Professional Documents
Culture Documents
ClasesVariable UniversidadGringa
ClasesVariable UniversidadGringa
Semester 1, 2012
Craig Westerland
Department of Mathematics and Statistics
University of Melbourne
Course Information
Lectures: This course consists of 36 one hour lectures (three per week) and 11 practice
classes (one per week). The lectures are at
2:15pm - 3:15pm on Tuesdays, in MSLE Lower Theatre
2:15pm - 3:15pm on Thursdays, in the Russell Love Theatre, Richard Berry, and
11am - 12pm on Fridays, in the Russell Love Theatre, Richard Berry.
Dr. Norman Do will teach the first week of classes. Practice classes start on the Monday of
Week 2. They are at 4:15pm - 5:15pm on Mondays, in the Russell Love Theatre, Richard
Berry. There is a non-teaching period from Monday 17 September to Sunday 30 September.
Assessment: Assessment will be based on a 3-hour end-of-semester written examination
[X marks out of 80] and three written assignments during semester [A marks out of 20].
The final mark M out of 100 will be M = X + A.
Problem Sheets: There are six problem sheets for this course. These relate to the basic
skills to be acquired from the course. It is important to do most of the problems on these
problem sheets. The problems marked with an asterisk are more difficult or theoretical.
Subject Web Page: A web page will be maintained for this subject on the LMS:
http://www.lms.unimelb.edu.au
Lecture notes and supplementary materials will be available from this site.
0-2
J. E. Marsden and M. J. Hoffman, Basic Complex Analysis Freeman (Third Edition) 1998.
Useful References:
Almost any book with Complex Analysis/Variables/Functions in the title including:
0-3
Complex analysis is a core subject in pure and applied mathematics, as well as the physical
and engineering sciences. While it is true that physical phenomena are given in terms of
real numbers and real variables, it is often too difficult and sometimes not possible, to solve
the algebraic and differential equations used to model these phenomena without introducing
complex numbers and complex variables and applying the powerful techniques of complex
analysis.
Topics include: the topology of the complex plane; convergence of complex sequences and
series; analytic functions, the Cauchy-Riemann equations, harmonic functions and applications; contour integrals and the Cauchy Integral Theorem; singularities, Laurent series, the
Residue Theorem, evaluation of integrals using contour integration, conformal mapping; and
aspects of the gamma function.
At the completion of this subject, students should understand the concepts of analytic
function and contour integral and should be able to:
apply the Cauchy-Riemann equations
Generic Skills
In addition to learning specific skills that will assist students in their future careers in science,
they will have the opportunity to develop generic skills that will assist them in any future
career path. These include:
problem-solving skills: the ability to engage with unfamiliar problems and identify relevant
solution strategies
analytical skills: the ability to construct and express logical arguments and to work in
abstract or general terms to increase the clarity and efficiency of analysis
collaborative skills: the ability to work in a team
time-management skills: the ability to meet regular deadlines while balancing competing
commitments.
0-5
0-6
0-8
Complex Numbers
Some simple equations do not admit solutions in the field of real numbers:
z 2 = 1,
zR
i = + 1,
i2 = 1,
i
/R
z = x + iy,
Im z = y R
Two complex numbers are equal if and only if they have the same real and imaginary parts:
z1 = z2
Re z1 = Re z2 and
x1 + iy1 = x2 + iy2
Im z1 = Im z2
x1 = x2 and y1 = y2
1-1
C := {x + iy : x, y R}
This is the set of numbers obtained by appending i to the real numbers. A number of the
form iy is called pure imaginary.
The field of complex numbers is the set C equipped with the arithmetic operations of
addition, subtraction, multiplication and division defined by
(x1 + iy1) + (x2 + iy2) := (x1 + x2) + i(y1 + y2),
x1 , x2 , y 1 , y 2 R
x1 , x2 , y 1 , y 2 R
(y1x2 x1y2)
x2 iy2
(x1x2 + y1y2)
+i
,
=
2
2
2
2
x2 iy2
x2 + y 2
x2 + y 2
x1 , x2 , y 1 , y 2 R
2
x2
2 + y2 6= 0
1-2
z1 , z2 , z3 C
0. Closure:
z1 + z2 C,
z1 z2 C
1 z = z,
for all z C
2. Commutative laws:
z1 + z2 = z2 + z1 ,
z1 z2 = z2 z1
3. Associative laws:
(z1 + z2) + z3 = z1 + (z2 + z3) = z1 + z2 + z3
(z1 z2)z3 = z1(z2 z3) = z1 z2 z3
4. Distributive laws:
z1(z2 + z3) = z1 z2 + z1 z3
5. Inverses:
z1 + z2 = 0
z2 = z1,
z1 z2 = 1
1
1
z2 = z1 = , z1 6= 0
z1
6. Zero factors:
z1 z2 = 0
z1 = 0 or z2 = 0
1-3
The complex numbers C can be constructed formally from the set of real numbers R:
Definition: A complex number is defined algebraically as the ordered pair [x, y] of real numbers x, y R. The complex operations on C := {[x, y] : x, y R} are defined by
1 Equality.
2 Addition.
3 Multiplication.
[x, y] = [x, y ]
x = x and y = y
[x, y] + [x, y ] := [x + x, y + y ]
[x, y][x, y ] := [xx yy , xy + xy]
Exercise: Show that the operations of addition and multiplication are commutative,
associative and distributive.
Numbers of the form [x, 0] behave like real numbers so we identify x [x, 0]:
1 Addition.
2 Multiplication.
We prove that
[x, 0] + [x, 0] := [x + x, 0]
[x, 0][x, 0] := [xx, 0]
[x, y]2 = [1, 0] 1
It follows that every complex number can be written in the form [x, y] x + iy since
[x, y] = [x, 0] + [0, y] = [x, 0] + [0, 1][y, 0] x + iy
1-4
Solving Equations
Example: Solve
z 2 = 1
z 2 = 1
Example: Solve
z = i
z 2 + 2z + 2 = 0
z 2 + 2z + 2 = 0
Check:
Example: Solve
LHS =
=
=
=
since
(i)2 = i2 = 1
2 4
z=
= 1 i
2
z 2 + 2z + 2
(1 i)(1 i) + 2(1 i) + 2
(1 2i 1) + (2 2i) + 2 = 0
RHS
z2 = z
z2 = z
or
(x + iy)2 = x iy
x2 y 2 + 2ixy = x iy
x2 y 2 = x and 2xy = y
1 and y 2 = 3
x = 2
4
y = 0 and x = 0, 1
3
1
z = 0, 1, i
2
2
1-5
Complex Plane
z1
z2
z1 +z2 z2
z1
0
z (x, y)
y
1-6
Polar Form
z = x + iy = r(cos + i sin )
x2 + y 2 = r
By Pythagoras, this is the distance of the point z or (x, y) from the origin 0.
where
cos =
x
,
|z|
sin =
y
|z|
k Z,
< 0
< 0
z = z;
2. z1 + z2 = z 1 + z 2;
3.
z1 z2 = z 1 z 2 ;
Re z | Re z| |z|;
|z|2 = z z;
z z = 2i Im z
z + z = 2 Re z;
z1
z2
z
= 1,
z2
Im z | Im z| |z|
|z1z2| = |z1||z2|;
then
z
|z1|
1
,
=
z2
|z2|
z2 6= 0
z2 6= 0
arg z = arg z
k = 0, 1
z1 6= 0, z2 6= 0
7. Unlike real numbers, the complex numbers are not ordered. So inequalities, such as z1 z2
or z1 > z2, only make sense if z1, z2 R.
1-8
Triangle Inequality
Triangle Inequality:
Proof:
|z1 + z2|2 = (z1 + z2)(z1 + z2) = (z1 + z2)(z 1 + z 2)
= z1z 1 + (z1z 2 + z2z 1) + z2z 2
= |z1|2 + (z1z 2 + z1z 2) + |z2|2
= (|z1| + |z2|)2
Now take positive square root.
|z1| |z2| |z1 z2|
1-9
(2 + i)(1 + 3i) 2 + 4i
(2 + i)(1 + 3i) 2 + 4i = (2 + i)(1 3i) 2 + 4i
Example: Simplify
i(3 2i)
1+i
2 + 3i 1 i
2 + 3 + 3i 2i
5+i
i(3 2i)
=
=
=
1+i
1+i 1i
1+1+ii
2
(3 4i)(2 i)
Example: Simplify
.
1 + 3i
(3 4i)(2 i)
|3 4i||2 i|
9 + 16 4 + 1
25 5
25
=
=
=
=
1 + 3i
|1 + 3i|
1+9
2
10
kZ
|z| = 4 + 4 = 8 = 2 2
y
1
2
sin =
= =
|z|
2 2
2
p
Complex Exponential
Definition: The complex exponential is defined as
ez := ex+iy = exeiy = ex(cos y + i sin y)
yR
|ei | = 1,
= arg z
ei/2 = i,
eki =
1,
1,
k even
k odd
Exponent Laws:
ez1+z2 = ez1 ez2 ,
(ez )n = enz ,
nZ
1-11
De Moivres Formula:
(cos + i sin )n = (ei )n = ein = cos n + i sin n,
Trigonometric Functions:
i
cos = Re e
R, n Z
ei + ei
=
,
2
sin = Im e
ei ei
=
2i
zC
eiz + eiz
,
cos z :=
2
ez + ez
cosh z :=
,
2
cos z = cosh iz
eiz eiz
,
sin z :=
2i
ez ez
sinh z :=
,
2
sin z = i sinh iz
zC
1-12
Binomial Theorem
Theorem 1 (Binomial Theorem) For z1, z2 C and n N
n n1
n nk k
n
n
(z1 + z2) = z1 +
z1 z2 + +
z1 z2 + + z2n
1
k
n
X
n nk k
=
z1 z2
k
k=0
0! := 1;
n!
,
n
(nk)! k!
:=
0,
k
0kn
otherwise
Example: Find (1 + i)6, (a) by using the binomial theorem and (b) by using exponential
polar form:
(1 + i)6 = 1 + 6i + 15i2 + 20i3 + 15i4 + 6i5 + i6
= 1 + 6i 15 20i + 15 + 6i 1 = 8i
i/4 6
) = 8e3i/2 = 8i
( 2e
1-13
n+1
X n
n
+
k
k1
k=0
z1n+1k z2k
n+1
X n + 1 n+1k
=
z2k
z1
k
k=0
1-14
Trigonometric Identity I
Exercise: Sum the geometric series
n
X
k=0
= 1 + z + + z
1 z n+1
,
=
1z
n N, z C, z 6= 1
cos k =
(n+1)
cos n
sin
2
2
sin 2
k=0
0 < < 2
cos k = Re
k=0
X
n
k=0
eki
1 e(n+1)i
= Re
1 ei
1 )i
(n+ 2
1 (e
(n+ 1 )i
Re
i/2
2
2i
e
e
=
= Re
i/2
i/2
e
e
1 sin(n + 1 ) + sin
2
2
2
=
sin 2
ei/2)
sin 2
1-15
Trigonometric Identity II
Example: Prove the trigonometric identity
1 (z z ) sin 1 (z + z ),
sin z1 + sin z2 = 2 cos 2
1
2
2
2 1
z1 , z2 C
1 i(z1z2)/2
i(z
z
)/2
i(z
+z
)/2
i(z
+z
)/2
1
2
1
2
1
2
RHS =
e
+e
e
e
2i
1 iz1
iz1
iz2
iz2
=
= LHS
e e
+e e
2i
In this way all trigonometric identities in z are reduced to algebraic identities in eiz .
z2 =
2
1-16
Roots of Unity
Roots of Unity: Solve z n = 1: Write z n and 1 in exponential polar form and equate
modulus and argument
z n = (|z|ei )n = |z|neni ,
|z|n = 1 and
|z| = 1 and
z = e2ki/n = k ,
1 = e2ki ,
kZ
arg(z n) = n arg z = n = 2k
2k
, k = 0, 1, . . . , n 1
Arg z = =
n
k = 0, 1, . . . , n 1
Example: Solve wn = z, that is, find w = z 1/n: Write w, z in exponential form and equate
modulus and argument
wn = |w|neni, z = |z|ei+2ki, k Z;
arg w = , arg z =
1-17
= 2
3
3
1/3
(cos
+ i sin
), 2
(cos
+ i sin
),
12
12
4
4
17
17
+ i sin
)
21/3(cos
12
12
1-18
Fractional Powers
Exercise: Show that the sets of numbers (z 1/n)m and (z m)1/n are the same. We denote this
common set by z m/n.
Fractional Powers:
by
k = 0, 1, . . . , n 1
i
i2/3 = (ei/2+2ki )2/3 = ei/3+4ki/3,
k = 0, 1, 2
1
3
= 1,
i
= (i2)1/3 = (1)1/3
2
2
1-19
Argand Diagrams
Example: Indicate graphically, on a single Argand diagram, the sets of values of z determined
by the following relations:
(a) Point
(c) Circle
z = 1 2i
|z 1 i| = 1
|z 1 i| < 1
(d) Disk
(f) Annulus
3 Re z 5
(h) Ray
1 |z + 3| 2
Arg z = 3/4
(b)
3i
2i
(f )
(c)
(g)
(d)
(e)
(a)
12i
(h)
1-20
|z z0| < r
inside the circle of radius r about z = z0 is called an open disk or neighbourhood of z0. The
set |z| < 1 is the open unit disk.
Open Sets: A point z0 in a set S C is an interior point of S if there is some open disk
about z0 which is completely contained in S. If every point of S is an interior point of S we
say S is open. The empty set and C are open sets.
|z z0| r
Closed Sets: A point z0 is said to be a boundary point of S C if every open disk about z0
contains at least one point in S and at least one point not in S. Note that a boundary point
z0 may or may not be in S. The set S of all boundary points of S is called the boundary of
S. A set which contains all of its boundary points is called closed. A set S is closed if and
only if its complement C\S is open. A point which is not an interior point or boundary point
of S is an exterior point. The empty set and C are both open and closed sets.
1-21
Bounded Sets: A set S C is called bounded if there exists a real number R such that
|z| < R for every z S. A set S C which is both closed and bounded is called compact.
1-22
Regions:
The Disk (d), Annulus (f) and Strip (g) are regions.
|z i| + |z 2i| < 2.
Boundaries:
The boundary of Disk (d) is the Circle (c). The boundary of Annulus (f) is the union of the
circles |z + 3| = 1 and |z + 3| = 2. The boundary of the Strip (g) is the union of the lines
Re z = 3 and Re z = 5.
f : z = x + iy
which assigns to each z in the domain D C a unique image w = f (z) in the range R C so
that f (z) is single-valued.
It is not possible to represent a complex function f (z) by a graph. The complex function
f (z) is however determined by the pair of real functions u(x, y), v(x, y) of two real variables
x and y. D and R are usually regions in C:
v
w = f (z)
2-1
2-2
zz0
Unlike a function of a real variable, z can approach z0 along many different paths in the
complex plane. If the limit exists, it is independent of the way in which z approaches z0.
