More solutions manual at www.DumbLittleDoctor.

com
Thanks to the William's work!

Exercise Solutions to Complex Analysis

Complex Analysis is not that complex! Real analysis is real Math!

Note: References refer to Greene & Krantz “Function Theory of One Complex
Variable.” Most problems come from there.

Exersize 1. Prove that for any complex numbers z and w,

2
|z + w|2 = |z|2 + |w| + 2Re(z · w)
2 2 2 2
|z + w| + |z − w| = 2 |z| + 2 |w|
Solution 1. This follows by direct conputation. Write, for convenience (and abuse
of notation) z = z + z̃i and w = w + w̃i, with z, z̃, w, w̃ ∈ R. Now we have, in the
first case,
2 2
|z + w| = |(z + w) + (z̃ + w̃)i| = (z + w)2 + (z̃ + w̃)2
Now, after expanding the squares, we get from above
(z 2 + z̃ 2 ) + (w2 + w̃2 ) + 2(zw + z̃ w̃)
We know that (z 2 + z̃ 2 ) = |z| and similarly for w. Hence from above we obtain
2 2
|z| + |w| + 2(zw + z̃ w̃)
On the other hand,
Re(z · w) = Re((z + z̃i)(w − w)) = Re(zw + z̃ w̃ + (z̃w − z w̃)i) = zw + z̃ w̃
The result now follows.

The second case now follows from the first, by noting that
2 2 2 2
|z − w| = |z + (−w)| = |z| + |−w| + 2Re(z · −w)
which equals
2 2
|z| + |w| − 2Re(z · w)
The result now follows.
Exersize 2. Prove that the function
1−z
φ(z) = i
1+z
maps the set D = {z ∈ C : |z| < 1} 1−to−1 and onto the set U = {z ∈ C : Imz > 0}.
Solution 2.
Exersize 3. Compute all fifth√roots of 1 + i, all cube roots of −i, all sixth roots
of −1 and all square roots of − 3/2 + i/2.
1
 More solutions manual at www.DumbLittleDoctor.com
Thanks to the William's work!

Solution 3.
Exersize 4. Compute each of the following derivatives:

∂ 2
(a) ∂z (x − y)
∂ 2
(b) ∂z (x + y )
4
∂
(c) ∂z∂z 3
(xy 2 )
2
∂ 2
(d) ∂z∂z (zz − z 3 z + 7z)
Solution 4.
Exersize 5. Prove that if f is holomorphic on U ⊂ C, then
 2
2  ∂f 
∆(|f | ) = 4  
∂z
Solution 5.
Exersize 6. Prove that if f is C 2 , holomorphic, and nonvanishing, then log |f | is
harmonic.
Solution 6.
Exersize 7. (a) Let γ be the boundary curve of the unit disc, equipped with coun-
terclockwise orientation. Give an example of a C 1 function f on a neighborhood of
D(0, 1) such that I
f (ζ)dζ = 0
γ
but such that f is not holomorphic on any open set.
(b) Suppose that f is a continuous function on the disc D(0, 1) and satisfies
I
f (ζ)dζ = 0
∂D(0,r)

for all 0 < r < 1. Must f be holomorphic on D(0, 1)?

Solution 7. (a) Consider f (z) = 1 − zz. If z = x + iy, then we see that
f (z) = 1 − (x2 + y 2 ). Clearly the line integral of this function over the unit circle
is zero. However, the function is not holomorphic on any open set (indeed, ∂/∂z of
f is never zero on an open set).

(b) The answer here is no . Consider, for example, f (z) = 1/z 2 . This function
is holomorphic on C \ {0}. Further, (as shown in example on p. 50), the integral of
f (z) over the unit circle is 0. Now, since ∂D(0, r) is homotopic to ∂D(0, 1) where
f is holomorphic, we must have
I I
0= f (z)dz = f (z)dz
∂D(0,1) ∂D(0,r)

But f is clearly not holomorphic on D(0, 1) (there is a singularity at the origin).

Exersize 8. Calculate explicitly the integrals
I
z 3 dz
∂D(8i,2)
I
(z − i)2 dz
∂D(6+i,3)
 More solutions manual at www.DumbLittleDoctor.com
Thanks to the William's work!

Solution 8. For the first integral, our parameterization is γ(t) = 8i + 2eit with
t ∈ [0, 2π] with t increasing. Then we have, by definition,
I Z 2π
3
z dz = (8i + 2eit )3 (2ieit )dt
∂D(8i,2) 0
it
By substitution with u = 8i + 2e , one immediately obtains
Z 2π
(8i + 2eit )3 (2ieit )dt = 0
0
as expected.
For the second integral, our parametarization is
γ(t) = 6 + i + 3eit with t ∈ [0, 2π]
Now, by definition, we have
I Z 2π
2
(z − i) dz = (6 + i + 3eit − i)2 (3ieit )dt
∂D(6+i,3) 0

After expanding and simplifying, one gets

Z 2π Z 2π Z 2π
3i(6 − 2i)2 eit dt + 6i(6 − 2i) dt + 27i eit dt
0 0 0
The only integrals requiring attention are those with the integrand eit . These are
computed by integrating the real and complex parts of eit = cos(t) + i sin(t).
Exersize 9. Calculate I
1 1
dζ
2πi ∂D(0,1) ζ + 2
I
1 1
dζ
2πi ∂D(0,2) ζ + 1
z
Solution 9. For the first integral, let f (z) = z+2 . It is clear that f is holomorphic
on D(0, 1). In fact, f is holomorphic on an open set containing D(0, 1). Now, with
this, we see that
I I
1 1 1 f (ζ)
dζ = dζ
2πi ∂D(0,1) ζ + 2 2πi ∂D(0,1) ζ − 0
And by Cauchy’s Integral Formula, the latter is equal to f (0) = 0.

