Math 115 (2006-2007) Yum-Tong Siu 1

Conformal Mappings and Application to Electrostatics

We will apply holomorphic functions (conformal mappings) to the prob-

lem of finding electrostatic potentials with prescribed constant boundary val-
ues for the two dimensional case. First we discuss the properties of certain
special kinds of conformal mappings, in particular, linear fractional transfor-
mations, and their mapping properties. Then we use a number of examples
to illustrate how conformal mappings are used for applications to problems
in electrostatics.

Linear Fractional Transformations. A linear fractional transformation is a

map of the form
az + b
z 7→ w := .
cz + d
Its derivative dw
dz
is given by
dw a(cz + d) − c(az + b) ac − bd
= = .
dz (cz + d)2 (cz + d)2
To be able to invert it, the necessary condition is that its derivative is not
identically zero, i.e., ad − bc 6= 0. This is also both necessary and sufficient
for invertibility on the extended complex plane C ∪ {∞} which adds one
point ∞ to C so that near the point ∞ we can use the coordinate system
ζ = z1 . Another way to describe the linear fractional transformation
az + b
z 7→ w :=
cz + d
is to introduce two complex variables (z1 , z2 ) so that z = zz21 and to introduce
w1
two complex variables (w1 , w2 ) so that w = w 2
and write
µ ¶ µ ¶µ ¶
w1 a b z1
= .
w2 c d z2
Since ad − bc is the determinant of the matrix
µ ¶
a b
,
c d
when ad − bc 6= 0 we can invert the matrix and get
µ ¶−1 µ ¶
a b 1 d −b
=
c d ad − bc −c a
Math 115 (2006-2007) Yum-Tong Siu 2

and µ ¶ µ ¶µ ¶
z1 d −b w1
= .
z2 −c a w2
Thus
dw − b
z= .
−cz + a
Of course, we get the same result by solving for w in terms of z and from
az + b
w=
cz + d
get
czw + dw = az + b
and
dw − b = −cwz + az = z (−cw + a)
and finally
dw − b
z= .
−cw + a
Note that though even without the condition ad − bc 6= 0 we can still write
down the map
dw − b
z= ,
−cz + a
but putting
az + b
w=
cz + d
into
dw − b
−cw + a
gives
d az+b
cz+d
−b adz+bd−bcz−bd
cz+d (ad − bc)z
az+b
= =
−c cz+d + a −acz−bc+acz+ad
cz+d
ad − bc
which gives z only when ad − bc 6= 0.
Math 115 (2006-2007) Yum-Tong Siu 3

Mapping Behavior of Linear Fractional Transformations. To understand the

mapping behavior of the linear fractional transformation
az + b
z 7→ w := ,
cz + d
we decompose it as the composite of simpler linear fractional transformations
and write
az + b a ad − bc 1
w= = −
cz + d c c cz + d
just by dividing az + b by cz + d by long division. This is the same as writing
the matrix µ ¶
a b
c d
as a product of simpler ones as follows.
µ ¶ µ a ¶ µ ad−bc ¶ µ ¶
a b 1 c − c 0 0 1
= .
c d 0 1 0 1 c d

The linear fractional transformation is a composite of the following three

simpler types of linear fractional transformations:
(i) A translation z 7→ z + a for some a ∈ C.

(ii) A homothety of multiplication by a nonzero constant z 7→ az for some

a ∈ C − {0}.

(i) The reciprocal map z 7→ z1 .

The mapping properties of the first two types are clear. Now we look at the
third one and verify that the reciprocal map maps the collection of circles
and straight lines to the collection of circles and straight lines. However,
it does not mean that it maps circles to circles and maps straight lines to
straight lines. It may happen that the reciprocal map maps some circle to
a straight line and maps some straight line to a circle. The equation of the
circle centered at a ∈ C with radius r > 0 is given by

|z − a|2 = r2 .

