NOTES ON RIEMANNS ZETA FUNCTION

DRAGAN MILI

CI

C
1. Gamma function
1.1. Denition of the Gamma function. The integral
(z) =
_

0
t
z1
e
t
dt
is well-dened and denes a holomorphic function in the right half-plane {z C |
Re z > 0}. This function is Eulers Gamma function.
First, by integration by parts
(z + 1) =
_

0
t
z
e
t
dt = t
z
e
t

0
+ z
_

0
t
z1
e
t
dt = z(z)
for any z in the right half-plane. In particular, for any positive integer n, we have
(n) = (n 1)(n 1) = (n 1)!(1).
On the other hand,
(1) =
_

0
e
t
dt = e
t

0
= 1;
and we have the following result.
1.1.1. Lemma.
(n) = (n 1)!
for any n Z.
Therefore, we can view the Gamma function as a extension of the factorial.
1.2. Meromorphic continuation. Now we want to show that extends to a
meromorphic function in C. We start with a technical lemma.
1.2.1. Lemma. Let c
n
, n Z
+
, be complex numbers such such that

n=0
|c
n
|
converges. Let
S = {n | n Z
+
and c
n
= 0}.
Then
f(z) =

n=0
c
n
z + n
converges absolutely for z CS and uniformly on bounded subsets of CS. The
function f is a meromorphic function on C with simple poles at the points in S and
Res(f, n) = c
n
for any n S.
1
2 D. MILI

CI

C
Proof. Clearly, if |z| < R, we have |z + n| |n R| for all n R. Therefore, we
have |
1
z+n
|
1
nR
for |z| < R and n R. It follows that for n
0
> R, we have

n=n0
c
n
z + n

n=n0
|c
n
|
|z + n|

n=n0
|c
n
|
n R

1
n
0
R

n=n0
|c
n
|.
Hence, the series

n>R
cn
z+n
converges absolutely and uniformly on the disk {z |
|z| < R} and denes there a holomorphic function. It follows that

n=0
cn
z+n
is a meromorphic function on that disk with simple poles at the points of S in
{z | |z| < R}. Therefore,

n=0
cn
z+n
is a meromorphic function with simple poles
at the points in S. Therefore, for any n S we have
f(z) =
c
n
z + n
+

mS{n}
c
m
z + m
=
c
n
z + n
+ g(z)
where g is holomorphic at n. This implies that Res(f, n) = c
n
.
Going back to , we have
(z) =
_

0
t
z1
e
t
dt =
_
1
0
t
z1
e
t
dt +
_

1
t
z1
e
t
dt.
Clearly, the second integral converges for any complex z and represents an entire
function. On the other hand, since the exponential function is entire, its Taylor
series converges uniformly on compact sets in C, and we have
_
1
0
t
z1
e
t
dt =
_
1
0
t
z1
_

p=0
(1)
p
p!
t
p
_
dt
=

p=0
(1)
p
p!
_
1
0
t
p+z1
dt =

p=0
(1)
p
p!
1
z + p
for any z C. Therefore,
(z) =
_

1
t
z1
e
t
dt +

p=0
(1)
p
p!
1
z + p
for any z in the right half-plane. By 1.2.1, the right side of this equation denes a
meromorphic function on the complex plane with simple poles at 0, 1, 2, 3, .
Hence, we have the following result.
1.2.2. Theorem. The function extends to a meromorphic function on the complex
plane. It has simple poles at 0, 1, 2, 3, . The residues of are p are given
by
Res(, p) =
(1)
p
p!
for any p Z
+
.
This result combined with the above calculation immediately implies the follow-
ing functional equation.
1.2.3. Proposition. For any z C we have
(z + 1) = z(z).
NOTES ON RIEMANNS ZETA FUNCTION 3
1.3. Another functional equation. Let Re z > 0. Then
(z) =
_

0
t
z1
e
t
dt.
Let Re p > 0 and Re q > 0. Then, by change of variable t = u
2
, we get
(p) =
_

