Suggested solution of HW3

z+a
Chapter 2 Q11: (a) By scaling, we assume R = 1. Consider g = f ◦ φ where φ(z) = .
1 + āz
By Cauchy formula, we get
˛ ˛
1 g(z) 1 f ◦ φ(z)
f (a) = g(0) = dz = dz.
2πi ∂B(1) z 2πi ∂B(1) z

We now perform change of coordinate w = φ(z). We have

dz 1 − |a|2 w−a
= 6= 0 and z = .
dw (1 − āw)2 1 − āw

Also, φ map ∂D to ∂D bijectively. Thus,

˛
1 f (w) 1 − |a|2
f (a) = · dw
2πi ∂B(1) w − a 1 − āw
ˆ 2π
1 f (eiθ ) 1 − |a|2 iθ
= · · e dθ
2π 0 eiθ − a 1 − āeiθ
ˆ 2π
1 f (eiθ )(1 − |a|2 )
= dθ
2π 0 (eiθ − a)(e−iθ − ā)
ˆ 2π  iθ 
1 e +a
= f (eiθ )Re iθ dθ.
2π 0 e −a

(b) Follows from direct computation.

∂u
Chapter 2 Q12: (a) Let g(z) = 2 . Since u is harmonic,
∂z

∂ ∂ ∂
u = 0 =⇒ g = 0.
∂ z̄ ∂z ∂ z̄

Hence, g(z) is holomorphic. As D is simply connected, g has a primitive F on D,

∂F ∂u
= g(z) = 2 . Since F is analytic, using the fact that ∂F∂ z̄ = 0, we have
∂z ∂z

∂F ∂F ∂F ∂u ∂u ∂u
= = −i =2 = −i .
∂z ∂x ∂y ∂z ∂x ∂y

From ∂x F = ∂x Re(F ) + i∂x Im(F ), we know that

∂x [Re(F ) − u] = 0

and similarly from ∂y F ,

∂y [Re(F ) − u] = 0.

That show that Re(F ) = u + C for some real constant C (viewing it as a function
on R2 ). By considering F − C, we may assume C = 0. Suppose f1 , f2 are two
analytic functions so that
Re(fi ) = u.

1
 Then h(z) = f1 (z) − f2 (z) is analytic and by CR equation
 
∇(Im(h)) = ∂x Im(h), ∂y Im(h) = 0.

Therefore, they only differ by a real constant.

(b) By part (a), there exists a holomorphic function f such that Re(f ) = u. Applying
the formula in Q11 on a slightly smaller ball B(r) and takes real part on both
side. The result follows by taking r → 1.
Remark : One can also show it by proving the Poisson integral formula agrees with
u on the boundary. By Maximum Principle, u agree with the Poisson formula in the
interior as well. It can be seen that the integral representation of f in Q11 can also
be obtained from the Poisson formula of its real part and imaginary part.
P∞
Chapter 3 Q15: (a) Writting f (z) = n=0 an z n . For n > k, by Cauchy formula,
˛
1 f (z)
an = dz.
2πi ∂B(R) z n+1

Thus,
ˆ 2π
1 ARk + B
|an | ≤ dθ → 0 as R → ∞.
2π 0 Rn

Thus, f is a polynomial of degree at most k

(d) By translation and scaling, we can assume f (C) ⊂ {z : 0 ≤ Re(z) ≤ 1}. Consider
g(z) = ef , g is a bounded entire function. Thus g is constant implying f is also
a constant function.

Chapter 3 Q19: (a) Assume u attain a maximum at p in Ω, there exists r > 0 such that B(p, r) ⊂ Ω.
By previous exercise, there exists a holomorphic function f on B(p, r) such that
Re(f ) = u. But it contradicts with open mapping theorem. So it is impossible.
(b) It is a direct consequence of part (a).
Remark : One can also prove it directly using the mean value property (Strong
Maximum Principle) of harmonic function.

Chapter 5 Q4: (a)

N
Y −1 ∞
Y
F (z) = (1 − e−2πnt e2πiz ) · (1 − e−2πnt e2πiz )
n=1 n=N

∞
X
Consider the series |e−2πnt e2πiz |. Choose N ≈ |z|/t yields
N +1

∞ ∞
X X e−2πtN 1
|e−2πnt e2πiz | ≤ e−2πIm(z) e−2πnt ≤ e2π|z| ≤ .
1 − e−2πt 1 − e−2πt
n=N n=N

2
 Q∞
Thus, n=N (1 − e−2πnt e2πiz ) is bounded above by a constant Ct . Now we esti-
mate the term with finite elements.
N −1
Y 2
(1 − e−2πnt e2πiz ) ≤ (1 + e2π|z| )N ≤ 2N e2π|z|N ≤ ec1 |z|+c2 |z|
n=1

for some constant c1 , c2 > 0. So F is of order ≤ 2.

