134 Chapter 6 Conformal Mappings

Solutions to Exercises 6.1

1. An analytic function f(z) is conformal where f

(z) = 0. If f(z) =
z
2
+1
e
z
, then f

(z) = e
z
(z
2
+
2z 1). We have
f

(z) = 0 z
2
2z + 1 = 0 z = 1.
Thus f is conformal at all z = 1.
5. The function f(z) = z +
1
z
is analytic at all z = 0, and then f

(z) = 1
1
z
2
. So
f

(z) = 0 1
1
z
2
= 0 z
2
= 1 z = 1.
Thus f(z) is conformal at all z = 0, 1.
9. The function sin z is entire. Its derivative cos z is nonzero except for z =

2
+k. Hence f(z) = sin z
is conformal at all z =

2
+k. In particular, f(z) is conformal at z = 0, +i a, i. At these points,
f(z) rotates by an angle arg cos z and scales by a factor of | cos z|. For z = 0, we have cos 0 = 1.
Thus f rotates by an angle arg 1 = 0 and scales by a factor of 1. Thus at z = 0, sin z acts like the
identity map z z.
For z = + i a, we have cos( + i a) = cos(ia) = cosh a. Thus f rotates by an angle
arg (cosh a) = and scales by a factor of cosh a.
For z = i, we have cos i = cosh . Thus f rotates by an angle arg (cosh ) = 0 and scales by
a factor of cosh .
13. The lines x = a and y = b are clearly orthogonal and they intersect at the point a + ib. Their
images by a mapping f(z) are two curves that intersect at the point f(a + ib). The image curves
will be orthogonal at the point f(a + ib) if f(z) is conformal at a + ib. Hence it is enough to check
that f(z) is analytic and f

(z) = 0 at z = a + ib, in order to conclude that the image curves are

orthogonal at f(a + ib). In the case of f(z) = e
z
, the image curves will always be orthogonal,
because f

(z) = 0 for all z. Indeed, the image of the line x = a (or z = a + iy, < y < )
is the circle w = e
a+iy
= e
a
e
iy
, with center at 0 and radius e
a
. The image of the line y = b (or
z = x + ib, < x < ) is the ray at angle b, w = e
x
e
ib
. The ray and the circle intersect at right
angle at the point e
a+ib
.
17. (a) We have
J(
1
z
) =
1
2
_
1
z
+
1
1
z
_
=
1
2
_
z +
1
z
_
= J(z).
(b) Fix R > 1 and let S
R
= {z : |z| = R, 0 Arg z}. Write the image of a point z = Re
i
on S
R
as w = u + iv. Then
w = J(Re
i
) =
1
2
_
Re
i
+
1
R
e
i
_
=
1
2
_
R(cos +i sin ) + +
1
R
(cos i sin )
_
=
1
2
_
R+
1
R
_
cos + i
1
2
_
R
1
R
_
sin .
So
u =
1
2
_
R+
1
R
_
cos and v =
1
2
_
R
1
R
_
sin ;
u
1
2
_
R+
1
R
_ = cos and
v
1
2
_
R
1
R
_ = sin.
Section 6.1 Basic Properties 135
As varies from 0 to , w traces the upper part of the ellipse
u
2
_
1
2
_
R+
1
R
_
2
+
v
2
_
1
2
_
R
1
R
_
2
= cos
2
+ sin
2
= 1.
21. Let be the slit plane, C \ {z : z 0}, f(z) = Log z, and z
n
= 1 + i
(1)
n
n
(n = 1, 2, . . .).
Note that f

(z) =
1
z
= 0 for all z . Also, is straightforward to check that f is one-to-one on .
However, as we now show, the sequence {z
n
} converges to a point on the boundary, namely 1, but
f(z
n
) does not converge. Indeed,
Log (z
n
) = Log
_
1 +i
(1)
n
n
_
= ln

1 + i
(1)
n
n

+i Arg
_
1 +i
(1)
n
n
_
= ln
_
1 +
1
n
2
+ i
n
.
If n = 2k is even, z
2k
approaches 1 from the upper half-plane. So Log (z
2k
) tends to ln(1)+i = i,
as k . If n = 2k + 1 is odd, z
2k+1
approaches 1 from the lower half-plane. So Log (z
2k+1
)
tends to ln(1) i = i, as k . So Log (z
n
) does not converge as n .
136 Chapter 6 Conformal Mappings
Solutions to Exercises 6.2
1. Consider the linear fractional transformation (LFT)
(z) = i
1 z
1 +z
.
(a) We have
(1) = 0, (0) = i, (i) = i
1 i
1 + i
= 1.
(b) Let L
1
denote the line through z
1
= 1 and z
2
= 0, and L
2
the line through z
2
= 0 and z
3
= i.
We know that the image of a line by a LFT is either a line or a circle. So to determine whether
[L
j
] is a line or a circle, it suces to check whether the images of three points on L
j
are colinear.
Another way to determine the image of a line by a LFT is to check whether the image is bounded
or unbounded. If it is unbounded, then it is necessarily a line. For example, the point z = 1 is on
L
1
. Since (1) = , we conclude that [L
1
] is a line through the points (z
1
) = 0 and (z
2
) = i;
that is, [L
1
] is the imaginary axis in the w-plane.
Since L
2
is perpendicular to L
1
and is conformal at z = 0, the point of intersection, [L
1
] and
[L
2
] must be orthogonal at [0], their point of intersection. Now [L
2
] goes through the points
i and 1. In order for [L
2
] to be perpendicular to the imaginary axis, [L
1
], it must be a circle
that goes through the points 1 and i. To get a third point on this circle, take z = i on L
2
. Then
(i) = i
1+i
1i
= 1.
5. The inverse of the LFT
w = (z) = i
1 z
1 +z
=
i iz
1 +z
=
iz + i
z + 1
is given by (2), where a = b = i, c = d = 1:
z = (w) =
z i
z + i
.
Let us check the images of w
j
:
(w
1
) = (0) = 1 = z
1
;
(w
2
) = (i) = 0 = z
2
;
(w
3
) = (1) = i = z
3
.
9. Suppose c = 0 and consider the equation
(z) = z or
az + b
cz + d
= z (z =
d
c
.
This equation is equivalent to
cz
2
+ (d a)z b = 0,
which has at most two solutions. Thus in this case we have at most two distinct xed points. If
c = 0, the LFT takes the form (z) = z +, where = 0. This linear function is either the identity
( = 1) or has one xed point, which is obtained by solving z = z + , z =