1
The situation is different for the second integral, since the singularity of ζ+1 lies
1
inside D(0, 2). Observe that f (ζ) = 1+ζ is holomorphic on D(0, 2) \ {−1}. Thus,
since ∂D(−1, 1) is homotopic to ∂D(0, 2) on D(0, 2) \ {−1}, we have
I I
1 1
f (ζ)dζ = f (ζ)dζ
2πi ∂D(0,2) 2πi ∂D(−1,1)
On the other hand, the integral is invariant under translation. Hence, we must have
I I I
1 1 1 1
f (ζ)dζ = f (ζ − 1)dζ = dζ
2πi ∂D(−1,1) 2πi ∂D(0,1) 2πi ∂D(0,1) ζ
And as we know, I
1
dζ = 2πi
∂D(0,1) ζ
 More solutions manual at www.DumbLittleDoctor.com
Thanks to the William's work!

Hence we have I
1 1
dζ = 1
2πi ∂D(0,2) ζ +1
Exersize 10. Let U = {z ∈ C : 1/2 < |z| < 2}. Consider the two paths γ1 (t) =
e2πit and γ2 (t) = e4πit , 0 ≤ t ≤ 1. Prove that γ1 and γ2 are not homotopic.
Solution 10. The fundamental group of the given annulus is isomorphic to the
fundamental group of the circle (the circle is a retract of the annulus). In turn, the
fundamental group of the circle is isomorphic to Z. It can be shown, by the covering
lemma (or the standard construction of the isomorphism of π1 (S 1 ) with Z), that
e4πt = 2e2πt when viewed as elements in Z. So clearly e4πt is not homotopic to e2πt
(intuitively, one is traced twise, the other is traced once).
Exersize 11. Let U1 ⊂ U2 ⊂ · · · ⊂ C be topologically simply connected open sets.
Define U = ∪j Uj . Prove that U is then topologically simply connected. What if
the hypothesis of openness is removed?
Solution 11. Say γ is a loop in U. Then γ ∗ ⊂ ∪j Uj , where by γ ∗ we mean the
trace of γ. But γ ∗ is compact, so there exists 1 ≤ m < ∞ such that γ ∗ ⊂ ∪m
j=1 Uj .
On the other hand, ∪m j=1 ⊂ Um+1 . So γ is a loop in Um+1 . Given that Um+1 is
simply connected, γ is homotopic to a point in Um+1 . Hence γ is homotopic to a
point in U. We’re done.
Exersize 12. Let S be a topologically simply connected subset of C and let f be a
complex-valued, continuous function on S. Does it follow that f (S) is topologically
simply connected?
Solution 12. The answer is no . Let S = C and let f (z) = ez . Clearly C is simply
connected and f (C) is not (it is missing the origin).
Exersize 13. Prove that the Cauchy integral formula is valid if we assume only
that F ∈ C 0 (D), and F is holomorphic on D.
Solution 13. We state and prove the claim as a theorem:

Suppose that f is holomorphic on the open disk D(z0 , r) and continuous on ∂D.
Let γ be ∂D, with counterclockwise orientation. Then for all z ∈ D(z0 , r), we have
I
1 f (ζ)
f (z) = dζ
2πi γ ζ − z

Proof: Let 0 < r0 < r, such that z ∈ D(z0 , r0 ). Let γs : [0, 1] be the standard
parametarization of ∂D(z0 , s) (counterclockwise). Now, since f is continuous on
the annulus A = {w : r0 ≤ w ≤ r}, then so is
f (ζ)
F (ζ) =
ζ −z
So F (ζ) is uniformly continuous on A. Hence for all  > 0, there exists r() with
r0 ≤ r() ≤ r such that
I  Z 
1
f (z0 + reit ) f (z0 + r()eit )
 I  
it it 

 F (ζ)dζ − F (ζ)dζ  ≤  z0 + reit − z rie − z0 + r()eit r()ie  dζ < 
  
 γ γr()  0
 More solutions manual at www.DumbLittleDoctor.com
Thanks to the William's work!

But by Cauchy’s Integral Formula, we have

I
F (ζ)dζ = 2πif (z)
γr()

Hence we must have I 

 
 F (ζ)dζ − 2πif (z) < 
 
γ
Since the left side is independent of , we conclude that
I
F (ζ)dζ = 2πif (z)
γ

as desired. 
Exersize 14. Let f be a continuous function on {z : |z| = 1}. Define, with γ =
the unit circle traversed counterclockwise,
(
f (z) if |z| = 1
F (z) = 1
H f (ζ)
2πi γ ζ=z dζ if |z| < 1

Is F continuous on D(0, 1)?

Solution 14. Take f (z) = z. Then we have
I I
1 f (ζ) 1 1
dζ = dζ
2πi γ ζ − z 2πi γ ζ(ζ − z)
If z = 0, then we obviously get
I
1
dζ = 0
γ ζ(ζ − z)
On the other hand, if z 6= 0, by partial fraction decompozition, we have
1 1/z 1/z
= −
ζ(ζ − z) ζ −z ζ
Hence we get
I I I
1 f (ζ) 1 1 1 1
dζ = dζ − dζ
2πi γ ζ −z 2πiz γ ζ −z 2πiz γ ζ
By Cauchy’s Integral Formula, both integrals above are equal to 2πi. Hence we get
0 above. So we have for all z in the open disk D(0, 1), F (z) = 0. On the other
hand, by definition, for |z| = 1, F (z) = f (z) = z 6= 0 Clearly then F (z) is not
continuous on D(0, 1). 
Exersize 15. Calculate
I
1 1
dζ
2πi γ (ζ − 1)(ζ − 2i)
where γ is the circle with center 0, radius 4, and counterclockwise orientation.
Solution 15. This integral is easily computed by partial fraction decomposition
and Cauchy’s Integral Formula. Indeed, there are constants A, B (we skip the
tedious computation) such that
1 A B
= +
(ζ − 1)(ζ − 2i) ζ − 1 ζ − 2i
 More solutions manual at www.DumbLittleDoctor.com
Thanks to the William's work!