Example: Show from the limit definition that
whenever
lim z 2 = 1:
zi
|z i|(|z i| + 2) < (1 + 2) =
3
|z i| < () = Min(1, 3 )
z
does not exist:
z0 z
For the limit to exist, it must be independent of the path along which z = x + iy approaches
z0 = 0. We show that the limits z 0 along the x- and y-axes are different
Example: Show that
lim
z0
y=0
lim
z
x
z
iy
= lim
= 1 6= lim = lim
= 1,
z0
x0 x
y0 (iy)
z
z
x=0
2-3
Limit Theorems
Theorem 2 (Limit Theorems) If a, b C are constants (independent of z) and lim f (z),
zz0
zz0
1. Linear:
2. Product:
3. Quotient:
lim
zz0
lim
zz0
f (z) g(z) =
zz0
zz0
lim f (z)
zz0
lim f (z)
f (z)
zz0
lim
if
=
zz0 g(z)
lim g(z)
zz0
lim g(z)
zz0
lim g(z) 6= 0
zz0
zz0
if and only if
lim u(x, y) = u0 and
xx0
yy0
lim v(x, y) = v0
xx0
yy0
Proof: Exercise using | Re w| |w|, | Im w| |w| and the triangle inequality. Note, for these
two variable limits to exist they must be independent of the path along which (x, y) (x0, y0)
in R2.
2-4
lim
z2
z2 + 3
lim
;
z2
iz
z2 + 3
iz
lim (z 2 + 3)
= z2
lim iz
z2
7
7
= i
2i
2
z2 + 1
;
lim 4
zi z 1
z2 + 1
(z 2 + 1)
1
1
lim 4
= lim 2
=
lim
=
2
zi z 1
zi (z 1)(z 2 + 1)
zi (z 2 1)
2-5
There are many directions or paths along which 1z can approach infinity as z 0. We
identify these limits with a single number and extend the complex plane by adding a point
= C {}.
at infinity denoted by the symbol . This forms the Riemann sphere C
The extended complex plane is closed and identified with the stereographic projection of
a sphere (Riemann sphere) of radius r = 1
2 onto the horizontal plane C passing through the
south pole (z = 0). A line drawn from the north pole of the sphere through the point z
on the sphere maps to the point z C. The equator maps onto the unit circle |z| = 1, the
southern hemisphere maps onto |z| < 1 and the northern hemisphere maps onto |z| > 1. The
south pole corresponds to the origin z = 0 and the north pole to the point z = at infinity.
Stereographic Projection:
1
=
z0 z
lim
2-6
Limit z
Definition: We define
1
lim f (z) := lim f
z
z0
z
|z| > M,
for M > 0.
Example:
i 1
lim (i z)
i
z
iz 2 z
2
z
z0
= lim z1
= i
=
= lim
lim 2
2
z z 1
z0 2 1
z0 1 z 2
lim (1 z )
z
z0
Example:
1 + i
z + iz 2
z 2 + iz
2
z
z
= lim 1
= lim
lim
z z 3 + 1
z0 3 + 1
z0 1 + z 3
z
lim (z + iz 2)
0
= z0
=
= 0
3
1
lim (1 + z )
z0
Example:
1 +1
z2 + 1
1
2
= lim z 1
lim
= lim ( + z) =
z
z0
z0 z
z
z
since
1
0
z
lim 1
=0
=
= lim
2
z0 + z
z0 1 + z
1+0
z
2-7
Continuity
Definition: Suppose the function f (z) is defined in an open disk about z = z0 and f (z0) is
defined at z = z0. Then f (z) is continuous at z0 if
lim f (z) = f (z0) = f
zz0
lim z
zz0
To be continuous at z0, (i) the function f (z) must be defined at z = z0, (ii) the limit
lim f (z) must exist and (iii) the limit must equal the function value f (z0).
zz0
a f (z) + bg(z)
f (z) g(z)
f (z)
provided g(z0) 6= 0
g(z)
f (g(z)) if f (w) is continuous at w0 = g(z0)
2-8
Continuity Examples
Example: Show that f (z) = ez is continuous in C:
f (z) = ez = exeiy = ex cos y + iex sin y
is continuous in C since
u(x, y) = ex cos y,
v(x, y) = ex sin y
are continuous functions of two real variables in R2 being products of the continuous functions
ex, cos y and sin y.
Example: Show that f (z) = |z| is continuous in C:
lim
lim f (z) = lim |z| = xx
zz0
zz0
0
yy0
x2 + y 2 =
1/2
2
2
lim
(x
+
y
)
xx
0
yy0
2
x2
0 + y0 = |z0| = f (z0 )
for any z0 C since the function of one real variable g(r) = r is continuous for r 0 where
here r = x2 + y 2.
2-9
The concept of an analytic function on an open region is very powerful, much more
powerful than the concept of an analytic real function on an open interval. For example, we
will show later, that if a function f (z) is analytic then its derivatives f (n)(z) of all orders exist
and are also analytic. This is not the case for real functions:
f (x) = sgn(x) x2,
f (x) = 2|x|,
x (, )
e1/x ,
0,
x>0
x0
is C (has derivatives of all orders) but is not analytic since f (n) (0) = 0 for all n and the
Taylor expansion about x = 0 converges to zero and not to f (x).
2-10
d n
z = nz n1.
dz
n1
n2(z)2 + + (z)n
nz
z + n
z
2
= nz
So
n1
n n2
+
z
(z) + + (z)n1
2
(z + z)n z n
d n
z = lim
= nz n1
z0
dz
z
Exercise: Show that the functions f (z) = Re z, f (z) = Im z and f (z) = |z| are continuous
but nowhere differentiable.
Koch Curve
Consider the curve defined iteratively starting with the real line segment [0, 1] according
to the replacement:
7
Iterating gives:
The limiting curve is a fractal called the Koch (snowflake) curve. This curve is continuous
but nowhere differentiable (there are no tangents to the curve). It has an infinite length but
the area under the curve is finite. Its fractal dimension is
log 4
dfractal =
1.26
log 3
Johan Thim, Continuous Nowhere Differentiable Functions, Masters Thesis, Lulea (2003).
2-12
Katsuura Function
Consider the function defined iteratively starting with the function K0(x) = x on [0, 1]
according to the replacements:
7
Iterating gives
K2(x)
K3(x)
K4(x)
K5(x)
The Katsuura function K(x) = lim Kn(x) is continuous but nowhere differentiable on [0, 1].
n
h0
f (z0 + h) f (z0)
lim h
h0
h0
h
lim
= f (z0) 0 = 0
This shows that
lim f (z0 + h) = f (z0)
h0
or
zz0
2-14
Differentiation Rules
Theorem 6 (Differentiation Rules) If a, b are constants and f (z) and g(z) are differentiable
in an open region R, then in this region
1. Linear:
d
af (z) + bg(z) = a f (z) + bg (z)
dz
2. Product:
d
f (z) g(z) = f (z) g(z) + f (z) g (z)
dz
3. Quotient:
4. Chain:
df dg
d
= f (g(z0))g (z0) if f (w) is
f (g(z)) =
dz
dg dz
d 1
1
f (w) =
with w = f (z), z = f 1(w)
dw
f (z)
2-15
2. Derivative of a product:
d
f (z) g(z)
dz
f (z + z)g(z + z) f (z)g(z)
= lim
z0
z
z0
2-16
aj , bj C
Since differentiable implies continuous, it follows that R(z) is continuous at any point z where
the denominator does not vanish.
2-17
LH
opitals Rule
Theorem 7 (LH
opitals Rule) If f (z) and g(z) are differentiable at z = z0 and f (z0) = 0,
g(z0) = 0 but g (z0) 6= 0 then
f (z)
f (z0)
lim
=
zz0 g(z)
g (z0)
Proof: Using the quotient limit theorem
f (z)f (z0 )
f (z)f (z0 )
lim
zz0
f (z)
f (z0)
zz0
zz0
=
=
lim
= lim
g(z)g(z0)
zz0 g(z)
zz0 g(z)g(z0)
g (z0)
lim
zz0
zz
zz0
0
Theorem 8 (General lH
opitals Rule)
If f (z) and g(z) are analytic at z = z0 and f (z), g(z) and their first n 1 derivatives all vanish
at z = z0 but the nth derivative g (n)(z0) 6= 0 then
f (n)(z)
f (n) (z0)
f (z)
= lim
=
lim
(n)
zz0 g (n)(z)
zz0 g(z)
g (z0)
Proof: Consequence of Taylors theorem proved later.
2-18
Examples of LH
opitals Rule
Example: Evaluate the limit
z 10 + 1
lim
zi z 6 + 1
ez 1 z
lim
z0
z2
d z
e = ez
dz
To prove this we need the Cauchy-Riemann theorem.
2-19
Cauchy-Riemann Equations
Differentiability of f (z) = u(x, y) + iv(x, y) implies strong constraints on u(x, y), v(x, y).
If f (z) is differentiable the limit
f (z + z) f (z)
z0
z
f (z) = lim
must exist independent of the path. We conclude that the limit along the x- and y-axes must
agree
f (z) = lim
"
= lim
"
x0
y=0
x=0
y0
Hence necessary conditions for f (z) to be differentiable are given by the Cauchy-Riemann
equations
v
u
v
u
+i
= i
+
f (z) =
x
x
y
y
u
v
=
,
x
y
v
u
=
x
y
2-20
Cauchy-Riemann Theorem
Theorem 9 (Cauchy-Riemann Theorem)
Suppose f (z) = u(x, y) + iv(x, y) is defined in an open region R containing z0. If u(x, y)
and v(x, y) and their first partial derivatives exist and are continuous at z0 (that is u(x, y)
and v(x, y) are C 1 at (x0, y0)) and satisfy the Cauchy-Riemann equations at z0, then f (z) is
differentiable at z0. Consequently, if u(x, y) and v(x, y) are C 1 and satisfy the Cauchy-Riemann
equations at all points of R then f (z) is analytic in R.
Proof: See a textbook such as Saff and Snider. The proof uses the mean-value theorem.
f
df z
df
=
=
,
x
dz x
dz
df
f
u
v
=
=
+i
,
dz
x
x
x
2-21
v
u
=
= ex sin y
x
y
v
u
+i
= ex(cos y + i sin y) = ez
x
x
Example: Discuss where the function f (z) = (x2 + y) + i(y 2 x) is (a) differentiable and
(b) analytic.
Solution: Since u(x, y) = x2 + y and v(x, y) = y 2 x, we have
u
= 2x,
x
v
= 2y,
y
v
u
=
= 1
x
y
These partial derivatives are continuous everywhere in C. They satisfy the Cauchy-Riemann
equations on the line y = x but not in any open region. It follows by the Cauchy-Riemann
theorem that f (z) is differentiable at each point on the line y = x but nowhere analytic.
2-22
Harmonic Functions
Definition: A real function of two variables (x, y) is harmonic in an open connected domain
D if it is C 2 (that is continuous with continuous first and second partial derivatives) in D and
satisfies the Laplace equation
2
2
:=
+
= 0
x2
y 2
2
The Laplace equation occurs in many areas of two-dimensional physics including continuum
and fluid mechanics, aerodynamics and the heat equation. We see that the solutions to these
equations (harmonic functions) are naturally associated with analytic functions.
Theorem 10 (Harmonic Functions)
If f (z) = u(x, y) + iv(x, y) is analytic in an open
connected domain D, then u(x, y) and v(x, y) are harmonic in D.
Proof: Since f (z) is analytic, u(x, y) and v(x, y) are C (possess continuous partial derivatives
of all orders). We will prove this later. In particular, since they are C 2, the mixed second
derivatives are equal
u
v
v
u
=
,
=
y x
x y
y x
x y
Substituting for the first partial derivatives from the Cauchy-Riemann equations give
2v
2v
= 2,
2
y
x
2u
2u
2 =
y
x2
2-23
Harmonic Conjugates
Theorem 11 (Harmonic Conjugates)
If u(x, y) is harmonic in a simply-connected open domain D then there exists another harmonic
function v(x, y) on D (called the harmonic conjugate of u) such that f (z) = u(x, y) + iv(x, y)
is analytic in D. The conjugate is obtained by solving the Cauchy-Riemann equations.
Example: Find an analytic function f (z) by finding the conjugate of the harmonic function
u(x, y) = x3 3xy 2 + y.
Solution: First check that u(x, y) is harmonic in C
2u
2u
u=
+
= 6x 6x = 0
2
2
x
y
2
v
u
=
= 6xy 1
x
y
Integrating gives
v(x, y) = 3x2y y 3 + g(x) = 3x2y x + h(y)
where the real functions g(x) and h(y) are arbitrary.
h(y) = y 3 + const, and
v(x, y) = 3x2y y 3 x + c,
2-24
Singular Points
Definition: A point at which f (z) fails to be analytic is called a singular point of f (z).
Types of singularities:
1. Isolated Singularities: The point z = z0 is an isolated singularity of f (z) if there is an open
neighbourhood that encloses no other singular point.
2. Poles: If there is a positive integer n such that
lim (z z0)nf (z) = 6= 0
zz0
2-25
2-26
Exponential Function
Definition: The complex exponential is defined by
exp : C C,
Unlike the real exponential ex, the complex exponential function ez is not one-to-one
ez = 1 z = 2ki,
kZ
5i
=e ,
complex period = 2i
3i
i
i
3i
5i
S2
S1
S0
S1
S2
3-1
Logarithm Function
Definition: For z 6= 0 we define the logarithm as the inverse
of the exponential function
z = ew
w = log z
kZ
5i
3i
i
i
3i
5i
S2
S1
S0
S1
S2
k = 0, 1
3-2
Derivative of Logarithm
Theorem 12 (Derivative of Log)
The function Log z is analytic in the cut plane C\(, 0] with derivative
1
d
Log z = ,
dz
z
z C\(, 0]
Proof:
Since Log z is the inverse of ew which is analytic in S0, Log z is analytic in the cut
plane C\(, 0]. There is a jump discontinuity across the branch cut along the negative real
axis:
branch point
branch cut
z 6= 0
3-3
Im(3z i) = 3 Im z 1 = 0
1
x 0, y =
3
i/3
3-4
Complex Powers
Definition: If a, z C and z 6= 0, we define the complex power
z a := ea log z
so that each value of log z gives a value of z a. The principal branch of z a is
z a := ea Log z ,
z 6= 0
d a
d a Log z
a
a
z =
e
= ea Log z = a e(a1) Log z = z a = az a1
dz
dz
z
z
provided the same branch is used for the logarithm defining z a and z a1.
Example: Find the principal value of z = i2i:
By definition with z = i and a = 2i, the principal value is
i2i = e2i Log i = e2i(Log |i| + i Arg i) = e2i(Log 1 + i/2) = e
a 6= z a z a :
Example: If z1 = 1 + i, z2 = i and a = 1
,
show
that
for
principal
values
(z
z
)
1
2
1 2
2
3i/4
We have z1z2 = 1 i = 2e
and
1/2
(z1z2)
1/2
= (1 i)
=e
1
2
Log(1 i)
=e
1
2
Log
2+
1
2
i Arg(1 i)
1
4
3-5
log
z
Log
|z|
+
i
arg
z
1/2
1/2
2
w= z=z
= e2
= |z|
e 2 i arg z
= e2
In particular, arg w = 1
2 arg z and
3
< arg w
for < arg z 3
2
2
Consequently, increasing arg z by 2 (which brings you back to the same point in the complex
z-plane) changes the sign of w
w = z 1/2 =
+|z|1/2 e 2 i Arg z ,
1
|z|1/2 e 2 i Arg z ,
< arg z
(branch 1)
These branches cover the complex plane C twice except at the branch point.
