A new approach towards solving the Riemann

hypothesis
Al-Mouid Mohammed Hasibul Haque

April 15, 2022

Abstract
In this paper, we create a new approach, using elementary tech-
niques from complex analysis, by which we attempt to prove the Rie-
mann hypothesis. We modify the gamma function and use that mod-
ified form to reformulate the Riemann xi function. Then using some
formulas that are similar to the Jacobi theta identity and analytic func-
tion manipulation techniques from complex analysis, we show that all
the non-trivial zeros of the Riemann zeta function must lie on the crit-
ical line. Then by recalling Hardy’s theorem we attempt to prove that
the Riemann hypothesis is true.
Keywords: Complex analysis, Riemann hypothesis, Riemann’s Func-
tional equation

1
1 Introduction
The Riemann hypothesis has been an unsolved mystery in the field of pure
mathematics for a long time.it is said to be the most daunting problem in
all of pure mathematics. It is an essential part of analytic number theory.
Hundreds of other theorems in analytic number theory depend on this hy-
pothesis being true. The heart of the problem is simple to understand. It
simply asks the question ‘Do all the non trivial zeros of the Riemann zeta
function lie on the critical line or are there some zeros that lie elsewhere
in the critical strip ?’ Or to put it in simple mathematical terms for all
<(s) ∈ (0, 1) is

ζ(s) = 0
only for <(s) = 21 . The ζ(s) function is a meromorphic function with residue
1. It has many forms but its most fundamental form is the following
∞
X 1
ζ(s) =
ns
n=1

Two functions that are very important in the study of this function are the
Gamma function and the xi function. These functions are the following
ˆ ∞
Γ(s) = xs−1 e−x dx
o

and
1 s s
ξ(s) = s(s − 1)π − 2 Γ( )ζ(s)
2 2
Euler was the first to notice that ζ(s) = 0 for s belonging to negative even
integers. These zeros are called the trivial zeros of the zeta function. Since
then many mathematicians have worked on this function and Riemann was
the first to prove that all the non-trivial zero of the zeta function lie on the
‘Critical strip’ which is the region where 0 < <(s) < 1. The distribution
of these zeros as well as various properties of the Riemann zeta function
and some consequences of the Riemann hypothesis are well documented in
E.C. Titchmarsh’s book[1]. Among the mathematicians who have proved
various theorems related to the Riemann hypothesis G.H. Hardy stands out
the most in terms of finding where the non-trivial zeros lie. G.H. Hardy has
proved that there are infinitely many non-trivial zeros in the ‘Critical line’
meaning the line where <(s) = 12 .

2
 In order to prove that the hypothesis is true, many mathematicians
have tried various different approaches. The most common approach is to
use analytic number theory to study the properties of the Riemann zeta
function and similar types of L-functions on the critical strip. From these
approaches we now know that more than two-fifths of the non-trivial zeros
of the zeta function lie on the critical line. This result can be found in J.
B. Conrey’s paper[2]. Our approach also uses complex analysis to get our
desired result. However the exact method that we use is different and hints
toward a more general condition by which we can test whether an L-function
has a Riemann hypothesis or not.
Now for the proof we will mainly use the method of contradiction. We
use the technique of contour integration on a function which is derived from
manipulating the definition of the ξ(s) function. Then we derive an inequal-
ity by manipulating the parts of the contour integral using some formulas
that are similar to the Jacobi theta identity and the assumption that the
value of ξ(s) is 0. By doing this we restrict ourselves to only the non-trivial
zeros of the ζ function. From this inequality we arrive at a contradiction
that occurs only if any non-trivial zero lies on the critical strip but not the
critical line. By doing so we prove that all the non-trivial zeros of the zeta
function must lie on the critical line. But before jumping into the proof let’s
prove two lemmas.

2 Preliminaries
These lemmas are as follows
s
Lemma 2.0.1. Let ξ(s) = 12 s(s−1)π − 2 Γ( 2s )ζ(s) be the Riemann xi function
which is an entire function. Since it is an entire function ξ(s) and ξ(s̄) must
exist. Then the following holds true

|ξ(s̄)| = |ξ(s)|

Proof. We can easily see that ξ(s) = |ξ(s)| because the output of ξ(s) is just
a complex number. So its complex conjugate must have the same modulus.
Since ξ(s) is an entire function, it can be expressed as an analytic function
in any region within the complex plane and the center of the region where
the function is analytic can be expressed as c, where c ∈ C or c ∈ R. For our
purposes we say that the region is an open disc D where D ⊂ C and we will
choose c ∈ R (Since the function is entire, it doesn’t matter which region we
choose or where the center is, because the function is by definition analytic

3
in the entire complex plane. So if we choose a real number in D to be the
center of our function then the function still remains analytic.) Now we can
write
X∞
ξ(s) = an (s − c)n
n=0
So
∞
X
ξ(s̄) = an (s̄ − c)n
n=0

and
∞
X
ξ(s) = an (s − c)n
n=0
∞
X
≤ |an | |(s̄ − c̄)|n
n=0

Because zn = (z̄)n where z ∈ C . Similarly

∞
X
|ξ(s̄)| = an (s̄ − c)n
n=0
∞
X
≤ |an ||(s̄ − c)|n
n=0
Now
∞
X ∞
X
n
ξ(s) − |ξ(s̄)| ≤ |an | |(s̄ − c̄)| − |an ||(s̄ − c)|n
n=0 n=0

and |an | = |an | where an ∈ C. So

∞
X
ξ(s) − |ξ(s̄)| ≤ |an | (|(s̄ − c̄)|n − |(s̄ − c)|n )
n=0

Recall that we chose c ∈ R so we have c̄ = c . Thus

∞
X
ξ(s) − |ξ(s̄)| ≤ |an | (|(s̄ − c)|n − |(s̄ − c)|n )
n=0

⇒ ξ(s) − |ξ(s̄)| ≤ 0

∴ ξ(s) ≤ |ξ(s̄)|

4
Similarly if we take
∞
X ∞
X
|ξ(s̄)| − ξ(s) ≤ |an ||(s̄ − c)| − n
|an | |(s̄ − c̄)|n
n=0 n=0

By a similar argument we arrive at

|ξ(s̄)| − ξ(s) ≤ 0

∴ |ξ(s̄)| ≤ ξ(s)

Now both of these inequalities must be true. And the only way both can be
true is if ξ(s) = |ξ(s̄)|. And by using the fact that ξ(s) = |ξ(s)| we can
clearly state that |ξ(s̄)| = |ξ(s)|. 