Hence we have
I I I
1 1 1 A 1 B
dζ = dζ + dζ
2πi γ (ζ − 1)(ζ − 2i) 2πi γ ζ − 1 2πi γ ζ − 2i
Setting f1 (ζ) = A and f2 (ζ) = B and employing Cauchy’s Integral Formula, we get
I I
1 A 1 B A B
dζ + dζ = +
2πi γ ζ − 1 2πi γ ζ − 2i 2πi 2πi

Exersize 16. Calculate
ζ2 + ζ
I
1
dζ
2πi γ (ζ − 2i)(ζ + 3)
where γ is the circle with center 1, radius 5, and counterclockwise orientation.
Solution 16. This too is easily computed by Cauchy’s integral formula. Indeed,
ζ2 + ζ ζ2 ζ
= +
(ζ − 2i)(ζ + 3) (ζ − 2i)(ζ + 3) (ζ − 2i)(ζ + 3)
Hence we have
ζ2 + ζ ζ2
I I I
1 1 1 ζ
dζ = dζ + dζ
2πi γ (ζ − 2i)(ζ + 3) 2πi γ (ζ − 2i)(ζ + 3) 2πi γ (ζ − 2i)(ζ + 3)
Now one can use partial fraction decomposition for both integrals on the right, and
employ Cauchy’s Integral Formula, much like in the previous solution. We perform
the computation for only one of the integrals. We have:
A(ζ + 3) + B(ζ − 2i) = ζ 2
This leads to
4 9
A=− and B = −
2i + 3 3 + 2i
Hence we get
ζ2
I I I
1 1 A 1 B
dζ = dζ + dζ
2πi γ (ζ − 2i)(ζ + 3) 2πi γ ζ − 2i 2πi γ ζ +3
with the values A and B determined above. Cauchy’s integral formula yields the
conclusion. 
Exersize 17. Verify that
I I
1 dζ 1 dζ
=
2πi γ1 (ζ − 1)(ζ + 1) 2πi γ2 (ζ − 1)(ζ + 1)
where γ1 is ∂D(1, 1) equipped with counterclockwise orientation and γ2 is ∂D(−1, 1)
equipped with clockwise orientation.
Solution 17. Let γ˜2 trace γ2 in counterclockwise direction. Then we have
I I
1 dζ 1 dζ
=−
2πi γ˜2 (ζ − 1)(ζ + 1) 2πi γ2 (ζ − 1)(ζ + 1)
By abuse of notation, write γ2 for γ˜2 . So we must verify that
I I
1 dζ 1 dζ
dζ = − dζ
2πi γ1 (ζ − 1)(ζ + 1) 2πi γ2 (ζ − 1)(ζ + 1)
where γ1 and γ2 both have counterclockwise orientation.
 More solutions manual at www.DumbLittleDoctor.com
Thanks to the William's work!

1 1
Well, set f1 (ζ) = ζ+1and f2 (ζ) = ζ−1 Then
I I
1 dζ 1 f1 (ζ)
dζ = dζ
2πi γ1 (ζ − 1)(ζ + 1) 2πi γ1 ζ − 1
Similarly
−f1 (ζ)
I I
1 dζ 1
− dζ = dζ
2πi γ2 (ζ − 1)(ζ + 1) 2πi γ2 ζ + 1
Now employing Cauchy’s Integral Formula, we get
I
1 f1 (ζ) 1
dζ =
2πi γ1 ζ − 1 πi
and
−f1 (ζ)
I
1 1
dζ =
2πi γ2 ζ +1 πi

Exersize 18. Let γ be the unit circle equipped with clockwise orientation. For
each real number λ, give an example of a nonconstant holomorphic function F on
the annulus A = {z : 1/2 < |z| < 2} such that
I
1
F (z)dz = λ
2πi γ

Solution 18. Set F (z) = λz . Then we have

I I
1 1 λ
F (z)dz = dz = λ
2πi γ 2πi γ z

aj z j is convergent
P
Exersize 19. TRUE or FALSE: Let aj > 0, j = 1, P 2, . . . . If
j
on D(0, r) and if  > 0 is sufficiently small, then (aj + )z is convergent on
D(0, r0 ) for some 0 < r0 < r.
Solution 19. The answer is yes, and follows directly from definitions. Indeed,
 
Xm m
X m
X
lim (aj + )z j = lim  aj z j +  zj 
m→∞ m→∞
j=1 j=1 j=1

0
Now if we let 0 < r < min {1, r}, we get
m
X
 lim zj ∈ C
m→∞
j=1

aj z j converges, we deduce that

P
Since by hypothesis j
m
X
lim (aj + )z j ∈ C
m→∞
j=1


Exersize 20. Let Ω = C \ {z : |z| ≤ 1}. Determine all biholomorphic self-maps of
Ω.
 More solutions manual at www.DumbLittleDoctor.com
Thanks to the William's work!

