Analyt

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Complex function of a complex variable

J. Cousteix

The objective is to remind a few notions wich are useful to understand the treat-
ment of potential two-dimensional flows by using analytic functions. Rigourous
presentations are available in mathematics books [1, 2].

1 Function of a complex variable


Let z be a complex number
z = x + iy,

where i is the pure imaginary number such as i2 = −1. In the complex plane C, this
number is represented by a point M whose coordinates are x and y, z is called the
affix of M.

y M

θ
x
0 x

Fig. 1: Representation of a complex number in the complex plane

Sometimes, it is interesting to use polar coordinates

z = r eiθ .
p
The modulus |z| of the complex number is |z| = r = x2 + y2 and its argument is θ
which can be defined in the domain [−π, π [, for example.
Let us consider a function Z of z whose value is complex

Z = P( x, y) + iQ( x, y).

The conjugate of Z is denoted Z

Z = P − iQ.

All the complex functions of a complexe variable are not interesting. Let us take for
example
P = x, Q = 2y.
2 TAS - AERODYNAMICS 1

It is possible to define a mapping from z to Z but Z cannot be expressed as a function


of z only. We have
3 1
Z = x + 2iy = z − z̄.
2 2
If we take
P = x 2 − y2 , Q = 2xy,

we have
Z = x2 − y2 + 2ixy = ( x + iy)2 ,

or,
Z = z2 .

In this case, we see that Z can be defined as a function of z. In both cases, it is possible
to define a mapping from z to Z but the difference is that, in the second casen it is
possible to express Z as function of z—here with an algebraic function—whereas
in the first case, this is not possible. The functions similar to the second case are
of particular interest. These functions are obtained by looking for complex valued
functions of a complex variable for which a derivative can be defined.

2 Derivative
Let Ω be an open set of C. We define f : Ω → C a complex valued function of z. In
general we have
f (z) = P( x, y) + iQ( x, y).

2.1 Definition
The function f : Ω → C has a derivative at a point z0 ∈ Ω if the ratio

f ( z ) − f ( z0 )
f 0 (z0 ) = lim ,
z → z0 z − z0

exists when z → z0 by following any path in the complex plane, the limit being
independent of the path. If the path is parallel to the real axis, we have z − z0 =
x − x0 , if the path is parallel to the imaginary axis we have z − z0 = i (y − y0 ), but
we can consider any other path to calculate f 0 (z0 ).
If the derivative exists in a neighborhood of z0 we say that the function is holo-
morphic at z0 . If the derivative exists at any point of Ω we say that the function is
holomorphic on Ω. A holomorphic function on Ω takes one value and only one at
any point of Ω. We call H(Ω) the set of holomorphic functions on Ω.
The classical rules of calculation of the derivatives of a real valued function of a
real variable apply to holomorphic functions. The functions which are expressed as
function of z only (and sufficiently regular) are holomorphic. A function which is
expressed as a function of z and of its conjugate z̄ = x − iy is not holomorphic.
TAS - AERODYNAMICS 1 3

2.1.1 Cauchy-Riemann Conditions


We consider a function f (z) = P( x, y) + iQ( x, y), we suppose that P and Q are
differentiable. It is shown that the function f has a derivative at point z0 if and only
if the Cauchy-Riemann conditions are satisfied

∂P ∂Q ∂P ∂Q
= , =− .
∂x ( x0 ,y0 ) ∂y ( x0 ,y0 ) ∂y ( x0 ,y0 ) ∂x ( x0 ,y0 )

If P and Q have second order continuous derivatives, we get

4 P = 0, 4 Q = 0.

We say that P and Q are harmonic conjugate functions.


If the path z − z0 is taken along the real axis, we have

∂P ∂Q
f 0 ( z0 ) = +i .
∂x ∂x
If the path z − z0 is taken along the imaginary axis, we have

∂Q ∂P
f 0 ( z0 ) = −i
∂y ∂y

By following these two paths, if the derivative exists, its value is the same. In fact,
we can follow any other path since f 0 (z0 ) does not depend on the path. We have also

∂P ∂P
f 0 ( z0 ) = −i
∂x ∂y
∂Q ∂Q
= +i
∂y ∂x

2.1.2 Hydrodynamical Interpretation of Cauchy-Riemann Conditions


We consider a two-dimensional flow. We take u = P and v = − Q. If the flow
is incompressible and irrotational, the function w = u − iv satisfies the Cauchy-
Riemann conditions. The complex velocity w(z) is a holomorphic function of z.

