Complex Analysis

Fall 2007 Homework 5: Solutions

1.5.22 If z = x + iy then
z 4 = (x4 − 6x2 y 2 + y 4 ) + i(4x3 y − 4xy 3 )

so that u = Re(z 4 ) = x4 − 6x2 y 2 + y 4 and v = Im(z 4 ) = 4x3 y − 4xy 3 . We find that

∂2u ∂ 3 2
= 12x2 − 12y 2


2
= 4x − 12xy
∂x ∂x
and
∂2u ∂
−12x2 y + 4y 3 = −12x2 + 12y 2


2
=
∂y ∂y
so that
∂2u ∂2u
∇2 u = + 2 = 12x2 − 12y 2 + −12x2 + 12y 2 = 0.



∂x 2 ∂y
Since u is a polynomial in x and y its second-order partials are continuous and ∇2 u = 0, u
is harmonic. We leave the analogous computation involving v to the student.

1.5.32 Claim: If u and v are functions defined on an open set A, u and v satisfy the
Cauchy-Riemann equations and u is harmonic on A, then v is a harmonic conjugate of u on
A.
We are given
∂u ∂v
=
∂x ∂y
∂u ∂v
= − .
∂y ∂x
Since u and v already satisfy the Cauchy-Riemann equations, to show that v is conjugate to
u it suffices to prove that v is also harmonic. Since u is harmonic, it’s second-order partial
derivatives are all continuous (by definition). The Cauchy-Riemann equations above then
imply the same is true of v’s second-order partials. So all we need to do to prove that v
is harmonic is verify that ∇2 v = 0. Appealing to the Cauchy-Riemann equations again we

1
have
2 ∂2v ∂v
∇v = 2
+ 2
∂x ∂y
   
∂ ∂v ∂ ∂v
= +
∂x ∂x ∂y ∂y
   
∂ ∂u ∂ ∂u
= − +
∂x ∂y ∂y ∂x
2 2
∂ u ∂ u
= − +−
∂x∂y ∂y∂x
= 0

since the continuity of the second-order mixed partials implies their equality. Hence, v is
harmonic. As noted above, this completes the proof.
Notice that the only facts we used were that u and v were related by the Cauchy-Riemann
equations and that u had continuous second-order partial derivatives. In fact, if these are the
only assumptions that we make then we can, in fact, prove that both u and v are harmonic
and that they are conjugate.
Now on to the problem at hand. Since we are given that u is harmonic on the disk the
claim shows that it is enough to verify that u and v satisfy the Cauchy-Riemann equations.
To (hopefully) clarify things slightly, let me alter the notation of the problem. Let’s define:
Z y Z x
∂u ∂u
v(x, y) = c + (x, t) dt − (s, y0 ) ds
y0 ∂x x0 ∂y

By the Fundamental Theorem of Calculus (the real-variable version) we have

Z y Z x
∂v ∂ ∂u ∂ ∂u
(x, y) = (x, t) dt − (s, y0 ) ds
∂y ∂y y0 ∂x ∂y x0 ∂y
∂u
= (x, y)
∂x
since the second expression is not a function of y. This is one of the Cauchy-Riemann
equations. Similar reasoning gives
Z y
∂v ∂ ∂u ∂u
(x, y) = (x, t) dt − (x, y0 ).
∂x ∂x y0 ∂x ∂y

Since ∂u
∂x
is continuous we can pass the partial differentiation under the integral sign1 to
obtain Z y 2
∂v ∂ u ∂u
(x, y) = 2
(x, t) dt − (x, y0 ).
∂x y0 ∂x ∂y
1 If you’d like to see a proof of this fact let me know and I can provide you with the details.