Solution 20. For convenience, set D̃ = D(0, 1) \ {0}. Observe that

f : D̃ → Ω
given by f (z) = 1/z is a conformal map from D̃ to Ω. This is easy to see: f is
bijective, holomorphic when viewed on D̃ or on Ω, and it is its own inverse.
Now, let Aut(D̃), Aut(Ω) be the set of biholomorphic self-maps on D̃ and Ω,
respectively. These form a group, and we know that f induces a group isomorphism
fˆ:
fˆ : Aut(D̃) → Aut(Ω)
via
fˆ : h 7→ f ◦ h
So, we only need to describe maps of Aut(D̃).
Let, for convenience, D = D(0, 1), the open unit disc centered at the origin. We
know that
˜
{φ ∈ Aut(D) : φ(0) = 0} = {ωz : D → D : |ω| = 1} ≡ Aut(D)
˜
We claim that Aut(D̃) = Aut(D). To prove this claim, observe that if φ ∈
Aut(D̃), then φ, and its inverse, φ−1 are holomorphic on D̃. Now, since D̃ is a
bounded domain, it is not the case that
lim |φ(z)| = ∞
z→0

Again, since D̃ is a bounded domain, it isn’t the case that φ has an essential
singularity at zero, since then |φ| will admit arbitrarily large values near zero. Thus
φ has a removable singularity at zero, hence φ can be extended to a holomorphic
function, say Φ, on D.
We prove that Φ(0) = 0. Indeed, since D is simply connected, Φ(D) = D̃ ∪ Φ(0)
must be simply connected. This is the case only if Φ(0) = 0. That is, Φ induces a
surjective homomorphism
Φ̂ : π1 (D) → π1 (Φ(D))
and since π1 (D) = {1}, we must have π1 (Φ(D)) = {1}.
˜
But then Φ ∈ Aut(D). This proves that φ is of promissed form. Thus all
biholomorphic maps on Ω are of the form z 7→ 1/(ω/z) with |ω| = 1. So we get
z 7→ z/ω = zω. Since ω is arbitrary, we get z 7→ ωz. That is, Aut(Ω) is the set of
rotations.
Exersize 21. Let Ω ⊂ C be a bounded, simply connected domain. Let φ1 : Ω → D
and φ2 : Ω → D be conformal maps. How are φ1 , φ2 related to each other?
Solution 21. Observe that φ1 ◦ φ−12 : D → D is a conformal map. Let
z−a
φa (z) =
1 − az
with a ∈ D = D(0, 1). Then we know that every conformal map from D to D is of
the form ωφa , with |ω| = 1 (ω and a depend on the map). Hence we must have
φ1 ◦ φ−1
2 = ωφa
for some ω, a ∈ C as above. Hence
φ2 − a
φ1 = ωφa ◦ φ2 = ω
1 − aφ2
 More solutions manual at www.DumbLittleDoctor.com
Thanks to the William's work!

Exersize 22. Let {fα } be a normal family of holomorphic functions on a domain

U . Prove that {fα0 } is a normal family.
Solution
 0 22. For convenience, let A = {fα } and A0 = {fα0 }. Suppose that
fj ⊂ A is a sequence in A0 . Now, since A is a normal family, we know that for
0

every sequence {fj } ⊂ A, there is a normally converging subsequence {fjk }. That

is, there is f0 defined on U such that fjk → f0 uniformly on compacta. But then
it follows from Theorem 3.5 and its corollary that fj0k → f00 uniformly on compacta
(the compact sets here are subsets of U , of course). Hence A0 is a normal family.
(Of course, it is taken here for granted that f0 is holomorphic - this follows
immediately, since fjk are holomorphic).
Exersize 23. Suppose that {fn } is a uniformly bounded family of holomorphic
∞
functions on a domain Ω. Let {zk }k=1 ⊂ Ω and suppose that zk → z0 ∈ Ω. Assume
that limn→∞ fn (zk ) exists for k = 1, 2, . . . . Prove that the full sequence {fn }
converges uniformly on compact subsets of Ω.
Solution 23. Since there is M > 0 such that for all n, we have |fn | ≤ M , by
Montel’s theorem we do know that there exists a normally converging subsequence,
say {fnm }, converging to g. We show that in fact {fnm } = {fn }.
For a contradiction let {fj } = {fn } \ {fnm } and assume that {fj } is infinite. But
then there is a subsequence {fjl } ⊂ {fj }, again by Montel’s theorem, that converges
normally, say to h. Assume also that g 6= h. Both are obviously holomorphic.
Now, for all k the limit
lim fn (zk )
n→∞
exists. Hence
g(zk ) = lim fnm (zk ) = lim fn (zk ) = lim fjl (zk ) = h(zk )
m→∞ n→∞ l→∞

But zk → z0 , so g and h agree on an accumulating sequence. Thus we must have

g = h. We’re done.
Exersize 24. Construct a linear fractional transformation that sends the unit disc
to the half plane that lies below the line x + 2y = 4.
Solution 24. Let
φ : D(0, 1) → CH+
(where CH+ is the complex upper half plane) be given by
i + iz
φ(z) =
1−z
This is the Cayley transform. Now let
ψ(z) = z + 2i
Finally, let
−1
σ(z) = e(2π−tan (1/2))i
z
−1
(that is, rotation counterclockwise by 2π − tan (1/2)). Now let
f =σ◦ψ◦φ
It is easily verified that f is the transform we need.
 More solutions manual at www.DumbLittleDoctor.com
Thanks to the William's work!

