ECE 6382

Fall 2020
David R. Jackson

Notes 3
Integration in the Complex Plane

Notes are adapted from D. R. Wilton, Dept. of ECE

1
Defining Line Integrals in the Complex Plane

 Define  n on C between zn 1 and zn

y

 Consider the sums

N b  zN z
N    n 
z N 1
C
n IN   f   n   zn  zn 1 
…
3 … zn n 1
 2 z2 z3
 Let the number of subdivisions N  
 1 z1
such that  z n   zn  zn 1   0 and define
a  z0 b
x I   f  z  dz  lim I N
N 
a
z
N    n 

b
 lim
N 
 f   n   zn  zn 1 
n 1
I   f  z  dz   f  z  dz (The result is independent of the
C a details of the path subdivision, for
reasonably well - behaved functions.)
2
 Equivalence Between Complex and
Real Line Integrals
Denote
b b
I   f  z  dz   u  x, y   iv  x, y    dx  idy 
a a

b   xN ,y N  b   xN ,yN 
  u  x, y  dx  v  x, y  dy  i  v  x, y  dx  u  x, y  dy
a   x0 ,y0  a   x0 ,y0 

 C u dx  v dy  i C v dx  u dy

I C u dx  v dy  i C v dx  u dy
The complex line integral is equivalent to two real line integrals on C.
3
 Review of Line Integral Evaluation
y

t f  tN  A line integral written as C u  x, y  dx  v  x, y  dy

C t N 1 is really a shorthand for
 x(t ), y (t )  tn … tf
…  dx dy 
t2
t3
  dt dt  dt
 u  v
t1 t0
t0 where t is some parameterization of C :
x
C : x  x  t , y  y  t , t0  t  t f
t
t0 t1 t2 t3 … tn … t N 1 t f  t N

The path C goes

 Example : parameterizations of the circle x 2  y 2  a 2 counterclockwise
around the circle.
t
1) x  a cos t, y  a sin t, 0  t (   )  2

2) x  t , y  a2  t2 , t0  a, t f   a, t
-a a
and -a a t

x  t , y   a 2  t 2 , t0  a, t f  a,
4
 Review of Line Integral Evaluation (cont.)
y

t f  tN
C t N 1
 x(t ), y (t )  tn …
t2 t3
…
t1
t0
x

 While it may be possible to parameterize C using x and / or y

as the independent parameter, it must be remembered that the other
variable is in general always a function of that parameter!

Illustration : C u  x, y  dx  v  x, y  dy
xf
 dy 
           dx  dx
u x, y x  v x, y x (if x is the independent parameter)
x0
yf
 dx 
       dy  v  x  y  , y 
u x y , y  dy (if y is the independent parameter)

y0
5
 Line Integral Example
y
1
Consider z  Evaluate I   dz : where
C z
a C

 C : x  a cos  , y  a sin  , 0    
x

Although it is easier to use polar coordinates (see the next example), we use
Cartesian coordinates to illustrate the previous Cartesian line integral form.

1 1 1  x  iy   x   y   x   y 
f  z        2 2
i 2 
2  2
i 2 
z x  iy x  iy  x  iy   x  y   x  y   a   a 

I C u dx  v dy  i C v dx  u dy
Hence
   u  iv  dx    v  iu  dy
x C C
u  z 
a2  I1  I 2
The red color denotes functional dependence.
y
v  z  2 a 0
a  x   y (x)    y  x(y )  
I1   
 2  2   dx
a
i
 a 
I 2    2  i  2   dy
 a 
a 0 a
6
 Line Integral Example (cont.)
y
Consider
C z
a
 x   y (x)  
a
I1   a

 2  2   dx
i
 a 
a
 a
x  x  1  2 2
   i
 2  2
a a 
a  x  dx

a
a
 1 
 0i 2 
a 
 a 2  x 2 dx
a
a
 1   x a  x 
2 2
a2 1  x 
 i 2   sin   
 a   2 2  a  
a

 1   a 2    
 i     
 a2   2  2 2  
 
 
 i 
2

7
 Line Integral Example (cont.)
0
 y  x(y )  
y I 2    2  i  2   dy
0 a  a 
Consider
C z 0
i
a  0  2  x(y ) dy
a 0

x   0   2    2   

 
a 0
i i
 2  a  y dy  2   a 2  y 2 dy
2 2
a 0 a a
a
i
2 2
a  a 2  y 2 dy
0
a
i  y a2  y2 a2 1  y 