2
 2 2
Since u is harmonic, ∂∂xu2 = − ∂∂yu2 . Substituting this into the above and again using the
Fundamental Theorem of Calculus we have
Z y 2
∂v ∂ u ∂u
(x, y) = 2
(x, t) dt − (x, y0 )
∂x y0 ∂x ∂y
Z y 2
∂ u ∂u
= − 2
(x, t) dt − (x, y0 )
y0 ∂y ∂y
t=y !
∂u  ∂u
= − (x, t) − (x, y0 )
∂y t=y0 ∂y
∂u ∂u ∂u
= − (x, y) + (x, y0 ) − (x, y0 )
∂y ∂y ∂y
∂u
= − (x, y)
∂y
which is the other half of the Cauchy-Riemann equations.

1.6.2

(a) Choose a branch of log w that is analytic on the real axis and define 3z = ez log 3 . Then
3z is analytic where the branch of log is and
d z
3 = (log 3)3z
dz
there.
(b) Choose a ray R emanating from the origin in the complex plane and let log w denote
any branch of the logarithm that is analytic on C \ R. Denote by the set R − 1 the
ray translated to the left one unit in the plane (so that it emanates from −1). Then
log(z + 1) is analytic on C \ (R − 1) and
d 1
log(1 + z) =
dz z+1
there.
(c) Choose a ray R emanating from the origin in the complex plane and let log z denote
any branch of the logarithm that is analytic on C \ R. Then z ( 1 + i) = e(1+i) log z is
analytic on C \ R and
d 1+i
z = (1 + i)z i = (1 + i)ei log z
dz
there

3
(d) Choose a ray R emanating from the origin in the complex plane and let log z denote
√ 1
any branch of the logarithm that is analytic on C \ R. Then z = z 1/2 = e 2 log z is
analytic on C \ R and
d√ 1 1
z = z −1/2 = √
dz 2 2 z
there.
(e) Choose a ray R emanating from the origin in the complex plane and let log z denote
√ 1
any branch of the logarithm that is analytic on C \ R. Then 3 z = z 1/3 = e 3 log z is
analytic on C \ R and
d√ 1 1 1
3
z = z −2/3 = 2/3 = √ 2
dz 3 3z 3 ( 3 z)
there.

1.6.4

(a) Let log z denote the principal branch of the logarithm so that log 1 = 0. Then log z is
analytic at z = 1 and
 
log z d  1 
lim = log z  =  = 1.
z→1 z − 1 dz z z=1
z=1

If, however, we chose any other branch of the logarithm that was defined at z = 1, then
log 1 = 2nπi for some nonzero integer n and this can be used to show that the limit
above does not exist in this case.
(b) If f (z) = z, then
z−1 f (z) − 1
lim = lim
z→1 z − 1 z→1 z − 1
0
which is the definition of f (1). However, we know that f (z) is not analytic anywhere
so that f 0 (1) does not exist. Hence, the limit in question does not exist.

1.6.8 For simplicity, let’s assume that both u and v have continuous first-order partial
derivatives. Then the function f = u + iv is analytic on A and so are both (f (z))2 and ef (z) .
One easily checks that

Re(f (z))2 = u2 − v 2 = u1
Im(f (z))2 = 2uv = v1

4
and
Re ef (z) = eu cos v = u2
Im ef (z) = eu sin v = v2 .
SInce they are the real and imaginary parts of analytic functions, we find that the pairs
u1 , v1 and u2 , v2 satisfy the Cauchy-Riemann equations on A.
A comment is in order here. The continuity assumption on the partial derivatives of u
and v is not really necessary. One can use the multi-variate chain rule to show that if g(z)
is analytic and u and v satisfy the Cauchy-Riemann equations then so, too, do the real and
imaginary parts of g(u + iv).

