Complex analysis
1.1
Analytic functions
iu0y :
Since the two versions of f 0 (z0 ) are equal if the limit exists, an analytic function
always fullls
u0x
u0y
= vy0
=
vx0 :
vx0 :
It is possible to prove that the implication also goes in the other direction,
if we add the requirement that all rst order partial derivatives are continuous:
(u0x = vy0 , u0y =
vx0 and u0x ; vy0 ; u0y ; vx0 are all continuous at z0 ) ) f (z) = u(z)+iv(z) analytic:
1.2
Complex integration
where the two last integrals are real contour integrals. Here C is a curve given
in the complex plane. If we have a parametrization of C as z(t) = x(t) + iy(t)
where t : a ! b; we can also dene the curve integral as
Z
Z b
f (z)dz =
(u(x(t) + iy(t)) + iv(x(t) + iy(t)))(x0 (t)dt + iy 0 (t)dt):
C
where C is the unit circle in the positive direction. One parametrization of the
unit circle is C : z(t) = eit ; t : 0 ! 2 : Inserting this parametrization in the
integral gives dz = ieit dt and
I
Z 2
Z 2
z n dz =
(eit )n ieit dt = i
eit(n+1) dt:
0
If n + 1 6= 0 we get
I
Z 2
z n dz = i
eit(n+1) dt = i[
while when n =
1
1
eit(n+1) ]20 =
(ei2
i(n + 1)
n+1
1 we get
I
Z
z n dz = i
C
e0 dt = i[t]20 = 2 i:
(n+1)
1) =
1
(1 1) = 0;
n+1
So
I
1: In this case
z n dz =
dz = 2 i:
1.3
In fact, if f (z) is analytic everywhere on and inside the curve C; we know that
I
f (z)dz = 0:
@P
)dxdy;
@y
ZZ
ZZ
@v
@x
ZZ
@u
@u @v
)dxdy + i
(
+
)dxdy
@y
@x @y
D
ZZ
(0)dxdy + i
(0)dxdy = 0
D
1.4
vx0 :
f (z)
dz
z z0
where z0 is inside C and f (z) is analytic everywhere inside C: Then we can add
a little circle C0 around z0 with radius " and one straight line L leading from
one point on C to one on the circle C0 that does not intersect z0 : Then, the line
L goes the other direction. This means that C + L L + C" is a closed curve,
so the Cauchy theorem is valid. Hence,
I
f (z)
dz:
z z0
C+L L+C"
Furthermore,
f (z)
dz +
z z0
Hence,
f (z)
dz = 0:
z z0
f (z)
dz =
z z0
f (z)
dz:
z z0
C"
f (z)
dz =
z z0
f (z)
dz !
z z0
C"
f (z0 )
dz =
z z0
C"
f (z0 ) it
"ie dz ! 2 if (z0 ):
"eit
C"
1
2 i
f (z)
dz:
z z0
1.5
= ::: + a2 (z z0 )2 + a1 (z z0 ) + a0 + a:1 (z
1
1
X
X
=
an (z z0 )n +
an (z z0 )n
n=0
z0 )
+ a:
2 (z
z0 )
+a
n= 1
z0 j < R2 g;
where R1 < R2 ; of course. Often R1 = 0; such as in the case f (z) = 1=z; which
is analytic everywhere except in one single point z = 0:
The principal part of the Laurent series is
1
X
n= 1
an (z
z0 )n =
a
z
z0
a
(z
z0 )2
a
(z
z0 )3
+ ::: .