2.2 Multiform functions


A function f (z) is multiform or multiple valued if f (z) can take different values for
the same value of z. The mapping from z to f (z) is not unique.
1
Let us consider the function Z = z m wher m is an integer. Using polar coordi-
nates, we have
1 θ +2kπ
Z = r m ei m ,
where k is an integer. For the same value of z, m different values of Z are obtained.
The function has m different branches according the choice of k. Let us follow z along
a closed curve surrounding the origin of coordinates. The starting point is z0 with
an argument θ0 . Let us suppose that k is fixed, for example k = 0. If θ increases from
4 TAS - AERODYNAMICS 1

θ = θ0 , the initial value of Z is obtained again when z has made m rounds around
the origin, that is, when θ has increased by 2mπ.
In this example, the path followed by z surrounds the origin. If the path of z does
not surround the origin, a single value of Z is obtained for z = z0 but this value
depends of the value chosen for k. We say that the origin is a branch point.
A multiple valued function can be rendered single valued if we decide that z
cannot follow a path which crosses a barrier in the complex plane. Such a barrier
(branch cut or cut line) is for example the line ∆, Fig. 2, which corresponds here to
θ = π. In this way, the domain of variation of θ is −π ≤ θ < π and the function Z is
defined by
1 θ
Z = r m ei m ; −π ≤ θ < π,
1
Then, the function z m is defined by its principal determination. In the complex plane
equipped with such a barrier, there exists no closed path surrounding the origin. A
single valued function is defined but the function is discontinuous across the barrier
since θ undergoes a jump of 2π.


x
0
Fig. 2: Branch cut in the complex plane

The function Z = ln z is also a multiple valued function. With the cut θ = π, Z is


defined by
Z = ln r + iθ ; −π ≤ θ < π.

which is the principal determination of the logarithm function.


We must notice that in both cases, in the complex plane equipped with a cut, the
origin is a singular point but this is not an isolated singularity since it is not possible
to find a neighborhood of the origin (N \ {O}) where the function is holomorphic.

2.3 Series Expansions


2.3.1 Taylor series

If a function f (z) is holomorphic at a point a, there exists a circle γ of center a within


which f (z) can be developed as


f (n) ( a )
f (z) = ∑ n!
(z − a)n , n integer
n =0
TAS - AERODYNAMICS 1 5

which is a Taylor series, f (n) ( a) is the derivative of order n of f (z) at point a. The
radius of convergence of the series is at least equal to the radius of the circle γ. The
radius of convergence of the series is equal to the distance between a and the singu-
larity of f (z) which is the closest to a.

2.3.2 Analytic function


The coefficients cn being constants, we define a function g(z) by the following power
series

g(z) = ∑ cn (z − a)n , n integer. (2-1)
n =0

Definition

A function f is analytic on an open set Ω in C if and only if, at every point of Ω, f can
be developed in a power series (2-1) wich is convergent on a disk whose radius is
non zero.
The following equivalence is proved [1]

f is holomorphic on Ω ⇐⇒ f is analytic on Ω.

An immediate consequence is that a holomorphic function on Ω is C ∞ on Ω, that is,


the function is indefinitely differentiable. Thus, we have a very strong result, the fact
that a function has a derivative at every point of Ω implies that f has derivatives of
any order.

2.3.3 Laurent series

C2 Let us suppose that f (z) is a holomorphic


function in an annular region, Fig. 3, between
z1 two concentric circles C1 and C2 of center a.
z2
a It is possible that the function f (z) has iso-
C1 lated singularities within the circle C1 . Singu-
lar points can exist within C1 but the center a
is not necessarily singular. The function f (z)
has a unique expansion in Laurent series

Fig. 3: Domain of convergence f (z) = ∑ An (z − a)n .
of a Laurent series. The points −∞
z1 and z2 are singular

The Laurent series is convergent in the annular region. The most extended domain
of convergence is obtained when there is a singularity on C1 and another one on C2 ,
Fig. 3.
6 TAS - AERODYNAMICS 1

It must be noticed that this theorem is not valid if the annular region is crossed
by a cut line.
f (n) ( a )
In general, the coefficients An cannot be identified with because f (z) is
n!
not necessarily holomorphic within circle C1 .
The Laurent series can be decomposed into two parts
−1 ∞
f (z) = ∑ An (z − a)n + ∑ An (z − a)n .
−∞ 0

The first part is called the principal part and the second part is called the regular
part.

2.3.4 Poles and Residues


Let f be a holomorphic function on an open neighborhood V of a except at point a.
We suppose that limz→ a | f (z)| exists and is +∞. We say that f has a pole at point a.
If | f | has no limit at point a we say that f has an essential singularity.
If f is a holomorphic function on an annular region R1 < |z − a| < R2 , f has a
unique expansion in Laurent series on this region

f = ∑ an (z − a)n .
n ∈Z

Function f has a pole at point a if the number of non zero coefficients an with a
negative index is finite. The absolute value of the smallest (negative) index is the
order of the pole. We define the residue of f at point a as the coefficient of the term
1
in
z−a
Res( f ; a) = a−1 .
If f has a simple pole at point a, the residue is

Res( f ; a) = lim [(z − a) f (z)] .


z→ a

If f has a pole of order k at point a, the residue of f at point a is

1 dk −1 h i
Res( f ; a) = lim k−1 (z − a)k f (z) .
(k − 1)! z→a dz

2.4 Integration
2.4.1 Definition
The integral of f (z) about a curve C is denoted
Z
I= f (z)dz, f = P + iQ
C

This integral can be defined by using a parametrization z = γ(t) of the curve C ,


where t is a real Z
I= f (γ(t))γ0 (t)dt
C
TAS - AERODYNAMICS 1 7

2.4.2 Cauchy’s Theorem


If f (z) is holomorphic on a simply connected domain D , the integral of f (z) about a
closed curve contained in D , Fig. 4, is zero
Z
f (z) dz = 0.
C