The transformation
1
z 7→ w :=
z
Math 115 (2006-2007) Yum-Tong Siu 4

would map it to ¯ ¯2
¯1 ¯
¯ − a¯ = r 2 ,
¯w ¯
which is
|1 − aw|2 = r2 |w|2
or
1 − aw − āw̄ + |a|2 |w|2 = r2 |w|2
or
|a|2 − r2 |w|2 − aw − āw̄ + 1 = 0.
¡ ¢

This is a straight line if and only if |a|2 − r2 = 0, i.e., the origin 0 is on the
original circle ¯ ¯2
¯1 ¯
¯ − a¯ = r 2 .
¯w ¯
Otherwise, the the original circle is mapped to another circle
¡ 2
|a| − r2 |w|2 − aw − āw̄ + 1 = 0.
¢

Let us see what a straight line is mapped to by this linear fractional trans-
formation. The equation of a general line is

αx + βy + γ = 0 with α, β, γ ∈ R.

We can rewrite it in terms of the complex variable z and get

z + z̄ z − z̄
α +β +γ =0
2 2i
or µ ¶ µ ¶
α β α β
− i z+ + i z̄ + γ = 0.
2 2 2 2
Let a = α
2
− β2 i ∈ C. Then the equation of the line becomes

az + āz̄ + γ = 0,

which is transformed by the reciprocal map to

a ā
+ +γ =0
w w̄
Math 115 (2006-2007) Yum-Tong Siu 5

or
aw̄ + w̄ + γ |w|2 = 0.
This is a circle if and only if γ 6= 0, i.e., if and only if the origin is not on the
original line
αx + βy + γ = 0,
otherwise the image is a circle. The reciprocal map is the composite of the
conjugation map z 7→ z̄ (which is the reflection with respect to the real axis)
and the inversion with respect to the unit circle studied in Euclidean plane
geometry
1
z 7→
z̄
iθ 1 −iθ
which sends z = re to r e and can be described as sending a point P to
another point Q on the line joining the origin to P so that the product of
the distance between the origin and P and the distance between the origin
and Q is 1.
The rule to remember when we have to determine whether the image of a
circle or a straight line should be a circle or a straight line is that a straight
line can be regarded as a circle which passes through ∞.

Cross-Ratios. A general linear fractional transformation

az + b
z 7→ w :=
cz + d
is defined by four complex numbers a, b, c, d. However, the simultaneous
multiplication of the four complex numbers by the same nonzero complex
number would not change the linear fractional transformation. So the degree
of freedom in the choice of a general linear fractional transformation is three
complex numbers. When we fix three distinct complex numbers z1 , z2 , z3 , we
expect to be able to find a linear fractional transformation for any prescribed
triple of distinct complex numbers w1 , w2 , w3 . However, when we want to
find a linear fractional transformation mapping a prescribed quadruple of
distinct complex numbers z1 , z2 , z3 , z4 to another prescribed quadruple of
distinct complex numbers w1 , w2 , w3 , w4 , we expect not to be able to do it
unless the two quadruples satisfy some relation. The relation turns out to be
the preservation of cross-ratios by any linear fractional transformation. For
Math 115 (2006-2007) Yum-Tong Siu 6

any given quadruple of distinct complex numbers z1 , z2 , z3 , z4 , its cross-ratio

(z1 , z2 , z3 , z4 ) is defined by
z1 −z3
z1 −z4
(z1 , z2 , z3 , z4 ) = z2 −z3 .
z2 −z4

The verification of the preservation of a cross-ratio by a linear fractional

transformation is as follows. We have
azj + b
wj = for 1 ≤ j ≤ 4.
czj + d
Then
az1 + b az3 + b
w1 − w3 = −
cz1 + d cz3 + d
(az1 + b) (cz3 + d) − (az3 + b) (cz1 + d)
=
(cz1 + d) (cz3 + d)
(adz1 + bcz3 ) − (adz3 + bcz1 )
(cz1 + d) (cz3 + d)
(ad − bc) (z1 − z3 )
= .
(cz1 + d) (cz3 + d)
Changing z3 to z4 and w3 to w4 and taking the quotient of the two equations,
we get
w1 − w3 z1 − z3 cz4 + d
= .
w1 − w4 z1 − z4 cz3 + d
Changing z1 to z2 and w1 to w2 and taking the quotient of the two equations,
we get
w1 −w3 z1 −z3
w1 −w4 z1 −z4
w2 −w3 = z2 −z3 .
w2 −w4 z2 −z4
This, in particular, tells us that the linear fractional transformation z 7→ w
which sends a prescribed triple of distinct complex numbers z1 , z2 , z3 to a
prescribed triple of distinct complex numbers w1 , w2 , w3 must preserve the
cross-ratios (z1 , z2 , z3 , z) and (w1 , w2 , w3 , w), which means that the linear
fractional transformation is given by
w1 −w3 z1 −z3
w1 −w z1 −z
w2 −w3 = z2 −z3 .
w2 −w z2 −z