Now we claim the reverse inequality. It can be observed that m − int are the
zeros of F . And for p ≤ 2,

∞ ∞ ∞ ∞
X X 1 X X 1
p
= 2 + n2 t2 )p/2
= +∞
n=1 m=−∞
|m − int| n=1 m=−∞
(m

ˆ ∞ ˆ ∞
1
by comparing this with dx dy.
1 1 (x2 + t2 y 2 )p/2

An alternative way is to estimate M (r), we argue as follows.

For |z| = r, where r is sufficiently large. Writing iz = x + iy, denote e−2πnt

by R, it can be checked that

|1 − e−2πnt e2πiz |2 = 1 − 2Re2πx cos y + R2 e4πx .

√
r 2 −π 2
So, the maximum is at least Re2π when r is large enough. Denote M (r)
to be the maximum of F (z) on B(r). We deduce that

∞
Y √ ∞
Y
r 2 −π 2
M (r) > (1 + e−2πnt e2π )≈ (1 + e−2πnt e2πr ).
n=1 n=1

Taking log gives,

∞
X N
X ∞
X
log M (r) ≥ log(1+e−2πnt e2πr ) = log(1+e−2πnt e2πr )+ log(1+e−2πnt e2πr )
n=1 n=1 n=N +1

Since log(1 + z) ≥ z/2 whenever 0 < z < 1/2, the second terms is bounded below
by Ct if we choose N ≈ (log 2 + 2πr)/2πt.
On the other hand,
N
X ˆ N
−2πnt 2πr
log(1 + e e )≥ log(1 + e2πr e−2πxt ) dx
n=1 1

ˆ N/2
> log(1 + e2πr e−2πxt ) dx
1
N/2 ˆ N/2
2πx · e2πr
= x log(1 + e2πr e−2πxt )

+ dx.
1 1 e2πxt+ e2πr

3
 e2πr
Noted that for x ∈ [1, N/2], is bounded below by 1/4 when r → ∞.
e2πxt + e2πr
So we have for some constant C = C(t) > 0,
N
X ˆ N
−2πnt 2πr
log(1 + e e )≥ log(1 + e2πr e−2πxt ) dx
n=1 1
2
eπr
  
log 2 + 2πr π log 2 + 2πr
> log 1 + √ + − C.
4πt 2 4 4πt

Combining all the inequalities, we can see that limr→∞ log rM2 (r) > C̃ for some
constant C̃ > 0. Thus, the order is at least 2.
(b) F (z) = 0 iff one of its factors is 0. So, it vanishes exactly when z = −int + m for
n ≥ 1, m, n are integers.

Chapter 5 Q5:
ˆ ∞
α
Fα (z) = e−|t| e2πizt dt
−∞

Noted that
α α/α−1
|e2πizt | = e−2πtIm(z) ≤ e2π|t||z| ≤ e|t| /α
· eC1 |z|

where C1 is a constant obtained from Young ‘s inequality.

ˆ ∞ ˆ ∞
−|t|α 2πizt C1 |z|α/α−1 α α α/α−1
Fα (z) = e e dt ≤ e e−|t| e|t| /α
dt = C2 eC1 |z| .
−∞ −∞

So, Ord(f ) ≤ α/α − 1.

To show the reverse inequality, we make use of the Jensen’s inequality.
"ˆ # ˆ
b b
1
i.e. log f (x) dx ≥ log((b − a)f (x)) dx, ∀ integrable f ≥ 0.
a b−a a

Put iz = x ∈ R, for any R > 0

"ˆ #
ˆ ∞  R
−tα 2πxt −tα 2πxt
log Fα (z) ≥ log e e dt ≥ log e e dt
0 0
ˆ R ˆ R
1 α 1
≥ log[R · e−t e2πxt ] dt = log R + (2πxt − tα ) dt
R 0 R 0

Rα
= log R + πxR − .
α+1

2 α−1
Choose R such that πx = α+1 R implying

1 1 α
log Fα (z) ≥ log R + Rα = log x + C α x α−1 + log C
α+1 α−1

where C = ( α+1
2 π)
1/(α−1)
.
This show that the order of Fα is at least α/(α − 1).