2 2   sin   
a  2 2  a  
 0

i  a2 1  a 

 2  sin   
a  2
2  a  
 
 i 
2
8
 Line Integral Example (cont.)

y
   
Consider z I  I1  I 2  i    i  
C 2 2
a

x
Hence

I  i
y
C z
Note: By symmetry (compare z and –z), a
we also have: 
x
1
 z dz  2 i
C

9
 Line Integral Example
y
Consider
 dz : where
n
C z z  Evaluate z
r C
 C : x  r cos  , y  r sin  , 0    2
x
 z  r cos   i r sin   rei ,
 dz  riei d ,

2

 
n
  z dz 
n
rei riei d
C 0
2
2 i  n 1
e
 ir n1  ei 
n 1
d  i r n 1 (n  1)
0
i  n  1
0

i 2   n 1
n 1 e 1  0 , n  1 This is a useful result and
 r   it is used to prove the
 n  1 2 i, n  1 “residue theorem”.

Note: For n = -1, we can use the result on slide 9 (or just evaluate the integral in  directly).
10
 Cauchy’s Theorem
y
Consider
R  a simply - connected region

A “simply-connected” region
C means that there are no “holes” in
the region. (Any closed path can
be shrunk down to zero size.)

Cauchy’s theorem:

If f  z  is analytic in R then  f  z  dz  0
C

11
 Proof of Cauchy’s Theorem
R  a simply connected region
y

 First, note that if f  z   w  u  iv, then

 f  z  dz   udx  vdy  i  vdx  udy ;

C

C C C
then use Stokes's theorem (see below).

 Construct 2D vectors A  uxˆ  vy,

ˆ B  vxˆ  uy,
ˆ dr  dx xˆ  dy yˆ in the xy  plane
and write the integral above as
Stokes's
Theorem
 f  z  dz   A  dr i  B  dr      A  zˆ dS i     B   zˆ dS , but
C C C interior of C interior of C

xˆ ˆy ˆz xˆ ˆy ˆz
C.R. C.R.
cond's cond's
   v u    u v
ẑ     A  ˆz     0, ˆz     B   ˆz     0
x y z x y x y z x y
u v 0 v u 0

  f  z  dz  0
C 12
 Proof of Cauchy’s Theorem (cont.)

Some comments:

 The proof using Stokes's theorem requires that u  x, y  ,v  x, y  have

continuous first derivatives and that C be smooth .

 The Goursat proof removes these restrictions; hence the theorem

is often called the Cauchy - Goursat theorem .

13
 Cauchy’s Theorem and Path Independence
f  z  is analytic   f  z  dz  0
Consider
C

 This implies that the line integral between any two points is independent of the path,
as long as the function is analytic in the region enclosed by the paths.

R  a simply - connected region R  a simply - connected region

y y

C2
C
C1

x x

 f  z  dz   f  z  dz   f  z  dz  0  f  z  dz   f  z  dz
C C1 C2 C1 C2

14
 Extension of Cauchy’s Theorem to
Multiply-Connected Regions
y y
R  a multiply - connected region R   simply - connected region

c2
c1

C2 C1 C2 C1

x x

 If f  z  is analytic in R then  f  z  dz  0 in general.

C1,2

 Introduce an infinitesimal - width "bridge" to make R into a simply connected region R 

 f  z  dz   f  z  dz   f  z  dz  0 since integrals along c1 , c2 are in

C1  C2  c1  c2 C1 C2

opposite directions and thus cancel   f  z  dz   f  z  dz

C1 C2

Unlike in Cauchy's theorem, now integrals are usually nonvanishing!

15
 Extension of Cauchy’s Theorem to
Multiply-Connected Regions

Summary: a multiply connected region

y
R  a multiply - connected region

C2 C1

 f  z  dz   f  z  dz
C1 C2

16
 Extension of Cauchy’s Theorem to
Multiply-Connected Regions (cont.)

 Example :

1
 z dz  2 i The integral around the arbitrary closed
C path C must give the same result as the
integral around the circle (and we already
know the answer for the circle).
y

17
 Cauchy’s Theorem, Revisited

Consider
If the function happens to be analytic everywhere within a simply connected region,
then we can shrink a closed path down to zero size, verifying that the line integral
around the closed path must be zero (since the integrand must be finite in the region).