If the principal part contains an innite number of terms we say that f (z) has
an essential singularity att z0 : An example is e1=z at z = 0: If the number of
terms is nite we say that f (z) has a pole of order n if an 6= 0 but ak = 0 for
all k > n: It has thus a simple pole (of order 1) if the principal part is
a
z0
3 (z
z0 )
+ :::
where a
part is
6= 0:The function f (z) has a double pole (of order 2) if the principal
a
z
where a
6= 0; but a
z0
a
(z
z0 )2
1
z zo
in the
1:
This can be shown from the Laurent series by three steps: 1. mutiply by
z)n ; 2. take n 1 derivatives, and 3. let z = z0 :
For example, if n = 2 we have
z0 )2 + a1 (z
f (z) = ::: + a2 (z
z0 ) + a0 + a:1 (z
z0 )
+ a:
2 (z
z0 )
z0 )2 = :::a1 (z
2;
z0 )3 + a0 (z
z0 )2 + a:1 (z
z0 ) + a:
2:
and we get
z0 )2 = :::3a1 (z
z0 )2 + 2a0 (z
z0 ) + a
:1 :
d
f (z)(z
dz
z0 )2 ]z=z0
= :::3a1 (z0
=
0+0+a
z0 )2 + 2a0 (z0
1
=a
z0 ) + a:
1:
For a simple pole there is a third way to calculate a residue, which often is
very practical. Assume that
p(z)
f (z) =
q(z)
where p(z0 ) 6= 0 but q(zo ) = 0: Then
Res(f (z); z0 ) =
p(z0 )
:
q 0 (z0 )
p(z)
(z
q(z)
p(z)
p(z0 )
p(z)
[ q(z) q(z ) ]z=z0 = 0
:
0
q (z0 )
z0 )]z=z0
]
q(z) z=z0
(z z0 )
= fuse q(z0 ) = 0g
(z z0 )
1.6
If we have several singular points of f (z) inside a curve C; we can divide integration along C to integration over small circles, one for each singularity of
f (z). Integrating
f (z) = :::+a2 (z z0 )2 +a1 (z z0 )+a0 +a:1 (z z0 )
+a:
2 (z
z0 )
+a
3 (z
z0 )
z n dz = 0 for all n 6=
dz = 2 i. So by a
f (z)dz = 2 i
n
X1
Res(f (z); zi );
i=0
where z0 ; :::; zn 1 are all singular points inside C for function f (z) which is
analytic everywhere inside C except at z0 ; :::; zn 1 . This statement is the Cauchy
Residue Theorem.
1.7
f (cos ; sin)d :
+:::
)=2:
Denote
g(z) = f ((z + z
)=2; (z
)=2i)=iz:
By the Cauchy residue theorem, we then only need to nd all residues of g(z)
inside the unit circle (ignoring those outside, of course). If g has singularities in
z1 ; :::; zn ; we have the value of the integral.
Z2
f (cos ; sin)d = 2 i
n
X
Res(g; zi ):
k=1
Example 1 Solve
Z2
1
d :
5 + sin
1
d = fz = ei g =
5 + 4 sin
1
5 + 4(z
1 )=2i
dz
:
iz
dz:
1
2
Now, where are the poles of 2z2 +5iz
2?
2 ; i.e., where are the zeros of 2z + 5iz
2
Inserting z = x + iy gives the two zeros i=2 and 2i; so 2z + 5iz 2 =
1
1
(2z + i)(z + 2i): So 2z2 +5iz
2 = (2z+i)(z+2i) :
We have the integration contour jzj = 1: One of the two poles of the function
1
i=2: The residue in this pole is
(2z+i)(z+2i) is inside the curve:
1
z + 2i
=
z= i=2
1
=
i=2 + 2i
2i=3:
1
d =
5 + 4 sin
1
2z 2 + 5iz
dz = 2 i( 2i=3) = 4 =3:
Example 2 Solve
Z2
cos 2
d and
(3 cos + 5)2
Z2
sin 2
d :
(3 cos + 5)2
Solution: As in the previous solution, the substitution z = ei gives integration over the circle C : jzj = 1 in the complex plane. Furthermore, d = dz=iz;
sin = (z z 1 )=2i and cos = (z + z 1 )=2. Now, the calculation is easier if
we use the following facts:
Z2
e2i
d ];
(3 cos + 5)
Z2
Re[
Z2
Z2
Im[
e2i
d ]
(3 cos + 5)
cos 2
d
(3 cos + 5)2
sin 2
d
(3 cos + 5)2
e2i
d
(3 cos + 5)2
4
i
z2
dz
(3(z + z 1 )=2 + 5)2 zi
I
z3
dz:
(3z + 1)2 (z + 3)2
1 d
z3
9 dz (z + 3)2
1 z2
(z + 9) ;
9 (z + 3)3
4iz 3
;
(3z + 1)2 (z + 3)2
( 13 )2
1
4i
)=
3
9 ( 1) + 3
3
8
1
+9
3
13i
:
4 144
So we have
Z2
e2i
4iz 3
d = 2 iRes(
;
2
(3 cos + 5)
(3z + 1)2 (z + 3)2
1
13
)=
;
3
288
and
Z2
cos 2
d
(3 cos + 5)2
13
and
288
Z2
sin 2
d
(3 cos + 5)2
0:
1.8
We will next observe that a real integral can be viewed upon as a complex
integral:
Z b
Z
f (x)dx =
f (z)dz
a
where C is the straight line from the complex number a + i0 to b + i0: If f (z)
is analytic (except in a few points) we can then close the curve and solve the
integral using residue calculus.