1
. In all case, the
LFT has at most two xed points.
13. The circle in the rst gure is centered at 1 +i and has radius 1. To map it to the circle shown
in the second gure, use the translation
f
1
(z) = z 1 i.
Section 6.2 Linear Fractional Transformations 137
To map the inside of the unit circle to its outside, and its outside to its inside, as shown in the
mapping from the second to the third gure, use an inversion
f
2
(w
1
) =
1
w
1
.
Note that if |z| = 1, then |f
2
(z)| = |1/z| = 1, so f
2
maps the unit circle to the unit circle. Since it
takes boundary to boundary, and 0 to , it will act as we claimed. Finally, to map the unit disk to
the upper half-plane, use the LFT
f
3
(w
2
) = i
1 w
2
1 +w
2
(see Example 1(a)). The desired mapping is
w = f
3
f
2
f
1
(z) = i
1 w
2
1 +w
2
= i
1
1
w1
1 +
1
w1
= i
w
1
1
w
1
+ 1
= i
z 1 i 1
z 1 i + 1
= i
z 2 i
z i
.
17. The rst mapping is f
1
(z) = sin z (see Example 4, Sec. 1.6.). The second mapping is the LFT
f
2
(w
1
) =
iw1
i+w1
(see Example 1(b)). Thus,
w = f
2
f
1
(z) =
i w
1
i + w
1
=
i sinz
i + sinz
.
21. Consider the LFT
f
1
(z) = i
1 z
1 +z
.
We know from Example 1(a) that f
1
takes the unit disk onto the upper half-plane. What does it do
to the upper semi-disk? The lower boundary of the semi-disk, the interval [1, 1] is perpendicular
to the upper semi-circle at the point 1. Since f
1
is conformal at z = 1, the images of the interval
[1, 1] and semi circle intersect at right angle. Since they both go through the point 1, which is
mapped to , we conclude that these images are perpendicular lines. Using f
1
(0) = i and f
1
(i) = 1,
we conclude that [1, 1] is mapped to the upper half of the imaginary axis, and the semi circle is
mapped to the right half of the real axis. Finally, testing the image of an interior point, say i/2,
we nd that f
1
(
i
2
) =
4
5
+
3
5
i, which is a point in the rst quadrant. With this we conclude that
the upper semi-disk is mapped to the rst quadrant, since boundary is mapped to boundary and
interior points to interior points.
To go from the rst quadrant to a horizontal strip, we use a logarithmic mapping, because the
logarithm maps the punctured plane onto a fundamental strip of width 2 (see Sec. 1.7.) For our
purpose, it is easy to check that Logz will work: The semi-line (0, infty is mapped to the line
(, ) and the semi-line, z = iy, 0 < y < , is mapped to Log (iy) = lny +i

2
, which describes
the horizontal line shown in the gure. The composition mapping is w = Log (i
1z
1+z
).
25. (a) Take 1 < < 1 and let

(z) =
z
1z
. Since is real, we have

(z) =
z
1z
. We know
from Example 3, Sec. 3.7, that

maps the unit disk onto itself, taking the unit circle onto itself. As
it is explained in Example 4, in order to center C
1
, it is enough to choose so that

(a) =

(b).
Equivalently,
a
1 a
=
b
1 b
;
(a )(1 b) = (b )(1 a);
(a +b)
2
2(1 +ab) + a +b = 0
2
2
1 +ab
a + b
+ 1 = 0.
138 Chapter 6 Conformal Mappings
Note that a + b = 0. Solving this quadratic equation in , we obtain with roots

1
=
1 + ab
a +b
+

_
1 + ab
a +b
_
2
1 and
2
=
1 +ab
a + b

_
1 +ab
a + b
_
2
1.
We next proceed to show that
1
> 1 and 0 <
2
< 1. So we must choose
2
in constructing