Now let’s move on to the second lemma. The way this lemma is proved
follows a similar approach to the proof of the classical Jacobi theta function
identity although with some modifications in the method (due to a slightly
different statement of the lemma). Now before we do this let’s define the
following two sets
C+ := {z ∈ C : =(z) > 0}
and
C− := {z ∈ C : =(z) < 0}

Lemma 2.0.2. For x ∈ C− − I the following holds true

πi
X
−n2 πx2 i e− 4 X n2 πi 12
e = e x
x
n∈Z n∈Z

Proof. For x ∈ C− − I we can clearly see that both the series in the L.H.S.
and R.H.S. converge. From Poisson’s summation formula we know
Xˆ ∞
−n2 πx2 i 2 2
X
e = e−y πx i e−2πiky dy
n∈Z k∈Z −∞

Xˆ ∞
2 x2 +2ky+( k )2 −( k )2 ]
= e−iπ[y x x dy
k∈Z −∞
Xˆ ∞ k 2 k 2
= e−iπ[(xy+ x ) −( x ) ]
dy
k∈Z −∞

5
 2
ˆ ∞
iπ k2 k 2
X
= e x e−iπ[(xy+ x ) ] dy
k∈Z −∞
ˆ ∞
k2 2 (y+ k )2
eiπ x2 e−iπx
X
= x2 dy
k∈Z −∞

k
Let y+ =z
x2
dy = dz
then as y → ∞, z → ∞ + ic
and as y → −∞, z → −∞ + ic
Where c is a positive constant which depends on =(x), <(x) and k. Now
ˆ ∞+ic ˆ ∞ ˆ −∞+ic ˆ ∞ 
−iπx2 z 2 −iπx2 z 2 −iπx2 z 2 −iπx2 z 2
e dz = e dz− e dz + e dz
−∞+ic −∞ −∞ ∞+ic

And here the latter two integrals have the same lengths but have opposite
orientations and for this reason after adding the two their effect on the sum
is nullified and thus
ˆ ∞+ic ˆ ∞
−iπx2 z 2 2 2
e dz = e−iπx z dz
−∞+ic −∞

We can also say that since the argument is the same thus we can just evaluate
the integral from −∞ to ∞ instead of −∞ + ic to ∞ + ic. Either way the
above equality remains the same. Then
X iπ k2 ˆ ∞
−n2 πx2 i 2 2
X
e = e x2 e−iπx z dz
n∈Z k∈Z −∞

2
r
X iπ k2 π
= e x
iπx2
k∈Z
1 X iπ k22
= √ e x
x i k∈Z
π
e−i 4 X iπ k22
= e x
x
k∈Z
πi
X
−n2 πx2 i e− 4 X n2 πi 12
∴ e = e x
x
n∈Z n∈Z


6
3 The Riemann hypothesis
The statement of the theorem is as follows

Theorem 3.1. Let ζ(s) = ∞ 1

P
n=1 ns be a meromorphic function that is an-
alytically continued to the entire complex plane except s = 1. Then all the
non-trivial zeros of ζ(s) for s∈ C lie on the line where <(s) = 12 .

This statement is slightly different from what Riemann originally hypoth-

esised and what Bombieri [4] stated in the description of this millennium
problem. But the statements are equivalent. Since G.H. Hardy already
proved that their are infinitely many zeros in the critical line we have an ad-
vantage here (A proof of Hardy’s theorem is found in [1, pp. 256-258]). If
we can simply prove that all the non-trivial zeros must lie on the line where
<(s) = 12 then it will be equivalent to proving the Riemann hypothesis.
In the proof we will use methods that are similar to the method first used
by Riemann [3] himself in order to define the Riemann xi function but our
method to define the xi function will be slightly different. Nonetheless the
Riemann’s functional equation will still apply and the Riemann xi function
will still be entire as we will see later on. Now let’s jump into the proof.

4 The Proof
Proof. Riemann’s functional equation states that

ξ(s) = ξ(1 − s)

We will need this later on. From the definition of the Gamma function we
have ˆ ∞
s s
Γ( ) = t 2 −1 e−t dt
2 0

By substituting t = n2 πx2 we get

ˆ ∞
s
s −1 2 2
Γ( ) = n2 πx2 2 e−n πx 2n2 πxdx
2 0
ˆ ∞
s s 2 2
⇒ Γ( ) = 2 ns π 2 xs−1 e−n πx dx
2 0
ˆ ∞
1 − s −s s 2 2
⇒ π n Γ( ) =
2 xs−1 e−n πx dx
2 2 0

7
By summing n from 1 to ∞ we get
∞ ∞ ˆ ∞
1 − s X −s s X 2 2
π 2 n Γ( ) = xs−1 e−n πx dx
2 2 0
n=1 n=1

The R.H.S. converges for <(s) > 1. So by convergence and using Fubini’s
theorem we have
ˆ ∞ ∞
1 −s s X 2 2
π ζ(s)Γ( ) =
2 x s−1
e−n πx dx
2 2 0 n=1