For integrals on the interval ( 1; 1) this is done by considering ( R; R) for
a large R; and this curve is closed by a circular arc CR in the upper half plane.
Then as R ! 1 all of the upper half plane becomes included, and typically the
integral over the circular arc tend to zero. So, in such a case the integral on
( 1; 1) is just 2 i times the sum of residues of f (z) in the upper half plane.
Example 3 Calculate
1
1
cos x
dx:
x2 + a2
1
1
eix
dx:
x2 + a2
This is a simplication since it is easier to calculate with eix than with cos x:
Note that the integrand has simple poles in ia only. Denote by IR the nite
straight curve on re real line z(t) = t : R ! R: Then
Z
R
R
eix
dx =
2
x + a2
LR
z2
eiz
dz:
+ a2
Now we add the circular arc CR that has radius R: z(t) = Reit ; t : 0 ! :
Then LR + CR is closed, so
I
z2
eiz
eiz
dz = 2 iRes( 2
; ia)
2
+a
z + a2
LR +CR
if R is large enough to include the pole ia: The pole ia is in the lower half
plane and never inside the curve LR + CR :
The circular arc gives no contribution when R ! 1 since
j
eiz
dzj = j
2
z + a2
LR
it
LR
R2
when R ! 1; since 0 t
; so e
Now the residue att ia is
eiRe
Reit dtj
2
R e2it + a2
1
R sin t
R sin t
2 !0
1:
2
Res(
So
eiz
eiz
ei a
; ia) = [ ]z=ia =
=
2
2
z +a
2z
2ia
e a
:
2a
eiz
eiz
dz
=
2
iRes(
; ia) = e
z 2 + a2
z 2 + a2
a
LR +CR
and
a
eiz
dz !
z 2 + a2
LR +CR
as R ! 1:
Finally,
1
1
cos x
dx = Re
x2 + a2
So
Example 4 Calculate
1
1
1
1
1
1
xm
dx
+1
xn
eix
dx
x2 + a2
eix
dx = e
x2 + a2
a
cos x
dx = e
x2 + a2
a
if n
zm
z n +1
zk = ei
2k+1
n
1.4
1.2
1.0
0.8
0.6
0.4
0.2
-0.4 -0.2
0.2
0.4
zm
dz
n
z +1
0.6
-0.2
0.8
1.0
1.2
1.4
-0.4
So:
C0
zm
dz +
n
z +1
CR
C1
zm
zm
dz = 2 iRes( n
; ei
+1
z +1
zn
1
n
):
So we need to take a closer look on these four integrals as well as the residue
1
at ei n .
