The Calculus of Residues

Using the Residue Theorem to evaluate integrals and sums

The residue theorem allows us to evaluate integrals without actually physically integrating i.e. it allows us to evaluate an integral just by knowing the residues contained inside a curve. In this section we shall see how to use the residue theorem to to evaluate certain real integrals which were not possible using real integration techniques from single variable calculus and how to nd the values of certain innite sums. In order to do this, we shall present a number of dierent types of integrals and sums and then the method of how to calculate them. We shall provide proofs for only certain integrals and sums.
P (x) 1. Integrals of the form Q(x) dx for a real variable x with Q(x) = 0.

Theorem 1.1. Suppose that P (x) and Q(x) are polynomials of a real variable x , Q(x) = 0 for any real x and deg (Q(x)) deg (P (x)) 2 P (x) (to guarantee the convergence of the integral Q(x) dx). Then

P (x) dx = 2i Q(x)

Res(P/Q; zk )
k

where the zi are the singularities in the upper half plane. Proof. We prove this through a number of steps. (i ) Let C denote the curve which consists of A, the semicircle of radius R centered at the origin and L the real line segment from R to R oriented counterclockwise (as illustrated).
A

(ii ) By the residue theorem, P (z) dz = 2i Q(z)

1

ResC (P/Q; zk )
k

where the zi are the singularities in the interior of C.

(iii ) Let A denote just the semicircle. Since deg (Q(x))deg (P (x)) 2, then for suciently large R, there exists a constant k such that |P (z)/Q(z)| k/R2 on A. Using the ML formula, it follows that P (z) dz Q(z) R k k = 0 2 R R

as R . (iv ) It follows that

P (x) dx = limR Q(x)

P (z) dz + Q(z)

P (z) dz Q(z)

= lim (2i
R k

ResC (P/Q; zk )) = 2i
k

Res(P/Q; zk )

where the zi are the singularities in the upper half plane.

Example 1.2. Evaluate

x2 + x dx. (x2 + 1)(x2 + 4)

By the previous result,

x2 + x dx = 2i (x2 + 1)(x2 + 4)

Res(P/Q; zk )
k

where the zi are the singularities in the upper half plane. The poles in the upper half plane are at z = i and z = 2i and both are simple poles, so we can evaluate them easily. Specically, if P (z) = z 2 + z and Q(z) = (z 2 + 1)(z 2 + 4), then we have Res(P (z)/Q(z); i) = and Res(P (z)/Q(z); 2i) = Therefore, we get

P (i) i2 + i i1 = = (i) 2 + 4) + 2i(i2 + 1) Q 2i(i 6i

4 + 2i 4 + 2i 2i P (2i) = = = . Q (2i) 4i((2i)2 + 4) + 4i((2i)2 + 1) 12i 6i

i1 2i 1 x2 + x dx = 2i( + ) = 2i = (x2 + 1)(x2 + 4) 6i 6i 6i 3

2. Integrals of the form

P (x) Q(x)

cos (x)dxfor a real variable x with Q(x) = 0.

P (x) Q(x)

sin (x)dx and

Theorem 2.1. Suppose that P (x) and Q(x) are polynomials of a real variable x , Q(x) = 0 for any real x and deg (Q(x)) deg (P (x)) 1 (to guarantee the convergence of the integral P (x) sin (x)dx and Q(x)
P (x) Q(x)

cos (x)dx). Then if R(z) = P (z)/Q(z), then

P (x) sin (x)dx = Im(2i Q(x) P (x) cos (x)dx = Re(2i Q(x)

Res(R(z)eiz ; zk ))
k

and

Res(R(z)eiz ; zk ))
k iz

where the zi are the singularities of R(z)e

in the upper half plane.

Proof. We proceed as with the last case but instead of integrating R(z) cos (z) or R(z) sin (z), we consider R(z)eiz . Let C denote the curve which consists of A, the semicircle of radius R centered at the origin and L the real line segment from R to R oriented counterclockwise (as illustrated).
A

Fix some real h > 0. Then we break up the curve A into the segments A+ = {z A|Im(z) h} and A = {z A|Im(z) < h}. We also observe the following: Since deg (Q) deg (P ) 1, we have |R(z)| K/|z| for some constant K and for suciently large z |eiz | = eRe(iz) = ey for z = x + iy. Using these observations, we have the following: (i ) eh iz R = Keh K R(z)e dz R A+ and K h R(z)eiz dz 4h = 4K R R A since clearly each segment making up A have lengths less than 2h.

