Complex Analysis, HW 4

Jacob Mansfield
February 2025

Problem 1

(10 points) Let a1 , . . . , ar be distinct complex numbers, and m1 , . . . , mr

be integers. Suppose that
r
Y
h(z) := Γ(z + ai )mi
i=1

is an entire function in z, with no zeroes or poles. Prove that h(z) = 1.

We begin by recalling the Gamma function has poles at zero and every
negative integer. Suppose we take z = −a1 . Then
r
Y r
Y
m1 mi mi
h(−a1 ) = Γ(−a1 + a1 ) Γ(z + ai ) = Γ(0) Γ(z + ai )mi
i=2 i=2

Which at first glance seems to also create a pole for h(z), unless m1 = 0.
However, if there is an aj such that a1 = aj + k for some positive integer k, and
mj = −m1 , then we could have
r
Y
h(z) = Γ(z + a1 )m1 Γ(z + aj )mj Γ(z + ai )mi (1)
i̸=1,j
r
Y
m1 mj
= Γ(z + aj + k) Γ(z + aj ) Γ(z + ai )mi (2)
i̸=1,j
r
Y
= [(z + aj + k − 1)m1 Γ(z + ak + k − 1)m1 ] Γ(z + aj )−m1 Γ(z + ai )mi
i̸=1,j
(3)

1
We may continue recursively applying the functional equation:
k
Y r
Y
h(z) = (z + aj + k − m)m1 Γ(z + aj )m1 Γ(z + aj )−m1 Γ(z + ai )mi (4)
m=1 i̸=1,j
k
Y r
Y
= (z + aj + k − m)m1 Γ(z + ai )mi (5)
m=1 i̸=1,j

and thus the two potential singularities cancel out given that the exponents are
equal and opposite.

So far, there are two kinds of al : those al where al − ai ∈

/ Z, ∀i, and then
al such that ∃ ai , al − ai ∈ Z. Let us proceed by considering each al in turn:
if it is of the first type, we know ml = 0. We can therefore reduce our product
form of h(z) to al exclusively of the second type. Getting fancy with it, we can
split all of these remaining al into equivalency classes
P via congruence mod Z.
As seen above, for a given equivalency class {ak }, k mk = 0, where mk is the
matching exponent of ak in the definition of h(z).

We will first pretend what is left belongs to a single equivalence class of size
l, where l < ∞ because l < r < ∞. This means we’re considering the expression
l
Y l
Y
Γ(z + ak )mk = Γ(z + a1 + nk )mk (6)
k=1 k=1

where without loss of generality, a1 < ak , ∀k, and thus all ak are some positive
nk ∈ Z away from a1 (where by definition n1 = 0). By the same recursive
application of the functional equation as in (4), we obtain that
l
Y n2
Y nl
Y Pl
Γ(z + a1 + nk )mk = (z + a1 + n2 − i)m2 · · · (z + a1 + nl − i)ml Γ(z + a1 ) k=1 mk

k=1 i=1 i=1

(7)
n2
Y nl
Y
= (z + a1 + n2 − i)m2 · · · (z + a1 + nl − i)ml Γ(z + a1 )0
i=1 i=1
(8)
n2
Y nl
Y
= (z + a1 + n2 − i)m2 · · · (z + a1 + nl − i)ml (9)
i=1 i=1

WLOG let nk+1 > nk , ∀k. Consider the largest n, namely nl . It will produce
(z + a1 + nl )ml . This has a zero at z = −a1 − nl , and as (z + a1 + nl )ml is a
factor of h(z), this would mean h(z) has a zero there as well. This is impossible
by the problem statement. There is no way to cancel this term out as we did
with the gamma functions: no matter what the exponents are, there is only one
contribution of (z + a1 + nl )p in the entire product, given that nl > nk ∀k, and

2
z + a1 is fixed. We conclude that ml = 0. We then apply the same argument
continuing downward from nl−1 to n2 , each time concluding that the exponent
mk = 0. Since all exponents are zero, the final product evaluates to 1.