∞ ∞ ˆ
s(s − 1) − s s X 2 2
⇒ π ζ(s)Γ( ) = s(s − 1)
2 x s−1
e−n πx dx
2 2 0 n=1
ˆ ∞ ∞
2 2
X
∴ ξ(s) = s(s − 1) xs−1 e−n πx dx
0 n=1
Since the L.H.S. is holomorphic for s ∈ C(meaning the function is entire),so
the R.H.S. is also holomorphic for s ∈ C.1
By substituting s with s̄, where s̄ is the complex conjugate of s, we get
ˆ ∞ ∞
2 2
X
ξ(s̄) = s̄(s̄ − 1) xs̄−1 e−n πx dx
0 n=1

Since ξ(s̄) = ξ(1 − s̄) from the functional equation we can write
ˆ ∞ ∞ ˆ ∞ ∞
−n2 πx2 2 2
X X
s̄(s̄ − 1) x s̄−1
e dx = s̄(s̄ − 1) x−s̄ e−n πx dx (1)
0 n=1 0 n=1

Now we write
ˆ ∞ ∞ ˆ ∞ ∞
X
−n2 πx2
X 2 πx2 ξ(s̄) ξ(s)
x s̄−1
e dx − xs−1 e−n dx = − (2)
0 0 s̄(s̄ − 1) s(s − 1)
n=1 n=1

Since ξ(s) and ξ(s̄) are entire and <(s) ∈ (0, 1), so the R.H.S. is holomorphic.
Moreover both the integrals in the L.H.S. are also analytic for <(s) ∈ (0, 1).2
By using equation (1) we write equation (2) as
ˆ ∞ ∞ ˆ ∞ ∞
−s̄
X
−n2 πx2
X 2 2 ξ(s̄) ξ(s)
x e dx − xs−1 e−n πx dx = −
0 0 s̄(s̄ − 1) s(s − 1)
n=1 n=1
1
By analytic continuation we can show that the R.H.S. is analytic for s ∈ C and further
more we can show that if s ∈ (0, 1) then the integral itself is holomorphic without the
s(s − 1) multiplicative factor. We will give a rigorous proof of this in the appendix.
2
This is a trivial consequence of corollary A.1.1. in the appendix.

8
 ˆ ∞ ∞ ∞
X 2 πx2
X 2 πx2 1 s̄ s̄ 1 s s
⇒ (x−s̄ e−n − xs−1 e−n )dx = π − 2 Γ( )ζ(s̄) − π − 2 Γ( )ζ(s)
0 2 2 2 2
n=1 n=1

ˆ∞ X
∞
2 πx2 1 s̄ s̄ 1 s s
∴ e−n (x−s̄ − xs−1 )dx = π − 2 Γ( )ζ(s̄) − π − 2 Γ( )ζ(s) (3)
2 2 2 2
0 n=1

Here we can do this because ,as we prove in a corollary in the appendix, the
integral in the L.H.S. of equation (3) can be analytic for <(s) ∈ (0, 1) via
the method of analytic continuation. So the above equation actually holds
for <(s) ∈ (0, ∞). Now let’s see what will happen to ζ(s̄) if ζ(s) = 0. For
this we will go back to the ξ(s) function. From lemma 1 we can state that

|ξ(s)| = |ξ(s̄)|

1 s s 1 s̄ s̄
⇒ s(s − 1)π − 2 Γ( )ζ(s) = s̄(s̄ − 1)π − 2 Γ( )ζ(s̄)
2 2 2 2
Since <(s) ∈ (0, 1) and Γ(s) has no zeros for <(s) ∈ (0, 1) we can divide
1 s̄ s̄
both sides by s̄(s̄ − 1)π − 2 Γ( ) . Recalling that the absolute value of a
2 2
product is the product of the absolute value and switching sides we get
Γ( 2s )
|ζ(s̄)| = |ζ(s)|
Γ( 2s̄ )
1 − 2s
2 s(s − 1)π
Because 1 s̄ = 1, we ignored it in the expression above. Now
2 s̄(s̄ − 1)π− 2
Γ( 2s̄ ) 6= 0 for all <(s̄) ∈ (0, 1) and same goes for Γ( 2s ) for all <(s) ∈ (0, 1).
So if ζ(s) = 0 then by the expression above ζ(s̄) = 0.
Now we assume ξ(s) = 0 and by doing this we consider ζ(s) = 0 only
when <(s) ∈ (0, 1).Then equation (3) becomes
ˆ ∞
∞X
2 πx2
e−n (x−s̄ − xs−1 )dx = 0 (4)
0 n=1

This expression is valid for <(s) ∈ (0, 1) 3 and the L.H.S. of the above equa-
tion becomes zero whenever ξ(s) = 0.
3
Here each integral can be analytically continued by corollary A.1.1 in the appendix.
And since the combination of analytic functions is analytic so equation (4) remains valid
for <(s) ∈ (0, 1)

9
 =(s)

π
Rei 4

B
γ

π
4

0
A R <(s)

Figure 1: visualization of the contour

Now let’s consider a contour c . It comprises three parts. The first is

the A line, then the γ arc and finally¸ Pthe B line. Now we will consider the
counter-clockwise contour integral c ∞ n=1 e
−n2 πx2 (x−s̄ − xs−1 )dx. Since

<(s) ∈ (0, 1) and <(s̄) ∈ (0, 1) for all =(s) ∈ R and =(s̄) ∈ R, the integral
in equation(2) is holomorphic for both s and s̄. Then the contour integral
becomes zero by Cauchy’s integral theorem. Meaning
˛ X
∞
2 πx2
e−n (x−s̄ − xs−1 )dx = 0
c n=1

Now let’s dissect this contour integral into three parts.