0
t
p1
e
t
dt = 2
_

0
e
u
2
u
2p1
du.
Analogously we have
(q) = 2
_

0
e
v
2
v
2q1
dv.
Hence, it follows that
(p)(q) = 4
_

0
_

0
e
(u
2
+v
2
)
u
2p1
v
2q1
du dv,
and by passing to the polar coordinates by u = r cos , v = r sin , we have
(p)(q) = 4
_

0
_
2
0
e
r
2
r
2(p+q)1
cos
2p1
sin
2q1
dr d
=
_
2
_

0
e
r
2
r
2(p+q)1
dr
_

_
2
_
2
0
cos
2p1
sin
2q1
d
_
= 2(p + q)
_
2
0
cos
2p1
sin
2q1
d.
We put s = sin
2
in the integral. Then we have
2
_
2
0
cos
2p1
sin
2q1
d =
_
1
0
s
q1
(1 s)
p1
ds.
If we dene
B(p, q) =
_
1
0
s
p1
(1 s)
q1
ds
for Re p > 0, Re q > 0, we get the identity
B(p, q) = B(q, p) =
(p)(q)
(p + q)
.
Here B is Eulers Beta function.
Let x (0, 1). Then we have
(x)(1 x) =
(x)(1 x)
(1)
= B(x, 1 x) =
_
1
0
s
x1
(1 s)
x
ds.
Moreover, if we change the variable s =
u
u+1
, the integral becomes
_
1
0
s
x1
(1 s)
x
ds =
_

0
u
x1
(u + 1)
x1
_
1
u
u + 1
_
x
du
(u + 1)
2
=
_

0
u
x1
1 + u
du.
1.3.1. Lemma. For y (0, 1) we have
_

0
u
y
1 + u
du =

sin y
.
4 D. MILI

CI

C
Proof. We cut the complex plane along the positive real axis. On this region we
dene the function z
z
y
1+z
with the argument of z
y
equal to 0 on the upper
side of the cut. This function has a rst order pole at z = 1 with the residue
e
iy
.
C
C
R

-1

If we integrate this function along a path which goes along the upper side of the
cut from > 0 to R, then along the circle C
R
of radius R centered at the origin,
then along the lower side of the cut from R to and nally around the origin along
the circle C

of radius , by the residue theorem we get

_
R

u
y
1 + u
du +
_
CR
z
y
1 + z
dz e
2iy
_
R

u
y
1 + u
du
_
C
z
y
1 + z
dz = 2ie
iy
.
First we remark that for z = 0 we have
|z
y
| = |e
y log z
| = e
y Re(log z)
= e
y log |z|
= |z|
y
.
Hence, the integrand in the second and fourth integral satises

z
y
1 + z

|z|
y
|1 + z|

|z|
y
|1 |z||
.
Hence, for small we have

_
C
z
y
1 + z
dz

2

1y
(1 )
;
and for large R we have

_
CR
z
y
1 + z
dz

2
R
1y
(R 1)
;
Clearly, this implies that
_
CR
0 as R and
_
C
0 as 0. This implies
that
_
1 e
2iy
_
_

0
u
y
1 + u
du = 2ie
iy
.
NOTES ON RIEMANNS ZETA FUNCTION 5
It follows that
_
e
iy
e
iy
_
_

0
u
y
1 + u
du = 2i
and nally
_

0
u
y
1 + u
du =

sin y
.

This lemma implies that

(x)(1 x) =

sin(1 x)
=

sin x
.
for x (0, 1). Since both sides are meromorphic, it follows that the following result
holds.
1.3.2. Proposition. For all z C, we have
(z)(1 z) =

sin z
.
Since z sin z is an entire function, the right side obviously has no zeros. So,
(z) = 0 is possible only at points where z (1 z) has poles. Since the poles
of are at 0, 1, 2, , it follows that poles of z (1 z) are at 1, 2, 3, .
At these points (n + 1) = n! = 0. Therefore, we proved the following result.
1.3.3. Theorem. The function has no zeros.
2. Zeta function
2.1. Meromorphic continuation. Riemanns zeta function is dened by
(z) =

n=1
1
n
z
for Re z > 1. In that region the series converges uniformly on compact sets and
represents a holomorphic function.
If we consider the expression for the Gamma function
(z) =
_