The principal value is the branch 1 value which has the same branch cut as the logarithm
on C\(, 0]
branch point
branch cut
3-6
Riemann Surfaces
The square root function is in fact single-valued and analytic on a Riemann surface of two
sheets
Riemann surface = C\{0} C\{0}
where it is understood that each time we cross the cut we move continuously from one to
the other branch (sheet).
branch point
branch cut
More generally, z m/n is analytic on a Riemann surface with n sheets and log z is analytic
on a Riemann surface with an infinite number of sheets (winding once anti-clockwise around
the branch point z = 0 increases the imaginary part by 2).
3-7
u(x, y) = Re z,
v(x, y) = Im z
3-8
z 2 + 1:
1 log(z + i)]
z 2 + 1 = (z i)1/2(z + i)1/2 = exp[ 1
log(z
i)
+
2
2
Clearly, there are branch points at z = i. Two possible choices of branch cuts are:
3-9
Trigonometric/Hyperbolic Derivatives
The trigonometric and hyperbolic functions are defined in terms of the complex exponential
eiz + eiz
,
cos z :=
2
ez + ez
cosh z :=
,
2
cos z = cosh iz
eiz eiz
,
sin z :=
2i
ez ez
sinh z :=
,
2
sin z = i sinh iz
These functions are all entire. Their derivatives are easily obtained using the derivative of
the exponential and the rules for differentiation. For example,
d
d eiz eiz
sin z =
dz
dz
2i
sin z
cos z
sinh z
cosh z
tan z =
, cot z =
,
tanh z =
,
coth z =
cos z
sin z
cosh z
sinh z
1
1
1
1
sec z =
, cosec z =
, sech z =
, cosech z =
cos z
sin z
cosh z
sinh z
These functions are not entire they are meromorphic (exhibit poles).
The familiar formulas for derivatives of these functions carry over to the complex functions.
For example, away from poles
d
tan z = sec2 z,
dz
d
sec z = sec z tan z
dz
3-10
Inverse Sine
C\{(, 1] [1, )}
Exercise: Show sin :
<
Re
z
<
2
2
Example: Show that the inverse sine function is the multi-valued function
arcsin z = i log[iz + (1 z 2)1/2]
1 iw
(e eiw )
z = sin w =
2i
e2iw 2iz eiw 1 = 0
eiw = iz + (1 z 2)1/2
1
2
Log(1 z 2 )
1
2
Log(1 z 2 )
are
1
3-11
Inverse Trigonometrics/Hyperbolics
The principal branch inverse trigonometric and hyperbolic functions are defined by
Arcsin z = i Log(iz +
1 z 2),
Arcsinh z = Log(z +
1 + z 2)
Arccos z = i Log(z + i 1 z ),
Arccosh z = Log(z +
Arctan z = 1
2 i [Log(1 iz) Log(1 + iz)],
z 1 z + 1)
The multi-valued inverse functions are obtained by replacing the Log with log and the
principal branch square root functions with with the multi-valued square root.
z 6= 1
z 6= 1
z 6= i
(b) arccosh z
(c) arctanh z
3-12
x (1, 1)
z 6= 1
3-13
Complex Sequences
Definition: We say that a complex sequence {an} = {an}
n=1 = {a1, a2, . . .} converges to a
limit L in C and write
lim a = L or
n n
an L as n
lim (an bn ) =
n
lim a
n n
lim a
an
n n
lim
=
if
n bn
lim bn
n
4. Continuity:
lim f (an ) = f
n
lim b
n n
lim b 6= 0
n n
lim a
n n
4-1
Log
, z 6= 0
Log |z|
Exercise: Show that a complex sequence converges if and only if the real and imaginary
parts converge, that is, if an = xn + iyn
lim a = L = A + iB
n n
lim x = A and
n n
lim y = B
n n
lim a = L
n n
4-2
Cauchy Convergence
Definition: A complex sequence an is a Cauchy sequence if for any > 0 there is an N ()
such that
|an am| < for all m, n > N ()
Terms of a Cauchy sequence are arbitrarily close together for n and m sufficiently large.
Theorem 14 (Completeness of R)
(i) A Cauchy sequence of real numbers converges to a limit in R. (ii) A bounded (|an | M )
monotonic sequence (an+1 an or an+1 an) of real numbers converges to a limit in R.
Theorem 15 (Completeness of C)
A complex sequence an converges to a limit in C if and only if it is Cauchy.
Proof: (i) Suppose an L as n then given > 0 there is an N () such that
|an L| <
whenever n > N ()
2
Hence an is Cauchy
|an am| = |(an L) (am L)| |an L| + |am L| < + = whenever m, n > N ()
2
2
(ii) Conversely, suppose an is Cauchy. Then using the inequality | Re z| |z|:
| Re an Re am| = | Re(an am)| |an am| <
4-3
Eulers Number
Theorem 16 (Eulers Number e)
X
1 n
1
1
1
e := lim (1 + ) =
=1+1+
+
+ :=
n
n
n!
2!
3!
n=0
Now an, bn are bounded increasing sequences, so they converge to limits e, with e .
(ii) Now suppose m < n and keep the first m terms of an
1
1
1
1
2
m1
an > 1 + 1 + (1 ) + +
(1 )(1 ) . . . (1
)
2!
n
m!
n
n
n
Holding m fixed and letting n gives
1
1
1
+
+ +
= bm
e1+1+
2!
3!
m!
Taking m we have both e and e so e = .
Complex Exponential
Theorem 17 (Complex Exponential)
n
w
,
ew = lim 1 +
n
n
wC
n
w
Log
(1+ w
n)
lim e
= lim (1 + )n
n
n
n
X
wn
w3
w2
+
+ =
= 1+w+
2!
3!
n=0 n!
The result is true. But this derivation cannot be justified it is not a proof!
We will prove this exponential series later when we consider complex Taylor series. But
first we need to consider the convergence of complex series.
4-5
Complex Series
n=1
n
X
k=1
ak = a1 + a2 + + an
an
n=1
if the sequence of partial sums converges to S. Otherwise, the series is said to diverge.
The index for a series need not start at n = 1. For example, the geometric series is
z n.
n=0
Example: (Geometric Series) Show, for |z| < 1, the geometric series converges with the sum
zn =
n=0
1
,
1z
|z| < 1
1
1 z n+1
=
1z
1z
as n
4-6
an and
n=1
n=1
1. Linear:
(Aan + Bbn ) = A
n=1
an = S
2. Parts:
n=1
an + B
Re an = Re S and
an is convergent then Sn =
n=1
n
X
k=1
an diverges.
n=1
ak S as n . So
an = Sn Sn1 S S = 0 as n
Example: The geometric series
Im an = Im S
n=1
n
n=1
n=1
If
bn
n=1
n=1
n=0
n
clear that |an| = |z| diverges. Similarly, for |z| = 1, |an| = |z|n = 1 6 0. Since limn an 6= 0
or does not exist in these cases we conclude from the divergence test that the series diverges.
In summary, the geometric series converges for |z| < 1 and diverges for |z| 1.
4-7
The comparison, ratio and root tests (familiar from real analysis) can be used to establish
convergence of complex series. See the supplementay material at the end of the slides for
this week.
X
zn
z
Example: Compare use of the ratio and root tests to show the exponential series e =
n=0 n!
converges for all z C:
Solution: Ratio test is easy
an+1
|z|
=
0 as n
an
n+1
n! 2n (n/e)n
|an|
1/n
|z|
e|z|
0 as n
=
1/n
n
(n!)
X
(2)3n+1z n
Example:
converges absolutely for all z C since:
n
n
n=1
23+1/n |z|
1/n
= lim |an|
= lim
=0<1
n
n
n
It is usually easier to use the ratio test (rather than the root test) when it applies.
4-8
1
n
n=0 (4 + 2z)
This is a geometric series
wn with w = (4 + 2z)1
n=0
4 + 2z
1
1
=
,
=
1
n
3 + 2z
1 4+2z
n=0 (4 + 2z)
since |w| < 1
|z + 2| > 1
2.
1
|z + 2| > 2
X
cos n
Example: Show that the series
is convergent for all
n
n=0 2
i
1 < 1 we have a convergent geometric series
Since | e2 | = 2
i n
X
e
n=0
1
i
1 e2
4 2 cos + 2i sin
2 ei
2
=
=
2 ei 2 ei
5 4 cos
i n
X
X
cos n
e
4 2 cos
=
Re
=
n
2
5 4 cos
n=0
n=0 2
4-9
Harmonic Series
Theorem 20 (Harmonic Series)
The harmonic series, given by the Riemann zeta
function (p) with p R, converges if p > 1 and diverges if p 1
X
1
,
(p) =
p
n=1 n
p>1
n
n
X
X
1
1
as n
>
Sn =
p
k
k
k=1
k=1
1
(iii) For p > 1 we have by the geometric series since p1 < 1
2
1
1
1
1
1
1
1
1
+ p +
+ p + p + p + +
+ + k
S2k 1 = 1 +
2p
3
4p
5
6
7
(2k1)p
(2 1)p
4
2k1
1
1
1
1
2
1 + p + p + + k1 p = 1 + p1 + p1 + + k1 p1 <
1
2
4
(2
)
2
4
(2
)
1 p1
X
2
1
,
=
(2) =
2
n
6
n=1
X
1
4
(4) =
,
=
4
n
90
n=1
etc.
4-10
Power Series
Definition: A series of the form
an(z) =
n=0
n=0
an (z a)n
A power series may converge at some, all or no points on the circle of convergence
|za| = R. The largest disk on which a power series converges is called its disk of convergence.
This is either an open or closed disk of the form |z a| < R or |z a| R.
4-11
Radius of Convergence
In practice, however, the radius of convergence is usually determined by applying the ratio or
root test:
an+1
an+1
an+1(z)
1
= |z a| lim
<1
= lim
= lim
n an
n an (z)
R n an
1
= lim |an|1/n
R n
zn
z2
z3
z
+
+
= 1+ +
2
4
9
16
n=0 (n + 1)
Solution: The convergence is determined by the ratio test with
2
an+1
(n
+
1)
1
= lim
= lim
=1
2
n
n
R
an
(n + 2)
4-12
X
f (n)(a)
n=0
n!
f (a)
(z a) = f (a) + f (a)(z a) +
(z a)2 +
2!
is called the Taylor series of f (z) around a. If a = 0, it is called the Maclaurin series for f (z).
Example: Find Maclaurin series for ez , eiz , cosh z, sinh z, cos z, sin z and Log(1 + z) giving
the disk of convergence:
Solution: These can be obtained using the formulas
dn z
= 1,
e
n
dz
z=0
dn iz
n
=
i
e
dz n
z=0
n
dn
d
z
z
1
1 [1 + (1)n ],
cosh
z
(e
+
e
)
=
=
2 dz n
2
dz n
z=0
z=0
dn
n1
=
(1)
(n 1)!
Log(1
+
z)
n
dz
z=0
etc
4-13
X
zn
z2
z3
zn
=1+z+
+
+ +
+
e =
n!
2!
3!
n!
n=0
X
z2
z3
(iz)n
(iz)n
iz
= 1 + iz
i
+ +
+
e =
n!
2!
3!
n!
n=0
z
|z| <
z 2n
z2
z4
z 2n
cosh z =
=1+
+
+ +
+
2!
4!
(2n)!
n=0 (2n)!
|z| <
|z| <
z3
z5
z 2n+1
z 2n+1
=z+
+
+ +
+
sinh z =
3!
5!
(2n + 1)!
n=0 (2n + 1)!
2
4
2n
2n
z
z
z
z
=1
+
+ (1)n
+
(1)n
cos z =
(2n)!
2!
4!
(2n)!
n=0
|z| <
|z| <
2n+1
z 2n+1
z3
z5
n z
(1)
sin z =
=z
+
+ (1)
+ |z| <
(2n + 1)!
3!
5!
(2n + 1)!
n=0
Log(1+z) =
(1)
n=1
n
z2
z3
n+1 z
=z
+
+ (1)
+
n
2
3
n
n
n+1 z
|z| < 1
4-14
Taylors Theorem
Theorem 23 (Taylors Theorem)
If f (z) is analytic in the disk |z a| < R then the Taylor series converges to f (z) for all z in
this disk. The convergence is uniform in any closed subdisk |z a| R < R. Specifically,
f (z) =
n
X
f (k)(z)
k=0
k!
(z a)k + Rn(z)
|za|R
|Rn(z)| 0
as n
w
z
a
R
R
R
4-15
General lH
opitals Rule
Theorem 24 (General lH
opitals Rule) If f (z) and g(z) are analytic at z = a and f (z),
g(z) and their first n 1 derivatives all vanish at z = a but the nth derivative g (n)(a) 6= 0 then
f (n)(a)
f (z)
f (n)(z)
=
lim
= lim
za g(z)
za g (n)(z)
g (n)(a)
Proof: Since f (z) and g(z) are both analytic in an open disk around z = a we can represent
them by their Taylor series
X
f (k) (a)
(z a)k
X
f (k) (a)
(z a)kn
X
f (k)(a)
(z a)kn
k!
k!
f (z)
k=0
k=n
= lim
= lim
lim
za X g (k)(a)
za X g (k)(a)
za g(z)
k
(z a)
(z a)kn
k!
k!
k=0
k=n
lim
za
=
lim
za
k!
k=n
X
g (k)(a)
k=n
k!
f (n)(a)
f (n)(z)
=
= lim
(n)
za g (n)(z)
g (a)
kn
(z a)
and use the quotient rule, uniform convergence and continuity of the series and continuity of
the nth derivatives.
4-16
Example: General lH
opitals Rule
2 cosh z 2z 2
1
Example: Evaluate (i) lim
, (ii) lim ( cot z) :
4
z0
z0 z
z
(i) The first three derivatives vanish at z = 0 so by lH
opital
2 cosh 0
f (z)
f (4) (a)
1
=
lim
=
=
za g(z)
4!
12
g (4)(a)
(ii) Alternatively you can use Taylor series
sin z z cos z
1
lim ( cot z) = lim
z0
z0 z
z sin z
z 3/3 +
(z z 3/6+ ) z(1 z 2/2 + )
= lim
= 0
= lim
3
2
z0 z +
z0
z(z z /6 + )
4-17
X
z z3! + z5!
z 2n
z2
z4
sin z
n
(1)
=
= 1
+
=
f (z) =
z
z
3!
5!
(2n + 1)!
n=0
The derivatives f (n)(z) are undefined at z = 0, so the coefficients in the Taylor series must
be interpreted as limits
f (n)(0)
f (n)(z)
7 lim
z0
n!
n!
Hence the Taylor coefficients are
sin z
=1
z0 z
z0
3
2
3
z
z
z
z(1 2 + )(z 6 + )
3 +
z cos z sin z
f (z)
(2 z 2) sin z 2z cos z
lim
= lim
z0 2!
z0
2z 3
2
3
3
z
z
z
2
3 +
(2 z )(z 6 + )2z(1 2 + )
1
1
= lim
= =
= lim
z0
z0
2z 3
2z 3
6
3!
and so on.
4-18
f (x) = e
Solution: Note that f (x) is C (it has continuous partial derivatives of all orders) but it is
not analytic it has an essential singularity at x = 0. Again since the derivatives f (n) (x)
are undefined at x = 0 we must interpret the coefficients in the Taylor series as limits
f (n) (x)
f (n)(0)
7 lim
x0
n!
n!