.
(b) We have |a| < 1, 0 < b < 1 and 0 |a| < b < 1. Since 1 b > 0, if 0 a < 1, then multiplying
1 b by a, we get 1 b > a(1 b). If 1 < a < 0, then 1 b > a(1 b), because the right
side is negative, while the left side is positive. In all cases, 1 + ab > a + b > 0 or 1 + ab > a + b.
Consequently,
_
1+ab
a+b
_
2
> 1; hence the discriminant of the quadratic equation in in part (a) is
positive and so we have two distinct roots,
1
and
2
.
(c) From the inequality 1 +ab > a +b > 0, we conclude that
1+ab
a+b
> 1 and so
1
> 1. The product
of
1

2
= 1, as can be checked directly or by using a well-known property that the product of the
roots of the quadratic equation ax
2
+bx+c = 0 is
c
a
. So since 1 <
1
, we conclude that 0 <
2
< 1,
in order to have the equality
1

2
= 1.
(d) Now arguing exactly as we did in Example 4, we conclude that (z) =
z
1z
with =
1+ab
a+b

_
_
1+ab
a+b
_
2
1 will map C
2
onto C
2
, C
1
onto a circle centered at the origin with radius r = (b),
and the region between C
2
and C
1
onto the annular region bounded by [C
1
] and the unit circle.
29. From the gure, the rst mapping is a translation:
f
1
(z) = z
3
2
i.
The second mapping is a linear fractional transformation that maps the shaded half plane to the
disk of radius 2 and center at the origin. This mapping is the inversion
f
2
(w
1
) =
1
w
1
.
To see this, recall that the mapping is conformal except at 0. So it will map perpendicular lines and
circles to the same. The image of a circle centered at the origin, with radius R, is another circle
centered at the origin with radius 1/R. Thus the circle in the second gure is mapped onto the circle
with center at the origin and radius 2. The region inside the circle is mapped outside the region of
the image circle and vice versa. In particular, the shaded half-plane is mapped inside the circle with
radius 2. We must show that the image of the lower half-plane, which is also mapped inside the
circle of radius 2, is the the smaller disk shown in the 3rd gure. To this end, we check the image of
the horizontal boundary of this lower half-plane, which is the line through
3
2
i. This line is mapped
to a circle (the image of the line is contained in a bounded region, so it has to be a circle and not a
line). Moreover, this circle makes an angle 0 with the real axis (the real axis is mapped to the real
axis by an inversion). Since f
2
() = 0 and f
2
(
3
2
i) =
2
3
i, we conclude that the image of the lower
clear half-plane is the clear smaller disk as shown in the 3rd gure.
To construct the last mapping of the sequence, we will appeal to the result of Exercise 25. Let us
prepare the ground for the application of this result, by rotating the picture in the 3rd gure by

2
,
and then scaling by
1
2
. This amounts to multiplying by
1
2
e

2
or
i
2
. So we introduce the mapping
f
3
(w
2
) =
1
2
e

2
w
2
. This will map the outer circle in the 3rd gure to the unit circle and the inner
circle to a circle of radius
1
6
and center at
1
6
. In the notation of Exercise 25, we have a = 0 and
b =
1
3
. According to Exercise 25, the mapping that will center the inner circle is

(w
3
) =
w3
1w3
,
where is the smaller of the roots of

2
2
1
1
3
+ 1 = 0
2
6 + 1 = 0.
Section 6.2 Linear Fractional Transformations 139
Thus = 3 2

2. Composing all the mappings together, we obtain:

f(z) =
w
3

1 w
3
=
i
2
w
2

1
i
2
w
2
=
iw
2
2
2 +iw
2
=
i
w1
2
2 + i

w1
=
i 2w
1
2w
1
+ i
=
i 2(z
3
2
i)
2(z
3
2
i) +i
=
2z i(1 3)
2z +i(3 +)
=
2(3 2

2)z i(1 3(3 2

2))
2z i2

2
=
(3 2

2)z + i(4 3

2)
z i

2
.
33. (a) We can reduce this problem to the one that is solved in Exercise 31 by rst mapping the unit
disk onto the upper half-space by using the linear fractional transformation (z) = i
1z
1+z
(Example
1(a)). If f is a one-to-one analytic mappings of the upper half-plane onto the unit disk, then f
is a one-to-one mapping of the unit disk onto itself. By Exercise 31, f = c

, where |c| = 1,
|| < 1, and

is the linear transformation given by (6). Consequently, f = c

1
, where
1
is the inverse of . By Proposition 1 (with a = i, b = i, c = d = 1), we nd
1
(z) =
zi
zi
=
iz
i+z
.
(This is the mapping in Example 1(b).) To get the explicit form of f, we compute
f(z) = c

1
(z) = c

1
(z)
1
1
(z)
= c
iz
i+z

1
iz
i+z
= c
i(1 ) z(1 + )
i(1 ) + z(1 + )
,
where || < 1 and |c| = 1. It is not dicult to show that every such mapping takes the upper half-
plane to the unit disk. Thus these are all the one-to-one analytic mappings of the upper half-plane
to the unit disk.
(b) Let (z) =
iz
i+z
be the mapping given by Example 1(b). Then maps the upper half-plane onto
the unit disk. If g is a one-to-one analytic mappings of the upper half-plane onto itself, then g
is a one-to-one analytic mappings of the upper half-plane onto the unit disk. By part (a), we have
g(z) = f(z) = c
i(1 ) z(1 + )
i(1 ) + z(1 + )
,
for some unimodular c and constant , with || < 1. Thus g(z) =
1
f(z). To get the explicit
formula of the LFT g, we recall that