It suffices to generalize this to the case of several equivalency classes. The

difficulty arises in ordering a subset of the complex numbers and not a subset
of the integers. However, we can simply take the smallest representative from
each equivalency class {ak }, and then order each representative by magnitude,
such that WLOG |ak1 | < |ak2 | < ... < |akn |. If any representatives have the
same magnitude, distinguish them by the magnitude of their real parts. Then
we have an ordered set of equivalency classes which are themselves ordered, as
seen above in the one equivalency class case. This creates an overall ordered
set. We may thus proceed downward as in the case of one equivalency class,
concluding that each exponent ml is zero and thus h(z) = 1. □

Problem 2

(10 points) Prove that

Z ∞
P1 (t) 1
dt = γ −
0 (1 + t)2 2
1
Where P1 (t) = t − ⌊t⌋ − 2

We will use a similar proof to that employed in lecture 10 for the proof of the
logarithmic form of the Stirling formula. We begin by considering the partial
Riemann-zeta sum:
n
X n−1
X
−s
ζn (s) = k = (1 + k)−s
k=1 k=0

We next recall the Euler formula seen in class:

m Z m Z m
X 1
f (i) = f (t)dt + (f (0) + f (m)) + P1 (t)f ′ (t)dt
i=0 0 2 0

With P1 (t) defined as in the problem statement. We apply the Euler formula
to the partial Riemann-zeta sum:
n−1 n−1 n−1
−s
Z Z
X
−s −s 1
(1+k) = (1+t) dt+ (1−s +(1+n−1)−s )+ P1 (t) dt
0 2 0 (1 + t)s+1
k=0

3
Passing to the limit as n → ∞, we have
∞ Z ∞ Z ∞
X 1 P1 (t)
(1 + k)−s = (1 + t)−s dt + −s dt (10)
0 20 (1 + t)s+1
k=0
∞
Z ∞
1 1 P1 (t)
ζ(s) = (1 + t)−s+1 + − s dt (11)
1−s 0 2 0 (1 + t)s+1
Z ∞
1 1 P1 (t)
ζ(s) = + −s dt (12)
s−1 2 0 (1 + t)s+1
Z ∞
P1 (t) 1 1
s s+1
dt = + − ζ(s) (13)
0 (1 + t) s−1 2

We take the limit as s → 1, obtaining

Z ∞  
P1 (t) 1 1
dt = + lim − ζ(s) (14)
0 (1 + t)2 2 s→1 s − 1

We will need the following:

Lemma: ζ(s) admits the following laurent series expansion about s = 1:

1
ζ(s) = + γ + O(s − 1)
s−1
Using the lemma, we are done:
Z ∞  
P1 (t) 1 1 1
dt = + lim − − γ − O(s − 1) (15)
0 (1 + t)2 2 s→1 s − 1 s − 1
Z ∞
P1 (t) 1
2
dt = − γ (16)
0 (1 + t) 2

Problem 3

Calculate Γ( 12 − n) where n is a positive integer. Show that Γ( 12 − n) → 0

when n → ∞. Then prove that Γ( 12 − n + it) → 0 uniformly for real t,
as n → ∞, where n is a positive integer.

We will simply keep applying the recursive functional equation for Γ(s),
giving us
    
1 1 1
Γ −n −n =Γ −n+1 (17)
2 2 2
     
1 1 1 1
Γ −n −n −n+1 =Γ −n+2 (18)
2 2 2 2
  n−1
Y 1   
1 1
Γ −n −n+k =Γ (19)
2 2 2
k=0

4
We land√ perfectly on Γ 12 because n is a positive integer. We recall that

Γ( 12 ) = π, and rewrite our expression:

  √
1 π
Γ −n = Qn−1 1
(20)
2


k=0 2−n+k
√
π
= n−1 (21)
(−1)n 2−n k=0 (2n − 1 − 2k)
Q
√
n 2n π
= (−1) (22)
(2n − 1)(2n − 3) . . . (3)(1)
n
√
n2 π · 2n−1 (n − 1)!
= (−1) (23)
(2n − 1)!
2n−1
√
n2 π(n − 1)!
= (−1) (24)
(2n − 1)!

It’s clear even from (20) that as n → ∞, the denominator tends to ∞ much
faster than the numerator and therefore Γ 12 − n → 0.