4
Chapter 5 Q10: (a)
 
ez − 1 = ez/2 ez/2 − e−z/2

= 2ez/2 sinh(z/2)

= −i2ez/2 sin(iz/2)

But we have
∞ 
z2
Y 
sin z = z 1− 2 2 .
n=1
π n

So
∞  ∞ 
z2 z2
 
iz Y Y
ez − 1 = −i2ez/2 ( ) 1 + 2 2 = ez/2 z 1+ 2 2 .
2 n=1 4π n n=1
4π n

(b) Since we have

∞ 
4z 2
Y 
sin 2z = 2z 1− 2 2
n=1
π n

and
sin 2z = 2 sin z cos z

We deduce that
∞   " Y∞  #−1
1 Y 4z 2 z2
cos z = · 2z 1− 2 2 · z 1− 2 2
2 n=1
π n n=1
π n
∞ 
4z 2 4z 2
Y   Y 
= 1− = 1− .
π 2 n2 n=0
π 2 (2n + 1)2
n is odd.

Chapter 5 Q11: By multiplication and substraction, we can assume f miss 0 and 1. By Hadamard’s
theorem, we have
f (z) = eP (z) = eQ(z) + 1

for some poly P (z) and Q(z) with same degree. If degree is non-zero, by fundamental
theorem of algebra, there exists z0 such that Q(z0 ) = iπ. Thus, eP (z0 ) = 0 which is
impossible. So both P (z) and Q(z) are constant function, implying f is constant.

Chapter 5 Q14: Assume F has finitely many zeros. By Hadamard’s factorization theorem,

N
Y
F (z) = eP (z) z m Ek (z/an )
n=1

for some a1 , a2 , ...aN , where P is a polynomial of degree k < ρ. But then F is of order
k instead of ρ. Contradict with our assumption.

5
Chapter 5 Q15: Suppose f has poles at {an }n∈N counting with multiplicity. Let g be a entire function
having zeros precisely at {an } by weierstrass factorization theorem. Hence f (z)g(z)
is entire.

Similarly, let f (z), g(z) be entire functions having zeros precisely at {an } and {bn }
respectively. Then the function f (z)/g(z) = h(z) is the desired entire function.

Chapter 6 Q5: Put s = 1/2 + it into the product formula of Γ(s), we have

π
Γ(1/2 + it)Γ(1/2 − it) = .
sin π(1/2 + it)
´∞
But Γ(z) = 0
e−t tz−1 dt, so Γ(z) = Γ(z̄).

Γ(1/2 + it)Γ(1/2 + it) = Γ(1/2 + it)Γ(1/2 − it)

π
=
sin π(1/2 + it)
2π
= .
eπt + e−πt

Chapter 6,Q12: (a) Noted that for all s ∈ C,

1 sin πs
= Γ(1 − s).
Γ(s) π

Put s = −k − 1/2, gives

1 1
= |Γ(k + 3/2)| .
|Γ(s)| π

By the functional equation of Γ, we get

√ √
Γ(k + 1)Γ(k + 3/2) = π2−2k−1 Γ(2k + 2) = π2−2k−1 (2k + 1)!.

Thus,
1 1
= |Γ(k + 3/2)|
|Γ(s)| π
(2k + 1)!
=√
π · k!22k+1
1 2k + 1 2k k+1
= √ · ···
2 π 4 4 4
 k
1 k
≥ √ .
2 π 4

Taking log, one can observe that

1
log ≥ C1 + k log k − k log 4.
|Γ(s)|

So the order of growth cannot be O(ec|s| ).

6
 (b) If we can find such entire function F . Clearly, the order of growth of F is 1. By
Hadamard’s factorization theorem,

∞ ∞
Y z Y z
F (z) = eAz+B z E1 (− ) = eAz+B z (1 + )e−z/n .
n=1
n n=1
n

Thus,
1
= e−B F (z)ez(−A+γ) .
Γ(z)

The right hand side is of order O(ec|z| ) while left hand side not. So contradiction
occurred.

Extra question: If a ∈ (0, 1], we consider the contour γ from c − iR → c + iR and followed by a
semi-circular arc L of radius R centered at c in clockwise direction. Therefore,
˛
1 az
dz = 0.
2πi γ z(z + 1)

On the other hand, as 0 < a ≤ 1, |az | ≤ aRe(z) ≤ 1 on L.

ˆ
az 1
| dz| ≤ O(R) · O( 2 ) → 0 as R → ∞.
L z(z + 1) R

Result follows.

If a > 1, consider the contour γ from c − iR → c + iR and followed by a semi-circular

arc L of radius R centered at c in anti-clockwise direction.
˛
1 az 1
dz = Res(f, 0) + Res(f, −1) = 1 − .
2πi γ z(z + 1) a

Argue as first case, we know that

ˆ
| f (z) dz| → 0 as R → ∞.
L∩{Rez≤0}

On L ∩ {Rez > 0}, by symmetry, the total integral is 0. Or you can use a better
contour consisting of {x + iR : x ∈ [0, c]}, {x − iR : x ∈ [0, c]} and a semi-circlar arc
centered at 0 with radius R. The computation is almost same.