1
(z) = i
1 z
1 +z
.
140 Chapter 6 Conformal Mappings
So
g(z) =
1
f(z) = i
1 f(z)
1 + f(z)
= i
1 c
i(1)z(1+)
i(1)+z(1+)
1 + c
i(1)z(1+)
i(1)+z(1+)
= i
i(1 c + c) +z(1 + +c +c)
i(1 + c c) +z(1 + c c)
= i
i( c) +z( + c)
i( +c) +z( c)
,
where = 1+ and = 1, |c| = 1, and || < 1. Conversely, any linear fractional transformation
g of this form can be written as g(z) =
1
f(z) (just reverse the preceding steps), and so, by the
mapping properties of f and
1
, g maps the upper half-space one-to-one and onto itself.
37. For z S, z = x +iy, we have
A(x
2
+ y
2
) +Bx +Cy + D = 0, where B
2
+C
2
4AD > 0.
Let
f(z) =
1
z
=
1
x + iy
=
x iy
x
2
+y
2
.
Then
f[S] =
_
u + iv : u =
x
x
2
+y
2
, v =
y
x
2
+ y
2
_
,
where x and y satisfy
A(x
2
+ y
2
) +Bx +Cy + D = 0, where B
2
+C
2
4AD > 0.
We want to prove that u and v satisfy a similar identity with possibly dierent values of the coe-
cients. Let us consider such a relationship with u and v:
E(u
2
+ v
2
) + Fu + Gv + H = E
x
2
+ y
2
_
x
2
+y
2
_
2
+
Fx
x
2
+y
2

Gy
x
2
+y
2
+ H
=
E
x
2
+y
2
+
Fx
x
2
+y
2

Gy
x
2
+ y
2
+
H(x
2
+ y
2
)
x
2
+ y
2
=
1
x
2
+y
2
_
H(x
2
+ y
2
) +Fx Gy +E
_
.
Choose H = A, F = B, G = C, and E = D. Then, for all (x, y) = (0, 0) in S, the last expression
is equal to 0. Hence (u, v) satisfy
D(u
2
+ v
2
) +Bu Cv +A = 0.
Moreover, B
2
+ C
2
4AD > 0.
41. (a) Consider a linear fractional transformation
f(z) =
az + b
cz + d
Section 6.2 Linear Fractional Transformations 141
and suppose that a = 0, c = 0, and ad bc = 0. We will not treat the easier cases in which a = 0
or c = 0 or ad bc = 0. We have
f(z) =
a
c
z +
b
a
z +
d
c
=
a
c
_
1 +
b
a

d
c
z +
d
c
_
= A+
B
z + C
,
where A, B, and C are constants with A = 0 and B = 0. Let T
1
(z) = z + C, (z) =
1
z
and
T
2
(z) = Bz + A. Then f = T
2
T
1
, where T
1
and T
2
are linear transformations, and is an
inversion.
(b) Since linear mappings and inversions map lines and circles to the same, it follows that a lin-
ear fractional transformation does the same, because it can be written as a composition of linear
mappings and an inversion.
142 Chapter 6 Conformal Mappings
Solutions to Exercises 6.3
1. Transform the problem to the upper half-plane as in Example 1, by using the linear fractional
transformation
(z) = i
1 z
1 +z
.
The LFT takes the unit disk onto the upper-half plane, the upper semi-circle onto the positive
real axis, and the lower semi-circle onto the negative real axis. Thus the problem in the upper half
of the w-plane becomes:
U = 0, U() = 70 if > 0, U() = 50 if < 0.
The solution in the w-plane is U(w) = a Arg (w) +b, where a and b must be chosen in according to
the boundary conditions: a Arg (w) +b = 70, when Arg = 0 and a Arg (w) +b = 50, when Arg = .
Thus b = 70 and a =
20

. Hence U(w) =
20

Arg w + 70 and so
u(z) = U((z)) =
20

Arg
_
i
1 z
1 + z
_
+ 70.
5. The mapping f
1
(z) = z
4
takes the shaded region in Fig. 14 onto the upper semi-disk of radius 1.
The upper semi-disk is mapped onto the rst quadrant by
f
2
(w
1
) = i
1 w
1
1 +w
1
.
Thus the mapping
w = (z) = i
1 z
4
1 + z
4
takes the shaded region in Fig. 14 onto the rst quadrant. It is also a conformal mapping, being the
composition of such mappings. We now determine the image of the boundary. The rays at angle 0
and /4 are mapped onto the interval [1, 1] by f
1
. The quarter of a circle is mapped to the upper
semi-circle by f
1
. The mapping f
2
takes the interval [1, 1] onto the upper half of the imaginary
axis, and the semi-circle onto the right half of the real axis.
Thus the problem in the rst quadrant of the w-plane becomes:
U = 0, U() = 100 if > 0, U(i) = 0 if > 0.
The solution in the w-plane is U(w) = a Arg (w) +b, where a and b must be chosen in according to
the boundary conditions: a Arg (w)+b = 100, when Arg = 0 and a Arg (w)+b = 0, when Arg =