R  a simply - connected region R  a simply - connected region

y y

C
C

x x

Shrink the path down.

18
 Fundamental Theorem of the
Calculus of Complex Variables
y

Consider zb
 zN
C z N 1
…
 z3 zn … This is an extension of the same
… theorem in calculus (for real functions)
 z2 z z3
2
 z1 z1 to complex functions.
za
x

Assume that f is analytic in a region R containing the path C.

dF
Suppose we can find F z  such that F  z    f  z :
dz
zb zb
dF
  f  z  dz   dz  F  zb   F  za 
za za
dz

19
 Fundamental Theorem of the
Calculus of Complex Variables (cont.)
y

Consider zb
Proof
 zN
C z N 1
… Recall :
 z3 zn …
… The integal is path independent
 z2 z z3
 z1 z1
2
if f  z  is analytic on paths from za to zb .
za
x
Choose a particular path C.

On this path:

zb N

 f  z  dz  Nlim

 f n   zn ,  z n  zn  zn 1 ,
za n 1

 n on C between z n1 and z n (definition of line integral).

20
 Fundamental Theorem of the
Calculus of Complex Variables (cont.)
y

Consider Proof (cont.)

 zN zb
C z N 1
…
 z3 zn …  Assume F  z  
dF
 f  z ;
… dz
 z2 z z3
F  zn   F  zn 1 
2
F
 z1 z1 Hence f   n   F  n   n 
zn zn
za
x

zb N N
F
  f  z  dz  lim
 z n 0
 f n   zn  lim
 z n 0
 z
n
 zn
za n 1 n 1 n

 lim  F  z1   F  za     F  z2   F  z1     F  z3   F  z2      
 z n 0      

    F  z N 1   F  z N  2     F  zb   F  z N 1  
   
 F  zb   F  za  (path independent if f  z  is analytic on paths from za to zb )

21
 Fundamental Theorem of the Calculus of
Complex Variables (cont.)

Fundamental Theorem of Calculus :

zb

 f  z  dz  F  zb   F  z a 
za

 This permits us to make use of indefinite integrals :

z n 1 e az
 z dz  n  1 ,  sin z dz e
n az
  cos z, dz  , etc.
a

y
1 i
Example:
1 i
1 i
C
 sin  z  dz   cos  z 
1
1
 cos  1  cos  1  i 
1
x

22
 Fundamental Theorem of the Calculus of
Complex Variables (cont.)
We have:
z z

 f  z  dz  F  z   F  z0   F  z    f  z  dz  F  z0 
z0 z0

Consider two different indefinite integrals:

z
F1  z    f    d  F1  z1  for arbitrary z1
z1
z
F2  z    f    d  F2  z2  for arbitrary z2
z2

z1
 F2  z   F1  z    f    d  F2  z2   F1  z1   constant
z2
Integral is a constant

All indefinite integrals can only differ by a constant.

23
 Cauchy Integral Formula
y R  a simply - connected region y R

C z0 C0
z0

x x

1
 f  z  is assumed analytic in R but we multiply by a factor that is
 z  z0 
analytic except at z0 and consider the following integral around C :
f  z
I   z  z0  dz
C

 To evaluate this, shrink that path to a small circular path path C0 as shown :
f  z f  z
 z  z0
dz   z  z0
dz
C C0

24
 Cauchy Integral Formula (cont.)

z
z0
We have: z

f  z f  z C0

 z  z0
dz   z  z0
dz C
C C0

 Evaluate the C0 integral on a small circular path, z  z0  rei , dz  rie i d :

f  z r 0 2
r i ei d
 dz  f  z0    2 i f  z0  for r  0
C0  z  z0  0 r e i

f  z 1 f z Cauchy Integral
  dz  2 i f  z0   f  z0   
 dz
C
z  z0 2 i C z  z0 Formula

 Note the remarkable result :

The value of f  z  at z0 is completely determined by its values on C !

25
 Cauchy Integral Formula (cont.)

Summary of Cauchy Integral Formula

z
1 f  z
f  z0   
 dz z0
2 i C z  z0
C

The Cauchy integral formula is useful for many purposes in complex variable theory.

