Suggested solution of HW2

Chapter 3 Q12: By Residue formula, for any N ≥ |u|

˛    
1 π cot πz X π cot πz π cot πz
dz = Res , n + Res , −u .
2πi |z|=N +1/2 (u + z)2 (u + z)2 (u + z)2
|n|≤N +1/2

At z = n,

d
sin πz  = π(−1)n .

dz n

Thus,
 
π cot πz 1
Res ,n = .
(u + z)2 (u + n)2
Also, direct checking yield

π2
 
π cot πz
Res , −u = − .
(u + z)2 sin2 πu

On the other hand, on the circle |z| = N + 1/2, write πz = x + iy, fix a small δ > 0.
If |x| > δ, we have

cos2 x + sinh2 y 1 + sinh2 y

| cot πz|2 = ≤ < C1 .
sin2 x + sinh2 y C + sinh2 y

If |x| ≤ δ,
cos2 x + sinh2 y 1 + sinh2 y
| cot πz|2 = 2 ≤ < 2.
2
sin x + sinh y sinh2 y
Thus,
˛
1 π cot πz
dz → 0 as N → ∞
2πi |z|=N +1/2 (u + z)2

which yield the desired result.

Chapter 4 Q1: (a) if ξ ∈ R,

ˆ
A(ξ) − B(ξ) = e2πit f (x)e−2πiξx dx = 0.
R

(b) We first claim that A(z) is differentiable at z0 ∈ H. For z = x + iy, y > 0,

ˆ t ˆ ∞
−2πi(x+iy)(ξ−t)
A(z) = f (ξ)e dξ = f (ξ)e2πi(x+iy)ξ dξ
−∞ 0
ˆ +∞
= f (ξ)e2πix e−2πyξ dξ.
0

Since f is of moderate decrease, and y > 0, A(z) define a holomorphic function

on H. And
ˆ +∞ ˆ +∞
−2πyξ A A
|A(z)| ≤ |f (ξ)|e dξ ≤ e−2πyξ dξ ≤ → 0 as y → ∞.
0 0 1 + ξ2 2πy

1
 Thus, A(z) is a bounded holomorphic function. Similar for B(z). By Morera’s
Theorem, F (z) is a bounded entire function, and thus is constant. By above
inequality, F (z) ≡ 0.
(c) Putting z = 0, we have for each t ∈ R
ˆ t
f (x) dx = 0.
−∞

By continuity, f ≡ 0.

Chapter 4 Q2: If f is analytic in Sa , obviously f (n) is analytic in Sb for any 0 ≤ b < a. It remains to
show that there exists Bn > 0 such that

Bn
|f (n) (x + iy)| ≤ for all x ∈ R and |y| < b.
1 + x2

Let δ = a − b, w ∈ Sb . By Cauchy formula,

 (n)  ˛ 
 f (w)  1  f (z) 
 n!  = 2π  dz
   
(z − w) n+1 
B(w,δ) 
ˆ 2π
1 |f (w + δeiθ )|
≤ dθ
2π 0 δn
ˆ 2π
1 A
≤ dθ
2πδ n 0 1 + [Re(w + δeiθ )]2

If |Re(w)| > 2δ,

 (n) 
 f (w)  A 1 A 1 4A 1
 n!  ≤ δ n · 1 − δ 2 + [Re(w)]2 /2 ≤ δ n 1 + (Re(w))2 /4 ≤ δ n 1 + (Re(w))2 .
 

If |Re(w)| ≤ 2δ,
 (n)
1 + 4δ 2

 f (w) 
 ≤ A ·
 n!  δ n 1 + (Re(w))2

An!(1 + 4δ 2 )
Choose Bn = .
δn

Chapter 4 Q3: If ξ < 0, let γ be the curve composed of the upper semi circle of radius R from R to
−R and the straight line from −R to R on the real axis. For R sufficiently large, by
Residue formula, one can obtain
˛
1 a 1
e−2πizξ dz = e2πξa .
2πi γ a2 +z 2 2i

On the other hand,

˛ ˆ R ˆ π
a a
f (z) dz = e−2πixξ dx + e−2πiRξ(cos θ+i sin θ) dθ
γ −R a + x2
2
0 a2 + R2 e2iθ