(ii ) Choosing h =

M , we have R(z)eiz dz Ke
R

A+

and R(z)eiz dz
A

4K

R 4K = R R 4K + 0 R R(x)eix dx.

so it follows that R(z)eiz dz

A

Ke

as R . Thus
R

lim

R(z)eiz dz =
C

(iii ) It follows that 2i

k

ResC (R(z)eiz ; zk ) = lim

R(z)eiz dz =
C L

R(x)eix dx R(x) sin (x)dx

L

=
L

R(x)(cos(x)+i sin (x))dx =

L

R(x) cos(x)dx+i

by the residue theorem where the zi are the singularities of R(z)eiz in the upper half plane. Hence

P (x) sin (x)dx = Im(2i Q(x) P (x) cos (x)dx = Re(2i Q(x)

Res(R(z)eiz ; zk ))
k

and

Res(R(z)eiz ; zk ))
k

We illustrate again with an example. Example 2.2. Evaluate

cos (x) dx x2 + 9 sin (x) dx. x2 + 9 eiz Res( 2 ; zk )) z +9

and

By the previous result,

cos (x) dx = Re(2i x2 + 9

where the zi are the singularities in the upper half plane. There is a single simple pole in the upper half plane at z = 3i so we can evaluate the residue easily. Specically, if P (z) = eiz and Q(z) = z 2 + 4, so Res(P (z)/Q(z); 3i) = Therefore, we get

P (i) e3 = . Q (i) 6i

cos (x) e3 dx = x2 + 9 3 sin (x) dx = 0. x2 + 9

and

P (x) 3. Integrals of the form 0 Q(x) dx for a real variable x with Q(x) = 0.

Otherwise we integrate the function R(z) log (z) where R(z) = P (z)/Q(z) over the following contour C and note that as R and 0 , we have 1 P (x) R(z) log (z)dz = lim dx R C 2i 0 Q(x) using calculations similar to the previous results.
R

First note that if P (x)/Q(x) is an even function, we can use the previous result and the fact that 1 P (x) P (x) dx = dx. Q(x) 2 Q(x) 0

Therefore, we get the following: Theorem 3.1. Suppose that P (x) and Q(x) are polynomials of a real variable x , Q(x) = 0 for any real x and deg (Q(x)) deg (P (x)) 2 (to guarantee the convergence of the integral 0 P (x) dx). Then Q(x)
0

P (x) dx = Q(x)

Res(
k

P (z) log (z); zk ) Q(z)

where the zi are the singularities in the the whole complex plane. Rather than providing a complete proof, we illustrate with an example.

Example 3.2. Evaluate

0

x4

x dx. +1 Res( z log (z); zk ) +1

Here we have
0

x4

x dx = +1

z4

where the zi are the singularities in the the whole complex plane. The poles are all simple and are at the solutions of z 4 = 1 i.e. z = ei/4 , e3i/4 , e5i/4 and e7i/4 . Evaluating the residues, we have P (eki/4 ) eki/4 log (eki/4 ) = 3ki/4 log (eki/4 ) Q (eki/4) 4e ki 1 ki = = ki/2 4e 4 16eki/2 for k = 1, 3, 5, 7. Thus we have i 3i 5i 7i x dx = + + + x4 + 1 16ei/2 16e3i/2 16e5i/2 16e7i/2 0 Res(P (z)/Q(z) log (z); eki/4 ) = = i 3i 5i 7i + 16i 16i 16i 16i

4 = 16 4

4. Integrals of the form 0 Px(x) dx for a real variable x with P (x) = 0 and with 0 < < 1. In order to integrate functions of this type, we use the same contour as the last example. Though it is a little more complicated to show, we get the following result: Theorem 4.1. Suppose is a number with 0 < < 1, P (x) is a polynomial of a real variable x , P (x) = 0 for any real x and deg (P (x)) > 1 (to guarantee the convergence of the integral 0 Px(x) dx). Then
0

x 2i dx = P (x) 1 e2i(1)

Res(
k

z 1 ; zk ) P (z)

where the zi are the zeros of P (z). We illustrate with an example. Example 4.2. Evaluate
0

x2

x dx. +1 z 2 Res( 2 ; zk ) (z + 1)
1

Using the theorem, we have

0

2i x2 dx = 2 +1 x 1 ei

where the zi are the zeros of z 2 + 1. The zeros of z 4 + 1 are z = i. They are both simple poles, so we get 1 1 5i z 2 2(1 i) i 2 ei/4 1 e4 Res( 2 ; i) = = i/2 = 3i = = z +1 P (i) 2e 2 4 2e 4 and 1 1 7i 2(1 i) (i) 2 e3i/4 1 z 2 e 4 ; i) = = 3i/2 = i = = . Res( 2 z +1 P (i) 2e 2 4 2e 4 Thus we have 1 1 z 2 2 2i 2 2i x2 Res( 2 dx = ; zk ) = i = 2 +1 i x 1e (z + 1) 4 2 0 k 5. Integrals of the form 0 R(cos (), sin ())d for a Rational function R and a real variable . In order to integrate functions of this type, rst observe that
2 2