0
t
z1
e
t
dt
for Re z > 0, and change the variable t into t = ns for n N, we get
(z) = n
z
_

0
s
z1
e
ns
ds,
i.e., we have
(z)
n
z
=
_

0
t
z1
e
nt
dt
for any n N and Re z > 0. This implies that, for Re z > 1, we have
(z)(z) =

n=1
_

0
t
z1
e
nt
dt
=
_

0
t
z1
_

n=1
e
nt
_
dt =
_

0
t
z1
e
t
1 e
t
dt =
_

0
t
z1
e
t
1
dt.
This establishes the following integral representation for the zeta function.
6 D. MILI

CI

C
2.1.1. Lemma. For Re z > 1 we have
(z)(z) =
_

0
t
z1
e
t
1
dt.
This integral can be split in to two parts
(z)(z) =
_

0
t
z1
e
t
1
dt =
_
1
0
t
z1
e
t
1
dt +
_

1
t
z1
e
t
1
dt.
Clearly, the second integral converges for all z C and therefore represents an
entire function which we denote by F.
On the other hand, the function z
1
e
z
1
has a simple pole at 0 with residue
1. Therefore,
1
e
z
1
=
1
z
+ G(z)
where G is an meromorphic function with rst order poles at 2mi with integer
m = 0. It follows that
1
e
z
1
=
1
z
+

n=0
c
n
z
n
for any |z| < 2. By Cauchys estimates, if we x 0 < r < 2, it follows that
|c
n
|
M
r
n
for some M > 0. In particular, there exists M > 0 such that |c
n
|
M
2
n
for all n Z
+
.
This implies that, for Re z > 1, since the above series converges uniformly on
[0, 1], we have
_
1
0
t
z1
e
t
1
dt =
_
1
0
t
z2
dt +
_
1
0
_

n=0
c
n
t
z+n1
_
dt
=
1
z 1
+

n=0
c
n
_
1
0
t
z+n1
dt =
1
z 1
+

n=0
c
n
z + n
.
Therefore, we have
(z)(z) = F(z) +
1
z 1
+

n=0
c
n
z + n
.
By 1.2.1, the right side is a meromorphic function, holomorphic for any complex z
dierent form z = 1, 0, 1, 2, . . . . Hence, z (z)(z) extends to a meromor-
phic function in the complex plane with simple pole at 1 and either simple poles or
removable singularities at z = 0, 1, 2, . . . depending if c
n
= 0 or c
n
= 0. Since
has no zeros, z
1
(z)
is an entire function. This implies that the zeta function
extends to a meromorphic function in the complex plane.
Since (1) = 1, has a simple pole at 1 with residue equal to 1. Moreover, since
has simple poles at z = 0, 1, 2, . . . by 1.2.2, the function
1

has simple zeros

there. It follows that has removable singularities at z = 0, 1, 2, . . . . Therefore,
we established the following result.
2.1.2. Theorem. The zeta function is a meromorphic function with simple pole
at 1. The residue at this pole is 1.
NOTES ON RIEMANNS ZETA FUNCTION 7
2.2. The functional equation. In this section we prove the following result.
2.2.1. Theorem. For any z C, we have
(z) =
z1
2
z
sin
_
z
2
_
(1 z)(1 z).
To establish this functional equation we rst establish a variant of 2.1.1.
Let C be a path sketched in the following gure.
C

2i
-2i
We cut the complex plane along positive real axis and consider the integral
_
C
w
z1
e
w
1
dw.
Here we x the argument of w to be 0 on the top side of the cut and 2 on the
bottom. Clearly, the integrand is holomorphic on the complement of the positive
real axes except at zeros of the denominator of the integrand, i.e., at the points
2mi for m Z.
If the radius of the arc C

is less than 2, it follows immediately from the

Cauchy theorem that this integral doesnt depend on and the distance of the
horizontal lines from the positive real axis. Therefore, to evaluate it we can let
and that distance tend to 0.
First we remark that for w = 0 we have
|w
z
| = |e
z log w
| = e
Re(z log w)
= e
Re z log |w|Imz arg(w)
= |w|
Re z
e
Imz arg(w)
.
Since the argument is in [0, 2], for a xed z, we see that |w
z
| < M|w|
Re z
for some
positive constant M.
We can estimate the integral over C

as

_
C
w
z1
e
w
1
dw

M
Re z
_
2
0
1
|e
e
i
1|
d.
Since w e
w
1 has a simple zero at the origin, e
w
1 = wg(w) where g is
holomorphic near the origin and g(0) = 1. It follows that
|e
w
1|
1
2
|w|
8 D. MILI

