circle by z() = e
i
(0 2 or any other interval of
length 2!), we see that on the unit circle cos() and sin() can be written in terms of z as
cos() =
e
i
+ e
i
2
=
z + z
1
2
and sin() =
e
i
e
i
2i
=
z z
1
2i
.
Therefore,
_
2
0
F(cos , sin ) d =
_
C
1
F(
z+z
1
2
,
zz
1
2i
) dz = 2i
n

k=1
Res
z=z
k
F(
z+z
1
2
,
zz
1
2i
).
Make sure the residues in the sum are only for the poles inside the unit circle!!
IV. Integrals that require dierent contours.
Notes: These are integrals for which the complex function we want to integrate has a pole or a
branch point somewhere on the real axis. In either cases, a contour that avoids going through
the pole or the branch cut is needed!
Examples:

_

0
sin x
x
dx =

2
This is similar to case II, except that the complex function g(z) =
e
iz
z
has a pole at
z = 0! Consider the following contour = L
1
(

) L
2

C
R
(where R > > 0):
The function is analytic inside the contour, so for any R > > 0,
_

e
iz
z
dz = 0.
Splitting the integral up, and parameterizing L
1
by z = r ( < r < R) and L
2
by
z = r ( < r < R),

_
R

e
ir
r
(dr)
_
b
C
e
iz
z
dz +
_
R

e
ir
r
dr +
_
b
C
R
e
iz
z
dz = 0.
Using Jordans inequality, we can show that lim
R
_
b
C
R
e
iz
z
dz = 0. Since z = 0 is a simple
pole of g, we will use the Laurent series of g around 0 to show that
lim
0
_
b
C
e
iz
z
dz = i Res
z=0
g(z).
The Laurent series of g(z) is
b
o
z
+

n=0
a
n
z
n
, where b
o
= Res
z=0
g(z). Then,
_
b
C
e
iz
z
dz =
_
b
C
b
o
z
dz +

n=0
a
n
_
b
C
z
n
dz =
_

0
ie
i
e
i
d +

n=0
a
n
_
b
C
z
n
dz
= i +

n=0
a
n
_
b
C
z
n
dz.
Simply show that the last sum tends to zero (notice that each of the integrals in the
sum can be bounded by
n+1
!). Using these limits (notice that b
o
= 1 in this case; also
e
ir
e
ir
= 2i sin(r)),
2i
_

0
sin r
r
dr = i.

_

0
1

x(x +1)
dx =
Notice that

x is simply the positive square root of a postive real number, but when
we consider the complex function f(z) =
z

1
2
(z + 1)
speciying the branch cut is necessary!
We shall use the branch 0 < arg z < 2; therefore, f(z) is analytic away from the branch
cut arg z = 0 and away from the pole at z = 1. We will use the contour

:
We will take the limit 0 before taking limits in and R; therefore, it is ne to
simplify by considering the contour as pictured on page 274 of your book where = 0.
The interesting part is what happens on the two lines: L
1
is parameterized by z(x) = x+i,
and L
2
is parameterized by z(x) = x i (with
_

2
x

R
2

2
of course,
this will just become x R when we take the limit 0.)
_
L
1
f(z) dz =
_

R
2

2
(x + i)

1
2
(x + i) + 1
dx and
_
L
2
f(z) dz =
_

R
2

2
(x i)

1
2
(x i) + 1
dx
Make sure to use the branch cut when evaluating (x i)

1
2
! Since x + i is in the rst
quadrant, but xi is in the fourth, when we write these in polar coordinates using their
arguments, we have
(x + i) =

x
2
+
2
e
i arg(x+i)
x e
0
as 0
(x i) =

x
2
+
2
e
i arg(xi)
x e
2i
as 0.
Therefore, as 0,
_
L
1
f(z) dz
_
R

x
(x + 1)
dx
_
L
2
f(z) dz
_
R

x
e
i
(x + 1)
dx.
In the book, these are given as equalities since they assume the limit has already been
taken that is, L
1
is the line on the real axis, but with all limits being taken from above
the line, and L
2
is the line on the real axis with the limits taken from below. Finally, let
0 and R (and estimate to show that the integrals of f on both

C

and

C
R
go
to zero!)
2
_

0
1

x(x + 1)
dx = 2i Res
z=1
f(z) = 2i
_
(1)

1
2
_
= 2i
_
e
i

2
_
= 2