This is an easy way to write down explicitly the linear fractional transforma-
tion.
Math 115 (2006-2007) Yum-Tong Siu 7

Conformality of Holomorphic Map With Nonzero Derivative. We now con-

sider the mapping behavior of a holomorphic map w = f (z) at a point
z0 where its derivative f ′ (z0 ) is nonzero. Take a smooth curve t 7→ z(t)
passing through the point z0 so that z (0) = z0 . The image of the curve
t 7→ w(t) := f (z(t)) passes through the point w0 := f (z0 ). By the chain rule
and the Cauchy-Riemann equation We have
dw ∂f dx ∂f dy
= +
dt ∂x dt ∂y dt
∂f dx ∂f dy
= +i
∂x µ
dt ∂x dt¶
∂f dx dy
= +i
∂x dt dt
∂f dz dz
= = f ′ (z) .
∂x dt dt
This means that the angle made between the tangent to the curve t 7→
z(t) at z0 and the real axis in the z variable is equal to the angle made
between the tangent to the curve t 7→ w(t) at w0 and the real axis in the w
variable plus the angle of the polar representation of f ′ (z0 ). In particular,
the angle made between two curves at z0 is the same as the angle at w0 made
between their images under the map f . When we have a family of curves
t 7→ γs (t) parametrized by a real variable s passing through z0 such that
their tangents at z0 rotates in the counterclockwise sense as s increases, the
tangent at w0 to the image curve t 7→ f (γs (t)) under f also rotates in the
counterclockwise sense. In other words, the holomorphic map z 7→ w := f (z)
is angle-preserving and orientation-preserving when f ′ (z0 ) 6= 0. We call a
mapping with the angle-preserving property a conformal mapping. Note that
the conjugation map z 7→ z̄ is conformal but reverses the orientation.

Example One. Consider a “lightning fence” of height h, represented by the

line-segment [0, hi] joining 0 and hi. Let Im z = 0 represent the ground. Let
D be the domain which is equal to the upper half-plane {Im z > 0} minus
the line-segment [0, hi]. Assume that there is an electric potential u on the
domain D whose value at Im z = 0 and line-segment [0, hi] is 0 and whose
value has limit ∞ as Im z → ∞. Find u and the electric field −grad u.
Solution. Consider the map z 7→ z1 := z 2 which maps the domain D to C
minus {Im z = 0, Re z ≥ −h2 }. Let z2 = z1 + h2 so that z 7→ z2 maps D
Math 115 (2006-2007) Yum-Tong Siu 8

√
to C minus {Im z = 0, Re z ≥ 0}. Let w be the branch of z2 defined by
the numerical value of the angle in the polar representation of
√ z2 in (0, 2π).
We can now set u to be the real part of w which is equal to z 2 + h2 . The
electric field −grad u is given by
µ ¶ µ ¶
∂u ∂u ∂u ∂v
− , =− ,−
∂x ∂y ∂x ∂x
∂(u − iv) dw z
=− =− = −√ .
∂x dz z + h2
2

Example Two. Let D be the domain in C which is obtained from C by

removing the two rays [1, ∞) and (−∞, −1]. Find a bounded harmonic
function u on the domain D such that the value of u is 1 on [1, ∞) and is
−1 on (−∞, −1].
Solution. First of all, consider the map
z+1
z 7→ z1 =
z−1
so that the domain D becomes C minus [0, ∞) with (−∞, −1] going to [0, 1]
√
and [1, ∞) going to itself. Apply z1 → z2 = z1 which is the branch of
√
z1 defined by the numerical value of the angle in the polar representation
of z1 in (0, 2π). In the map z1 7→ z2 the domain C minus [0, ∞) becomes
the upper half-plane and the interval [1, ∞) goes to the union of [1, ∞) and
(−∞, −1] and the interval [0, 1] goes to [−1, 1]. We now apply the map
z2 7→ z3 = zz22 −1
+1
which maps the upper half-plane to the lower half-plane,
because ad − bc = −2 < 0 and the interval [−1, 1] goes to (−∞, 0] and the
union of [1, ∞) and (−∞, −1] goes to [0, ∞). We now take z3 7→ w = log z3
which is the branch of log z3 defined by the numerical value of the angle in
the polar representation of z3 in (−π, 0). Finally we set u = 1 + π2 Im w
which is obtained by solving for α and β so that α Im w + β becomes −1
when Im w = −π and becomes 1 when Im w = 0.