We can start by observing that
Z
zm
dzj ! 0
j
n
CR z + 1
as R ! 1; since
Z
Z
zm
Rm eimt
j
dzj
=
j
Rieit dtj
n
n int + 1
CR z + 1
CR R e
Z
Rm eimt
= R
j n int
jdt = fuse jz + wj
+1
CR R e
Z
Rm
Rm+1
dt = 2 n
!0
= R
n
1
R
1
CR R
11
jzj
jwjg
as R ! 1 since n 2 m:
Simultaneously, since C0 = t for t : 0 ! R, we have also
Z
Z 1
zm
xm
dz
!
dx
n
xn + 1
C0 z + 1
0
as R ! 1: But what happens to
Z
C1
zm
dz?
zn + 1
zm
dz =
n
z +1
C1
2
n
for t : 0 ! R gives
2
n
(tei
(tei
2
n
)m
)n
+1
2
n
ei
dt
C0
zm
dz +
zn + 1
CR
zm
dz
zn + 1
C1
zm
dz
zn + 1
(1
xm
dx + 0
xn + 1
Z 1
2i m+1
n
e
)
0
zm
; ei
+1
zn
1
n
Hence,
(1
2i
m+1
n
)=[
Z
zm
] i
nz n 1 z=e
1
n
1 i
e
n
xm
2 i i
dx =
e
xn + 1
n
(m
(m
n+1)
n
n+1)
n
e2i
(m
m+1
n
n+1)
n
= ei
=ei
m+1
n
(m+1)
n
(e
m+1
n
ei
m+1
n
2iei
)=
12
m+1
n
sin
xm
dx:
xn + 1
2+2m
n
(m + 1)
;
n
tm
dt
tn + 1
1.9
Z
f (x)dx = lim (
!0
x0
f (x)dx +
f (x)dx):
x0 +
1
1
1
dx = 0;
x
1
dx
1 x
although
is divergent. This allows a version of the Cauchy Residue
theorem with poles on the curve. For a function f (z) which is analytic except
at z1 ; :::; zn inside C and except w1 ; :::; wm on the curve C; we have
PV
f (z)dz = 2 i
Res(f (z); zi ) + i
i=1
1.10
n
X
m
X
Res(f (z); wi ):
i=1
Multi-valued functions
We can
p are the funcp also integrate multi-valued functions, where two examples
tions z and ln z: Normally we have rei = rei( +2 ) : Since z = z 1=2 and
ln z = jzj + i arg z we get for these functions
p i =2
(rei )1=2 =
re
but
p i =2 i
p i =2
(rei( +2 ) )1=2 =
re
e =
re
which are not the same, and
ln rei
i( +2 )
ln re
= jrj + i but
= jrj + i( + 2 );
p
ln x x
dx:
(x + 1)2
Solution: We here integrate in a large circle Reit and in a small eit around
the branch point, and connect them with two straight lines L+ and L :
13
We dene L+ as the line from to R on the real line where the argument
arg z = 0, and L as the same but taken in the opposite direction and with
argument arg z = 2 : This argument on L is necessary by following the path
and keeping the multivalued functions continuous.
This makes L+ + CR + L + C a closed path, and the only pole inside the
curve is z = 1: Hence,
p
p
I
ln z z
ln z z
dz = 2 iRes(
; 1)
(z + 1)2
(z + 1)2
L+ +CR +L +C
p
p
(ln )2 + (2 )2
2
(1
)2
!0
as
! 0:
Also, as R ! 1 and
Z
! 0; we have
p
p
Z 1
ln x x
ln z z
dz
!
dx
2
(x + 1)2
0
L+ (z + 1)
which is the integral we are interested in. Now only the integral over L remains.
Here we have z = te2 i ; so
p
ln z z = (t + i2 )( t):
Hence,
p
ln z z
dz !
(z + 1)2
p
ln t t
dt
(t + 1)2
i2
t
dt:
(t + 1)2
p
ln z z
dz
(z + 1)2
2 i( i +
L+ +CR +L +C
! 0, we get
p
p
Z 1
Z 1 p
ln x x
ln x x
x
dx
+
dx
+
i2
dx
2
(x + 1)2
(x
+
1)
(x
+
1)2
0
0
2
)=2 +i
So on the left side we get two copies of the requested integral. Taking real parts
and imaginary parts of the equality gives two integrals:
p
Z 1
ln x x
Re :
dx =
(x + 1)2
0
Z 1 p
2
x
Im :
dx
=
;
(x + 1)2
2
0
where the rst is the requested integral.
15