2
 And
ˆ π  ˆ π
a −2πiRξ(cos θ+i sin θ) a
e2πRξ sin θ dθ → 0 as R → ∞.
 
 e dθ ≤
 a2 + R2 e2iθ  R 2 − a2
0 0

Result follows when we take R tends to +∞. If ξ ≥ 0, instead we consider the curve
composed of lower semi cirle from −R to R and the straight line from R to −R. And
then argue as same as before.
The inversion can be checked by direct integration.
ˆ +∞ ˆ 0
1 1
e−2πa|ξ| e2πiξx dξ = and e−2πa|ξ| e2πiξx dξ =
0 2π(a − ix) −∞ 2π(a + ix)

Summing up yield the result.

a
Chapter 4 Q6: Follows from applying Poisson summation formula to f = and the result in
a2 + x2
Q3.
∞
X ∞
X 0
X ∞
X
e−2πa|n| = e−2πan + e2πan = 1 + 2 e−2πan
n=−∞ n=1 n=−∞ n=1

1 + e−2πa
= = coth πa.
1 − e−2πa

Chapter 4 Q7: (a) If ξ ≤ 0, using contour integral as in the first part of Q3, as the pole is in the
lower half plane, we get
˛ ˆ R ˆ π
1 1
0= f (z)e−2πizξ dz = e−2πixξ dx + e−2πiξR(cos θ+i sin θ) dθ.
γ −R (x + τ )k 0 (τ + Reiθ )k

while
ˆ π  ˆ π
1 −2πiξR(cos θ+i sin θ)  1
e2πiξ sin θ dθ → 0.
 

 (τ + Reiθ )k e dθ ≤ k
0

0 (R − |τ |)

Thus, fˆ(ξ) = 0 if ξ ≤ 0.

If ξ > 0, using the contour as in the latter part of Q3. we have

˛
1 1
e−2πizξ dz = Res−τ (f (z)e−2πizξ ).
2πi γ (z + τ )k

At z = −τ ,
dk−1 −2πizξ 

k−1 2πiτ ξ
e  = (−2πiξ) e .
dz k−1 −τ

Thus,
(−2πiξ)k−1 e2πiτ ξ
Res−τ (f (z)e−2πizξ ) =
(k − 1)!
in which
˛
1 −2πizξ (−2πi)k ξ k−1 e2πiτ ξ
e dz = − .
γ (z + τ )k (k − 1)!

3
 But
˛ ˆ −R
1 1 1
e−2πizξ dz = e−2πixξ dx + O( k ).
γ (z + τ )k R (x + τ )k R

Takeing R → ∞ yields

(−2πi)k k−1 2πiξτ

fˆ(ξ) = ξ e .
(k − 1)!

Then we apply Poisson summation formula to get the desired equality.

(b) Putting k = 2, then
∞ ∞
X 1 2
X
2
= −4π me2πimτ
n=−∞
(τ + n) m=1

Using the equality

∞
X z
mz m = for |z| < 1.
m=1
(1 − z)2

Since Im(τ ) > 0, we can substitute z = e2πiτ into the above equation. Thus,
∞
X e2πiτ π2
−4π 2 me2πimτ = −4π 2 2πiτ 2
= 2 .
m=1
(1 − e ) sin (πτ )

∞
π2 X 1
(c) Yes. Since both and define a meromorphic function on
sin2 (πz) n=−∞
(z + n)2
C with pole at integers, and equal in value on the upper half plane. By identity
theorem, they are equal.

For sake of completeness, I prove the identity theorem for meromorphic function
here. It suffices to prove that they are equal on any compact set Ω ⊂ C. If f and
g are two meromorphic functions on Ω such that f (z) = g(z) on a nonempty open
region U . As poles are isolated, there exists holomorphic function f1 , f2 , g1 , g2
on interior of Ω such that

f1 g1
f (z) = , g(z) = on int(Ω).
f2 g2

Define h = f1 g2 − f2 g1 which is holomorphic on int(Ω). h = 0 on U . By identity

theorem, h = 0 on Ω. Thus, f ≡ g on Ω.