CI

C
for small w. Hence, we have

_
C
w
z1
e
w
1
dw

4M
Re z1
for small . In particular, if Re z > 1, we see that the integral over C

tends to 0
as 0.
Hence, by taking the limit as goes to zero, and the horizontal lines tend to the
real axis, we get
_
C
w
z1
e
w
1
dw =
_

0
t
z1
e
t
1
dt e
2(z1)i
_

0
t
z1
e
t
1
dt
=
_
1 e
2iz
_
_

0
t
z1
e
t
1
dt = 2ie
iz
sin(z)
_

0
t
z1
e
t
1
dt
= 2ie
iz
sin(z)(z)(z)
by 2.1.1. This implies that
2ie
iz
sin(z)(z)(z) =
_
C
w
z1
e
w
1
dw
for any path C we considered above and Re z > 1. Clearly, the right integral
makes sense for arbitrary complex z, hence this equality is an equality of mero-
morphic functions on C. This yields the following result, which is another integral
representation of as a meromorphic function.
2.2.2. Lemma. For any z C, we have
sin(z)(z)(z) =
i
2
e
iz
_
C
w
z1
e
w
1
dw.
On the other hand, we can calculate using the residue theorem the integral
along the path drawn in the following gure. There the outside square passes
thru points (2m + 1)i and (2m + 1)i.

2i
-2i
2mi
2(m+1)i
-2mi
-2(m+1)i

NOTES ON RIEMANNS ZETA FUNCTION 9

We have
_

w
z1
e
w
1
dw = 2i
m

n=1
_
(2n)
z1
e
i
(z1)
2
+ (2n)
z1
e
3i
(z1)
2
_
= 2
z

z
ie
i(z1)
_
e
i
(z1)
2
+ e
i
(z1)
2
_
m

n=1
n
z1
= 2
z+1

z
ie
iz
cos
_
(z 1)
2
_
m

n=1
1
n
1z
= 2
z+1

z
ie
iz
sin
_
z
2
_
m

n=1
1
n
1z
.
Now we want to estimate the integral along the sides of the square for Re z < 0.
For a xed z, we see as before that there exists M > 0, such that on the right
vertical side of the square the integrand satises the estimate

w
z1
e
w
1

M
|w|
Re z1
|e
w
1|
.
Moreover, we have
|e
w
1| ||e
w
| 1| = |e
Re w
1|
for any w, hence we have

w
z1
e
w
1

M
|w|
Re z1
|e
Re w
1|
M
| Re w|
Re z1
|e
Re w
1|
and the right side of the square. This expression clearly tends to zero faster than
1
Re w
as Re w +. Hence, the integral over the right side tends to zero as the
square grows.
On the left side of the square we have the same estimate

w
z1
e
w
1

M
| Re w|
Re z1
|e
Re w
1|
.
Since Re w is negative in this case, for large | Re w| we have

w
z1
e
w
1

2M| Re w|
Re z1
.
Since Re z < 0, this bound goes to zero faster that
1
| Re w|
. On the other hand, the
length of the side of the square is 2| Re w|. Hence, the integral along the left side
also tends to zero as the square grows.
It remains to treat the top and bottom side. As before, we see that on these
sides we have

w
z1
e
w
1

M
|w|
Re z1
|e
w
1|
M
| Imw|
Re z1
|e
w
1|
.
On the other hand, the function w |e
w
1| on the horizontal lines tends to
as Re w + and to 1 as Re w . Hence it is bounded from below.
Moreover, since it is periodic with period 2i, shifting the line up and down by 2i
doesnt change that bound. Hence, in our estimates we can assume that
1
|e
w
1|
C
10 D. MILI