Hence the Taylor coefficients are
1/x2
x0
lim
x0
x0
f (x)
2!
= lim
x0
= 0,
2
2
1/x
(4 6x )e
2x6
= 0,
2e1/x
=0
lim f (x) = lim
3
x0
x0
x
2
2
4
1/x
4(2 9x + 6x )e
f (x)
= lim
lim
=0
9
x0
x0 3!
6x
1/x2
n
2n+2
n
e
|x|
(n+1)!
x
|x|
=
0 as x 0
2 (1/x2 )n+1 =
n
n
1/x
x
|x|
e
(n+1)!
We conclude that the Taylor series of f (x) about x = 0 vanishes identically and so it does
not converge to f (x) in any open neighbourhood of x = 0. Existence of the Taylor series is
not sufficient to guarantee convergence to the function we need analyticity!
4-19
Manipulation of Series
Theorem 25 (Adding Series)
Let A and B be constants and let f (z), g(z) be analytic with Taylor series
f (z) =
n=0
an(z a)n,
g(z) =
n=0
bn(z a)n
n=0
dn
dz
(n)
(n)
[Af
(z)
+
Bg(z)]
=
Af
(z)
+
Bg
(z).
n
f (z) =
n=0
an(z a)n,
g(z) =
n=0
bn(z a)n
then the Taylor series of f (z)g(z) around a is given by the Cauchy product
f (z)g(z) =
n=0
cn(z a) ,
cn =
n
X
ank bk
k=0
4-20
z1n
f (n) (z)
=
an =
n!
n!
z=0
z2n
g (n)(z)
=
bn =
n! z=0
n!
X
X
zn
zn
2 zn
1 zn
f (z)g(z) = ezz1 ezz2 =
n=0 n!
n=0 n!
X
cnz n = ez(z1+z2)
=
n=0
n
X
The result follows by taking z 1 in ezz1 ezz2 = ez(z1 +z2) using continuity which follows from
uniform convergence of the series on |z| r for any r > 0.
4-21
z3
2z 5
anz = z
tanh z =
+
3
15
n=0
n
Solution:
z2 z4
cosh z tanh z = (1+ + + )(a0 +a1z +a2z 2 +a3z 3 + )
2! 4!
a1 3
a2 a0 4
a3 a1 5
a0 2
= a0 + a1z + (a2 + )z + (a3 + )z + (a4 + + )z + (a5 + + )z +
2
2
2
4!
2
4!
z5
z3
+
+
= sinh z = z +
3!
5!
1
2
a0 = 0, a1 = 1, a2 = 0, a3 = , a4 = 0, a5 =
, ...
3
15
4-22
Limits of Integrals
Lemma 27 (Limits of Integrals)
Let {an(z)} be a sequence of continuous functions on a region S and let be a curve inside
S. If an(z) converges uniformly to a(z) on then
Z
lim
a (z)dz =
lim an(z)dz = a(z)dz
n n
n
Z
Z
Z
an(z)dz a(z)dz = (an (z) a(z))dz
Z b
|an(z(t)) a(z(t))||z (t)|dt
a
Z b
sup |an(z) a(z)|
|z (t)|dt
z
a
4-23
Term-By-Term Integration
Theorem 28 (Term-By-Term Integration)
Let {an(z)} be a sequence of continuous functions on a region S and let be a curve inside
S. If a(z) =
n=1
a(z)dz =
n=1
an (z)dz =
Z
X
n=1
an(z)dz
Example: Use term-by-term integration to find the Maclaurin series of Log(1 + z):
|z| r < 1 we have
n+1
z X
z
X
X
dw
n
n n
nz
n
=
(1)
(1) w dw =
(1)
w dw =
Log(1+z) =
n+1
0 n=0
0
0 1+w
n=0
n=0
Uniform and absolute convergence follows by Weierstrass test.
Z z
On
4-24
Term-by-Term Differentiation
Lemma 29 (Term-by-Term Differentiation)
If an(z) exists in S C,
n=1
n=1
X
d X
an(z) for all z S
an(z) =
dz n=1
n=1
Proof: See text similar to term-by-term integration.
X
f (n)(a)
n=0
n!
(z a)n
the Taylor series for f (z) is given by term-by-term differentiation of the Taylor series for f (z).
Proof: The derived series obtained by term-by-term differentiation is
f (z) =
X
f (n)(a)
n=1 (n 1)!
(z a)n1
Since this is the Taylor series of the analytic function f (z) it converges uniformly in any
closed subdisk of the disk of convergence. So the result follows from the lemma.
The derived series for f (z) converges with the same disk of convergence as for f (z).
Example: Use term-by-term differentiation to find the Maclaurin series of (1 z)2:
1
d
1
=
(1 z)2
dz 1 z
X
X
d
d X
zn =
n z n1,
=
(z n) =
dz n=0
n=0 dz
n=1
|z| r < 1
4-25
The remaining slides of Week 4 are supplementary material on complex series. For the
most part, this material involves results and concepts obtained by relatively straightforward
extensions of the corresponding results and concepts for real series previously encountered in
real analysis.
Knowledge of most of this material will be assumed for completing problems on the problem
sheets and assignments as well as in the exams. In particular, you should be familiar with the
following topics:
This material we not be included in lectures. Please see the recommended textbooks for
more details on this material.
4-26
an is absolutely convergent if
n=1
n=1
|an| is convergent.
complex series that is convergent but not absolutely convergent is called conditionally convergent.
n=1
Proof: Let Sn =
n
X
ak , Tn =
k=1
n
X
k=1
sequence of partial sums Sn is also Cauchy and therefore converges since for n > m > N ()
|Sn Sm| =
n
X
k=m+1
ak
n
X
k=m+1
4-27
X
(1)n1
converges conditionally:
n=1
1/2
S2
S4
X
1
S6 S8
n=1 n
S7 S5
1
S1
S3
diverges.
(ii) The even and odd partial sums are monotone and bounded
1
1
1 1
1 1
1
S2n = 1
+
+
+ +
2
3 4
5 6
2n 1 2n
1
1
1 1
1
1
<1
1
= 1
2 3
2n 2 2n 1
2n
2n
1 1
1
1 1
1
S2n1 = 1
2 3
4 5
2n 2 2n 1
1 1
1
1
1
1
1
+
+ +
+
=
2
3 4
2n 3 2n 2
2n 1
2
By completeness, they converge, and to the same limit since
lim (S
S2n) = lim
n 2n1
n
1
=0
2n
lim S
= lim S2n
n 2n1
n
4-28
Rearranging Series
Theorem 32 (Rearranging Absolutely Convergent Series)
The terms of an absolutely convergent series can be arbitrarily rearranged.
Proof: See text.
Since the previous series is absolutely convergent for |z| < 1, we can rearrange its terms
Log(1 + z) =
nz
(1)
n=0
n+1
= z + z3 z2 + z5 + z7 z4 +
n+1
4n3
X
z
|z| < 1
z 4n1
z 2n
+
=
, |z| < 1
4n 1
2n
n=1 4n 3
But this is not a power series, Abels theorem does not apply and the result does not hold
at z = 1. In fact for n 1 we have the identity
1 + 1 1 =
4n3
4n1
2n
1 1
1 1 + 1 1 +1
4n3
4n2
4n1
4n
2 2n1
2n
1
1
1
+
4n 1 2n
n=1 4n 3
1 1
1
1 1
1 Log 2 = 3 Log 2
= 1 + + + + = Log 2 + 2
2
3 2
5
7 4
Comparison Test
Theorem 33 (Comparison Test)
(i) If |an| bn for n n0 and
bn = T is convergent, then
n=n0
n
X
k=n0
cn diverges then
n=n0
n
X
|ak |, Tn =
an is absolutely convergent.
n=n0
dn diverges.
n=n0
bk . Then
k=n0
Sn Tn < T
and the increasing sequence of partial sums Sn converges by the completeness of R. So
n=n0
an =
nX
01
an +
an
n=n0
n=1
n=1
n
X
k=n0
So Un as n since
n
X
k=n0
ck .
dk
n
X
ck
k=n0
4-30
Ratio Test
Theorem 34 (Ratio Test) Suppose that
an+1
= lim
n an
n=1
nN
m>0
and the series is absolutely convergent by comparison with the geometric series.
(ii) If > 1 and r is such that 1 < r < then
|an+1| > |an|r > |an|,
nN
X
(1)n
, the conditionally convergent series
(iii) = 1 for the absolutely convergent series
2
n=1 n
X
X
1
(1)n
n=1
n=1 n
4-31
zn
is absolutely convergent for |z| 1 :
Example: Show that the series
n=1 n(n + 1)
n
zn
|z|
1
1
=
= bn
2
n(n + 1)
n(n + 1)
n(n + 1)
n
So the series converges absolutely for |z| 1 by comparison to the convergent harmonic series
with p = 2 > 1.
|an| =
X
(3)n
converges absolutely (ii)
diverges since:
Example: (i)
2
n=1 n
n=1 n!
= 0 < 1 (i)
lim
n n + 1
an+1
=
= lim
n an
3n2
= 3 > 1 (ii)
lim
n
2
(n + 1)
X
(3)n
X
(z + 2)n1
Example: Show that the series
converges absolutely for all |z + 2| 4:
n
3
n=1 4 (n + 1)
3
an+1
|z
+
2|
(n
+
1)
|z + 2|
= lim
= lim
<1
=
3
n an
n
4
(n + 2)
4
so the series converges absolutely for |z + 2| < 4. If |z + 2| = 4, the series converges absolutely
by comparison to the harmonic series with p = 3 > 1:
|z + 2|n1
1
1
|an| = n
=
4 (n + 1)3
4(n + 1)3
n3
4-32
Root Test
Theorem 35 (Root Test) Suppose that
= lim |an|1/n
n
n=1
n=N
|an|
nN
rn
n=N
So the series is absolutely convergent by comparison with the geometric series. (ii) and (iii)
are similar to the ratio test.
4-33
Uniform Convergence
Definition: The sequence {an(z)} converges uniformly to a(z) on a subset S C if for any
> 0 there is an N () (independent of z S) such that
sup |an(z) a(z)| < whenever n > N ()
zS
The series
n=1
uniformly to a(z) on S.
n=1
converges pointwise on S.
Proof: (i) We need to show that for each z S
|an(z) a(z)| < whenever n > N (, z)
4-34
an (z) converges uniformly on S and an(z) is continuous for each n then S(z)
n=1
zz0
an(z) =
n=1
n=1
lim a (z) =
zz0 n
an(z0),
n=1
z0 S
Proof: (i) We are given that (a) {an(z)} converges uniformly to a(z) on S, (b) an (z) is
continuous at z0, that is,
whenever n > N ()
(a)
sup |an(z)a(z)| <
3
zS
zS
<
+ + = whenever |z z0| < (, z0) by (a) and (b)
3
3
3
(ii) Follows by applying (i) to the sequence of continuous partial sums.
4-35
a(x) = lim x =
n
0,
1,
x [0, 1)
x=1
The sequence converges uniformly and absolutely on any closed interval [0, r] with r < 1 but
does not converge uniformly on the closed interval [0, 1]:
Solution: On the interval [0, r] the convergence is uniform
and absolute
0.8
x[0,r]
x[0,r]
0.6
x[0,1]
x[0,1]
0.4
0.2
0.2
0.4
0.6
0.8
Exercise: Discuss the convergence, uniform or otherwise, of an(z) = |z|n on the square
0 Re z 1, 0 Im z 1.
4-36
Weierstrass M -Test
Theorem 38 (Weierstrass M -Test)
Suppose (i) |an(z)| Mn for all z S and (ii)
uniformly and absolutely on S.
Mn converges. Then
the series T =
n=1
an(z) converges
n=1
n=1
n=1
n
X
ak (z),
Tn =
Mk
k=1
k=1
|Sm(z) Sn(z)| =
n
X
m
X
k=n+1
m
X
k=n+1
ak (z)
m
X
k=n+1
|ak (z)|
Mk = |Tm Tn|
4-37
X
1
converges uniformly and absolutely for
Example: The Riemann zeta function (z) =
z
n
n=1
Re z p > 1 :
(x+iy)Log n
z
|an(z)| = |n | = e
= ex Log n
ep Log n = np = Mn
X
1
Since the harmonic series
is convergent for p > 1, the result follows from the Weierp
n=1 n
strass M -test.
X
zn
converges uniformly and absolutely on |z| 1.
Example: Show that the series
2
n=1 n
= Mn
2
2
n
n
n=1
4-38
X
zn
n=1 n
r n = Mn
|an(z)| =
n
n
n=1
X
zn
n=1 n
X
zn
= Log(1 z),
|z| < 1
4-39
n=0
an (z a)n
has the disk of convergence |z a| < R and converges at z = w on the circle of convergence
|z a| = R. Then
f (w) = lim f (z) =
zw
n=0
an (w a)n
nz
(1)
n=0
n+1
n+1
|z| < R = 1
X
(1)n
X
1
1
Log 2 = lim Log(1 + z) =
=
z1
2n
n=0 n + 1
n=1 2n 1
|z|<1
= 1
1 1
1
+ +
2
3 4
4-40
Complex Integrals
Definition: Consider a complex-valued function of a real variable t
z : [a, b] C
where x(t), y(t) are continuous on [a, b]. Then we define the complex integral
Z b
a
z(t)dt :=
Z b
a
(x(t) + iy(t))dt =
Z b
a
x(t)dt + i
Z b
a
y(t)dt
where the real integrals giving the real and imaginary parts
Re
Z b
a
z(t)dt =
Z b
a
Re z(t)dt,
Im
Z b
a
z(t)dt =
Z b
a
Im z(t)dt
exist by continuity as the limits of Riemann sums where {tj } is a partition of [a, b] with t0 = a,
tn = b and tj = tj tj1
Z b
a
x(t)dt := n
lim
tj 0
n
X
x(tj ) tj
j=1
5-1
Complex Primitives
Theorem 40 (Complex Primitives)
If the complex-valued function z(t) is continuous on [a, b] and has a primitive Z(t) such that
Z (t) = z(t) then
Z b
a
(t2 + it) dt
(t2 + it) dt =
Z 1
eit dt =
Z 2
0
Z 2
0
"
t3
3
+i
#1
2
t
2 0
1
1
+ i
3
2
eit dt
2
=0
Or better
Z 2
0
2
1 2i
1
it
it
e
1) = 0
= (e
e dt =
i
i
0
5-2
Primitives
Definition: If f (z) and F (z) are analytic in an open connected domain D and F (z) = f (z)
then F (z) is called a primitive (antiderivative or indefinite integral) of f (z) and denoted
F (z) =
f (z) dz + constant
Theorem 41 (Primitives) If F (z), G(z) are both primitives of f (z) in a connected open
domain D so that F (z) = G(z) = f (z) everywhere in D, then F (z) G(z) is constant in D.
Proof: Let H(z) = F (z)G(z) = u+iv. Since H (z) = 0 it follows from the Cauchy-Riemann
equations that in D
u
u
v
v
=
=
=
=0
x
y
x
y
Integrating gives
u = const,
v = const
H = u + iv = const in D
1,
0,
|z| < 1
|z| > 1
is analytic and f (z) = 0 on its domain of definition D but f is not constant on the disconnected domain D.
5-3
Smooth Curves
We want to define contour integrals along curves or contours which generalize straight line
segments and arcs of circles. To do this we define the notion of smooth curves.