Chapter 5 Q2: (a) Let p(z) be a polynomial of degree m. There exists a constant C > 0 such that

|p(z)| ≤ C(|z|m + 1)

∀ > 0, there exists A > 0 such that

|z|m ≤ A exp(|z| )

Thus, p(z) is of order less than or equal to , for any  > 0. So, it is of order 0.

4
 (b) If z = |z|eiθ , b = |b|eiφ ,
n n n
|ebz | = e|z| |b| cos(nθ+φ)
≤ e|z| |b|

So it is of order less than or equal to n. Put z = x ∈ R to see that the order is

exactly n.
(c) Put z = x ∈ R. Since for any s ∈ R, there exists R > 0 s.t. ex > xs for all x > R.
z
The function ee has infinity order of growth.

Chapter 5 Q3:
 
X  X
iπn2 τ 2πinz 
e e ≤ e−πn·Im(nτ +2z)


 
n∈N n∈N
X 2
≤ e−πn ·Im(τ )
· e2π|n||z|
n∈N
2 X 2
≤ e2π|z| e−πn ·Im(τ )/2
(using AM-GM inequality)
n∈N
2
≤ Ce2π|z|

for some constant C depends on τ only. Thus it is of order ≤ 2. It remains to show

that the order is exactly 2. For each r and m ∈ N, we have
2
Θ(r + mτ |τ ) = e−iπm r−2imπr
Θ(r|τ ).

Thus,
2
|Θ(r + mτ |τ )| = etπm |Θ(r|τ )|.

Choose r such that Right hand side is non-zero. Thus the growth of order is at least
2.

Chapter 5 Q6: Using the product formula for the sine function. Putting z = 1/2, we get
∞ ∞
2 Y 1 Y (2n + 1)(2n − 1)
= (1 − 2 ) = .
π n=1 4n n=1
(2n)2

Chapter 5 Q7: (a) Clearly, an 6= −1 for all n ∈ N, otherwise the conclusion fail. Since |an | → 0, we
may assume |an | ≤ 1/2 for all n ∈ N. Hence, log(1 + an ) is well defined if we
choose the principle branch. By the power series expansion of log, we know that
for all |z| ≤ 1/2,
|log(1 + z) − z| ≤ C|z|2 .

Therefore, for any m ≥ n ≥ 1,

 
Xm m
X  m
X
log(1 + ak ) − ak  ≤ C |ak |2
 

 
k=n k=n k=n

The conclusion follows immediately from cauchy criterion.

(b) Take an = einπ/4 will suffices.

5
 (c) Take a sequence an which contains a constant subsequence ank = −1.

Chapter 5 Q8: The product converge since

N
Y z sin z h z i−1 sin z
cos k
= N
sin N
→ as N → ∞.
2 2 2 z
k=1

Chapter 5 Q9:
N
Y k N +1
(1 + z 2 )(1 − z) = (1 + z 2 ) → 1 as N → ∞.
k=0

Extra question: Consider the contour suggested. Choose the log using Principle branch and use
Residue theorem, we obtain
˛
1 z a−1
dz = eiπ(a−1) .
2πi γ 1+z

On γ1 ,
ˆ ˆ R ˆ R
z a−1 e(a−1) log(t+iδ) ta−1
dz = dt → dt as δ → 0.
γ1 1+z  1 + t + iδ  1+t

Similarly,
ˆ ˆ R (a−1) log(t−iδ)
z a−1 e
dz = dt
γ3 1 + z  1 + t − iδ
ˆ R
e(a−1) log (t+iδ)
= dt
 1 + t + iδ
ˆ R (a−1)log(t+iδ) ˆ R a−1
(a−1)2πi e (a−1)2iπ t
=e dt → e dt as δ → 0.
 1 + t + iδ  1+t

Therefore,
ˆ ˆ R
!
z a−1 ta−1
dz = dt (1 − e(a−1)2π )
γ1 −γ3 1+z  1+t

while when R → ∞ and  → 0,

ˆ
z a−1 

1
dz  ≤ CR · Ra−1 ·

 →0

γ2 1 + z 1 + R

and
ˆ
z a−1 

1
dz  ≤ C · a−1 ·

 → 0.

γ2 1+z 1−

Hence,
ˆ ∞
ta−1 2πieiπ(a−1) π
dt = = .
0 1+t (1 − e(a−1)2πi ) sin(πa)