10
 We can write
˛ X
∞ ˆ X
∞ ˆ X
∞
2 2 2 2 2 2
e−n πx (x−s̄ −xs−1 )dx = e−n πx (x−s̄ −xs−1 )dx+ e−n πx (x−s̄ −xs−1 )dx
c n=1 A n=1 γ n=1

ˆ X
∞
2 πx2
+ e−n (x−s̄ − xs−1 )dx
B n=1

We define the line A to start from 0 and go to R where R → ∞ and define

π
B to start from Rei 4 and go to 0 where R → ∞. Now
˛ X ∞ ˆ RX∞
2 2 2 2
e−n πx (x−s̄ − xs−1 )dx = lim e−n πx (x−s̄ − xs−1 )dx+
c n=1 R→∞ 0
n=1
ˆ X
∞ ˆ X
∞
2 πx2 2 πx2
e−n (x−s̄ − xs−1 )dx + e−n (x−s̄ − xs−1 )dx
B n=1 γ n=1

ˆ ∞
RX ˆ X
∞
2 πx2 2 πx2
∴ 0 = lim e−n (x−s̄ − xs−1 )dx + e−n (x−s̄ − xs−1 )dx+
R→∞ 0 B n=1
n=1

ˆ X
∞
2 πx2
e−n (x−s̄ − xs−1 )dx (5)
γ n=1

π
The angle for arc γ is 4. Now by using the ML inequality we can write

ˆ X
∞
2 πx2
e−n (x−s̄ − xs−1 )dx ≤ length of the arc times max value of the integrand
γ n=1

ˆ X
∞ ∞
2 πx2 πR −s̄ X 2 2
∴ e−n (x−s̄ − xs−1 )dx ≤ (R − Rs−1 ) e−n πR
γ n=1 4
n=1

Now R > Rs̄

and R > R1−s ,
so if we divide R by or by R1−s in both Rs̄
cases R will be larger and gives a value for the function which is greater
2 2 2 2
than 1. But R << en πR , so if we divide R byPen πR then we will find a
fraction less than 1. So if we multiply R with ∞ n=1 e
−n2 πR2 then we find
2 2
that R is multiplied with O(e−n πR ) . And as R → ∞ the product tends
2 2 2 2
to zero, because R << O(e−n πR ) and as R → ∞, O(e−n πR ) → 0.
ˆ X ∞
2 2
∴ e−n πx (x−s̄ − xs−1 )dx ≤ 0
γ n=1

11
 ˆ X
∞
2 πx2
∴ e−n (x−s̄ − xs−1 )dx = 0
γ n=1
´ P
Now we can write B ∞ n=1 e
−n2 πx2 (x−s̄ − xs−1 )dx as the following by sub-
π
stituting x with tei 4 and integrating with respect to t
ˆ X
∞ ˆ X
∞ ˆ X
∞
2 πx2 2 πx2 2 πx2
e−n (x−s̄ −xs−1 )dx = e−n x−s̄ dx− e−n xs−1 dx
B n=1 B n=1 B n=1

ˆ 0 ∞ ˆ 0 ∞
2 πx2 2 πx2
X X
= lim e−n x−s̄ dx − e−n xs−1 dx
R→∞ Rei π4 Re iπ
4
n=1 n=1
ˆ ∞
0X ˆ ∞
0X
π 2 πt2 π 2 πt2 π
= ei 4 lim e−in t−s̄ e−is̄ 4 dt − lim e−in ts−1 eis 4 dt
R→∞ R R→∞ R
n=1 n=1

Since t is a dummy variable we can change it with x. Then the expression

becomes
ˆ X ∞ ˆ 0X
∞
2 2 π 2 2 π
e−n πx (x−s̄ − xs−1 )dx = ei 4 lim e−in πx x−s̄ e−is̄ 4 dx−
B n=1 R→∞ R
n=1

ˆ ∞
0X
2 πx2 π
lim e−in xs−1 eis 4 dx
R→∞ R
n=1

If we let x → 0 but x 6= 0 then we can take a segment from B line that

πi
goes from e a to 0 where  → 0, then we find the integral in that small
segment to be
ˆ 0 X ∞
2 2
lim π
e−n πx (x−s̄ − xs−1 )dx
→0 ei 4
n=1
´0 P
We can then easily show that lim→0 ei π4 ∞
n=1 e
−n2 πx2 (x−s̄ −xs−1 )dx = 0.4
π
So we can easily disregard that segment and let the B line go from Rei 4 to
π
ei 4 , where R → ∞ and  → 0. Then the above expression becomes
ˆ X
∞ ˆ X
∞ ˆ  ∞
!
−n2 πx2 i π4 (1−s̄) −n2 πx2 i is π4 −n2 πx2 i
X
e (x−s̄ −xs−1 )dx = lim e e x−s̄ dx − e e xs−1 dx
B n=1 R→∞ R n=1 R
→0 n=1

4
This can be easily done using a length based argument. We will prove a theorem in
the appendix that will show that this is the case.