Definition: A smooth curve is a function z : [a, b] R C such that z(t) = x(t) + iy(t)
where x(t), y(t) are C 1 and hence differentiable. The curve is simple (non-self-intersecting)
if z(t1) 6= z(t2) for a t1 < t2 < b and is closed if z(a) = z(b). A smooth curve is directed or
oriented with initial point z(a) and endpoint z(b).
n
Definition: A contour =
j=1 j consists of a finite sequence of smooth curves j ,
j = 1, 2, . . . , n with endpoints joined to initial points. A contour is thus piecewise smooth
(differentiable) and allows for self-intersections, corners and cusps (corners with common
tangents).
t=b
smooth
curve
2
not
simple
t=a
cusp
1
2
5
4 closed
contour
3
5-4
Jordan Theorem
Theorem 42 (Jordan Theorem) A simple closed curve divides the plane C into two open
regions having the curve as their common boundary. The bounded region is the interior of
and the unbounded region the exterior. The curve is positively oriented if the interior
always lies to the left.
Proof: Very difficult see textbook.
t=b
smooth
curve
2
not
simple
t=a
cusp
1
2
5
4 closed
contour
5-5
Contour Integrals
Definition: Let f (z) = u(x, y) + iv(x, y) be continuous in an open region containing the
smooth curve . Then we define the contour integral of f (z) along by
b
dz
dt =
(u + iv)(x(t) + iy (t))dt
f (z)dz :=
f (z(t))
dt
a
Z b
Z b
a
Z b
a
The real Riemann integrals exist because the integrands are continuous. In terms of line
integrals in vector analysis
Z
f (z)dz =
(u + iv)(dx + idy) =
udx vdy + i
Pn
j=1 j is defined by
f (z)dz :=
n Z
X
j=1 j
vdx + udy
f (z)dz
A contour integral over a simple closed contour in the positive (counter-clockwise) sense is
denoted
I
5-6
Z
Af (z) + Bg(z) dz = A
Z
2. Orientation:
3. Additivity:
f (z)dz =
f (z)dz =
f (z)dz + B
g(z)dz
f (z)dz
f (z)dz +
1
1+2
Z
Z b
|f (z)||z (t)|dt,
4. Bound: f (z)dz
a
5. Change of Variables:
f (z)dz
: z = z(t), t [a, b]
f (z)dz =
f (g(w))g (w)dw
where z = g(w), is the image of under the change of variables and g(w) is analytic in a
region containing .
5-7
1 + i:
Z 1
dz
dt =
z
dt
0
zdz =
= 2i
Z 1
0
Z 1
0
tdt = 2i
(1 + i)t(1 + i)dt
2 1
t
= i
2 0
dz
where is the circle |z| = 1 traversed
z
Solution: We parametrize the circle |z| = 1 as z(t) = eit with t [0, 2] so that
z(0) = z(2) = 1 and z (t) = ieit
I
dz
=
z
dz
dz
dz
+
= 2
|z|=1 z
|z|=1 z
|z|=1 z
= 2
Z 2
= 2i
2
1 dz
z
dt = 2
eit(ieit )dt
dt
0
0
Z 2
0
dt = 4i
5-8
Path Dependence
Example: Evaluate the line integral
1+i
1 1
(a) We parametrize the straight line segment by z = (1 + i)t with z (t) = 1 + i, z = (1 i)t
and t [0, 1]
Z
2 1
Z 1
Z 1
dz
t
dt =
=1
z dz =
z
(1 i)t(1 + i)dt = 2
tdt = 2
dt
2 0
0
0
0
Z 1
(b) We parametrize (i) 1 : z = t with z (t) = 1 and t [0, 1] and (ii) 2 : z = 1 + it with
z (t) = i and t [0, 1]
Z
2 1
Z 1
dz
t
1
z dz =
z
dt =
=2
tdt =
dt
2 0
0
1
0
1
1
dz
1+i
dt =
(1 it) i dt =
(t + i)dt = 2
z dz =
z
dt
2
0
0
0
Z 1
Z 1
1 +2
z dz =
1 + i) = 1 + i
z dz = 1
+
(
2
2
2
z dz =
1 + i) 1 = i
z dz = 1
+
(
2
2
5-9
Path Independence
Z
1 1
(a) Parametrize the quarter circle |z| = 1 by z = eit with z (t) = ieit and t [0, /2]
Z
Z /2
/2
/2
2
2 dz
2it it
z dz =
z
dt =
e ie dt = i
e3it dt
dt
0
0
0
1 3i/2
1
1 3it /2
e
1) = (1 + i)
= (e
= i
3i
3
3
0
(b) We parametrize (i) 1 : z = t with z (t) = 1 and t [0, 1] and (ii) 2 : z = it with
z (t) = i and t [0, 1]
3 1
Z 1
dz
t
1
2
2
2
2
z dz =
z dz =
dt =
z
t dt =
=
dt
3 0
3
1
1
0
0
Z 1
Z 1
1
1
dz
i
2
2
2
z dz =
z
dt =
(it) i dt = i
t2dt =
dt
3
2
0
0
0
1
1
i
z dz =
z dz = = (1 + i)
3 3
3
1 +2
2
1
z 2 dz =
1
1
z 2 dz = (1 + i) + (1 + i) = 0
3
3
2
5-10
Greens Theorem
Theorem 44 (Greens Theorem) If R is a closed region of the xy plane bounded by a
simple closed curve and if M and N are C1 in R then
I
M dx + N dy =
ZZ
R
=R
N
M
x
y
dxdy
z dz =
xdx+ydy + i
5-11
M dx =
Z b
a
M (x, Y1(x))dx +
Z a
b
=
=
N dy =
Z f
e
y=Y1 (x)
N (X2, y)dy +
Z e
f
x=X1 (y)
dy dx =
ZZ
R
M
dxdy
y
N (X1, y)dy
M (x, Y2(x))dx
Z b
a
dx dy =
ZZ
R
N
dxdy
x
Adding gives the required result. If R is not convex (or is multi-connected), we can subdivide
R into two or more convex regions by cuts and use additivity.
5-12
Cauchys Theorem
Theorem 45 (Cauchys Theorem) If f (z) is analytic in a simply-connected open domain
D and f (z) is continuous in D then for any simple closed curve in D
I
f (z)dz = 0
Proof: Let R be the union of and its interior so that = R. The result then follows from
Greens theorem (which requires simple and f (z) continuous)
I
f (z)dz =
=
(u+iv)(dx+idy) =
ZZ
R
udxvdy + i
v u
dxdy + i
x y
ZZ
R
vdx+udy
u v
dxdy = 0
x y
v
u
=
x
y
f (z)dz = 0
Proof: Difficult see text. Removes requirements that is simple and f (z) is continuous in
R. Note = R must be positively oriented so that R is always on the left. Also the region R
need not be simply-connected provided all of the boundary of R is included in = R.
5-13
1
Z
Z
Z
cos z dz =
Z 1
cos z dz = i
cos z dz =
cos z dz = i
cos t dt = sin t
0
Z 1
0
Z 1
= sin 1
1
cos(1+it)dt = sin(1+it)
1
cos(it) dt = sin(it)
cos z dz =
= sin(1+i)sin 1
1
cos(i+t) dt = sin(i+t)
0
Z 1
0
1
3
i
= sin(1+i) sin i
1+i
2
= sin i
cos z dz
ez dz
:
Example: Evaluate the contour integral
2
|z|=2 (z 9)
I
The integrand is analytic except at the poles z = 3, so it is analytic in and on the circle
|z| = 2. The closed contour integral therefore vanishes by Cauchys theorem.
5-14
Proof: Recall that, in vector analysis, a C 1 vector field A on a simply-connected open domain
D is conservative and can be written as A = if and only its curl, A, vanishes. So let
A1 = (u, v, 0), A2 = (v, u, 0) and use the Cauchy-Riemann equations to show that in R3
j
i
A1 = x y
u v
i
A2 = x
v
k
u
v
z = y + x k =
0
k
u
v
z = x y k =
0
Note that if dr = (dx, dy, dz) then since A1 and A2 are conservative
Z
f (z)dz =
=
Z
Z
(u + iv)(dx + idy) =
(u, v, 0) dr + i
u dx v dy + i
(v, u, 0) dr =
v dx + u dy
A1 dr + i
A2 dr
z
2
f (z)dz = F (z) = F (z2) F (z1)
z1
Z z
z0
f (w)dw,
z0 D
zD
6-1
f (z)dz =
1 2
f (z)dz =
f (z)dz
f (z)dz
Since f (z) must be analytic everywhere between 1 and 2, this region must be simply
connected so that 1 can be continuously deformed into 2. By path independence, F (z) is
well-defined in D with derivative
F (z + z) F (z)
F (z) = lim
z0
z
1
= lim
z0 z
Z z+z
z0
Z z
z0
f (w)dw
z+z
1
f (w)dw = f (z)
= lim
z0 z z
This last limit follows, since f (z) is continuous, by the following Lemma.
6-2
Estimating Integrals
Lemma 49 (Continuity Lemma) If f (z) is continuous in an open domain containing z and
z + z then
z+z
1
lim
f (w)dw = f (z)
z0 z z
whenever
|w z| < ()
1
Length() =
f (w) f (z)
dt
dt =
|z| t
dt
|z| t
dt
|z|
Here we used the path independence of the contour integral to choose to be the straight
line segment between z and z + z so that Length() = |z|.
Theorem 50 (Integral Estimate) If f (z) is continuous in an open domain containing and
|f (z)| M on , then
Z
Z t
dz
2
f (z)dz
f (z) dt M Length()
dt
t1
6-3
Z z
2
z1
|z|=1
z ndz:
z
2
n+1
,
Z z
n+1
2 n
z1
z dz =
h
iz
z1
2
,
Log z
z1
|z|=1
z ndz = lim
Z z
2
0 z1
n 6= 1
z2
n = 1
z ndz =
0,
2i,
|z| = 1
n 6= 1
n = 1
z1
h
iz
dz
2
=0
= lim Log(z)
z
z
z
2
1
1
|z+2|=1 z
6-4
Trigonometric Integrals
Example: Show that
Z 2
0
cos2n t dt =
1 3 5 (2n 1)
2,
2 4 (2n)
n = 1, 2, 3, . . .
1) and
(z
+
z
Solution: Let z = eit so that dz = ieitdt = izdt and dt = dz/iz. Then cos t = 1
2
by the binomial theorem
Z 2
0
cos2n t dt =
I
2n
i2n dz
X
1
2n
2n2k dz
1 (z + z 1)
=
z
iz
22n i |z|=1 k=0 k
z
|z|=1 2
2n
= 2n
2 i n
=
I
1
2n
dz
= 2n
2 i n
|z|=1 z
2i
(2n)(2n 1)(2n 2) 1
1 (2n)!
2
=
2
2n
n
n
2
n! n!
(2 n!)(2 n!)
(2n 1)(2n 3) 3 1
2
=
(2n)(2n 2) 4 2
|z|=1
z ndz =
0,
2i,
n 6= 1
n = 1
Exercise: Establish the previous cosine integrals using the methods of real analysis, that is,
integration by parts and recursion relations.
6-5
a
1
0,
dz
=
2i,
za
a outside
a inside
1
is analytic in C\{a}. Hence if a lies outside the integral
za
vanishes by Cauchys theorem. If a lies inside , the contour can be deformed into a small
circle of radius > 0 centered on a with z a = eit, z (t) = ieit and 0 t 2 so that
Proof: The integrand f (z) =
2 i eit
2
dz
dz
=
=
dt = i
dt = 2i
eit
za
|za|= z a
0
0
Note that the analyticity domain of f (z) between and is not simply-connected but that
introducing a cut 1 gives a simply-connected domain R. Then Cauchys theorem implies
I
f (z)dz =
f (z)dz =
f (z)dz
f (z)dz = 0
6-6
1 2
R
3z 2
2
1
Solution: By partial fractions, f (z) = 2
= +
is analytic in C\{0, 1}.
z z
z
z1
We introduce small circles centered at z = 0, 1 and two cuts 1, 2 so that the region R
between and the circles is simply-connected. Cauchys theorem then tells us that we can
deform the contour
I
f (z)dz =
f (z)dz =
f (z)dz =
f (z)dz = 0
2
1
+
dz
z
z1
= 2(2i) + 0 + 2 0 + 2i = 6i
where each of the four integrals is evaluated using the previous theorem.
This method easily generalizes to evaluate closed contour integrals of any rational function
of z with only simple poles.
6-7
a inside
a
1
R
Proof: The integrand is analytic in D\{a}. We introduce a small circle of radius > 0
centered on z = a and a cut 1 so that the region R between and is simply-connected.
We parametrize by z a = eit with z (t) = ieit. By deforming the contour we find
f (z)
f (z)
dz =
dz
za
za
Since the LHS is independent of we can take 0
I
2
f (z)
f (z)
dz = lim
dz = i lim
f (a + eit)dt
0 z a
0 0
za
= i
Z 2
0
lim f (a + eit)dt = i
Z 2
0
6-8
A Trigonometric Integral
|z| = 1
Example: Evaluate the integral I =
Z 2
0
d
:
3 2 cos
z+
dz
d
= i
I =
|z|=1 z(3 z z 1)
0 3 2 cos
I
I
dz
dz
= i
= i
2
|z|=1 (z z+)(z z )
|z|=1 z 3z + 1
I
f (z) = (z z+)1
which is analytic inside |z| = 1, we have by the Cauchy integral formula
2
2
f (z)
=
dz = (i)(2i)f (z ) =
I = i
z z+
|z|=1 (z z )
5
I
Exercise: Evaluate I using the methods of systematic integration for real integrals, that is,
by using the substitution t = tan(/2).
6-9
Proof:
a inside , n = 0, 1, 2, . . .
dn
1
1
f (z)
n f (z)
dz =
dz
f (a) = n
n
n
da
da 2i z a
2i a z a
"
ez dz
:
Example: Evaluate the integral
2
|z|=4 (z 2)
I
Let f (z) = ez which is entire, then by the general Cauchy integral formula with n = 1
ez dz
2
=
2if
(2)
=
2ie
|z|=4 (z 2)2
6-10
Taylors Theorem
Theorem 55 (Taylors Theorem)
If f (z) is analytic in the disk |z a| < R then the Taylor series converges to f (z) for all z in
this disk. The convergence is uniform in any closed subdisk |z a| R < R. Specifically,
f (z) =
n
X
f (k)(z)
k=0
k!
(z a)k + Rn(z)
|za|R
|Rn(z)| 0
as n
w
z
a
R
R
R
6-11
1
=
wa
(za)n+1
2
n
a
(z
a)
(z
a)
(wa)n+1
1 +
+
+
+
+
za
wa
(w a)2
(w a)n
1 wa
into the first formula, integrate term-by-term and use the second formula we find
f (z) =
n
X
f (k) (a)
k=0
k!