2
.
Thus b = 100 and a =
200

. Hence U(w) =
200

Arg w + 100 and so

u(z) = U((z)) =
200

Arg
_
i
1 z
4
1 +z
4
_
+ 100.
9. To solve the given problem, we can proceed as in Example 6 and make the necessary changes.
A much quicker way is based is to use the solution in Example 6 and superposition, as follows.
Let u
1
(z) denote the solution in Example 6. Let u
2
(z) = 100. It is clear that u
2
is harmonic
for all z. Thus u
2
u
1
is harmonic in the shaded region of Fig. 18. On the real axis, we have
u
2
(z) u
1
(z) = 100 0 = 100. On the upper semi-circle, we have u
2
(z) u
1
(z) = 100 100 = 0.
Thus 100 u
1
is the solution, where u
1
is the solution in Example 6.
13. This problem is very similar to Example 3. The rst step is to map the region to an annular
region bounded by concentric circles. This can be done by using the linear fractional transformation
(z) =
4z 1
4 z
.
Section 6.3 Solving Dirichlet Problems with Conformal Mappings 143
To derive this mapping, use the result of Exercise 27, Sec. 6.2, with a = 0, b =
8
17
, and =
1
4
. Then
(z) = 1
4
(z) =
z
1
4
1
1
4
=
4z 1
4 z
.
Using , we map the outer circle to the unit circle and the inner circle to the circle of radius
1
4
and
center at 0. As in Example 3, the solution is
U(w) = 100 + 100
ln|w| + ln 4
ln(1/4)
= 100 100
ln|w| + ln 4
ln 4
= 100
ln|w|
ln 4
.
Thus the solution is
u(z) =
100
ln 4
ln

4z 1
4 z

.
17. (a) The solution of the Dirichlet Problem in Figure 25(a) is obtained by applying the Poisson
formula (3), with f(s) = s if 1 < s < 1, f(s) = 1 if s < 1 and f(s) = 1 if s > 1:
u(x + iy) =
y

_
1
1
s
(x s)
2
+ y
2
ds +
y

_
1

1
(x s)
2
+ y
2
ds +
y

_

1
ds
(x s)
2
+ y
2
= I
1
+I
2
+ I
3
.
We compute each integral separately:
I
1
=
y

_
1
1
x s
(x s)
2
+ y
2
ds +
y

_
1
1
x
(x s)
2
+ y
2
ds
=
y

1
2
ln
_
(x s)
2
+ y
2
_

1
1
+
xy

_
1
1
ds
(s x)
2
+y
2
=
y
2
_
ln
_
(x + 1)
2
+ y
2
_
ln
_
(x 1)
2
+ y
2
_
+
x
y
_
1
1
ds
_
sx
y
_
2
+ 1
=
y
2
ln
(x + 1)
2
+ y
2
(x 1)
2
+ y
2
+
x

_
1
s=1
du
u
2
+ 1
(Let u =
s x
y
.)
=
y
2
ln
(x + 1)
2
+ y
2
(x 1)
2
+ y
2
+
x

tan
1
_
s x
y
_

1
s=1
=
y
2
ln
(x + 1)
2
+ y
2
(x 1)
2
+ y
2
+
x

_
tan
1
_
1 + x
y
_
+ tan
1
_
1 x
y
_
_
.
Similarly, we nd
I
2
=
y

_
1

1
(x s)
2
+y
2
ds
=
1
2
+
y

tan
1
_
1 + x
y
_
;
I
3
=
y

_

1
ds
(x s)
2
+ y
2
=
1
2

y

tan
1
_
1 x
y
_
.
Thus
u(x + iy) =
y
2
ln
(x + 1)
2
+ y
2
(x 1)
2
+ y
2
+
1 +x

tan
1
_
1 +x
y
_
+
1 +x

tan
1
_
1 x
y
_
.
144 Chapter 6 Conformal Mappings
(b) To solve the problem in Fig. 25(b), we reduce to the problem in Fig. 25(a) by using the mapping
sin z. This yields the solution
u(x +iy) =
Im(sin z)

ln
( Re (sin z) + 1)
2
+ ( Im(sin z))
2
( Re (sin z) 1)
2
+ ( Im(sin z))
2
+
1 + Re (sin z)

tan
1
_
1 + Re (sin z)
Im(sin z)
_
+
1 + Re (sin z)

tan
1
_
1 Re (sin z)
Im(sin z)
_
=
cos xsinhy

ln
(sin xcosh y + 1)
2
+ (cos xsinh y)
2
sin xcosh y 1)
2
+ (cos xsinh y)
2
+
1 + sinxcosh y

tan
1
_
1 + sin xcosh y
cos xsinhy
_
+
1 + sin xcosh y

tan
1
_
1 sin xcosh y
cos xsinhy
_
,
where we have used the fact that Re sin z = sin xcosh y and Im sin z = cos xsinhy.
25. To solve the Dirichlet problem in the upper half-plane with boundary function
f(x) =
1
x
4
+ 1
,
we appeal to the Poisson formula, which gives: for y > 0 and < x < ,
u(x, y) =
y

1
1 +s
4
1
(s x)
2
+ y
2
ds.
We evaluate the integral by using the residue techniques of Sec. 5.3. Let
f(z) =
1
1 + z
4
1
(z x)
2
+y
2
.
By Proposition 2, Sec. 5.3, we have
u(x, y) =
y