1.6.10
√
(a) We let w denote the principal branch of the square root. That is, we choose the
√ 1
branch of the argument with −π ≤ arg w < π and set w = e 2 log w . With this choice
√
w is analytic on the set√C \{u+iv | u ≤ 0, v = 0}. If we let R = {u+iv | u ≤ 0, v = 0}
and f (z) = z 3 − 1 then z 3 − 1 is analytic on f −1 (C \ R) = C \ f −1 (R). Our goal is to
determine f −1 (R).
We have z ∈ f −1 (R) iff f (z) = z 3 − 1 ∈ R. This can happen iff z 3 ∈ Rp +1 =
{u + iv | u ≤ 1, v = 0} = R ∪ [0, 1]. If x ∈ R then z = x has the solutions z = |x|eiθ
3 3

where θ = π/3, π, 5π/3. It follows that z 3 belongs to the ray R iff z belongs to one of
the rays emanating from 0 with angle π/3, π or 5π/3 relative to the real axis. SImilarly,
if x ∈ [0, 1] then z 3 = x implies that z belongs to one of the line segments of length 1
making an angle of 0, 2π/3 or 4π/3 with the real axis. If you sketch these six regions
together you will see that f −1 (R) can be described as follows: it consists of the rays
emanating from 0 with angle π/3, π or 5π/3 relative to the real axis together with√ their
length 1 extensions across the origin. If we call the set just described S, then z 3 − 1
is analytic on C \ S. Now go draw a picture of this set.
√
(b) Dealing with
√ this function is much√easier. Since sin w is entire, sin z is analytic
wherever z is. So, for example, if z is √ a branch analytic on C \ R (where R is any
ray emanating from the origin) then sin z is analytic on the same set.

2.1.2

(a) This is the hardest of the bunch. The curve γ consists of two line segments which are
parameterized by
γ1 (t) = it, t ∈ [0, 1]
and
γ2 (t) = (1 − t)i + t(i + 2), t ∈ [0, 1].

5
 We have
x(γ1 (t)) = Re γ1 (t) = 0
γ10 (t) = i

and
x(γ2 (t)) = Re γ2 (t) = 2t
γ20 (t) = −i + i + 2 = 2
so that
Z Z Z
x dz = x dz + x dz
γ γ1 γ2
Z 1 Z 1
= (0)(i) dt + (2t)(2) dt
0 0
Z 1
= 4 t dt
0
= 2.

(b) This is the easiest of the bunch. A complex antiderivative for z 2 + 2z + 3 is F (z) =
z3
+z 2 +3z so that by the fundamental theorem of calculus (for complex path integrals)
3
Z
29 32 13 16 32
z 2 + 2z + 3 dz = F (2 + i) − F (1) = + i− = + i.
γ 3 3 3 3 3

(c) We can parametrize γ by

γ(t) = 1 + 2eit , t ∈ [0, 2π].
Since γ 0 (t) = 2ieit we have
Z Z 2π Z 2π
1 1 it
dz = 2ie dt = i dt = 2πi.
γ z−1 0 (1 + 2eit ) − 1 0

2.1.4 We begin by noting that

1 1 −1/2 1/2
= = + .
z2 − 2z z(z − 2) z z−2
Consequently Z Z Z
1 1 1 1 1
2
dz = − dz + dz.
γ z − 2z 2 γ z 2 γ z−2

6
Since γ lies entirely in the first and fourth quadrants, we can choose a branch of log z that
is analytic on γ. Since any such branch has 1/z as its derivative, the first integral above is
zero by the Fundamental Theorem of Calculus for path integrals. On the other hand, if we
parameterize γ by γ(t) = 2 + eit , t ∈ [0, 2π] then
Z Z 2π Z 2π
1 1 it
dz = ie dt = i dt = 2πi.
γ z −2 0 (2 + eit ) − 2 0

Combining this with our earlier observations we have

Z
1 1 1
2
dz = − (0) + 2πi = πi.
γ z − 2z 2 2

2.1.6 For convenience we assume that all paths involved are C 1 . For a complete proof, one
must break each path in question into C 1 pieces and apply the special cases that we prove
below.