R(cos (), sin ())d =

0 |z|=1

R(cos (), sin ())

dz iz

1 1 1 1 dz (z + ), (z + ) 2 z 2i z iz |z|=1 1 1 i on |z| = 1 since if we put z = e , then cos () = 2 (z + z ) and 1 1 sin () = 2i (z + z ) and d = dz/iz. This implies the following method to calculate such integrals: = R Theorem 5.1.
2 1 1 R( 2 (z + 1 ), 2i (z + 1 )) z z ; zk Res z

R(cos (), sin ())d = 2

0 k

where the zk are the poles in the circle |z| We illustrate with an example. Example 5.2. Evaluate

1.

1 d a + cos () 0 where a > 1 (so that the integrand is always nite). We evaluate using the theorem. We have
2

R(cos (), sin ())d = 2

0 k

Res

1 1 a+ 1 (z+ z ) 2

; zk .

However, observe that

1 1 1 a+ 2 (z+ z )

z2

2 + 2az + 1

which has poles at z = a a2 1, which are both on the real axis. Next note that z = a + a2 1 is inside the unit circle since a 1 < 2 2 a 1 < a, so 1 < a + a 1 < 0 and similarly, z = a a2 1 is outside the unit circle. Thus we have
2 0

R(cos (), sin ())d = 2Res

2 ; a + a2 1 z 2 + 2az + 1 2 2 = 2 = . 21 2 a a2 1 6. Sums of Series

The last application of the residue theorem is to the number theoretical problem of summing series. In order to evaluate sums of series, we shall construct functions whose residues agree with the terms of the series and then integrate over a appropriately chosen contour and apply the residue theorem. Most of the integrals we shall perform will be over the square CN centered at the origin with vertices (see illustration below). ((N + 1/2), (N + 1/2))

In order to do this, we need the following two results. Lemma 6.1. For any N and 0, for any z on CN we have coth (/2) 1. | cot (z)|

|cosec(z)|

Proof. Putting z = x + iy and using the trigonometric sum formulas and basic identities, we have and It follows that | cos (z)|2 = cos2 (x) + sinh2 (y) | sin (z)|2 = cosh2 (y) cos2 (x). | cot (z)|2 = cos2 (x) + sinh2 (y) . cosh2 (y) cos2 (x)

On the vertical sides of CN , we have 1 x = (N + ) 2 giving so 1 cos ((N + )) = 0 2 1.

| cot (z)| = | tanh (y)| On the horizontal sides, 0 so | cot (z)|2 Then we have | cot (z)| cos2 (x) 1

cosh2 (y) sinh2 (y) + 1 = = coth2 (y). cosh2 (y) 1 sinh2 (y)

1 | coth (z)| = coth ((N + )) coth ( ). 2 2 Hence on CN , we have | coth (z)| max(1, coth ( )) = coth ( ). 2 2 We use a similar argument with cosec(z) to show that on the vertical sides we have 1 |cosec(z)| 1 | cosh (y)| and on the horizonal sides 1 |cosec(z)| 1 | sinh ( )| 2 giving the result. Lemma 6.2.

(i )

cosec(z) has a simple pole at z = n with residue (1)n for all integers n. (ii ) coth (z) has a simple pole at z = n with residue 1 for all integers n. Proof. Clearly cosec(z) = and coth (z) = sin (z)

cosh (z) sinh (z)

10

have isolated singularities at z = n for each integer n and through direct calculation, we can show that these are the only singularities of these functions (by simply nding when sin (z) = 0). Therefore, we just need to determine the residues of these singularities. For any integer n, put w = z n. Then we have cos (z) = cos (w + n) = cos (w) cos (n) sin (w) sin (n) = cos (w) cos (n) and sin (z) = sin (w + n) = sin (w) cos (n) + cos (w) sin (n) = sin (w) cos (n). Next observe that cot (z) = = (1 cos (w) cos (w) cos (n) cos (w) = = sin (w) sin (w) cos (n) sin (w) + +
(w)4 4! (w)5 5!

(w Note that

(w)2 2! (w)3 3!

+ ...)

1 1 = w 1 + ...)
(w)2 2! (w)2 3!

(w)2 2! (w)2 3!

+ +

(w)4 4! (w)4 5!

+ ... + ...

1 + + ... is an analytic function in w, so has a Taylor expansion around w = 0 with rst term 1 (since plugging w = 0 yields 1). Thus we have 1 1 w 1
(w)2 2! (w)2 3!

(w)4 4! (w)4 5!