1
1 + s
4
1
(s x)
2
+y
2
ds = 2i
y

j
Res (f, z
j
) = 2iy

j
Res (f, z
j
),
where the sum extends over the residues of f in the upper half-plane. We have
z
4
= 1 z
4
= e
i
z
1
= e
i

4
, z
2
= e
i
3
4
, z
3
= e
i
5
4
, z
4
= e
i
7
4
.
Only z
1
and z
2
are in the upper half-plane. At these points, f has simple poles and the residues
there are computed with the help of Proposition 1(ii) and (iii), Sec. 5.1: We have
Res (f, z
1
) =
1
4z
3

z=z1
1
(z x)
2
+y
2

z=z1
=
1
4e
i
3
4
1
(e
i

4
x)
2
+ y
2
= R
1
;
Res (f, z
2
) =
1
4z
3

z=z2
1
(z x)
2
+ y
2

z=z2
=
1
4e
i
9
4
1
(e
i
3
4
x)
2
+ y
2
=
1
4e
i

4
1
(e
i
3
4
x)
2
+ y
2
= R
2
.
The equation (z x)
2
+ y
2
= 0 has two roots z = x iy, as you can check. Only x + iy is in the
upper half-plane, since y > 0. The residue of f at x +iy is:
Res (f, x +iy) =
1
z
4
+ 1

z=x+iy
1
2(z x)

z=x+iy
=
1
(x +iy)
4
+ 1
1
2iy
= R
3
.
Section 6.3 Solving Dirichlet Problems with Conformal Mappings 145
Thus,
u(x, y) = 2iy
_
1
4e
i
3
4
1
(e
i

4 x)
2
+y
2
+
1
4e
i

4
1
(e
i
3
4 x)
2
+ y
2
+
1
(x +iy)
4
+ 1
1
2iy
_
.
This answer should be real-valued, because the Poisson integral involves real-valued functions only.
This is not obvious, but the integral can be simplied using Mathematica to yield an expression that
is obviously real-valued.
29. We appeal to the Poisson formula (the second version in (3), which gives: for y > 0 and
< x < ,
u(x, y) =
y

sin a(x s)
1
s
2
+ y
2
ds.
This integral follows from Example 1, Sec. 5.4, because
sin a(x s)
1
s
2
+ y
2
=
sin axcos as
s
2
+ y
2

cos axsinas
s
2
+ y
2
.
Thus
_

sin a(x s)
1
s
2
+ y
2
ds =
_

sin axcos as
s
2
+ y
2
ds
_

cos axsin as
s
2
+ y
2
ds
= sin ax
_

cos as
s
2
+ y
2
ds cos ax
=0
..
_

sin as
s
2
+ y
2
ds
= sin ax
_

cos as
s
2
+ y
2
ds,
where one of the integrals is 0 because the integrand is an odd function. We now appeal to Example
1, Sec. 5.4 (do the cases a > 0 and a < 0 separately):
u(x, y) =
y

sin ax
_

cos as
s
2
+ y
2
ds =
y sin ax

y
e
|a|y
= e
|a|y
sin ax.
33. The conformal mapping that takes the rst quadrant to the upper half-plane is w = (z) = z
2
.
It also maps the interval [0, 1] in the z-plane to the interval [0, 1] in the w-plane. To determine
the boundary values in the w-plane, we compose the inverse of with the boundary values in the
z-plane. This gives the value 0 for all points on the real axis outside the interval [0, 1], and for the
w in the interval [0, 1], the boundary value is (

w)
2
= w. We thus obtain the boundary values in
the w-plane: f(w) = w if 0 < w < 1 and 0 otherwise. The solution in the w-plane is obtained by
applying the Poisson integral function. We will use coordinates (, ) in the w-plane. Thus
U(, ) =

f(s)
( s)
2
+
2
ds =

_
1
0
s
( s)
2
+
2
ds
=

tan
1
_
s

_
+
1
2
ln(
2
+ (s )
2
)
_
1
0
=

_
tan
1
_
1

_
+ tan
1
_

_
_
+

2
ln

2
+ (1 )
2

2
+
2
,
where the integral in s was evaluated by similar methods as in Exercise 17. The solution in the
z-plane is obtained by replacing by Re (z) = Re z
2
= x
2
y
2
and by Im(z) = Imz
2
= 2xy.
Thus
u(x, y) =
x
2
y
2

_
tan
1
_
1 (x
2
y
2
)
2xy
_
+ tan
1
_
x
2
y
2
2xy
_
_
+
xy

ln
(2xy)
2
+
_
1 (x
2
y
2
)
_
2
(2xy)
2
+ (x
2
y
2
)
2
146 Chapter 6 Conformal Mappings
Exercises 6.4
1. The outside angles at w
1
, w
2
, and w
3
are
1
=

2
,
2
= , and
3
=

2
, respectively. By (7), we
have (z 1)
1
2
(z + 1)
1
2
= i(1 z
2
)
1
2
. Applying (2), we nd
f = A
_
(z 1)

1
2
(z + 1)

1
2
z dz +B = A
_
1
(z 1)
1
2 (z + 1)
1
2
z dz +B
= A
_
1
i(1 z
2
)
1
2
z dz +B =
A
2i
_
1 z
2
+ B.
Using f(1) = 0, we obtain B = 0. Using f(0) = 1, we obtain A = 2i. Thus f(z) =