12
 ´ P ´ P∞ −n2 πx2 −s̄
Substituting the value of B ∞
n=1 e
−n2 πx2 (x−s̄ −xs−1 )dx and
γ n=1 e (x −
s−1
x )dx into equation(5) we get
ˆ ∞
RX
2 πx2
0 = lim e−n (x−s̄ − xs−1 )dx + 0
R→0 0
n=1

ˆ X
∞ ˆ X
∞
!
i π4 (1−s̄) −n2 πx2 i is π4 −n2 πx2 i
+ lim e e x−s̄ dx − e e xs−1 dx
R→∞ R n=1 R n=1
→0

Putting the value of the integral from equation(4) in the above equation
gives us
ˆ X
∞ ˆ X
∞
!
i π4 (1−s̄) −n2 πx2 i −s̄ is π4 −n2 πx2 i s−1
0 = lim e e x dx − e e x dx
R→∞ R n=1 R n=1
→0

ˆ ∞
RX ˆ ∞
RX
!
i π4 (1−s̄) −n2 πx2 i is π4 −n2 πx2 i
⇒ 0 = lim e e x−s̄ dx − e e xs−1 dx
R→∞  
→0 n=1 n=1

ˆ ∞
RX ˆ ∞
RX
i π4 (1−s̄) −n2 πx2 i −s̄ is π4 2 πx2 i
⇒ lim e e x dx = lim e e−n xs−1 dx
R→∞  R→∞ 
→0 n=1 →0 n=1

Writing s = σ + iy, where σ ∈ (0, 1) and y ∈ R, we get

ˆ ∞
RX ˆ ∞
RX
i π4 {1−(σ+iy)} −n2 πx2 i −(σ+iy) i(σ+iy) π4 2 πx2 i
lim e e x dx = lim e e−n xσ+iy−1 dx
R→∞  R→∞ 
→0 n=1 →0 n=1

Switching sides and rewriting the complex conjugates in the expression gives
ˆ ∞
RX ˆ ∞
RX
i(σ+iy) π4 −n2 πx2 i σ+iy−1 i π4 (1−σ+iy) 2 πx2 i
lim e e x dx = lim e e−n x−σ+iy dx
R→∞  R→∞ 
→0 n=1 →0 n=1

ˆ ∞
RX ˆ ∞
RX
i(2σ−1) π4 −n2 πx2 i σ+iy−1 2 πx2 i
⇒ lim e e x dx = lim e−n x−σ+iy dx
R→∞  R→∞ 
→0 n=1 →0 n=1

ˆ ∞
RX ˆ ∞
RX
2 πx2 +(2σ−1) π )i 2 πx2 i
∴ lim e(−n 4 xσ+iy−1 dx = lim e−n x−σ+iy dx
R→∞  R→∞ 
→0 n=1 →0 n=1

13
 (6)

Taking the L.H.S. of equation(6) and writing R → ∞ as ∞ (we can

do this because it doesn’t change the improper integral ) and using Euler’s
formula to write the exponential terms into terms of sines and cosines, we
get
ˆ ∞X ∞
π π
lim [cos(−n2 πx2 +(2σ −1) )+i sin(−n2 πx2 +(2σ −1) )]xσ+iy−1 dx
→0  4 4
n=1

Meanwhile R.H.S of equation(6) is

ˆ ∞X ∞
lim [cos(−n2 πx2 ) + i sin(−n2 πx2 )]x−σ+iy dx
→0 
n=1
 π ´ R P∞ −n2 πx2 i −s̄
Both the L.H.S. and the R.H.S. of equation (6) are part of limR→∞ ei 4 (1−s̄)  n=1 e x dx
π ´ R P∞
 →0
−n2 πx2 i xs−1 dx which, by the assumption of equation (4),
−eis 4  n=1 e
is equal to to zero as shown above. Since this entire integral is zero, so its
parts do not diverge. Also after using the Euler formula to express both
the integrals in terms of functions of sines and cosines as shown above,
we can obviously say that the above expressions also don’t diverge and as a
consequence the individual real and imaginary parts also don’t diverge. Fur-
thermore any manipulation or linear combination of these newly expressed
integrals also don’t diverge because of the same reason.
From L.H.S. of equation(6) and using the fact that xiy = eilnx we can
write
ˆ ∞X ∞
π π
lim [cos(−n2 πx2 +(2σ−1) )+i sin(−n2 πx2 +(2σ−1) )][cos(y ln x)+i sin(y ln x)]xσ−1 dx
→0  4 4
n=1

ˆ ∞
∞X
π π
= lim [cos(n2 πx2 −(2σ−1) )−i sin(n2 πx2 −(2σ−1) )][cos(y ln x)+i sin(y ln x)]xσ−1 dx
→0  4 4
n=1

Now letting n2 πx2 − (2σ − 1) π4 = A we can write the above expression as

ˆ ∞
∞X
lim [cos(A) − i sin(A)][cos(y ln x) + i sin(y ln x)]xσ−1 dx
→0 
n=1
ˆ ∞
∞X
= lim [cos A cos(y ln x)+sin A sin(y ln x)+i{cos A sin(y ln x)−sin A cos(y ln x)}]xσ−1 dx
→0 
n=1

14
 ˆ ∞
∞X
= lim [cos(A − y ln x) − i sin(A − y ln x)]xσ−1 dx
→0 
n=1

Similarly by letting n2 πx2 = B in the R.H.S. we get

ˆ ∞
∞X
lim [cos(−B) + i sin(−B)][cos(y ln x) + i sin(y ln x)]x−σ dx
→0 
n=1

ˆ ∞
∞X
= lim [cos(B − y ln x) − i sin(B − y ln x)]x−σ dx
→0 
n=1

Now letting A − y ln x = A0 and B − y ln x = B 0 , and then equating the

L.H.S. and R.H.S. of equation(6) we get

ˆ ∞
∞X ˆ ∞
∞X
0 0
lim (cos A − i sin A )x σ−1
dx = lim (cos B 0 − i sin B 0 )x−σ dx
→0  →0 
n=1 n=1

(7)

Since equation (7) is a simple manipulation of equation (6), so no term in

equation (7) diverges. Equating the real and imaginary parts of equation
(7) we get
ˆ ∞
∞X ˆ ∞
∞X
lim cos A0 xσ−1 dx = lim cos B 0 x−σ dx
→0  →0 
n=1 n=1

(8)

And ˆ ˆ
∞
∞X ∞
∞X
0 σ−1
lim sin A x dx = lim sin B 0 x−σ dx
→0  →0 
n=1 n=1

(9)