26
 Cauchy Integral Formula (cont.)

 Note that if z0 is outside C, the integrand is analytic inside C ;

hence by the Cauchy's theorem, we have

1 f  z
 z  z0 dz  0
2 i 
C

z0 is inside C z0 is outside C

z0 z0

C
C

1 f  z 1 f  z
f  z0    z  z0 dz 
 dz  0
2 i 
C
2 i C z  z0

27
 Cauchy Integral Formula (cont.)

Summary of Both Cases

C
z0

f  z  2 i f  z0  , z0 inside C
 z  z0
dz  
 0, z0 outside C
C

Application:
This gives us a numerical way to
determine if a point is inside of a
region: just set f (z) = 1.

28
 Derivative Formulas
Since f is analytic inside C, we can start with
z0
the Cauchy integral formula:

1 f  z
Analytic f  z0  
2 i C z  z0
C dz (Note z0 is inside C .)

1 f  z
f  z0  z    z  z0  z dz
2 i 
C

f  z0  z   f  z0  1  f  z f  z 

z
   z  z0  z z  z0
2 iz 
  dz

C

1  z 
  f  z   dz
2 i z   z  z  z   z  z 
C  0 0 
f  z0  z   f  z0  1  1 
 lim  lim  f  z    dz
z 0 z z 0 2 i 
C
  z  z  z   z  z 
 0 0 

1 f  z
f  z0   
 dz Note: We have proved that you can
2 i C  z  z0  2
differentiate w.r.t. z0 under the integral sign!
29
 Derivative Formulas (cont.)
Similarly (derivation omitted):
2 f  z
f   z0     z  z  3 dz
2 i 
C 0
z0

In general:
1 dn  1 
f    z0  
C
f  z n 
2 i C
n
 dz
dz0  z  z 0

f  z  analytic in a simply
connected region containing C
n! f  z
f    z0   
n
 n 1
dz
2 i C  z  z0 

If f is analytic in a region, then all of the derivatives exist, and

hence all of the derivatives are analytic as well.

Note: All derivatives can be determined from f on the boundary!

30
 Morera’s Theorem
Ιf a function f  z  is continuous in a simply - connected
R region R and

z  f  z  dz  0
 C
z0 for every closed contour C within R , then f  z  is analytic
throughout R .

Note: f will be analytic if we can prove its derivative exists!

Proof
z z

 f  z  dz  0   f    d is path independent, so define F  z    f    d .

C z0 z0

Note that
z
F  z   F  z0  1
z
F  z   F  z0   F  z    f    d    f    d
z0
z  z0 z  z0 z0

Note : We can chose a small straight - line path between the two points since the
integral is path independent. Along this small path, f is almost constant (from continuity of f ).
F  z   F  z0  1
z z  z0 f  z0  z
f  z0 
  f    d   d   z  z0   f  z0 
z  z0 z  z0 z0
z  z0 z0
z  z0

 F   z0   f  z0  . Hence F is analytic at any z0 in R . But then so are all its derivatives, including F   f .
Hence, f is analytic at z0 .
31
 Comparing Cauchy’s and Morera’s Theorems

 Cauchy's Theorem : If f  z  is analytic in a simply - connected

region R then

 f  z  dz  0, C in R .
C

 Morera's Theorem : Ιf a function f  z  is continuous in a simply -

connected region R and  f  z  dz  0 for every closed contour
C
C within R , then f  z  is analytic throughout R .

 These theorems are converses of one another!

32
 Properties of Analytic Functions

Existence of f   z 
Definition

 f  z  dz  0 Cauchy-Riemann
conditions
C

Analyticity

Path Existence of derivatives

independence of all orders

33
 Cauchy’s Inequality
y
 Suppose f  z  :
(a) is bounded ( f  z   M ) on the circle of radius R
C
R
(b) has a convergent power series representation x

f  z    an z n (see note below)
n 0
inside (and on) the circle of radius R. M
Then an  n
.
R
1 f  z 1 
1
2  z m 1
dz 
2
 an  z n  m1dz 
2
2 iam (This is from the previous line integral example,
where n-m-1 = -1.)
z R n 0 z R