We now need to show Γ( 21 − n + it) → 0 uniformly as n → ∞. We will use

the Stirling formula for Γ, which yields
√ 1
Γ(z) = 2πz z− 2 e−z+µ(z)

Where µ(z) is a small corrective term whose expansion we omit for brevity. We
will now replace this with Γ(x + it), mixing polar and cartesian notation to
obtain:
√ hp t
ix+it− 12
Γ(x + it) = 2π ( x2 + t2 )ei arctan( x ) e−x−it+µ(x+it) (25)
hp ix+it− 12 1
lim Γ(x + it) = x2 + t2 eiπ(x+it− 2 ) e−x−it (26)
x→−∞
hp ix
∼ x2 + t2 e−tπ−x (27)

we’ve gotten rid of all terms of the form piq for p, q ∈ R as these√
will simplyx
induce rotation without largely changing the magnitude. Clearly, x2 + t 2
goes to 0 much faster than ex goes to infinity, as x → −∞. The only concern is
the e−tπ contribution, as we want uniform convergence for all t ∈ R. However,
we note that this is compensated by the t2 under the square root, which is
also affected by a x exponent, so any increases in t do not affect the function
converging to zero. □.

5
 Problem 4

(10 = 2 · 5 points) Is there a holomorphic function f (z) whose real part

is given by the following functions:
y p y
1)3x2 y−y 3 , 2)x−xy, 3) 2 , 4) log x2 + y 2 , 5)
x + y2 (x − t)2 + y 2

Either prove it does not exist, or find f (z)

1. u(x, y) = 3x2 y − y 3

We recall from class that there is a well-defined bijection between the set of
holomorphic functions and the set of real-valued harmonic functions. Therefore,
we will show that the given function is harmonic, and then find its matching
holomorphic function by the construction used in class.
∆(3x2 y − y 3 ) = 6y − 6y = 0
Therefore the function is harmonic. Let us define
Z (x,y)
v(x, y) : = −uy dx + ux dy (28)
(0,0)
Z (x,y)
= −(3x2 − 3y 2 )dx + (6xy)dy (29)
(0,0)

= −x + 3y 2 x
3
(30)
And we will define our candidate function as
f (x, y) : = u(x, y) + iv(x, y) (31)
= 3x2 y − y 3 + i(3y 2 x − x3 ) (32)

as seen in class, this function will satisfy the C-R equations by construction and
hence be holomorphic.

2. u(x, y) = x − xy

We will once more check if this is harmonic:

d d
∆(x − xy) = (1 − y) + (−x) = 0 + 0 = 0
dx dy
So we may proceed, defining
Z (x,y)
v(x, y) : = xdx + (1 − y)dy (33)
(0,0)
2
x y2
= +y− (34)
2 2

6
And then our f (z), holomorphic by construction, is

x2 y2
f (x, y) := x − xy + i( +y− )
2 2
y
3. u(x, y) = x2 +y 2

We check harmonicity:
 2
x − y2
    
y d −2xy d
∆ = + (35)
x2 + y 2 dx (x2 + y 2 )2 dy (x2 + y 2 )2
−2y(x + y ) + 2xy(2 · 2x(x2 + y 2 ))
2 2 2
= (36)
(x2 + y 2 )4
−2y(x + y 2 )2 − (x2 − y 2 )(2 · 2y(x2 + y 2 ))
2
+ (37)
(x2 + y 2 )4
(x2 + y 2 )[−4y(x2 + y 2 ) + 8x2 y − 4y(x2 − y 2 )]
= (38)
(x2 + y 2 )4
−4yx2 − 4y 3 + 8x2 y − 4yx2 + 4y 3
= (39)
(x2 + y 2 )3
8x2 y − 8x2 y + 4y 3 − 4y 3
= =0 (40)
(x2 + y 2 )3

This is harmonic, so we define

(x,y)
y 2 − x2
Z
2xy
v(x, y) : = 2 2 2
dx − 2 dy (41)
(0,0) (x + y ) (x + y 2 )2
x
= (42)
x2 + y 2

Therefore our desired f (x, y) is

y + ix
f (x, y) =
x2 + y 2
p
4. u(x, y) = log x2 + y 2

Checking harmonicity:
d 1 1 d 1 1
∆(u(x, y)) = ( · 2x 2 )+ ( · 2y 2 ) (43)
dx 2 x + y2 dy 2 x + y2
x2 + y 2 − 2x2 x2 + y 2 − 2y 2
= 2 2 2
+ (44)
(x + y ) (x2 + y 2 )2
−x2 + y 2 + x2 − y 2
= =0 (45)
(x2 + y 2 )2