CI

C
on top and bottom sides of all squares. This implies that

w
z1
e
w
1

MC| Imw|
Re z1
on top and bottom sides of all squares. Since Re z < 0 and the length of these sides
is 2| Imw|, we see that the integrals over these sides tend to zero as the squares
grow.
Hence, for Re z < 0, as the squares grow, i.e., as m tends to innity, the integral
over the sides of the square tends to 0 and the integral along converges to the
integral along the path C from 2.2.2. In addition, the sum on the right side of the
above expression converges to the series for (1 z), i.e., we have
2ie
iz
sin(z)(z)(z) = 2
z+1

z
ie
iz
sin
_
z
2
_
(z 1).
Hence, we have
sin(z)(z)(z) = 2
z

z
sin
_
z
2
_
(1 z).
This yields
2 sin
_
z
2
_
cos
_
z
2
_
(z)(z) = 2
z

z
sin
_
z
2
_
(1 z).
and
cos
_
z
2
_
(z)(z) = 2
z1

z
(1 z).
for all z C. By substituting 1 z for z we get
2
z

1z
(z) = cos
_
(1 z)
2
_
(1 z)(1 z)
and nally
(z) = 2
z

z1
sin
_
z
2
_
(1 z)(1 z)
what completes the proof of the theorem.
2.3. Euler product formula. Let P be the set of all prime numbers in N. Assume
that P = {p
1
, p
2
, . . . } written in the natural ordering. For any m N, let S
m
be the
subset of N consisting of all integers which are not divisible by primes p
1
, p
2
, . . . , p
m
.
Then we claim that, for Re z > 1, we have
(z)
m

k=1
_
1
1
p
z
k
_
=

nSm
1
n
z
.
Clearly, p
1
= 2, and we have
_
1
1
2
z
_
(z) =

n=1
1
n
z

n=1
1
(2n)
z
=

nS1
1
n
z
.
So, the statement holds for m = 1.
NOTES ON RIEMANNS ZETA FUNCTION 11
Assume that the statement holds for m. Then, by the induction assumption, we
have
_
1
1
p
z
m+1
_
(z)
m

k=1
_
1
1
p
z
k
_
=
_
1
1
p
z
m+1
_

nSm
1
n
z
=

nSm
1
n
z

nSm
1
(p
m+1
n)
z
=

nSm+1
1
n
z
.
This establishes the above claim. Therefore, we see that as m tends to we get
the formula
(z)

k=1
_
1
1
p
z
k
_
= 1,
i.e., the following result holds.
2.3.1. Theorem (Euler product). For Re z > 1 we have
(z) =

kP
1
_
1
1
p
z
_.
The factors in Euler products are nonzero and holomorphic for Re z > 0.
The above observation lead Euler to a proof that the set P is innite. The
niteness of P would imply that the Euler product is holomorphic for Re z > 0,
contradicting the fact that has a pole at 1.
1
The Euler product formula implies that has no zeros for Re z > 1.
2.4. Riemann hypothesis. Since has no zeros for Re z > 1, and has no zeros
at all by 1.3.3, we see from the functional equation 2.2.1 that the only zeros of
for Re z < 0 come from zeros of the function z sin
_
z
2
_
at points 2, 4, . . . .
These are called the trivial zeros of .
It follows that all other zeros of have to lie in the strip 0 Re z 1. It is
called the critical strip.
1
Actually, Euler didnt use complex variables. One can show that niteness of primes would
contradict the fact that harmonic series diverge.
12 D. MILI

CI

Critical Strip
C
r
i
t
i
c
a
l

l
i
n
e

1
Simple pole
Im z
Re z
-2
zero
-1 0
Riemann conjectured that all zeros of in the critical strip lie on the critical line
Re z =
1
2
. This is the Riemann hypothesis.
Hadamard and de la Valee-Poussin proved that there are no zeros of on the
boundary of the critical strip (i.e., for Re z = 0 and Re z = 1). This implies the
Prime Number Theorem conjectured by Gauss.