(z a)k + Rn(z)
n+1
2R
R
max |f (w)| 0 as n
wC
2(R R ) R
6-12
Evaluation of Integrals
Example: Show that
eztdz
1
= sin t,
2
2i |z|=2 z + 1
I
tC
|z| = 2
1
2
i
Solution: The integrand is analytic except at the simple poles z = i. We place circles 1
and 2 of radius around these poles so that by deformation of contours
eztdz
=
|z|=2 z 2 + 1
eztdz
+
=
2 z 2 + 1
1
f1(z)dz
f2(z)dz
+
1 z i
2 z + i
I
where
ezt
,
f1(z) =
z+i
ezt
f2(z) =
zi
6-13
8dz
d
= i
I =
|z|=1 z(4 + z + z 1)3
0 (2 + cos )3
I
I
f (z)dz
8z 2 dz
=
i
= i
|z|=1 (z z+)3
|z|=1 (z 2 + 4z + 1)3
|z| = 1
z
z+
8z 2
f (z) =
(z z)3
z+ z = 2 3,
z+ z = 1
=
16
=
I = (i)(2i)f (z+)/2! =
(z z)5
(2 3)5
3
z=z+
6-14
Order m Zeros
Definition: A point z = a is a zero of order m of f (z) if f (z) is analytic at z = a and f (z)
and its first m 1 derivatives vanish at z = a but f (m) (a) 6= 0. A zero of order 1 is a simple
zero.
Lemma 56 (Order m Zero)
Let f (z) be analytic at z = a. Then f (z) has a zero of order m at z = a if and only if
f (z) = (z a)mg(z)
where g(z) is analytic at z = a with g(a) 6= 0.
Proof: Using Taylor series
X
f (n) (a)
f (z) =
n=m
with
g(z) =
n!
(z a)n
X
f (n+m) (a)
n=0 (n + m)!
(z a)n
analytic at z = a
Conversely,
g(z) =
n=0
an(z a)
f (z) = (z a)
n=0
an (z a)n
7-1
Isolated Zeros
Theorem 57 (Isolated Zeros)
Zeros of an analytic function which is not identically zero are isolated, that is, if z = a is a
zero then there are no zeros in a punctured disk about z = a.
Proof: The Taylor series about z = a with Taylor coefficients an converges to f (z) on some
open disk about z = a. If an = 0 for all n then f (z) 0. Otherwise, there is a smallest m 1
such that am 6= 0 and f (z) = (z a)mg(z) has an order m zero. Now g(z) is analytic and
therefore continuous at z = a with g(a) 6= 0. Hence g(z) 6= 0 in an open disk about z = a
and the result follows.
7-2
Isolated Singularities
Definition: Let f (z) have an isolated singularity at z = a and Laurent expansion
f (z) =
n=
an(z a)n ,
0 < |z a| < R
1
1
1
=1+ +
+
+
z
2! z 2
3! z 3
Proof: See text: (i) is easy, (ii) is similar to previous theorem, (iii) is hard (Picard).
7-3
Laurent Series
Definition: A series of the form
n=
an(z a)n =
n=0
an(z a)n +
n=1
an(z a)n
convergent in some open annulus r < |z a| < R is called a Laurent series around z = a.
A Laurent series is the sum of two Taylor series, the first in positive powers of w = z a
and the second in positive powers of w = (z a)1, that is, in negative powers of z a. The
1
first series converges for |z a| < R and the second series converges for |z a| > r or za
< 1
r.
The Laurent series therefore converges in the intersection of these two regions given by the
open annulus or ring r < |z a| < R.
In practice, a Laurent series is usually obtained by combining suitable Taylor series. For
example, the Laurent series for f (z) = z 2e1/z about z = 0 is
z
1
1
1
1
1
1
2
1+ +
+
+
+
+
=
z
+
z
+
+
2
3
2
z
2!z
3!z
2!
3!z
4!z
The inner and outer radii of convergence r, R are defined by the Cauchy-Hadamard formulas
r = lim sup |an|
n
1/n
1
= lim sup |an|1/n
n
R
In practice, however, r and R are usually determined by applying the ratio or root test.
7-4
Laurent Theorem
Theorem 59 (Laurent Theorem)
(i) Suppose f (z) is analytic in an open annulus r < |z a| < R and let C be any circle with
center at a lying in this annulus. Then the Laurent series with coefficients
f (w)
1
dw,
an =
2i C (w a)n+1
I
n = 0, 1, 2, . . . ,
a is inside C
n=0
n=1
then there is a unique function f (z) analytic in r < |z a| < R with the Laurent series
n=
an(z a)n .
7-5
X
X
1
zn
1
1
1
=
n+1
n+1
(z 1)(z 2)
(z 2) (z 1)
2
n=0
n=0 z
since, by the geometric series, for |z| < 2 or 2z < 1
X
1
zn
1 1
=
=
n+1
(z 2)
2 1 2z
n=0 2
and for 1 < |z| or 1z < 1
X
1
1
1 1
=
=
n+1
(z 1)
z 1 1z
n=0 z
7-6
(z1)
1
1
2
+
+
1+
= (z 1) +
(z 1)
(z 1)
(z 1)2
1
1
+
+
= (z 1) + 1 +
2
(z 1)
(z 1)
2
2
+
+
+
2
(z 1)
(z 1)
X
3
= (z 1) + 1 +
n
n=1 (z 1)
1
z(z 1)
in each of the regions (i) 0 < |z| < 1, (ii) |z| > 1, (iii) 0 < |z 1| < 1 and (iv) |z 1| > 1.
7-7
Residues
Definition: If f (z) has an isolated singularity at the point z = a, so it is analytic in a punctured
neighbourhood of a, then the coefficient a1 of (z a)1 in the Laurent series for f (z) around
a is called the residue of f (z) at a and denoted Res(f ; a)
f (z) =
n=
an(z a)n
Res(a) = Res(f ; a) = a1
If f (z) is analytic at the point z = a then f (z) has a Taylor expansion around a and
Res(a) = Res(f ; a) = a1 = 0
Lemma 60 (Order m Pole)
dm1
1
m
[(z
a)
f (z)]
Res(f ; a) = lim
za (m 1)! dz m1
In particular, if f (z) has a simple pole (m = 1) at z = a then
Res(f ; a) = lim (z a)f (z)
za
+
+
m
2
(z a)
(z a)
(z a)
dm1
m
[(z
a)
f (z)] = (m 1)! a1 + m! a0(z a) + (m 1)! a1 as z a
dz m1
If f (z) has an essential singularity then the residue is obtained from the Laurent series of
f (z) in the punctured neighbourhood of a using Res(f ; a) = a1.
7-8
Examples of Residues
z2
Example: Find the residues of f (z) =
:
2
(1 z) (2 z)
Solution: There is a simple pole at z = 2 and a double pole at z = 1
z 2
z2
= lim
= 4
Res(2) = lim (z 2)
2
2
z2
z2
(1 z) (2 z)
(1 z)
2
2
d
z
d
z
Res(1) = lim
(z 1)2
= lim
z1 dz
z1 dz (2 z)
(1 z)2(2 z)
z2
2z
=2+1=3
+
= lim
z1 2 z
(2 z)2
Example: If f (z) = P (z)/Q(z) is rational and the polynomial Q(z) has a simple zero at z = a,
use lH
opitals rule to show that
P (a)
,
Res(f ; a) =
Q (a)
Q(a) 6= 0
Solution: Using lH
opitals rule
P (z) + (z a)P (z)
P (a)
(z a)P (z)
= lim
=
,
Res(f ; a) = lim
za
za
Q(z)
Q (z)
Q (a)
Q(a) 6= 0
7-9
More Residues
Example: Find the residue
sin z
;1
Res 2
z z
sin z
sin z
= sin 1
Res 2
;1 =
z z
2z 1 z=1
z
Solution: There is no pole, rather z = 0 is an essential singularity. The Laurent expansion is
1
1
1
z2
+
3
5
z
3!z
5!z
=z
1
1
+
3
3!z
5!z
Hence
1
1
2
Res z sin ; 0 = a1 =
7-10
Residues as Integrals
Exercise: If C is any closed contour and m Z, use deformation of contours and a change
of integration variable to evaluate the contour integral to show that
1
m
m
(z a) dz = Res (z a) ; a = m,1,
2i C
I
a inside C
The contour integral does not exist if m < 0 and the contour passes through a.
Example: If f (z) is analytic at z = a or has an isolated singularity at z = a, show that
1
f (z)dz
Res(f ; a) = lim
0 2i |za|=
X
1
an(z a)ndz
= lim
0 2i |za|= n=
X
I
an
(z a)n dz = a1 = Res(f ; a)
= lim
0 n= 2i |za|=
The term-by-term integration is justified because the Laurent series is absolutely and uniformly convergent in any closed sub-annulus within the open annulus of convergence. If z = a
is an isolated singularity, the limit 0 ensures that there are no other singularities inside
the circle |z a| = . If f (z) is analytic at z = a, the integral vanishes for sufficiently small
by Cauchys theorem.
7-11
Meromorphic Functions
Definition: A function f (z) is meromorphic in the domain D if at every point of D it is either
analytic or has a pole.
Clearly, a meromorphic function has zeros and (isolated) poles but no other singularities.
We have seen that an entire (analytic everywhere) function can be expanded into an infinite
Taylor series. In this sense an entire function is like an infinite polynomial it has zeros but
no poles. Typical entire functions are cos z and sin z. A meromorphic function, such as cot z,
can always be written as the ratio of two entire functions its zeros are the zeros of the
numerator and its poles are the zeros of the denominator. Thus
cos z
cot z =
sin z
In this sense a meromorhic function is a generalization of a rational function allowing for
infinite polynomials in the numerator and denominator and thus an infinite number of zeros
and poles.
Exercise: Find all of the zeros and poles of the meromorphic functions cot z and tan z.
8-1
Residue Theorem
Theorem 61 (Residue Theorem)
If is a simple closed contour and f (z) is analytic inside and on except at the isolated
points z1, z2, . . . , zm inside then
I
f (z)dz = 2i
m
X
Res(f ; zk )
k=1
Proof: Let Ck be small circles about each isolated singularity z = zk so that on Ck we have
the Laurent expansion
f (z) =
(k)
n=
an (z zk )n
f (z)dz =
=
m I
X
f (z)dz =
k=1 Ck
m
X
X
k=1 n=
m
X
(k)
an
Ck
k=1 Ck n=
n
(k)
an (z zk )ndz
(z zk ) dz = 2i
Res(f ; zk )
= 2i
m I
X
m
X
k=1
(k)
a1
k=1
The term-by-term integration is justified by the uniform convergence of the Laurent expansions.
8-2
Z 2
0
2
dt
dt
1
=2
I=
0 2 cos t
0 2 cos t
1) and
Solution: Let z = eit so that dz = ieitdt = izdt and dt = dz/iz. Then cos t = 1
2 (z + z
dz
2
dz
=
2I =
i |z|=1 z 2 4z + 1
|z|=1 2 1 (z + z 1) iz
2
The integrand has simple poles at z = 2 3 but only z lies inside the unit circle with
residue
1
1
1
(z z)
= lim
=
zz (z z )
zz (z z)(z z )
2 3
+
+
Res(z) = lim
Hence
1
2
2
2i
2I =
=
i
2 3
3
I=
3
8-3
Improper Integrals
Definition: If f (x) is continuous over [a, ), its improper integral is defined by
Z
a
f (x)dx := lim
Z R
R a
f (x)dx
provided this limit exists. If f (x) is continuous on (, ) we define the double improper
integral
Z
f (x)dx := lim
Z R
R 0
f (x)dx + lim
Z 0
R R
f (x)dx
f (x)dx :=
lim
R,R
Z R
f (x)dx
PV
f (x)dx := lim
Z R
R R
f (x)dx
provided this limit exists. If the double improper integral exists it must equal its principal
value, but the principal value integral can exist when the double integral does not exist.
8-4
Z
0
e2xdx :
2x R
e
e2xdx = lim
e2xdx = lim
R 0
R
2 0
0
e2R
1
1
= lim
+
=2
R
2
2
Z R
PV
x dx = lim
Z R
R R
Z
x dx = lim
xdx :
2 R
x
2 R
= lim 0 = 0
R
8-5
then
|z| large
Z
f (z)dz = 0
lim
R CR
CR
z1
8-6
PV
f (x)dx =
lim
Z R
R R
m
X
f (x)dx = lim
R R
f (z)dz
Res(f ; zk )
= 2i
k=1
by closing the contour in the upper half plane and summing over residues.
CR
z1
8-7
Solution: The integrand f (z) = P (z)/Q(z) is rational with degree P = 2, degree Q = 4 and
double poles at z = i. The previous theorem therefore applies. The residue at z = i is
d
2z 2
2
1
2z
2i
d
z2
2
= lim
= 2+ 3=
Res(f ; i) = lim [(z i) f (z)] = lim
zi dz (z + i)2
zi (z + i)2
zi dz
(z + i)3
4i
8i
4i
Therefore
and so
lim
R R
f (z)dz = 2i
f (z)dz = lim
hZ R
=
4i
2
f (x)dx +
The previous
lemma
applies since, for |z|
1 |z|2 =
|z 2 + 1| |z|2 1 = |z|2 1 |z|2 2
|f (z)| =
CR
f (z)dz =
f (x)dx =
2
2
4
|z|
|z|
z2
=
2
1
2
2
2
2
2
2
(z + 1)
|z + 1|
|z|
( 2 |z| )
Note that in these cases the double improper integral exists so that
x2 dx
x2 dx
x2 dx
=
=
2
=
PV
2
(x2 + 1)2
0 (x2 + 1)2
(x2 + 1)2
8-8
It is often useful to expand rational functions into a finite number of partial fractions. For
example
1
1
2z
=
1 z2
1z 1+z
Similarly, meromorphic functions can be expanded into an infinite number of partial fractions.
Exercise: Establish the partial fraction expansions of the following meromorphic functions
X
1
1
cot z =
+ 2z
2 n2
z
z
n=1
X
1
(1)n
cosec z =
+ 2z
2
2
z
n=1 z n
X
(1)n1(2n 1)
sec z =
2n1 )2 z 2
(
n=1
2
X
1
tan z = 2z
tanh z = 2z
n=1 (
2n1 )2 z 2
2
2 + 2n1 )2
(z
n=1
2
X
1
1
+ 2z
coth z =
2 + n2
z
z
n=1
8-9
on circles CN : |z| = RN as N
where the circles do not pass through any poles and M is independent of N . Then
1
1
+
bn
f (z) = f (0) +
z an
an
n=1
a3
a2
a1
0
a4
CN
8-10
f (z)
with residues
zw
bn
f (z)
=
zan
zw
an w
f (z)
Res(F ; w) = lim (z w)
= f (w)
zw
zw
So using the residue theorem and subtracting
lim (z an)
Res(F ; an) =
X
1
f (z)dz
bn
= f (w) +
2i CN z w
a w
poles an in CN n
I
X
1
f (z)dz
bn
= f (0) +
2i CN
z
an
poles a in C
I
f (w)f (0) +
poles an in CN
where
bn
bn
an w an
f (z)dz
w
=
2i CN z(z w)
I
2RN M
f (z)dz
0 as N
RN (RN |w|)
CN z(z w)
8-11
n
n
n6=0
1
has simple poles at z = an = n =
z
z cos z sin z
(z n) z cos z sin z
bn = lim (z n)
= 1
= lim
zn
zn
z sin z
sin z
z
By lH
opital, f (z) has a removable singularity at z = 0
1
lim cot z
z0
z
z cos z sin z
= lim
z0
z sin z
2 sin z
= lim
z0 sin z/z + cos z
2z sin z
= lim
z0 sin z + z cos z
= 0 = f (0)
The circles CN of the previous theorem can be replaced with the squares N with vertices at
z = (N + 1
2 )(1 i) on which we can show
| cot z| M = coth(/2),
independent of N
8-12
Bound on Squares N
1 )(1 i).