1 f  z 1 f  z 1
2
M M
 am 
2  z m1
dz 
2  z m 1
dz 
2  R m 1
R d 
Rm
z R z R 0

m n M
 an  , M  max f  z 
Rn z R

Note: If a function is analytic within and on the circle, then it must have a convergent
power (Taylor) series expansion within and on the circle (proven later).
34
 Liouville’s Theorem
If f (z) is analytic and bounded in the entire complex plane, it is a constant.
Because it is analytic in the entire complex plane, f (z) will
have a power (Taylor) series that converges everywhere (proven later).
Proof


By the Cauchy Inequality, f  z    an z n
M n 0
an  n , for any R (so let R  )
R
 an  0 , n  0
 f  z   a0

 No "interesting" (i.e., non - constant) function f  z  is analytic and bounded everywhere!

 z n is analytic everywhere but unbounded at infinity

 sin z, e z are analytic everywhere but unbounded at infinity
1
 is bounded at infinity but is not analytic at z  z0
 z  z0  n

35
 The Fundamental Theorem of Algebra

(This theorem is due to Gauss*.)

N
The N th degree polynomial, PN  z    an z n , a N  0, N  0, has N (complex) roots.
n 0

Carl Friedrich Gauss

As a corollary, an Nth degree polynomial can be
written in factored form as

PN  z   aN  z  z1   z  z2    z  z N 

* The theorem was first proven in Gauss’s doctoral

dissertation in 1799 using an algebraic method. The
present proof, based on Liouville’s theorem, was given
by him later, in 1816.

http://en.wikipedia.org/wiki/Carl_Friedrich_Gauss (his signature)

36
 The Fundamental Theorem of Algebra (cont.)

Proof (by contradiction)

1
 First assume the polynomial has no roots. Then is analytic everywhere and
PN  z 
1 1
bounded at  since lim  lim  0.
z  PN  z  z  aN z
N

1
 But by Liouville's theorem, we then have that is a constant, contrary to our
PN  z 
assumption N  0.
 Hence PN  z  must have at least one root, say at z  z1 . We can then write
PN  z    z  z1  PN 1 ( z ) (see note below).
 Repeat the above procedure N  1 times (a total of N times) until we arrive at the conclusion that
PN  z    z  z1   z  z2    z  z N  P0
where P0  z   P0 is a constant .

Note:
To verify this, we can use the method of “polynomial division” to construct the polynomial PN-1 (z) in terms of an.
An example is given on the next slide.

37
The Fundamental Theorem of Algebra (cont.)

 Example (polynomial division)

Assume a third - order polynomial P  z   z 3  a2 z 2  a1z  a0 has a root at z  z1.


Show that P  z    z  z1  z 2  b1z  b0 . 
 Equate coefficients other than z 0 :
z 2 : b1  a2  z1

z1: b0  a1  b1z1

 
 The difference between P  z  and  z  z1  z 2  b1z  b0 must then be a constant.

 Since both terms vanish at z  z1, this constant must be zero.

38
 Numerical Integration in the Complex Plane
Here we give some tips about numerically integrating in the complex plane.
b
I   f  z  dz
a

One way is to parameterize the integral using z = z(t): We assume a given

function and a given path.
tb y
 dz  t a  t0
I  f  z  t     dt b
ta  dt  tb  t f t f  tN
C t N 1
so tb  x(t ), y (t )  tn …
…
I   F  t  dt t2
t3

ta a t1
t0
x
where
 dz  z  z  t
F  t   f  z  t      FR  t   iFI  t 
 dt 
tb tb
I   FR  t  dt  i  FI  t  dt
ta ta
39
Numerical Integration in the Complex Plane (cont.)
Summary

b tb tb tb
I   f  z  dz   F  t  dt   FR  t  dt  i  FI  t  dt
a ta ta ta

Note:
where It is not always necessary to treat
the real and imaginary parts of F
separately; we can just allow F to
 dz 
F  t   f  z  t      FR  t   iFI  t  be complex in many software.
 dt 

The integrals in t can be preformed in the usual way, using any convenient
scheme for integrating functions of a real variable (Simpson’s rule, Gaussian
Quadrature, Romberg method, etc.).

Note: The variable t can be x, or , or any other convenient variable.