Exersize 25. Let L be a partial differential operator of the form

∂2 ∂2 ∂2
L=a + b + c
∂x2 ∂y 2 ∂x∂y
with a, b, c constants. Assume that L commutes with rotations in the sense that
whenever f is a C 2 function on C, then Lf ◦ ρθ = L(f ◦ ρθ ) for any rotation
ρθ = eiθ z. Prove that L must be a constant multiple of the Laplacian.
Solution 25.
Exersize 26. Compute a formula analogous to the Poisson integral formula, for
the region U = {z : Imz > 0} (the upper half plane), by mapping U conformally to
the unit disc.
Solution 26. Let
z−i
φ(z) =
z+i
It’s easy to confirm that φ : HC+ → D(0, 1), where HC+ denotes the upper half
plane, is a conformal map. We already have the Poisson integral formula on the
unit disc, given by
Z 2π
1 1 − |a|2
u(a) = u(eiθ ) dθ
2π 0 |a − eiθ |2
Now, using φ−1 : D(0, 1) → HC+ , we obtain, for u harmonic on HC+ ,
u(a) = (u ◦ φ−1 )(φ(a)) for all a ∈ HC+
So we only need to apply the Poisson integral formula to (u ◦ φ−1 ).
So we obtain
2π
1 − |φ(a)|2
Z
1
u(a) = (u ◦ φ−1 )(φ(a)) = (u ◦ φ−1 )(eiθ ) dθ
2π 0 |φ(a) − eiθ |2
Now, since
i(1 + w)
φ−1 (w) =
1−w
we obtain  2
Z 2π   1 −  a−i 
iθ
1 i(1 + e ) a+i
u(a) = u 2 dθ
2π 1 − eiθ

0  a−i iθ
− e

 a+i 

Exersize 27. Compute a Possion integral formula for U = {z : |z| < 1, Imz > 0}
by mapping U conformally to the disc.
Solution 27. Let ψ1 and ψ2 be given by
 2
i+z z−i
ψ1 (z) = and ψ2 (z) =
i−z z+i
Then it is easily verified that ψ1 : U → HC+ is conformal and, as we already know
from the previous exercise, ψ2 : HC+ → D(0, 1) is conformal. So let
φ = ψ2 ◦ ψ1
be a conformal map from U to D(0, 1). Now apply the formula from the previous
exercise to this φ.
 More solutions manual at www.DumbLittleDoctor.com
Thanks to the William's work!

Exersize 28. If u is a continuous real-valued function on the closure of the unit

disc which is harmonic on the interior of the disc, then prove that u is the real part
of the holomorphic function
Z 2π iθ
1 e +z
h(z) = u(eiθ )dθ
2π 0 eiθ − z
Solution 28. Observe that
2π
eiθ + z
Z  
1 iθ
Re(h(z)) = Re u(e ) dθ
2π 0 eiθ − z
Since u is real-valued, we have
 iθ   iθ 
e +z e +z
Re iθ u(eiθ ) = Re iθ u(eiθ )
e −z e −z
An easy computation shows that
eiθ + z 1 − |z|2 + (eiθ z − eiθ z)
iθ
=
e −z |z − eiθ |2
Observe that for a, b ∈ C, it is true (as is verified by direct computation) that
ab − ab is pure imaginary. Hence from above we get
 iθ 
e +z 1 + |z|
Re iθ =
e −z |z − eiθ |2
Hence we get
2π
1 + |z|
Z
1
Re(h(z)) = u(eiθ ) dθ
2π 0 |z − eiθ |2
From the Poisson integral formula we know that this is precisely u(z).
Exersize 29. Let u be a continuous function on an open set U ⊂ C and let P ∈ U .
Suppose that u is harmonic on U \ {P }. Prove that u is in fact harmonic on all of
U.
Solution 29. Assume that U is connected (if not, we can restrict our attention to
the connected component of U that contains P ). Let R be a square centered at P ,
so small that R ⊂ U . We can cover R by four rectangles Ri , 1 ≤ i ≤ 4, such that
P ∈ ∂Ri for all i. Further, Rj ∩ Rj+1 6= ∅ for 1 ≤ j ≤ 3 and R1 ∩ R4 6= ∅.
Now, u is harmonic on each Rj . Thus, for each j, u on Rj is the real part of
some holomorphic hj = u + ivj . But because of the intersection property of Rj s
(above), Re(hj ) = Re(hj+1 ) on an open set. Hence, because of the open mapping
theorem, hj − hj+1 is an imaginary constant. We may take this constant to be zero.
So we get hi = hj for all 1 ≤ i, j ≤ 4. Hence u is the real part of a holomorphic
function h on R \ {P }. On the other hand, by construction of vj (see Theorem
1.5.1), vj are bounded. So, since u is continuous, h is bounded on R \ {P }. So
h can be extended to a holomorphic function h̃ on all of R. By continuity of u,
u = Re(h̃) on R. So u is harmonic on all of R.
Exersize 30. If h is entire, the restriction of h to the real axis is real, and the
restriction of h to the imaginary axis is imaginary, then prove that h(−z) = −h(z)
for all z.
 More solutions manual at www.DumbLittleDoctor.com
Thanks to the William's work!