(a) If γ is parameterized by γ : [a, b] → C then, using properties of real definite integrals,

we have
Z Z b
(c1 f + c2 g) = (c1 f (γ(t)) + c2 g(γ(t)))γ 0 (t) dt
γ a
Z b Z b
0
= c1 f (γ(t))γ (t) dt + c2 g(γ(t))γ 0 (t) dt
a a
Z Z
= c1 f + c2 g.
γ γ

(b) If γ is parameterized by γ : [a, b] → R then −γ is given, on the same interval, by

(−γ)(t) = γ(a + b − t). By the chain rule (−γ)0 (t) = −γ 0 (a + b − t) and therefore
Z Z b
f = f (−γ(t))(−γ)0 (t) dt
−γ a
Z b
= − f (γ(a + b − t))γ 0 (a + b − t) dt.
a

If we make the (real) change of variable s = a + b − t then ds = −dt and

Z b Z a
0
− f (γ(a + b − t))γ (a + b − t) dt = f (γ(s))γ 0 (s) ds
a b
Z b
= − f (γ(s))γ 0 (s) ds
Za
= − f
γ

7
 which is what we wanted to show.
(c) If γ1 is parameterized by γ1 : [a, b] → C and γ2 by γ2 : [b, c] → C then we know that
(
γ1 (t) , t ∈ [a, b]
(γ1 + γ2 )(t) =
γ2 (t) , t ∈ [b, c]

and hence (
0 γ10 (t) , t ∈ [a, b]
(γ1 + γ2 ) (t) =
γ20 (t) , t ∈ [b, c].
It follows that
Z Z c
f = f ((γ1 + γ2 )(t))(γ1 + γ2 )0 (t) dt
γ1 +γ2 a
Z b Z c
0
= f ((γ1 + γ2 )(t))(γ1 + γ2 ) (t) dt + f ((γ1 + γ2 )(t))(γ1 + γ2 )0 (t) dt
a b
Z b Z c
= f (γ1 (t))γ10 (t) dt + f (γ2 (t))γ20 (t) dt
Za Z b

= f+ f.
γ1 γ2

2.1.8

(a) We can parameterize the line segment from 0 to 1+i by γ(t) = t(1 + i), t ∈ [0, 1]. In
2
this case we find that γ 0 (t) = 1 + i and γ(t) = t2 (1 − i)2 so that
Z Z 1
2
z dz = t2 (1 − i)2 (1 + i) dt
γ 0
Z 1
2
= 2(1 − i) t2 dt = (1 − i)
0 3

(b) Since the path in question consists of two line segments, we must parameterize and
integrate over each separately. The first segment is given by γ1 (t) = t, t ∈ [0, 1].
Therefore Z Z 1
2 1
z dz = t2 dt = .
γ1 0 3

8
 The second segment can be parameterized as γ2 (t) = 1 + it, t ∈ [0, 1] and hence
Z Z 1
2
z dz = (1 − it)2 i dt
γ2 0
Z 1
= 2t + i(1 − t2 ) dt
0
1
t2
 
2

= t +i t− 
2 
0
i
= 1+ .
2
Finally, we have Z Z Z
2 2 4 i
z dz = z dz + z 2 dz = + .
γ γ1 γ2 3 2

If z 2 were the derivative of an analytic function then, by the Fundamental Theorem, the
values of all path integrals of z 2 beginning at 0 and ending at 1 + i would be the same. Since
the computation above shows that this is not the case, we conclude that z 2 has no complex
antiderivative.

2.1.10 If |z| = 2 then by the reverse triangle inequality we have

|z 2 + 1| ≥ |z 2 | − |1| = |z|2 − 1 = 22 − 1 = 3
so that  
 1  |1| 1
 z 2 + 1  = |z 2 + 1| ≤ 3 .
 

Hence, if C is the arc of the circle |z| = 2 lying in the first quadrant then
Z 
 dz  1 π
 2 + z 2  ≤ 3 `(C) = 3 .
 
C

2.1.12 On the set C\{z | Re z ≤ 0} the principal branch of the logarithm is an antiderivative
of 1/z. By the Fundamental Theorem of Calculus it follows immediately that
Z
1
dz = 0
γ z

for any closed curve γ in C \ {z | Re z ≤ 0}.