Example Three. Find a harmonic function u on the unit upper half-disk

which assumes the value 1 on the upper half-circle and the values 0 on the
interval [−1, 1].
Solution. First use a linear fractional transformation
z+1
w=
z−1
Math 115 (2006-2007) Yum-Tong Siu 9

sending the point 1 to infinity and the point 1 to the origin so that the upper
half-disk becomes the third quadrant. Thus the interval [−1, 1] is mapped
to the negative real axis and the upper half-circle is mapped to the negative
2
imaginary axis. We can set u = − 3π (−π + Im log w), where the branch of
the log is defined by the condition that the numerical value of the angle in
the polar representation of the variable is in (−π, π].

Bridge Between Exponential Map and Sine and Cosine. The map
µ ¶
1 1
w= z+
2 z

serves as the bridge between the exponential map and the sine and cosine,
hyperbolic sine, and hyperbolic cosine functions. To understand the mapping
behavior of the map, we use polar representations for z. Write z = reiθ . Then
we have
µ ¶ µ ¶ µ ¶
1 iθ 1 −iθ 1 1 1 1
w= re + re = r+ cos θ + r− sin θ.
2 r 2 r 2 r

When we use the Cartesian representation w = u + iv for w, we get

 ¡ ¢
 u = 12 r + 1r cos θ
1
¡ 1
¢
v= r− sin θ.

2 r

The circle |z| = r for 0 < r < 1 is mapped to the ellipse

u2 v2
+ =1
α2 β 2
¡ ¢ ¡ ¢
with semi-major axis α = 12 r + 1r and semi-minor axis β = − 12 r − 1r .
When the complex variable z travels along the circle |z| = r in the counter-
clockwise sense as θ goes from 0 to 2π, the complex variable w travels along
the ellipse
u2 v2
+ =1
α2 β 2
clockwise sense as θ goes from 0 to 2π.
Math 115 (2006-2007) Yum-Tong Siu 10

Now consider the circle |z| = r with r > 1. The circle |z| = r for r > 0 is
mapped to the ellipse
u2 v2
+ =1
α2 γ 2
¡ ¢ ¡ ¢
with semi-major axis α = 12 r + 1r and semi-minor axis γ = 12 r − 1r =
−β. When the complex variable z travels along the circle |z| = r in the
counterclockwise sense as θ goes from 0 to 2π, the complex variable w travels
along the ellipse
u2 v2
+ =1
α2 γ 2
also in the counterclockwise sense as θ goes from 0 to 2π.
The unit circle |z| = 1 is mapped to the closed interval [−1, 1] twice when
the ellipse
u2 v2
+ =1
α2 β 2
degenerates to [−1, 1] as β → 1 from < 1.
From the above discussion we know that the map
µ ¶
1 1
w= z+
2 z
(i) maps the upper half-disk to the lower half-plan,
(ii) maps the lower half-disk to the upper half-plan,
(iii) maps the upper half-plane minus upper half-disk to the upper half-plan,
(iv) maps the lower half-plane minus lower half-disk to the lower half-plan,
(v) maps the unit circle to the interval [−1, 1].
We now compose the map with the exponential map. Since the exponential
map z 7→ ez maps the half strip {x < 0, 0 < y < π} to the upper half-circle,
it follows that the map z 7→ eiz maps the half strip {y > 0, 0 < x < π} to the
upper half-circle. Thus
1 ¡ iz ¢
cos z = e + e−iz
2
maps the half strip {y > 0, 0 < x < π} to the lower half-plane so that
Math 115 (2006-2007) Yum-Tong Siu 11

(i) {0 < x < π, y > 0} goes to [1, ∞),

(ii) {0 < x < π, y = 0} goes to [−1, 1], and

(iii) {x = π, y > 0} goes to (−∞, −1].