7
We thus define
Z (x,y)
y x
v(x, y) : = − dx + 2 dy (46)
(0,0) +y 2 x2 x + y2
 
x
= − arctan (47)
y
So our desired f (x, y) is
 
p x
f (x, y) = log x2 + y 2 − i arctan
y
y
5. u(x, y) = (x−t)2 +y 2

We check harmonicity:
! !
d 2y(x − t) d (x − t)2 − y 2
∆u(x, y) = − 2 + 2 (48)
dx [(x − t)2 + y 2 ] dy [(x − t)2 + y 2 ]
2y(−3t2 + 6tx − 3x2 + y 2 ) 2y(−3t2 + 6tx − 3x2 + y 2 )
=− + (49)
(t2 − 2tx + x2 + y 2 )3 (t2 − 2tx + x2 + y 2 )3
=0 (50)
This means we define our v(x, y) as usual:
Z (x,y)
(x − t)2 − y 2 2y(x − t)
v(x, y) : = − 2 dx − 2 dy (51)
2
[(x − t) + y ]2 [(x − t)2 + y 2 ]
(0,0)
t−x
=− (52)
(x − t)2 + y 2
And therefore our f (x, y) is

y + i(x − t)
f (x, y) =
(x − t)2 + y 2

Problem 5

(10=5+5 points)
a) Let f1 (z), . . . , fn (z) be holomorphic functions on the unit disk D1 .
Suppose we have
Xn
|fi (z)|2 = 1, ∀z ∈ D1
i=1

Show all functions are constants.

b) Let H : Cn → R be a strictly convex function. Using the method of

solution in a), show that if H(f1 (z), ..., fn (z)) is constant on D1 , then
each function fi (z) is constant.

8
 a) We will begin by recalling that if f (x, y) = u(x, y) + iv(x, y) holomorphic
for two real-valued functions u, v, then both u, v must be harmonic. That is,
2 2
∆u(x, y) = ddxu2 + ddyu2 = 0 = ∆v(x, y). We will now expand the equality given
in the problem statement, and take the Laplacian of both sides:
n
X n
X
1= |fi (z)|2 = u2i (x, y) + vi2 (x, y) (53)
i=1 i=1
n
X
∆1 = ∆ u2i (x, y) + vi2 (x, y) (54)
i=1
n        
X d dui d dui d dvi d dvi
0= 2 ui + 2 ui + 2 vi + 2 vi (55)
i=1
dx dx dy dy dx dx dy dy
n 2 2
d2 ui d2 ui
     
X dui dui
0= 2ui + + 2 + 2 (56)
i=1
dx2 dy 2 dx dy
 2 2 2
d2 vi
   
d vi dvi dvi
+ 2vi + +2 +2 (57)
dx2 dy 2 dx dy
n
"  2  2  2  2 #
X dui dui dvi dvi
0= 2ui · 0 + 2vi · 0 + 2 + + + (58)
i=1
dx dy dx dy

Given that the sum of all the first derivatives, each squared (and thus ≥ 0),
yields 0, it means each of the first derivatives must be zero, and therefore both
ui (x, y) and vi (x, y) are constant functions for any i, and thus fi (x, y) is con-
stant.

b) We define the function G(z) = H(f1 (z), ..., fn (z)) and note by the prob-
lem statement this is a constant function. Assume that the domain of H is
parametrized by n complex numbers wi . H : C → R and so we once more apply
the Laplacian to both sides, obtaining

n
X ∂2H
∆G(z) = ∆(fi (z), fj (z)) (59)
i,j
∂wi ∂wj
n
X ∂2H
0= 4∇fi · ∇fj (60)
i,j
∂wi ∂wj

We eliminated the terms that came from ∆fi because these are all holomorphic
functions and hence these expressions would vanish. If H strictly convex, that
means it must have a positive definite Hessian matrix, therefore
n
X ∂2H
v i vj > 0
i,j
∂wi ∂wj

9
for (v1 , ..., vn ) nonzero vectors. Because the equation in (60) is stricly zero, this
means the divergence of each function fi is zero, which means each fi (x) is
constant on ∆ and we are done. □

10