Let N be the square with vertices at z = (N + 2
Then on N
| cot z| M = coth(/2),
independent of N
1 respectively
(i) If z = x + iy and y > 1
or
y
<
2
2
2iz
2y + 1
e
+
1
e
1
+
e
| cot z| = 2iz
= coth(/2)
2y
e
1
|e
1|
1e
2iz
2y
2y + 1
e
+
1
1
+
e
e
1
+
e
=
| cot z| = 2iz
2y
2y
e
1
|e
1|
1e
1 e
1 ) and 1 y 1 then
(ii) If x = (N + 2
2
2
N1 N
1 0
N+1
N
The Contour N :
(showing poles of
cot z)
8-13
X
X
1
1
1
1
1
cot z = +
+
= + 2z
2
2
z
z
n
n
z
n=1 z n
n6=0
z
n
n
n=1 z n
n=N z n
X
X
2z
1
1
1
1
+ lim
+
=
= +
2
2
N
z
z
n
z
+
n
z
n=1
n=1 z n
This is allowed because the double series is absolutely (and uniformly) convergent in |z| R
by the Weierstrass M -test
1
2R
R
1
z
+ =
= Mn ,
2
zn
n
n(n z)
n(n R)
n
X 1
X
1
since
=2
2
2
n
n=1 n
n6=0
n 2R
8-14
z2
1 2
sin z = z
n
n=1
Solution: Since the partial fraction expansion converges absolutely and uniformly we can
integrate term-by-term
X
2z
1
cot z =
2
2
z
n=1 z n
X
Y
sin z
z2
z2
Log
Log(1 2 ) = Log
=
(1 2 )
z
n
n
n=1
n=1
where the constant of integration vanishes. The result follows by taking exponentials.
Y
2n
2n
2 2 4 4 6 6
=
=
2
1 3 3 5 5 7
n=1 2n 1 2n + 1
4z 2
cos z =
1
(2n 1)2
n=1
8-15
Z 2
0
2
dt
dt
1
=2
I=
0 2 cos t
0 2 cos t
1) and
Solution: Let z = eit so that dz = ieitdt = izdt and dt = dz/iz. Then cos t = 1
2 (z + z
dz
2
dz
=
2I =
i |z|=1 z 2 4z + 1
|z|=1 2 1 (z + z 1) iz
2
The integrand has simple poles at z = 2 3 but only z lies inside the unit circle with
residue
1
1
1
(z z)
= lim
=
zz (z z )
zz (z z)(z z )
2 3
+
+
Res(z) = lim
Hence
1
2
2
2i
2I =
=
i
2 3
3
I=
3
9-1
Z 2
d
:
0 3 2 cos
This integral was evaluated previously using the
Cauchy integral formula.
Example: Evaluate the integral I =
z+
dz
d
= i
I =
|z|=1 z(3 z z 1)
0 3 2 cos
I
I
dz
dz
= i
= i
2
|z|=1 (z z+)(z z )
|z|=1 z 3z + 1
I
f (z) = (z z+)1
which is analytic inside |z| = 1, we have by the Cauchy integral formula
2
2
f (z)dz
= (i)(2i)f (z ) =
=
I = i
(z
z
)
z
z
|z|=1
5
+
I
We obtain the same result using residue calculus and summing over the poles inside |z| = 1
f (z)
; zk
I = (i)(2i)
Res
(z
z
)
k
X
= (i)(2i)f (z ) =
5
9-2
Z 2
0
dt
1 2p cos t + p2
1) and
Solution: Let z = eit so that dz = ieitdt = izdt and dt = dz/iz. Then cos t = 1
(z
+
z
2
i
dz
=
I =
1
2
p
|z|=1 (z p)(z 1 )
|z|=1 iz[1 + p p(z + )]
z
p
There are now two situations to consider:
dz
(i) |p| > 1: In this case there is a simple pole inside |z| = 1 at z = 1/p. So by the residue
theorem
1
2
i
2 1
=
I = 2i Res
=
p
p
p 1p p
p2 1
(ii) |p| < 1: In this case there is a simple pole at z = p so
I =
i
2
2 1
=
2i Res(p) =
p
p p 1p
1 p2
(iii) If |p| = 1, that is, p = 1, the integral does not exist since there is a pole on the contour
of integration.
2
dt
=
1 2p cos t + p2
|1 p2|
when
p 6= 1
9-3
Uniformity on an Arc
Definition: (Uniformity on an Arc) If along a circular arc of radius R, |f (z)| MR where
MR does not depend on (the polar angle) , and MR 0 as R (or R 0) we say that
f (z) tends uniformly to zero on CR = {z : |z| = R} as R (or R 0).
Example: Consider the function
z
f (z) = 2
z +1
on
CR = {z : |z| = R}
We deduce that
R
, R>1
2
R 1
|f (z)|
, R<1
2
1R
It follows that f (z) tends uniformly to zero when either R or R 0.
In general any rational function whose denominator is of higher degree than the numerator
tends uniformly to zero as R .
max
z=Rei
0 0+
|f (z)| 0
as R
9-4
Limiting Contours I
Theorem 65 (Limiting Contours I) If on CR, zf (z) tends uniformly to zero as R then
lim
R CR
f (z) dz = 0
0 0 + }
Proof: We have
|zf (z)| = R|f (z)| MR
where MR is independent of and MR 0 as R . It follows that
Z
Z
Z
MR
|f (z)| |dz|
f (z) dz
|dz| = MR 0
0
CR
R CR
CR
as
CR
|dz| =
Z +
0
0
R d = R
9-5
eikz f (z) dz = 0
eikz f (z) dz = 0
ekz f (z) dz = 0
ekz f (z) dz = 0
R CR
Z
lim
R CR
Z
lim
R CR
Z
lim
R CR
Proof: We prove the first case. The other cases are similar. Note that on CR
|dz| = R d
and
|f (z)| MR
It follows that
Z
0
Z
Z
1 kR sin
ikz
ikz
e f (z) dz
e
d
|e ||f (z)||dz| RMR
CR
CR
0
Z
Z /2
RMR
ekR sin d = 2RMR
ekR sin d
0
0
Z /2
MR
2kR/
(1 ekR) 0
e
d =
2RMR
k
0
as
)
=
sin(
2
2
9-6
r0 Cr
f (z) dz = 0
Cr
Z
M
r
f (z) dz
Cr
|dz| = Mr
9-7
Limiting Contours IV
Theorem 68 (Limiting Contours IV) If f (z) has a simple pole at z = a with residue Res(a)
and if Cr is a circular arc of radius r and centre a subtending an angle at z = a then
lim
r0 Cr
f (z) dz = i Res(a)
z = a + rei
where
0 0 +
9-8
Fourier Example
Exercise: Evaluate the Fourier integral
I=
a > 0,
kR
Solution: The function f (z) = (a2 + z 2)1 has simple poles at z = ia.
(i) k 0: Close the contour in the upper half plane. Provided R > a, we have
Z
ikz
e f (z) dz
CR
CR
= R(R2 a2)1 0
CR
|dz|
as R
since |eikz | = |eik(x+iy)| = eky 1 in the upper half plane. It follows that
ikz
e
ikx 2
2 1
ikz
e (a + x ) dx = 2i Res(e f (z); ia) = 2i
z + ia
ka
=
e
a
z=ia
ikz
ka
e
ikx 2
2 1
ikz
=
e (a + x ) dx = 2i Res(e f (z); ia) = 2i
e
z ia z=ia
a
|k|a
e
e (a + x ) dx =
a
2 1
Z
|k|a
cos(kx)
e
,
dx
=
2
2
a
a + x
Z
sin(kx)
dx = 0
2
2
a + x
9-9
Indented Contours I
Example: Consider the contour integrals
I =
Z
sin x
dx =
Im
ix
e
eiz
dz
I =
C z
I
dx,
eiz
dz =
C z
R eix
eiz
eiz
dx +
dz +
dx +
dz
x
Cr z
CR z
R x
r
Z r ix
e
eiz
lim
dz = 0
R CR z
Z
By Jordans Lemma
CR
Cr
-r
where the negative sign comes from the fact that Cr is traversed clockwise. It follows that
lim
r0, R
(Z
r eix
R x
dx +
Z R ix
e
r
dx
Z ix
e
dx = i
where the limit defines the Cauchy principal value. Equating real and imaginary parts gives
Z
cos x
dx = 0,
Z
sin x
dx =
Z
sin x
dx =
x
x
x
The integral I is a conditionally (not absolutely) convergent improper integral.
9-10
xa>0
then
Z
f (x)
dx
p
x
a
converges
for p > 0
+
p
dx
lim
p
p
p+1
p
p+1
R
R x
R a
x
a
a x
a x
a
Z
dx
F (x) =
So
Z
sin x
0
f (x) dx =
xp+1
Z
sin x
1
dx = 2
dx is a convergent improper integral.
x
sin x dx = cos x,
Z
sin x
dx =
Z a
|F (x)| 1,
Z
sin x
xa>0
dx
x
x
0
a
The first integral exists as a Riemann integral and the second is a convergent improper
integral by the theorem.
0
9-11
Indented Contours II
Example: Evaluate the contour integral
I =
eiz dz
C 2 4z 2
where C is the standard upper-half-plane contour indented by small semi-circles around the
singularities at z = /2.
0 =
+
Res
i
Res
2
2
2 4x2
Equating real parts gives
Z
cos x
dx =
2
2
4x
i
i
eix dx
=
4 4
2 4x2
Z
1
cos x
dx =
2
2
4x
2
eix dx
The Cauchy principal value of
must be retained since otherwise the integral
2
2
4x
fails to exist due to divergences at x = /2.
Z
The technique of indented contours only works for simple poles on the real axis! For higher
order poles the Cauchy principal values do not exist.
9-12
f (n) =
n=
poles zj of f (z)
X
n
(1) f (n) =
Res( csc z f (z); zj )
n=
poles zj of f (z)
2n + 1
f
2
n=
2n + 1
n
(1) f
2
n=
N1 N
poles zj of f (z)
poles zj of f (z)
where the residues are summed only over the poles zj of f (z).
The methods apply provided f (z) is analytic except at isolated
poles (zj
/ Z) and decays sufficiently rapidly as |z| , for
example, |f (z)| K/|z|2.
1 0
N+1
N
The Contour N :
(showing poles of
cot z)
9-13
(z n)
= lim
cos z f (z) = f (n)
zn sin z
independent of N
1 respectively
or
y
<
2iz
2y + 1
e
+
1
e
1
+
e
| cot z| = 2iz
= coth(/2)
2y
e
1
|e
1|
1e
2iz
2y + 1
2y
e
+
1
e
1
+
e
1
+
e
| cot z| = 2iz
=
2y
2y
e
1
|e
1|
1e
1 e
1 ) and 1 y 1 then
(ii) If x = (N + 2
2
2
cot
z
f
(z)dz
(8N + 4) 0 as N
2
N
N
9-14
N N
cot z f (z)dz =
N
X
= lim
f (n) +
n=N
poles zj of f (z)
f (n) +
n=
poles zj of f (z)
9-15
1
= coth a :
Example: Show that for a > 0,
2
2
a
n= n + a
1
The series is absolutely convergent for a > 0. Let f (z) = 2
with simple poles at z = ai.
2
z +a
The residues of
cot z
F (z) = cot z f (z) = 2
z + a2
at z = ai are then
(z ai) cot z
cot ai
(z + ai) cot z
=
= coth a
zai (z + ai)(z ai)
2ai
2a
1
coth a
=
Res(
cot
z
f
(z);
z
)
=
j
2
2
a
n= n + a
poles zj of f (z)
X
1
Example: Sum the series
2
n=1 n
1
:
2
2
n=1 n + a
The second series is absolutely and uniformly convergent for 0 a R < by the Weierstrass
M -test. Using uniform convergence, continuity and lH
opitals rule
X
2
X
X
1
1
1
1
a
coth
a
1
1
=
lim
=
lim
2 = lim
=
2
2 + a2
2
a0
a0 2 n= n2 + a2
a0
n
n
a
2a
6
n=1
n=1
9-16
10
Z 2
Z 2
1
1
|f (a)| =
f (a + reit)dt
|f (a + reit)|dt
2 0
2 0
()
Proceed by contradiction. Suppose maximum is |f (a)| for a inside so that, on some circle
C about a and inside , |f (a + reit)| |f (a)| and (if f (z) is not constant)
|f (a + reit )| < |f (a)|,
for some t
Then, by continuity, this holds in some interval t1 < t < t2 and so the mean value satisfies
Z
1 2
|f (a + reit)|dt < |f (a)|
2 0
which contradicts (). Hence either f (z) is constant or the maximum occurs on .
10-1
|z| = 1
1, 1
r=2
|z| = 1
1
|z| = 2
1
2
(Hint: Find
Exercise: Let f (z) = 1/z. Find the point or points where |f (z)| takes its maximum and
minimum values on the annulus 1 |z| 2.
10-2
Liouvilles Theorem
Lemma 73 (Cauchys Inequality)
If f (z) is analytic inside and on the circle C given by |z a| = r and |f (z)| M on C, then
|f
(n)
M n!
(a)| n ,
r
n = 0, 1, 2, . . .
Proof: On C we have z = a + reit and z (t) = ireit. So using Cauchys integral formula
n! 2
f (z)
f (z)
dt
dz
z
(t)
2i C (z a)n+1
2 0 (z a)n+1
I
n!
|f (n)(a)| =
M n!
n! 2 f (a + reit) ireit
dt
=
2 0 rn+1e(n+1)it
rn
Z
Note that for this theorem to apply f (z) must be bounded as |z| .
10-3
10-4
a0, a1, . . . , an C
|z|
f (z) is bounded on C. So, by Liouvilles theorem, f (z) and Pn(z) are constant. But this is a
contradiction. We conclude that Pn(z) = 0 must have at least one root.
(ii) Suppose z = z1 is a root of Pn(z) = 0. Then we apply polynomial division
Pn(z)Pn(z1) = (anz n + an1z n1 + + a1z + a0) (anz1n + an1z1n1 + + a1z1 + a0)
= an(z n z1n)+an1(z n1 z1n1) + + a1(z z1)
= (z z1)Qn1(z) = Pn(z)
where Qn1(z) is a polynomial of degree n 1. By iterating, Pn(z) has exactly n roots.
Note that the n roots need not be distinct. For example, the polynomial P2(z) = (z 1)2
has a double root at z = 1.
10-5
Factorization of Polynomials
Corollary 76 (Factorization of Polynomials)
Every polynomial of degree n 1
a
zj = n1 ,
an
j=1
i<j
zi zj =
an2
,
an
a
z1z2 . . . zn = (1)n 0
an
...
10-6
1 + z + z 2 + z 3 = 0.
(z + 1)(z i)(z + i) = 0
(z + 1)(z 2 + 1) = 0
z3 + z2 + z + 1 = 0
Corollary 77 (Roots in Z)
Suppose a0, a1, . . . , an Z and Pn(z) factors as
where z1 Z and b0, b1, . . . , bn1 Z. Then the integer root z1 must be an integer factor of
a0 = z1b0.