Fig. 4: Application of the Cauchy theorem. The function f (z) is holomorphic on D

2.4.3 Cauchy’s Integral


We suppose that f (z) is holomorphic on a domain D and on its boundary ∂D . Let a
be a point of D and C a closed curve in D surrounding a, Fig 5. The theorem of
Cauchy’s integral states that
1 f (z)
Z
f ( a) = dz.
2iπ C z−a
It must be noted that the curve C is directed in the positive direction (anti-clockwise)
and only one round about C is performed.
This theorem is valid if the domain is simply connected but can be extended to
the case of a multiply connected domain.

a
C

Fig. 5: Theorem of the Cauchy integral. The function f (z) is holomorphic in D .

A consequence of this theorem is that if f (z) is holomorphic on a domain D


bounded by a curve C , all the successive derivatives of f (z) exist on D and it is
shown that
n! f (z)
Z
f (n) ( a ) = dz.
2iπ C (z − a)n+1
8 TAS - AERODYNAMICS 1

2.4.4 Theorem of residues


Let Γ be a closed curve in the complex plane. Here, we suppose that the curve is
simple, the curve does not cut itself. Let f (z) be a function having residues Res( f ; Pi )
at N points Pi inside Γ, not on Γ. The curve is positively oriented (that is in the
counterclockwise direction). For one round on Γ we have
Z N
f (z)dz = 2iπ ∑ Res( f ; Pi ).
Γ i =1

Let us notice that in the case considered here, the curve winds around every point Pi
one time and only one time in the positive direction.

Γ
z2
z1
zk
zn

Fig. 6: Theorem of residues. The function f (z) is holomorphic on a domain bounded


by a curve C except at a finite number of isolated singular points indicated by the
crosses

Application We consider the integral


eαz
Z
I= dz,
C z2
eαz
where C is a simple closed curve. The function is holomorphic on the complex
z2
plane except at z = 0. The function eαz can be developed as
α2 z2 (αz)n
eαz = 1 + αz + +···+ +···,
2 n!
The radius of convergence is infinite. We obtain
eαz 1 α α2 αn
= + + + · · · + +···
z2 z2 z 2 n!z2−n
The point z = 0 is singular and we see that the residue is α. Then, we have
(
0 if C does not surround the origin
I=
2iπα if C surrounds the origin

2.4.5 Jordan Lemmas


These lemmas are often used to calculate definite integrals in the complex plane. We
consider a function f (z) continuous in the neighborhood of z0 except possibly at z0 .
We suppose that
lim |(z − z0 ) f (z)| = 0,
|z−z0 |→0
TAS - AERODYNAMICS 1 9

then, Z
lim f (z) dz = 0,
r →0 γ

where γ is a circle arc of radius r and of center z0 , Fig. 7.


Γ

γ
R
r

z0 z0

r→0

R→∞
Fig. 7: Jordan lemmas

We consider a function f (z) continuous outside a circle of center z0 and such that

lim |(z − z0 ) f (z)| = 0,


|z−z0 |→∞

then, Z
lim f (z) dz = 0,
R→∞ Γ
where Γ is a circle arc of radius R and of center z0 , Fig. 7.
These lemmas can be used, for example, to evaluate an integral on the real axis.
The path is completed by a circle arc whose radius tends to infinity and the theorem
of residues is applied.
Z ∞
1
Application We want to calculate the integral dx.
−∞ 1 + x4

We consider the integral


Γ
1
Z
z2 z1 I= dz.
C 1 + z4
x
−R O R The curve C is composed of the segment
[− R, R] on the real axis and of the semi-
Fig. 8: Application of Jordan circle Γ of radius R and center O. The inte-
lemma gral I is calculated when R → ∞.

The function 1 + z4 is equal to 0 for z1 = eiπ/4 , z2 = ei3π/4 , z3 = ei5π/4 , z4 =


1
ei7π/4 . The function f (z) = has 4 simple poles at z1 , z2 , z3 , z4 . If zk is a pole,
1 + z4
the value of the residue Rk at zk is
 " # −1
d(1 + z4 )
 
1 1 z
Rk = lim (z − zk ) 4
= = 3 = k4
z →0 1+z dz z=zk 4zk 4zk
10 TAS - AERODYNAMICS 1

With z4k = −1, we have


zk
Rk = −
4
Now, on Γ, we have
1
lim R =0
R→∞ 1 + z4
Thus, we can apply the Jordan lemma. From the theorem of residues, we obtain
Z ∞
1 2iπ
dx = 2iπ ( R1 + R2 ) = − ( z1 + z2 )
−∞ 1 + x4 4
that is Z ∞ √
1 2
dx = π
−∞ 1 + x4 2

References
[1] W. A PPEL. Mathématiques pour la physique et les physiciens. H&K Éditions, Paris,
2005. 3e édition.

[2] J. B ASS. Cours de mathématiques. Masson & Cie, Paris, 1964.

You might also like