Since equation (7) doesn’t have any terms that diverge so both equation (8)
and equation (9) don’t have any terms that diverge, because both expressions
in the L.H.S. and R.H.S. of equation(7) don’t diverge and since as a whole
the expressions don’t diverge so their individual real and imaginary parts

15
don’t diverge. Putting the value of A0 and B 0 in equation (8) and (9),
multiplying equation (9) with i and adding both equations together gives
ˆ ∞
∞X
π π
lim [cos(n2 πx2 −(2σ−1) −y ln x)+i sin(n2 πx2 −(2σ−1) −y ln x)]xσ−1 dx
→0  4 4
n=1
ˆ ∞
∞X
= lim [cos(n2 πx2 − y ln x) + i sin(n2 πx2 − y ln x)]x−σ dx
→0 
n=1
ˆ ∞
∞X ˆ ∞
∞X
(n2 πx2 i− πσi + iπ −iy ln x) σ−1 2 πx2 i−iy ln x)
⇒ lim e 2 4 x dx = lim e(n x−σ dx
→0  →0 
n=1 n=1
ˆ ∞
∞X ˆ ∞
∞X
iπ 2 πx2 i iπσ 2 πx2 i
⇒ lim e 4 en xσ−1−iy dx = lim e 2 en x−σ−iy dx
→0  →0 
n=1 n=1

ˆ ∞
∞X ˆ ∞
∞X
iπ iπσ
n2 πx2 i σ−1−iy 2 πx2 i
∴ 2e 4 lim e x dx = 2e 2 lim en x−σ−iy dx
→0  →0 
n=1 n=1

(10)

Since the L.H.S. and R.H.S. of equation (10) is a combination of the L.H.S.
and R.H.S. of equation (8) and equation (9) respectively and both equations
don’t have any terms that diverge, so both the L.H.S. and R.H.S. of equation
(10) don’t diverge.
Moreover we can write equation(6) as the following
ˆ ∞
∞X ˆ ∞
∞X
− iπ −n2 πx2 i σ−1+iy − iπσ 2 πx2 i
2e 4 lim e x dx = 2e 2 lim e−n x−σ−iy dx
→0  →0 
n=1 n=1
ˆ ∞
∞X ˆ ∞
∞X
iπσ iπ
−n2 πx2 i σ−1+iy 2 πx2 i
∴ 2e 2 lim e x dx = 2e 4 lim e−n x−σ−iy dx
→0  →0 
n=1 n=1

(11)

Now we can simply shift the line of integration to be from  → R to  + iδ →

R + iδ , where δ → 0, and by doing this the value of the integral doesn’t
change as the argument of both lines of integration are zero. A similar logic
applies when we consider the path to go from  − iδ → R − iδ , where δ → 0.
So we can write

16
 ˆ ∞
∞X ˆ ∞
∞−iδ X
−n2 πx2 i σ−1+iy 2 πx2 i
lim e x dx = lim e−n xσ−1+iy dx
→0  δ→0 −iδ
n=1 →0 n=1

(12)

ˆ ∞
∞X ˆ ∞
∞−iδ X
2 πx2 i 2 πx2 i
lim e−n x−σ−iy dx = lim e−n x−σ−iy dx
→0  δ→0 −iδ
n=1 →0 n=1

(13)

ˆ ∞
∞X ˆ ∞
∞+iδ X
n2 πx2 i −σ−iy 2 πx2 i
lim e x dx = lim en x−σ−iy dx
→0  δ→0 +iδ
n=1 →0 n=1

(14)
and
ˆ ∞
∞X ˆ ∞
∞+iδ X
n2 πx2 i σ−1−iy 2 πx2 i
lim e x dx = lim en xσ−1−iy dx
→0  δ→0 +iδ
n=1 →0 n=1

(15)
So from equation (10) and equation (11) we can write
ˆ ∞
∞−iδ X ˆ ∞
∞−iδ X
−n2 πx2 i −σ−iy 2 πx2 i
lim e x dx = lim e−n xσ−1+iy dx
δ→0 −iδ δ→0 −iδ
→0 n=1 →0 n=1

(16)
and
ˆ ∞
∞+iδ X ˆ ∞
∞+iδ X
n2 πx2 i −σ−iy 2 πx2 i
lim e x dx = lim en xσ−1−iy dx
δ→0 +iδ δ→0 +iδ
→0 n=1 →0 n=1

(17)
P∞ n2 πx2 i
P∞ −n2 πx2 i
P n2 πx2 i
Now n=1 e and n=1 e are part of the sums n∈Z e and
−n2 2
e πx i
P
n∈Z respectively, both of which abide by the second lemma stated

17
in the preliminary. This is because here x ∈ C− − I so lemma 2.0.2 can be
applied to the R.H.S. of the above expression. So we can write the following
equations from lemma 2.0.2
iπ
X
−n2 πx2 i e− 4 X n2 π 12 i
e = e x (18)
x
n∈Z n∈Z

And
2π 1 i iπ 2 πx2 i
e−n
X X
x2 = e− 4 x en (19)
n∈Z n∈Z

where the last equation can be easily proved the same way lemma 2.0.2 was
proved which holds for x ∈ C+ − I.
From equation (18) we have
∞ iπ ∞
!
X
−n2 πx2 i e− 4 X n2 π 12 i
1+2 e = 2 e x +1
x
n=1 n=1