Solution 30. By the Schwartz’s reflection principle we know that h(z) is holomor-
phic. Hence
h̃(z) = h(z) − h(z)
is also holomorphic. Since h is real-valued on R, we see that h̃(z) is identically zero
on R. So on C we get
h(z) = −h(z)
On the other hand, since h is pure imaginary for all pure imaginary values z, we
see that for these values of z, we have
h(−z) = −h(z)
or
h(−z) + h(z) = 0
But
z 7→ h(−z) + h(z)
is holomorphic, which is identically zero on the imaginary axis, hence is zero every-
where. Thus
h(−z) = −h(z)
for all z ∈ C.
Exersize 31. Let G and H be functions that are continuous on D and holomorphic
on D. Prove that if ReG and ReH agree on ∂D, then F −G is an imaginary constant
on D.
Solution 31. Set uG = Re(G) and uH = Re(H). Then uG , uH are harmonic on D
and continuous on D. Hence uG − uH is harmonic on D and continuous on D. By
the min/max principle for harmonic functions, uG − uH attains min/max on ∂D.
But uG = uH on ∂D. Hence min/max of uG − uH on D = 0. So uG = uH on D.
But then
F − H : D → iR
and by the open mapping theorem we conclude that F − H is identically an imag-
inary constant.
Exersize 32. If U is a domain, the boundary of which is the disjoint union of
finitely many simple closed C 1 curves, each with the property that the domain lies
only on one side of the curve at each point, then prove that every point of ∂U has
a barrier.
Solution 32. U is open, and ∂U = ∪m 1
j=1 γj where γj is a C simple, closed curve.
Hence U some m̃ (not necessarily equal to m) connected components, each with C 1
smooth, simple, closed boundary.
Now, let P ∈ γk , for some 1 ≤ k ≤ m. From the assumptions we immediately
deduce that there exists Q ∈ U c and a closed line segment I ∈ U c connected P and
Q such that I ∩ ∂U = {P }.
Now, following verbatim Example 7.7.11 (we don’t retype it here - there are no
modifications, everything follows word-for-word), one reduces the problem to the
disc D(0, 1). Now following verbatim example 7.7.9, we get the existence of barrier
at P .
Of course, this has been done only for the connected component bounded by γk .
We can extend the barrier b at P to all of U by simply defining b = −1 all the other
(closures of) connected components of U , since the closures of these are disjoint.
 More solutions manual at www.DumbLittleDoctor.com
Thanks to the William's work!

Exersize 33. Give an example of a domain U ⊂ C, U 6= C, such that no boundary

point of U has a barrier. But prove that, for any bounded domain U , there is a
point in ∂U that has a barrier.
Solution 33. To do: Develop the idea in red notepad... complete later.

On the other hand, for bounded domains U , one can always find P ∈ ∂U and
Q ∈ U c such that a closed segment I connects P and Q, I ⊂ U c and I ∩ ∂U = {P }.
Now one can follow the technique from solution 1.
Exersize 34. Let U be a bounded, holomorphically simply connected domain
with the property that each P ∈ ∂U has a barrier. Let φ be a positive, continuous
function on ∂U . Prove that there is a holomorphic function f on U such that |f |
is continuous on U and |f | restricted to ∂U equals φ.
Solution 34. Let ψ = log(φ). Then ψ is a continuous function on ∂U (since φ is
positive). Given that U has a barrier for all P ∈ ∂U , we can use Theorem 7.8.1
to conclude that there is a harmonic function h on U , continuous on U , such that
h|∂U = ψ. Since U is bounded and connected, we can cover U by finitely many
(closed), simply connected sets Bi in such a way, that the interior of each Bi lies
entirely in U . Clearly h is harmonic on each Bi ◦ . But then h is a real part of some
holomorphic h̃i on Bi ◦ (h is locally a real part of a holomorphic function on U ,
so take Bi small enough to insure this local property). Where Bi ◦ intersects Bj ◦ ,
h̃i − h̃j is a complex constant (by the open mapping theorem) which we may take
to be zero. Hence h can be regarded as the real part of some holomorphic function
h̃ on U , continuous on ∂U . Let H = eh̃ . Then H is also holomorphic on U and
continuous on ∂U . Furthermore,
|H| = eh
so on ∂U we have
|H|∂U | = eh |∂U = eh|∂U = eψ = φ
Exersize 35. Solve each of the following Dirichlet problems by using a conformal
mapping to transform the problem to one on the disc.

(a) Ω is the first quadrant. The boundary function φ equals 0 on the positive
real axis and y on the positive imaginary axis.

(b) Ω is the upper half of the unit disc. The boundary function is
φ(eiθ ) = θ, 0 ≤ θ ≤ π
1−x
φ(x) = π , −1 ≤ x ≤ 1
2

(c) Ω is the strip {z : 0 < Rez < 1}. The boundary function is identically 0 on
the imaginary axis and is identically 1 on the line {z : Rez = 1}.
Solution 35. We can transform Ω to D(0, 1) conformally via
z2 − i
z 7→ z 2 7→
z2 + i
 More solutions manual at www.DumbLittleDoctor.com
Thanks to the William's work!

It is easy to verify that this map is conformal. We can accordingly transform the
boundary conditions to get
r
i(1 + w)
ũ(w) = Im
1−w
for w ∈ D(0, 1). One can verify then that ũ corresponds to u(x, y) = y on Ω via
our conformal map above.
Observe that w = 1 corresponds to y = ∞ on the boundary of Ω.
Of course, all of this is too complicated. One could have simply found u(x, y) = y
by observation.

(b) To do: complete this later, follow computation in red notepad.

(c) This is essentially the same as (a). u(x, y) = x will do. One can find it ei-
ther by observation, or by mapping conformally the strip to the punctured disc,
and the boundary conditions accordingly.
Exersize 36. Prove that if f is a continuous, real-valued function on the interval
[a, b], a < b, and if φ is a convex function on R, then
Z b ! Z b
1 1
φ f (x)dx ≤ (φ ◦ f )(x)dx
b−a a b−a a
.
Solution 36. Let m be linear, say m(x) = αx + β. Then
Z b ! Z b
1 1
m f (t)dt = α f (t)dt + β
b−a a b−a a
since Z b
1
βdt = β
b−a a
from above we have
!
Z b Z b
1 1
m f (t)dt = m(f (t))dt
b−a a b−a a

Now, by convexity of φ, φ(x), is the supremum over all of its supporting tangent
lines m evaluated at x. Hence we have
Z b ! Z b ! Z b
1 1 1
φ f (t)dt ≥ m f (t)dt = m(f (t))dt
b−a a b−a a b−a a
where m is a supporting tangent line. Taking the supremum over all such m we
obtain immediately
Z b !
1 1
φ f (t)dt ≥ φ(f (t))dt
b−a a b−a
Exersize 37. If |z| < R, then prove that
∞  2n n 
Y R + z2 R
=
n=0
R2n R−z
Solution 37.
 More solutions manual at www.DumbLittleDoctor.com
Thanks to the William's work!