+ ...

+ +

(w)4 4! (w)4 5!

+ ... + ...

1 1 + positive powers of w w

In particular, the Laurent expanion of z around z = n will have a simple pole with residue 1. A similar argument holds for coth (z). Lemma 6.3. The rst three terms of the Laurent expansion of cosec(z) = 1/ sin (z) around z = 0 are z 2 7z 4 1 1 1+ = + + ... . sin (z) z 6 360 Proof. Since the Taylor series of sin (z) around z = 0 is sin (z) = z we have 1 1 = sin (z) z 1
z2 3!

z3 z5 z2 z4 + + = z 1 + + ... 3! 5! 3! 5! 1 4 + z + ... 5! 1 1 z 1

11

z2 z4 + .... 3! 5! Next observe that for suciently small z (say |z| < ), || < 1 (since it is a continuous function of z which is 0 at 0). Thus for |z| < we have = 1 1 = sin (z) z 1
2 4 z2 3!

where

1 4 + z + ... 5!

1 1 z 1
2

z z z2 z4 1 1+ + ... + + ... + ... z 3! 5! 3! 5! 1 z2 z4 z4 z 2 7z 4 1 = 1+ + 2 + higher order terms = 1+ + +. . . . z 3! 5! 3! z 6 360 = With these results in consideration, we can use them to evaluate certain special types of sums. Specically, for a rational function f (z), we do the following: (i ) To evaluate the sum
n=1

f (n)

we consider the integral f (z) cot (z)dz

CN

and let N (ii ) To evaluate the sum

n=1

(1)n f (n)

we consider the integral f (z)cosec(z)dz

CN

and let N We illustrate with some examples. Example 6.4. (i ) Show that
n=1

2 1 = . n2 6 cot (z) dz. z2

We consider the integral

N

lim

CN

12

By the residue theorem, we have

N

lim

CN

cot (z) dz = 2i z2

( residues in CN ) = 2i
n=N

( residues at z = n)

so we need to evaluate the residues of cot (z) z2 in CN . First observe that 1/z 2 is analytic away from z = 0, so for z = n, cot (z) z2 will have a simple pole (since cot (z) has a simple pole). To calculate the value of the residue, since cot (z) has a simple pole at z = n with residue 1 because the value of 1/z 2 at z = n is 1, the power series representation will be cot (z) = z2 1 + higher powers of (z n) n2 1 + higher powers of (z n) zn

which has residue 1. Thus the sum of the residues for z = 0 will be N N 1 1 =2 . 2 n n2 n=1 n=N,n=0 At z = 0, we have cot (z) cos (z) = 2 = cos (z)cosec(z) z2 z sin (z)

(z)2 (z)4 2 z 2 7 4 z 4 1 + ... 1+ + ... . = 3 1 z 2! 4! 6 360 Expanding, we see that the residue is 2 2 2 + = . 2 6 3 Thus using the residue theorem, we have
N

lim

CN

cot (z) dz = 2i z2

2 1 . ( residues in CN ) = 2i +2 3 n2 n=1 cot (z) dz. z2 coth ( ) 2 (N + 1 )2 2

Finally, we consider the integral

CN

Observe that at any point on the boundary, we have cot (z) z2

13

so using the ML-formula, we have

CN

cot (z) dz z2

coth ( ) 1 2 1 2 8 N + 2 (N + 2 )

8 coth ( ) 2 0 = N+1 2

as N . Thus

2 1 2i +2 3 n2 n=1 2 = 6
n=1

=0

or

1 . n2

Example 6.5. Evaluate

n=1

(1)n+1

1 . n4

For this case, we observe that

n=1

(1)

n+1

1 1 = (1)n 4 , 4 n n n=1 cosec(z) dz z4 8 0 (N + 1 )3 2

so we shall consider the integral

N

lim

CN

and negate the answer. As with the previous case, we see that
CN

cosec(z) dz z4

1 1 48 N + 2 (N + 2 )

as N , so it follows that ( residues at z = n) = 0.

n=

For each integer n = 0, the residue is (1)n /n4 , so it follows that the sum of all residues except at z = 0 will be (1)n (1)n =2 . n4 n4 n=1 n=N,n=0 At z = 0 we have 2 z 2 7 3 z 4 1 1 cosec(z) 1+ = 4 + . . . z4 z z 6 360 so the residue is 7 4 . 360 2 z 2 7 4 z 4 1 1+ + . . . z5 6 360
N N

14

It follows that

so

7 4 (1)n +2 =0 360 n4 n=1 7 4 = 720

N

n=1

(1)n n4

giving 7 4 = 720
N n=1

(1)n+1 . n4

Homework: Questions from pages 146-148: 1,3,5,6