∞ iπ iπ ∞
X 2 πx2 i e− 4 e− 4 X n2 π 12 i
⇒2 e−n = +2 e x −1
x x
n=1 n=1

∞ ∞
iπ X 2 πx2 i 1 2 X n2 π 12 i iπ
∴ 2e 4 e−n = + e x −e4 (20)
x x
n=1 n=1

From equation(19) we have

∞ ∞
!
−n2 π 12 i iπ
n2 πx2 i
X X
1+2 e x = −ie x 1 + 2
4 e
n=1 n=1

1
P∞ −n2 π i ∞
1 n=1 e
x2 X 2 πx2 i
⇒ iπ +2 iπ =1+2 en
x(−ie ) 4 x(−ie ) 4
n=1
P∞ −n2 π 12 i ∞
i i n=1 e
x X 2 πx2 i
⇒ iπ +2 iπ == 1 + 2 en
x(e )
4 x(e ) 4
n=1
P∞ −n2 π 12 i ∞
i i n=1 e
x iπ iπ X 2 πx2 i
⇒ +2 = e 4 + 2e 4 en
x x
n=1

18
 1
∞ P∞ −n2 π i
iπ X
n2 πx2 i i i n=1 e
x2 iπ
∴ 2e 4 e = +2 −e4 (21)
x x
n=1

Using equation (21) on the L.H.S. of equation (17) we get

"ˆ ˆ ∞+iδ X∞ ˆ ∞+iδ #
∞+iδ
i σ−1−iy −n2 π 12 i σ−2−iy iπ
L.H.S. = lim x dx + 2i e x x dx − e 4 xσ−1−iy dx
δ→0 +iδ x +iδ +iδ
→0 n=1

´ ∞+iδi σ−1−iy
´ ∞+iδ P∞ −n2 π 12 i σ−2−iy
Writing limδ→0 +iδ x x dx = I,2i limδ→0 +iδ n=1 e
x x dx =
iπ ´ ∞+iδ
→0 →0
E and limδ→0 e 4 +iδ xσ−1−iy dx = J we get,
→0

L.H.S. = I + E − J (22)

Now ˆ ∞
∞+iδ X 2π 1 i
E = 2i lim e−n x2 xσ−2−iy dx
δ→0 +iδ
→0 n=1

By substituting
1
x=
t
dx = −t−2 dt
While changing the limits we notice, as x →  + iδ and  → 0, δ → 0 then
1
t → +iδ 1
and +iδ → ∞ − iδ 0 . And since δ is infinitesimally small so δ 0 = δ
1
and thus +iδ → ∞ − iδ . By a similar argument as x → ∞ + iδ t →  − iδ.
ˆ ∞
−iδ X
2 πt2 i
∴ E = −2i lim e−n t2−σ+iy t−2 dt
δ→0 ∞−iδ
→0 n=1

ˆ ∞
∞−iδ X
2 πt2 i
= lim 2i e−n t−σ+iy dt
δ→0 −iδ
→0 n=1

Now since t is a dummy variable we can switch it with x

ˆ ∞
∞−iδ X
2 πx2 i
∴ E = 2i lim e−n x−σ+iy dx
δ→0 −iδ
→0 n=1

19
 iπ
Now multiplying e 4 to both sides of equation (16) gives us
ˆ ∞
∞−iδ X ˆ ∞
∞−iδ X
iπσ
+ iπ −n2 πx2 i σ−1+iy πi 2 πx2 i
2e 2 4 lim e x dx = 2e 2 lim e−n x−σ+iy dx
δ→0 −iδ δ→0 −iδ
→0 n=1 →0 n=1

ˆ ∞
∞+iδ X ˆ ∞
∞+iδ X
iπσ
+ iπ −n2 πx2 i σ−1+iy 2 πx2 i
∴ 2e 2 4 lim e x dx = 2i lim e−n x−σ+iy dx
δ→0 +iδ δ→0 +iδ
→0 n=1 →0 n=1

(23)

The R.H.S. of this equation is equal to E. So, we can substitute E with

the value of the L.H.S. of the above equation. From equation (23) we take
the L.H.S. and get
ˆ ∞
∞−iδ X
iπσ
+ iπ 2 πx2 i
L.H.S. = 2e 2 4 lim e−n xσ−1+iy dx
δ→0 −iδ
→0 n=1

ˆ ∞−iδ ∞
iπσ iπ X 2 πx2 i
=e 2 lim 2e 4 e−n xσ−1+iy dx
δ→0 −iδ
→0 n=1

" ˆ ∞−iδ ˆ ∞
∞−iδ X ˆ ∞−iδ
#
iπσ iπσ iπσ iπ
−n2 πx2 i
= lim e 2 xσ−2+iy dx + 2e 2 e xσ−2+iy dx − e 2 e xσ−1+iy dx
4
δ→0 −iδ −iδ −iδ
→0 n=1

The last expression comes from equation (20). Now let’s observe
ˆ ∞
∞−iδ X
iπσ 2π 1 i
2e 2 lim e−n x2 xσ−2+iy dx
δ→0 −iδ
→0 n=1

Substituting
1
x=
t
∴ dx = −t−2 dt
And the upper and lower limits become  + iδ and ∞ + iδ respectively. Since
the same type of change of variables that was done on E, the same logic for
change in the upper and lower limits apply here as well. Then we get
ˆ ∞
+iδ X
iπσ 2 πt2 i
2e 2 lim en t2−σ−iy (−t−2 )dt
δ→0 ∞+iδ
→0 n=1

20
 ˆ ∞
∞+iδ X
iπσ 2 πt2 i
= 2e 2 lim en t−σ−iy dt
δ→0 +iδ
→0 n=1

[By the same argument used before for E]

Since t is a dummy variable we can write the above expression as
ˆ ∞
∞+iδ X
iπσ 2 πx2 i
2e 2 lim en x−σ−iy dx
δ→0 +iδ
→0 n=1

iπσ ´ ∞−iδ iπσ ´ ∞−iδ iπ

Writing e 2 limδ→0 −iδ xσ−2+iy dx = I1 and e 2 limδ→0 −iδ e 4 xσ−1+iy dx =
→0 →0
J1 , we find the L.H.S. of equation (23) to be
ˆ ∞
∞+iδ X
iπσ 2 πx2 i
I1 + 2e 2 lim en x−σ−iy dx − J1
δ→0 +iδ
→0 n=1