This Corollary may yield integer roots of polynomials with integer coefficients:
P3(z) = (z 2)(z 2 + z + 1)
3
1
z = 2, i
2
2
by polynomial division
10-7
n1
+ + a2z 2
anz n
1 + an1z 1
1 + a1z 1 + a0 = Pn (z 1) = 0
Example: Expand into linear and quadratic factors the degree six polynomial
P6(z) = z 6 z 5 3z 4 + z 3 + 3z 2 z 2 = 0
Solution:
Try z = 1, 2 as roots:
P6(1) = P6(2) = 0 (z + 1), (z 2) are factors
1i 3
2
z =
= ei/3 z = ei/6
2
i
i
i
6
i
e 6 )(z
i
6
+e
)(z
e
P6(z) = (z + 1)(z 2)(z +
10-8
Analytic Continuation
Theorem 79 (Identity Theorem)
If f1(z) and f2(z) are analytic in an open connected region R and f1(z) = f2(z) on some
open neighbourhood of z, then f1(z) = f2(z) on R.
f (z),
1
f (z) =
f (z),
2
z R1
z R2
R1
R2
2n
n
has a singularity at each root of unity satisfying z 2 = 1.
n=0
10-9
If the function f1(z) = Log z, analytic in |z 1| < 1, is extended using Taylor series to the
negative real axis by analytic continuation clockwise and anti-clockwise around the branch
point at z = 0, then the two analytic continuations differ by 2i.
Exercise: Verify this using R1 : |z 1| < 1, R2 : |z i| < 1, R3 : |z + 1| < 1 for the clockwise
: |z 1| < 1, R : |z + i| < 1, R : |z + 1| < 1 for the anti-clockwise path.
path and R1
2
3
Definition: If analytic continuation of the function f (z) full circle around a point z = z0
brings you back to a different branch, then z = z0 is a branch point of the multi-valued
function f (z). This is called non-trivial monodromy. Note that analytic continuation around
a point on a branch cut always produces trivial monodromy.
10-10
Conformal Maps
Definition: A conformal map f : U V is a function which preserves angles. More specifically,
f is conformal at a point if the angle between any two C 1 curves through the point is preserved
under the mapping.
Theorem 81 (Analytic Maps are Conformal)
A function f (z) analytic in an open
neighbourhood of a with f (a) 6= 0 is conformal at z = a. Since, by continuity, f (z) 6= 0 in
an open neighbourhood of a, it follows that f (z) is conformal in a neighbourhood of a.
Proof: If : [0, 1] C is a C 1 curve and f ((t)) its image then the tangent slopes are
arg( (t)),
(t) 6= 0;
arg(f ((t)) ),
The tangents to the curves at z = a are 1 (t1) and 2 (t2) and the angle between them is
arg(2 (t2)) arg(1 (t1)),
1 (t1) 6= 0,
2 (t2) 6= 0
f (a)2 (t2)
2 (t2)
f (2(t2))2 (t2)
f (2 (t2))
=
=
=
f (1 (t1))
f (1(t1))1 (t1)
f (a)1 (t1)
1(t1)
The result follows by taking the argument on the left and right since
z
arg 2 = arg(z2) arg(z1)
z1
It follows that the group of conformal maps in two dimensions is infinite dimensional. By
contrast, this group is finite dimensional in three or higher dimensions.
11-1
maps
open
11-2
M
obius Transformations
Definition: A M
obius transformation (linear fractional or bilinear transformation) is any non of the form
constant function on C
w = f (z) =
az + b
,
cz + d
a, b, c, d C
a
d
= , f () =
(c 6= 0);
f () = (c = 0)
f
c
c
A M
obius transformation is conformal at every point except at its pole z = d/c since
f (z) =
ad 6= bc,
ad bc
6= 0
2
(cz + d)
M
obius transformations form a group under compositions with the identity I(z) = z.
a b
A=
,
c d
a b
=
6 0
det A =
c d
More precisely, w = f (z) is associated with the set A with 6= 0. Show that the matrix
associated with the composition f1(f2(z)) of M
obius transformations is
A1 A2 =
a1 b1
c1 d1
a2 b2
,
c2 d2
Also show that the inverse function f 1 satisfying f (f 1 (z)) = f 1(f (z)) = I(z) is given by
z=f
dw b
(w) =
,
cw + a
A1 =
d b
c a
11-3
Elementary M
obius Maps
M
obius transformations include translations, rescalings, rotations and inversions.
1. Translation:
w = z + z0
This is translation in the Argand plane by the vector (x0, y0) R2.
2. Rescaling:
w = z
w = ei0 z
w=
The rotation is
1
z
This is point-by-point inversion of the plane through the point at the origin z = 0 (z 7
followed by reflection in the x-axis (complex conjugation).
1
)
z
11-4
M
obius Compositions
Theorem 83 (M
obius Compositions)
If f (z) is a M
obius transformation then:
(i) f (z) is the composition of a finite sequence of elementry maps in the form of translations,
rescalings, rotations and inversions.
one-to-one onto itself.
(ii) f (z) maps C
(iii) f (z) maps the class of circles and lines to itself. Note that a circle is uniquely determined
by 3 distinct noncollinear points. A line is uniquely determined by 2 distinct points to which
we can add the point at on the Riemann sphere. A line is just a (great) circle which passes
through .
Proof: (i) We have
az + c
= f4(f3(f2(f1(z))))
f (z) =
cz + d
where
d
f1(z) = z + ,
c
(ii)-(iii) See textbook.
f2(z) =
1
,
z
f3(z) =
ad bc
z,
2
c
f4(z) = z +
a
c
11-5
Crossratio
z2 z3
= cross-ratio
z2 z1
w1
w2
z z
1
z2
w3
z3
11-6
Constructing M
obius Transformations
Example: Find a M
obius transformation that maps the unit disk |z| < 1 onto the right halfplane Re w > 0 such that f (1) = and f (1) = 0.
Solution: A M
obius transformation takes the form
w = f (z) =
az + b
cz + d
Since f (1) = and f (1) = 0, we see that d = c to ensure the divergence at z = 1 and
b = a to ensure the vanishing at z = 1. Hence
1+z
a1+z
=
c 1z
1z
To ensure f (i) lie on the imaginary w-axis we find
a
f (i) = i,
= R
c
From the properties of M
obius transformations, we conclude that the unit circle is mapped
into the line Re w = 0. The image f (0) of the point z = 0 at the centre of the disk
a
f (0) = = > 0
c
for this image to be in the right half-plane Re w > 0. Choosing the positive rescaling factor
= 1 gives the M
obius transformation
f (z) =
1+z
1z
The answer is not unique since a rescaling by 6= 1 leaves the right half plane invariant.
w = f (z) =
11-7
Three-Point Uniqueness
The general M
obius transformation
az + b
cz + d
appears to involve four complex parameters a, b, c, d. However, since ad 6= bc, either a 6= 0 or
c 6= 0 or both a and c are non-zero. We can therefore express the transformation with three
unkown coefficients
w = f (z) =
z + ab
w = f (z) = c
d
z
+
a
a
or
az+b
c
w = f (z) = c
z + dc
z2 z3
w w1 w2 w3
=
z2 z1
w w3 w2 w1
If the line or circles are oriented by the order of z1, z2, z3 and w1, w2, w3 respectively, then the
M
obius transformation maps the region to the left (left-region) of z1, z2, z3 onto the region
to the left (left-region) of w1, w2, w3.
Exercise: Show, using cross-ratios, that the unique M
obius transformation w = f (z) such
that f (1) = 0, f (0) = a, f (1) = is
f (z) = a
1+z
1z
11-8
Circles to Circles
Example: Find a M
obius transformation that maps the circle through the distinct noncollinear
points z1, z2, z3 to the circle through the distinct noncollinear points w1, w2, w3.
Solution: Suppose the M
obius transformation w = T (z) is such that T (z1) = 0, T (z2) = 1
and T (z3) = , that is, T (z) maps the circle onto the real axis Im w = 0. Following the
previous example, we find
T (z) =
z z1
z z3
z2 z3
= (z, z1; z2, z3) = cross-ratio
z2 z1
S(w) = T (z)
because
f (z1) = S 1(T (z1)) = S 1(0) = w1
f (z2) = S 1(T (z2)) = S 1(1) = w2
f (z3) = S 1(T (z3)) = S 1() = w3
The order of the points is important. Specifically, if the circles are oriented such that the
points are traversed in the given order, then the left-region is mapped onto the left region
where each left-region is an interior or exterior of the associated circle.
11-9
y = G(t),
t [a, b]
F, G analytic
maps the real axis of the w-plane onto the curve C in the z-plane.
Proof: See textbook. For closed curves, this often gives map between bounded regions.
Exercise: Find a conformal transformation that maps the ellipse
y2
x2
+ 2,
a, b > 0
a2
b
onto the interval [0, 2] along the real axis of the w-plane and the interior of the ellipse onto
a suitable rectangle in the strip 0 < Re w < 2 within the upper-half w-plane.
11-10
or
dh
= normal derivative to = 0
dn
H(x, y) = h0 R
or
dH
= normal derivative to C = 0
dN
then along C
Proof: (i) Let w = f (z) = u + iv and let k = k(u, v) be the harmonic conjugate of h = h(u, v).
Then g(w) = h + ik is analytic. So
H(x, y) = h(u(x, y), v(x, y)) = Re g(w) = Re g(f (z)) = harmonic function of x, y
because the composite function g(f (z)) is analytic.
Heat Flow:
In heat flow, Laplaces equation governs the temperature distribution (x, y) and the curves
= constant are isotherms. Typically, the temperature is fixed on parts of the boundary
= D. The rest of the boundary is assumed to be insulating so that the normal derivative
to vanishes
= 0.
n
Electrostatics:
In electrostatics, (x, y) is the electric potential with electric field E = and E = 0.
The curves = constant are equipotentials. Typically, one specifies either the potential or
the normal component of E on the boundary = D.
Fluid Flow:
In fluid flow, (x, y) is the stream function and the curves = constant are streamlines. For
flow around a nonporous body, the perimeter must be part of a streamline.
11-12
Cylindrical Capacitor
Example: Find the electrostatic
potential (z) between two long cylindrical conductors such
3 = 3 and = 1 on |z| = 1.
that = 0 on z 10
10
Solution: Since Log w = Log |w| + i Arg(w), w 6= 0 is analytic in an annulus (with a cut),
Log |w| is harmonic. So the radially symmetric solution we seek for two concentric cylinders
of radii r, R with 0 < r < R is
Log(|w|/r)
(w) =
,
r < |w| < R
Log(R/r)
We need to find a conformal mapping to relate the two geometries. The required M
obius
transformation and its inverse are
w = f (z) =
1
z3
z = f 1(w) =
1
3w + 3
z3
w+1
It is verified that the inner and outer circles in the z-plane are mapped onto concentric circles
in the w-plane
3
1
i
|w| = f
(1 + e ) = = r,
10
9
The solution in the w-plane is thus
|w| = |f (ei )| =
Log 9|w|
(w) =
,
Log 3
1
= R,
3
[0, 2]
1
1
< |w| <
9
3
z1
3
z3
11-13
Joukowsky Airfoil
In 1908, the mathematician Joukowsky considered the flow around an off-centre cylinder
Miraculously, this cylinder is mapped onto the Joukowsky airfoil under the conformal
Joukowsky mapping
1
w = J(z) = z +
z
It is therefore possible to obtain the airflow around the Joukowsky airfoil by studying the
airflow around a cylinder. This technique is of major importance in aerodynamics!
11-14
Gamma Function
Definition: The gamma function is defined by
(p) =
Z
0
xp1ex dx,
Re p > 0
From this integral expression, (p) has no branch points and is analytic for Re p > 0.
d x
p1
x
p1
(e ) dx = x
e
+ (p 1)
xp2exdx
(p) =
x
0
dx
0
0
= (p 1) (p 1), provided Re p > 1
Z
Z
0
1/2 x
dx = 2
Z
0
y 2
dy = ,
x = y2
(p) =
(p + n + 1)
,
p(p + 1)(p + 2) (p + n)
p 6= 0, 1, 2, 3, . . .
12-1
-4
-2
-2
-4
12-2
Beta Function
Definition: A related function is the beta function B(r, s) defined by
B(r, s) =
Z 1
0
ur1(1 u)s1 du
Z
0
xr1ex dx
Z
0
y s1ey dy
as a double integral over the first quadrant in the x-y plane. Substituting x + y = u we find
(r)(s) =
=
0
Z
0
eu
u
Z u
xr1(u x)s1 dx du
0
r+s1 u
du
Z 1
0
tr1(1 t)s1 dt
= (r + s)B(r, s)
z p-1 =x p-1
(p)(1 p) = B(p, 1 p) =
Z 1
up1(1 u)p du
0
Or, after substituting u = x(1 + x)1,
Z
(p)(1 p) =
xp1(1 + x)1 dx
0
branch cut
zp-1 =x p-1 e
2ip
Reflection Formula
Using residues and limiting contours theorems I/III (Uniformity on an Arc) we find
I
Z
0
xp1(1 + x)1 dx +
= (e2ip 1)
After rearranging we find
Z
0
p1
(1 + x)
Z 0
xp1e2ip(1 + x)1 dx
xp1(1 + x)1 dx
dx =
sin p
,
(p)(1 p) =
sin p
0 < Re p < 1
Although we derived the reflection formula for 0 < Re p < 1, it extends straightforwardly
to all p
/ Z using the recursion relation for the function.
Using the fact that / sin(z) has simple poles at z = n, n = 0, 1, 2 . . . with residues
(1)n , the reflection formula shows that (p) has simple poles at p = n, n = 0, 1, 2, . . . with
residues
(1)n
(1)n
=
(1 + n)
n!
12-4
Special Values
Z
0
xp1ex dx,
Re p > 0
(1) = 1
(1/2) =
Recurrence Relation
(p) = (p 1)(p 1)
Reflection Formula
(p)(1 p) = / sin(p)
Singularities
Simple poles at p = 0, 1, 2,
Residues
n = 0, 1, 2,
12-5
Zeta Function
X
1
(p) =
,
p
n=1 n
Re p > 1
X
(1)n1
(p) =
,
p
n
n=1
Re p > 0,
p 6= 1
with (p) = (p) for Re p > 1. The eta function (p) is absolutely convergent for Re p > 1
and conditionally convergent for 0 < Re p < 1 by a convergence test for Dirichlet series
n
X
ak bounded for large n
k=1
X an
< ,
p
n
n=1
an C,
Re p > 0;
This formula is not obvious and requires an integral representation of the eta function
xp1
1
(p) =
dx,
Re p > 0
x
(p) 0 e + 1
It follows that (p) has no branch points and is an analytic function in the complex p-plane
except for a simple pole at p = 1 with residue equal to unity.
12-6
Riemann/Euler Formulas
(p) =
n=1
Special values
1
,
=
p
n prime 1 n
Y
Re p > 1
(1) = 1/12
(0) = 1/2
(2) = 2/6
(4) = 4/90
Riemann Relation
Singularity
Zeros
Trivial zeros at p = 2, 4, 6, 8, . . .
Non-trivial zeros on the line Re p = 1
2
12-7
0.02
6
0.01
4
-10
-8
-6
-4
-2
2
-0.01
-0.02
-1.0
0.5
-0.5
1.0
1.5
2.0
2.5
-2
-0.03
-4
-0.04
-6
-0.05
12-8
3.0
2.5
2.0
1.5
1.0
0.5
-40
-20
20
40