40
Numerical Integration in the Complex Plane (cont.)

Numerically Integrating Functions in a Region where they are Analytic

If the function f is analytic, then the

integral is path independent. We y
can choose a straight line path!
t 1

We explore how this works. C b

a
Note: If the path is piecewise linear, we simply add
up the results from each linear part of the path. t 0
x

z  a  (b  a)t, 0  t  1
dz
 b  a  arg  dz   arg  b  a  
dt

b 1 1
dz
I   f  z  dz   f  z  t   dt   b  a   f  z  t   dt
a 0
dt 0

41
Numerical Integration in the Complex Plane (cont.)

1
I   b  a   f  z  t   dt
0
y

We can use any method that we t 1

wish to evaluate this integral along C b
the real axis. a z

t 0
x
We can also break up the integral into
real and imaginary parts if we wish:

1 1
I   b  a   Re f  z  t   dt  i  b  a   Im f  z  t   dt
0 0

Note: The term b-a is complex!

42
Numerical Integration in the Complex Plane (cont.)
1
I   b  a   f  z  t   dt
0
y
Uniform Partitioning t  1/ N
t 1

Note that if we partition uniformly in t, C b

then we are really partitioning a z
uniformly along the line with z.
t0
x
z  (b  a)t  (b  a) / N

(We don’t have to partition uniformly, but we can if we wish.)

t
t
0 t1 t2 tn 1

43
 Numerical Integration in the Complex Plane (cont.)
1
I   b  a   f  z  t   dt
0
t tnmid
t
Uniform sampling (Midpoint rule): 0 t1 t2 tn 1

N
 
I   b  a   f znmid t
n 1 z  a  ( b  a )t
 1
znmid  a  (b  a )tnmid , tnmid  t  n   y
 2 N intervals
zN
Using z  (b  a )t  t  z / (b  a ) C b  zN
znmid
z0 z
we then have a  z0
N
z
 
x
I   b  a  f znmid
n 1 ba
44
Numerical Integration in the Complex Plane (cont.)
N

I  z  f znmid 
n 1

where

znmid  a  (b  a )tnmid
y
 1
 a  ( b  a ) t  n   N intervals
 2 zN
C b  zN
1  1
 a  (b  a )  n   znmid
N  2 z0 z

 1 a  z0
 a  z  n  
 2 x

ba
Note : z  (b  a)t 
N
45
Numerical Integration in the Complex Plane (cont.)
“Complex Midpoint rule”
N

I  z  f znmid 
n 1

ba y
z 
N N intervals
zN
C b  zN
 1 znmid
znmid  a  z  n   z0 z
 2
a  z0
x

This is the same formula that we usually use for integrating a

function along the real axis using the midpoint rule!
(We essentially have a just rotation of the axis.)

46
Numerical Integration in the Complex Plane (cont.)

“Complex Simpson’s rule”

z
I
3
 f  z0   4 f  z1   2 f  z2   4 f  z3   2 f  z4    4 f  z N 1   f  z N  

N = number of segments y
= even number
N intervals
zN
b  zN
Number of sample points = N +1 C
zn z
z0
a  z0
zn  a   z  n x

ba
z 
N

47
Numerical Integration in the Complex Plane (cont.)
“Complex Gaussian Quadrature”
N Note:
I   In The Gaussian quadrature formulas are usually
n 1 given for integrating a function over (-1,1). We
translate these into integrating over a given
interval (interval n) of length z.
 z  6  z 
In     f  znmid  xi  wi
 2  i 1  2 
N intervals
y
(6-point Gaussian Quadrature) Six sample points are used
within each of the N intervals.

znmid  a   z   n  1 / 2  zN b  zN
C
znmid
z0
z
a  z0
Sample points Weights
x
x1 = -0.9324695 w1 = 0.1713245
x2 = -0.6612094 w2 = 0.3607616
x3 = -0.2386192 w3 = 0.4679139 ba
x4 = 0.2386192 w4 = 0.4679139 z 
x5 = 0.6612094 w5 = 0.3607616 N
x6 = 0.9324695 w6 = 0.1713245
48
Numerical Integration in the Complex Plane (cont.)
Here is an example of a piecewise linear path, used to calculate the
electromagnetic field of a dipole source over the earth (Sommerfeld problem).

Gaussian quadrature could be used on each of the linear segments of the path
(breaking each one up into N intervals as necessary to get convergence).

kti

k1
C Sommerfeld path

k0
k0
ktr
k1

49