Exersize 38. If bn > 1 for all n, then prove that

Y
bn
n

converges if and only if

X
log(bn ) < ∞
n

Solution 38. Suppose first that

X
log(bn ) < ∞
n

Observe then that !

X
exp log(bn ) <∞
n
hence Y
ebn = M < ∞
n
It is clearly true, for bn > 1, that
ebn > bn
hence we have
N
Y N
Y
M> ebn > bn
n n
for all N , proving convergence of the product.

Conversely, suppose that the product

Y Y
(1 + (bn − 1)) = bn
n n

converges. Observe that if for infinitely many n, bn > 2 then, keeping in mind that
for all n bn > 1, the product
Y
bn
n
will diverge. Hence for all n > N , N large enough, bn < 2. Without loss of
generality we may thus assume that for all n we have bn < 2. Then bn − 1 < 1,
hence we know that !
N N
1X Y
exp (bn − 1) ≤ bn
2 n n
proving that
X
(bn − 1)
n
must also converge. Observe that x − 1 − log(x) is an increasing function on [1, ∞),
and is equal to zero at x = 1. This shows that for all bn , bn − 1 ≥ log(bn ), proving
convergence of
X
log(bn )
n
 More solutions manual at www.DumbLittleDoctor.com
Thanks to the William's work!

Exersize 39. Determine whether

∞  
Y 1
1 + (−1)n
n=2
n

converges. Do the same for

∞  
Y 1
1 + (−1)n √
n=2
n

Solution 39. Suppose that N is even, say N = 2K. For convenience, write
 
1
bn = 1 + (−1)n
n
Observe that
N
Y
bn = b2 b3 × b4 b5 × · · · × bK−1 bK
n
A simple calculation shows that
bn bn+1 = 1 for all n ≥ 2
Hence
N
Y
bn = 1
n=2
On the other hand,
N
Y +1 N
Y
bn = bN +1 bn = bN +1 → 1 as N → ∞
n=2 n=2

This shows that the odd and the even subsequences both converge to 1, showing
that the product converges to 1.

For the second calculation, we’ll follow the same technique. Denote by PN the
even subsequence of the partial products (that is, N = 2K). Hence PN +1 is the
odd subsequence. We show that both converge to the same number.

A simple calculation will show that

  
1 1
1+ √ 1− √ <1
2n 2n + 1
for all n ≥ 2. Hence the even subsequence, PN is decreasing in N , thus converges.
Also,
PN − PN +1 = PN (1 − bN +1 ) ≤ 1 − bN +1 → 0 as N → ∞
 
where bn = 1 + (−1)n √1n . Hence

PN − PN +1 → 0
and we already know that PN converges. Hence so does PN +1 , hence so does the
product.
 More solutions manual at www.DumbLittleDoctor.com
Thanks to the William's work!

Exersize 40. Calculate

∞  
Y 1
1− 2
n=2
n
explicitly.
Solution 40.
P
Exersize 41. Prove that if j |aj | < ∞ and if σ is any permutation of the positive
integers, then
∞
Y ∞
Y
(1 + aj ) = (1 + aσ(j) )
j=1 j=1
P P
Solution 41. Since n |aj | converges, we know that n |aσ(j) | converges. But
then Y
(1 + |aσ(j) |)
n
converges and hence the product
Y
(1 + aσ(j) )
n

also converges. But then for any given  > 0 we have

     
 ∞ ∞   ∞ ∞ 
Y Y   Y Y 

 (1 + aj ) − (1 + aσ(j) ) = 
   (1 + aj ) − 1 −
  (1 + aσ(j) ) − 1 

j=N j=M   j=N j=M 
   
 ∞   ∞ 
Y  Y 
≤ (1 + aj ) − 1 + 
  (1 + aσ(j) ) − 1 < 2
j=N  j=M 
for N, M large (this follows from convergence of both products). Now, for fixed
M by taking N so large that {1, 2, . . . , N } includes {σ(1), σ(2), . . . , σ(M )}, we’re
done.
Exersize
P 42. Prove that, in general, the result of Exercise
P 8 fails if the hypothesis
j |aj | < ∞ is replaced by a weaker condition like aj < ∞.

Solution 42. Consider aj = (−1)j 1j for j ≥ 2. Let σ be such that σ(j) gives
the jth even number if j is not a multiple of, say, 5, and σ(j) gives the j/5th odd
number if j is a multiple of 5; keeping in mind that the ”first” even number is 2
and the ”first” odd number is 3 (since j ≥ 2). So, for instance:
σ(2) = 2; σ(3) = 4; σ(4) = 6; σ(5) = 3; σ(6) = 10; . . .
It isn’t hard to see that in this case
Y
(1 + aσ(j) )
j

diverges. On the other hand, as we have see above,

Y
(1 + aj )
j
converges.
 More solutions manual at www.DumbLittleDoctor.com
Thanks to the William's work!