1 z
2
.
5. The outside angles at w
1
and w
2
are
1
=

2
and
2
=

2
, respectively. By (7), we have
(z 1)
1
2
(z + 1)
1
2
= i(1 z
2
)
1
2
. Applying (2), we nd
f = A
_
(z 1)
1
2
(z + 1)
1
2
dz + B = iA
_
(1 z
2
)
1
2
dz + B
=
iA
2
_
z
_
1 z
2
+ sin
1
z
_
+B,
where in evaluating the last integral we used integration by parts. Using f(1) = 1, we obtain
iA
2

2
+ B = 1. Using f(1) = 1, we obtain
iA
2

2
+ B = 1. Thus B = 0 and A = i
4

. So
f(z) =
2

_
z
_
1 z
2
+ sin
1
z
_
.
9. The solution of this problem is identical to the solution of Example 7, where instead of the
angle

4
you should use an arbitrary angle . Thus the streamlines are the images by the mapping
f(z) = z

of the streamlines in the upper half-plane. The streamlines are thus parametrized by
w = (x +iy
0
)

, where y
0
is arbitrary but xed positive number, and < x < . The graph of
the streamlines is like the one shown in Fig.15.
Section 6.5 Greens Functions 147
Solutions to Exercises 6.5
1. According to (10), Greens function for the shaded region is
G(z, ) = ln

() (z)
1 (z)()

,
where is a one-to-one conformal mapping of the region onto the unit disk. Instead of doing the
computations from scratch, we use our knowledge of Greens function for the upper-half plane, from
Example 2. The rotation (z) = e
i

2
z takes onto the upper-half plane. In Cartesian coordinates,
this rotation is expressed as follows:
(x + iy) = i(x +iy) = y + ix.
Thus rho will map the point (x, y) to the point (y, x) and the point (s, t) to the point (t, s).
Applying this rotation then Greens function from Example 2, we obtain Greens function for :
G(z, ) =
1
2
ln

(y +t)
2
+ (x s)
2
(y +t)
2
+ (x +s)
2

=
1
2
ln

(x s)
2
+ (y t)
2
(x +s)
2
+ (y t)
2

,
where z = x + iy, = s +it, x, s > 0.
5. As in Exercise 1, we will use Greens function for the upper half-plane, but rst we must map
onto the upper half-plane using a one-to-one conformal mapping. It is straightforward to check
that f(z) = e
z
takes onto the upper-half plane. In Cartesian coordinates:
f(x +iy) = e
x
cos y +ie
x
sin y.
Thus f takes the point (x, y) to the point (e
x
cos y, e
x
sin y) and the point (s, t) to the point
(e
s
cos t, e
s
sin t). Applying f then Greens function from Example 2, we obtain Greens function for
:
G(z, ) =
1
2
ln

(e
x
cos y e
s
cos t)
2
+ (e
x
sin y e
s
sin t)
2
(e
x
cos y e
s
cos t)
2
+ (e
x
sin y +e
s
sin t)
2

,
where z = x + iy, = s +it, < x, s < , 0 < y, t < .
9. (a) Greens function in the rst quadrant, , is derived in Exercise 3: For z and in ,
G(z, ) = ln

|z | |z +|
|z | |z +|

.
Write z = x +iy, = s + it, where x, y, s, and t are positive. Then
G(z, ) =
1
2
ln
_
(x s)
2
+ (y t)
2
__
(x +s)
2
+ (y +t)
2
_
_
(x s)
2
+ (y t)
2
__
(x +s)
2
+ (y +t)
2
_
=
1
2
ln
_
(x s)
2
+ (y t)
2
_
+
1
2
ln
_
(x +s)
2
+ (y + t)
2
_

1
2
ln
_
(x s)
2
+ (y +t)
2
_

1
2
ln
_
(x +s)
2
+ (y +t)
2
_
.
The boundary of consists of two half-lines: the positive real axis, and the upper part of the
imaginary axis. On the imaginary axis, the normal derivative is minus the derivative with respect
148 Chapter 6 Conformal Mappings
to s. Thus, on the imaginary axis,

n
G(z, ) =

s
G(z, )

s=0
=
1
2

s
ln
_
(x s)
2
+ (y t)
2
_

1
2

s
ln
_
(x + s)
2
+ (y +t)
2
_
+
1
2

s
ln
_
(x s)
2
+ (y + t)
2
_
+
1
2

s
ln
_
(x + s)
2
+ (y + t)
2
_

s=0
=
(x s)
(x s)
2
+ (y t)
2

(x +s)
(x +s)
2
+ (y +t)
2

(x s)
(x s)
2
+ (y + t)
2
+
(x + s)
(x + s)
2
+ (y + t)
2

s=0
=
x
x
2
+ (y t)
2

x
x
2
+ (y +t)
2

x
x
2
+ (y +t)
2
+
x
x
2
+ (y + t)
2
=
2x
x
2
+ (y t)
2

2x
x
2
+ (y +t)
2
.
By a similar argument, we nd that, on the real axis,

n
G(z, ) =

t
G(z, )

t=0
=
2y
(x s)
2
+y
2

2y
(x + s)
2
+ y
2
.
From (9), we have
u(z) =
1
2
_

u()

n
G(z, )ds.
We have to gure our the symbols in this integral. For the part of the boundary that is on the
imaginary axis, ds = dt, and u() = g(t). For the part of the boundary that is on the real axis,
u() = f(s), and ds = ds (bad notation: on the left, ds stands for element of arc length; on the
right ds stands for element of integration with respect to the s variables.) Using the values of the
normal derivative, we get
u(z) =
y