Exersize 43. For which z does

∞
Y n
(1 + z 3 )
n=1
converge?
Solution 43. Clearly the product converges for all |z| < 1. If z is real and |z| > 1,
the product clearly diverges. If z = −1, the limit doesn’t exist. If z = 1 the product
diverges. Now, a simple calculation will show that
n
(1 + z 3 ) = (1 + z)(1 − z + z 2 )n
hence Y n Y
(1 + z 3 ) = (1 + z) (1 − z + z 2 )n
n n
and the latter product converges only when 0 6= |1 − z + z 2 | < 1.
Exersize 44. If U is a domain, the boundary of which is the disjoint union of
finitely many simple closed C 1 curves, each with the property that the domain lies
only on one side of the curve at each point, then prove that every point of ∂U has
a barrier.
Solution 44. U is open, and ∂U = ∪m 1
j=1 γj where γj is a C simple, closed curve.
Hence U some m̃ (not necessarily equal to m) connected components, each with C 1
smooth, simple, closed boundary.
Now, let P ∈ γk , for some 1 ≤ k ≤ m. From the assumptions we immediately
deduce that there exists Q ∈ U c and a closed line segment I ∈ U c connected P and
Q such that I ∩ ∂U = {P }.
Now, following verbatim Example 7.7.11 (we don’t retype it here - there are no
modifications, everything follows word-for-word), one reduces the problem to the
disc D(0, 1). Now following verbatim example 7.7.9, we get the existence of barrier
at P .
Of course, this has been done only for the connected component bounded by γk .
We can extend the barrier b at P to all of U by simply defining b = −1 all the other
(closures of) connected components of U , since the closures of these are disjoint.
Exersize 45. Give an example of a domain U ⊂ C, U 6= C, such that no boundary
point of U has a barrier. But prove that, for any bounded domain U , there is a
point in ∂U that has a barrier.
Solution 45. To do: Develop the idea in red notepad... complete later.

On the other hand, for bounded domains U , one can always find P ∈ ∂U and
Q ∈ U c such that a closed segment I connects P and Q, I ⊂ U c and I ∩ ∂U = {P }.
Now one can follow the technique from solution 1.
Exersize 46. Let U be a bounded, holomorphically simply connected domain
with the property that each P ∈ ∂U has a barrier. Let φ be a positive, continuous
function on ∂U . Prove that there is a holomorphic function f on U such that |f |
is continuous on U and |f | restricted to ∂U equals φ.
Solution 46. Let ψ = log(φ). Then ψ is a continuous function on ∂U (since φ is
positive). Given that U has a barrier for all P ∈ ∂U , we can use Theorem 7.8.1
 More solutions manual at www.DumbLittleDoctor.com
Thanks to the William's work!

to conclude that there is a harmonic function h on U , continuous on U , such that

h|∂U = ψ. Since U is bounded and connected, we can cover U by finitely many
(closed), simply connected sets Bi in such a way, that the interior of each Bi lies
entirely in U . Clearly h is harmonic on each Bi ◦ . But then h is a real part of some
holomorphic h̃i on Bi ◦ (h is locally a real part of a holomorphic function on U ,
so take Bi small enough to insure this local property). Where Bi ◦ intersects Bj ◦ ,
h̃i − h̃j is a complex constant (by the open mapping theorem) which we may take
to be zero. Hence h can be regarded as the real part of some holomorphic function
h̃ on U , continuous on ∂U . Let H = eh̃ . Then H is also holomorphic on U and
continuous on ∂U . Furthermore,
|H| = eh
so on ∂U we have
|H|∂U | = eh |∂U = eh|∂U = eψ = φ
Exersize 47. Solve each of the following Dirichlet problems by using a conformal
mapping to transform the problem to one on the disc.

(a) Ω is the first quadrant. The boundary function φ equals 0 on the positive
real axis and y on the positive imaginary axis.

(b) Ω is the upper half of the unit disc. The boundary function is
φ(eiθ ) = θ, 0 ≤ θ ≤ π
1−x
φ(x) = π , −1 ≤ x ≤ 1
2

(c) Ω is the strip {z : 0 < Rez < 1}. The boundary function is identically 0 on
the imaginary axis and is identically 1 on the line {z : Rez = 1}.
Solution 47. We can transform Ω to D(0, 1) conformally via
z2 − i
z 7→ z 2 7→
z2 + i
It is easy to verify that this map is conformal. We can accordingly transform the
boundary conditions to get
r
i(1 + w)
ũ(w) = Im
1−w
for w ∈ D(0, 1). One can verify then that ũ corresponds to u(x, y) = y on Ω via
our conformal map above.
Observe that w = 1 corresponds to y = ∞ on the boundary of Ω.
Of course, all of this is too complicated. One could have simply found u(x, y) = y
by observation.

(b) To do: complete this later, follow computation in red notepad.

(c) This is essentially the same as (a). u(x, y) = x will do. One can find it ei-
ther by observation, or by mapping conformally the strip to the punctured disc,
and the boundary conditions accordingly.
 More solutions manual at www.DumbLittleDoctor.com
Thanks to the William's work!

Exersize 48. Prove that if f is a continuous, real-valued function on the interval

[a, b], a < b, and if φ is a convex function on R, then
Z b ! Z b
1 1
φ f (x)dx ≤ (φ ◦ f )(x)dx
b−a a b−a a
.
Solution 48. Let m be linear, say m(x) = αx + β. Then
Z b ! Z b
1 1
m f (t)dt = α f (t)dt + β
b−a a b−a a
since Z b
1
βdt = β
b−a a
from above we have
!
Z b Z b
1 1
m f (t)dt = m(f (t))dt
b−a a b−a a

Now, by convexity of φ, φ(x), is the supremum over all of its supporting tangent
lines m evaluated at x. Hence we have
Z b ! Z b ! Z b
1 1 1
φ f (t)dt ≥ m f (t)dt = m(f (t))dt
b−a a b−a a b−a a
where m is a supporting tangent line. Taking the supremum over all such m we
obtain immediately
Z b !
1 1
φ f (t)dt ≥ φ(f (t))dt
b−a a b−a