_

0
f(s)
_
1
(x s)
2
+ y
2

1
(x + s)
2
+y
2
_
ds
+
x

_

0
g(t)
_
1
x
2
+ (y t)
2

1
x
2
+ (y + t)
2
_
dt.
(b) Consider the special case in which g(t) = 0. Call the solution in this u
1
. From part (a), we have
u
1
(z) =
y

_

0
f(s)
_
1
(x s)
2
+y
2

1
(x + s)
2
+ y
2
_
ds, z = x + iy.
We will show how to derive this solution by reducing the problem to a Dirichlet problem in the
upper half-plane. Indeed, consider the Dirichlet problem in the upper half-plane with boundary
values u(x, 0) = f(x) if x > 0 and u(x, 0) = f(x) if x < 0. Thus the boundary function is the
odd extension of f(x) to the entire real line. We will use the same notation for the odd extension
as for the function f. Then, by the Poisson integral formula on the real line,
u(x, y) =
y

f(s)
(x s)
2
+ y
2
ds, y > 0, < x < .
Section 6.5 Greens Functions 149
To determine the values of u
1
on the upper part of the imaginary axis, we set x = 0 and get
u(0, y) =
y

f(s)
s
2
+ y
2
ds = 0,
because f is odd. So u is harmonic in the upper half-plane; is equal to 0 on the imaginary axis; and
is equal to f(x) on the x-axis. Therefore, its restriction to the rst quadrant solves the Dirichlet
problem with boundary values 0 on the imaginary axis and f(x) on the positive real axis. Thus u
agrees with u
1
in the rst quadrant. Because f is odd, we can write
u(x, y) =
y

f(s)
(x s)
2
+ y
2
ds
=
y

_
0

f(s)
(x s)
2
+ y
2
ds +
y

_

0
f(s)
(x s)
2
+y
2
ds
=
y

_
0

f(s)
(x +s)
2
+y
2
ds +
y

_

0
f(s)
(x s)
2
+ y
2
ds
=
y

_

0
f(s)
(x + s)
2
+y
2
ds +
y

_

0
f(s)
(x s)
2
+ y
2
ds
=
y

_

0
f(s)
_
1
(x + s)
2
+ y
2
+
1
(x s)
2
+ y
2
_
ds,
which is precisely the formula for u
1
that we obtained earlier by using Greens function in the rst
quadrant.
(c) The case in Figure 19 in which f(s) = 0 is very similar to the case that we treated in (b). You
just have to interchange x and y, s and t. The solution in this case is
u
2
(x, y) =
x

_

0
g(t)
_
1
(y + t)
2
+x
2
+
1
(y t)
2
+ x
2
_
dt,
which is again the formula that you obtain by setting f = 0 in part (a).
(d) The Dirichlet problem in Fig. 19 can be written as the sum of two Dirichlet problems: the
problem in which g = 0, and the problem in which f = 0. The rst problem is solved in (b)
(the solution is u
1
), and the second problem is solved in (c) (the solution is u
2
). It is easy to
straightforward to verify that u = u
1
+u
2
is the solution of the problem in Fig. 19. Adding the two
formulas that we have for u
1
and u
2
, we obtain the formula for u that we derived in (a).
The point of this problem is that we were able to solve the Dirichlet problem in the rst quadrant
by reducing the problem to two problem in the upper half-plane and then using the Poisson integral
formula on the real line.
150 Chapter 6 Conformal Mappings
Solutions to Exercises 6.6
1. To determine a Neumann function for an unbounded region, we will apply Proposition 3. For
that purpose, we must describe a one-to-one conformal mapping of the region onto the upper
half-plane. It is clear that a translation by 1, followed by a rotation by an angle of

2
will do the
job here. So (z) = e

2
i
(z 1) = i(z 1) and the Neumann function is
N(z, ) = ln|(z) ()| + ln |(z) ()|
= ln|i(z 1) i( 1)| + ln | i(z 1) i( 1)|
= ln|i(z )| + ln|iz +i i + i|
= ln|z | + ln |z + 2|
=
1
2
ln
_
(x s)
2
+ (y t)
2
_
+
1
2
ln
_
(y +s 2)
2
+ (x +t)
2
_
,
where z = x + iy, = s +it, x, s > 1, and < y, t < .
5. We apply Proposition 3, with (z) = sin z (see Example 4, Sec. 1.6). We also use the fact that
sin(x + iy) = sin xcosh y + i cos xsinhy (see (16), Sec. 1.6). Then
N(z, ) = ln|(z) ()| + ln |(z) ()|
= ln| sin z sin| + ln|sin z sin |
=
1
2
ln
_
(sin xcosh y sin s cosh t)
2
+ (cos xsinhy cos s sinh t)
2
_
+
1
2
ln
_
(cos xsinh y + sin s cosh t)
2
+ (sin xcosh y cos s sinh t)
2
_
,
where z = x + iy, = s +it, 1 < x, s < 1, and 0 < y, t.