Substituting E with the L.H.S. of equation (23) into equation (22) and
rearranging we get
ˆ ∞
∞+iδ X ˆ ∞
∞+δ X
iπσ iπσ
n2 πx2 i −σ−iy 2 πx2 i
I+I1 +2e 2 lim e x dx−J1 −J = 2e 2 lim en x−σ−iy dx
δ→0 +iδ δ→0 +iδ
→0 n=1 →0 n=1

⇒ I + I1 − J − J1 = 0
⇒ 0 = I + I1 − J − J1
ˆ ∞+iδ ˆ ∞−iδ ˆ ∞+iδ
i σ−1−iy iπσ
σ−2+iy iπ
⇒ 0 = lim x dx + e x
2 dx − e 4 xσ−1−iy dx
δ→0 +iδ x −iδ +iδ
→0
ˆ ∞−iδ 
iπσ iπ
σ−1+iy
−e 2 e x 4 dx
−iδ
 ˆ ∞+iδ ˆ ∞−iδ ˆ ∞+iδ
iπ iπσ iπ
⇒ 0 = lim e 2 xσ−2+iy dx + e 2 xσ−2+iy dx − e 4 xσ−1−iy dx
δ→0 +iδ −iδ +iδ
→0
ˆ ∞+iδ 
iπσ iπ
−e 2 e 4 xσ−1+iy dx
+iδ
Now in order to calculate the integrals we are going to consider the line of
integration to go from  → ∞ instead of  − iδ → ∞ − iδ or  + iδ → ∞ + iδ
. This is because the values of the integrals don’t change (Since all of
the complex integrals above have lines of integration that are parallel to

21
the real number line, so it doesn’t matter whether we integrate over the
complex numbers in the given limits of integration or we integrate over the
real numbers with limits of integration going from  → ∞ ). So we can write
  σ−1−iy ∞  σ−1+iy ∞  σ−iy ∞  σ+iy ∞ 
iπ x iπσ x iπ x iπ
+ iπσ x
0 = lim e 2 +e 2 −e4 −e4 2
→0 σ − 1 − iy  σ − 1 + iy  σ − iy  σ + iy 
 σ−1−iy   σ−1+iy    σ−iy 
iπ  iπσ  iπ x
⇒ 0 = 0 − e lim
2 +0−e 2 lim − e lim
4 −0
→0 σ − 1 − iy →0 σ − 1 + iy x→∞ σ − iy
 σ+iy  
iπ
+ iπσ x
+e 4 2 lim −0
x→∞ σ + iy

These sort of limits exist due to the fact that σ ∈ (0, 1). Now since x → 
and  → 0 then i simply write x → 0 directly. Because σ − 1 <
we can
h σ−1±iy
x
0 so limx→∞ σ−1±iy = 0 and no added singularities arise in the entire
expression due to this substitution and furthermore the value of the integral
does not change.
Now
 σ−1−iy   σ−1+iy   σ−iy   σ+iy 
iπ x iπσ x iπ x iπ
+ iπσ x
0 = −e lim
2 −e 2 lim −e lim
4 −e 4 2 lim
x→0 σ − 1 − iy x→0 σ − 1 + iy x→∞ σ − iy x→∞ σ + iy

xσ−1−iy
   σ−1+iy   σ−iy   σ+iy 
iπ iπσ x iπ x iπ iπσ x
⇒ 0 = e 2 lim + e 2 lim − e 4 lim − e 4 + 2 lim
x→0 1 − σ + iy x→0 1 − σ − iy x→∞ σ − iy x→∞ σ + iy

  1−σ+iy   1−σ−iy 
iπ x iπσ x
⇒ 0 = e 2 lim + e 2 lim
x→∞ 1 − σ + iy x→∞ 1 − σ − iy
  σ−iy   σ+iy 
iπ x iπ
+ iπσ x
− e lim
4 +e 4 2 lim
x→∞ σ − iy x→∞ σ + iy

" πi iπσ
#
e 2 x1−σ+iy (1 − σ − iy) + e 2 x1−σ−iy (1 − σ + iy)
⇒ 0 = lim
x→∞ (1 − σ + iy)(1 − σ − iy)
" πi iπσ iπ
#
e 2 xσ−iy (σ + iy) + e 2 + 4 xσ+iy (σ − iy)
− lim
x→∞ (σ − iy)(σ + iy)

πi iπσ
[e 2 x1−σ+iy (1 − σ − iy) + e 2 x1−σ−iy (1 − σ + iy)](σ 2 + y 2 )
⇒ 0 = lim
x→∞ [(1 − σ)2 + y 2 ](σ 2 + y 2 )

22
 iπ iπσ iπ
[e 4 xσ−iy (σ + iy) + e 2 + 4 xσ+iy (σ − iy)]{(1 − σ)2 + y 2 }
− lim
x→∞ [(1 − σ)2 + y 2 ](σ 2 + y 2 )
Taking the absolute value of both sides yields
πi iπσ
[e 2 x1−σ+iy (1 − σ − iy) + e 2 x1−σ−iy (1 − σ + iy)](σ 2 + y 2 )
0 = lim
x→∞ [(1 − σ)2 + y 2 ](σ 2 + y 2 )
iπ iπσ iπ
[e 4 xσ−iy (σ + iy) + e 2 + 4 xσ+iy (σ − iy)]{(1 − σ)2 + y 2 }
− lim
x→∞ [(1 − σ)2 + y 2 ](σ 2 + y 2 )
πi iπσ
[e 2 x1−σ+iy (1 − σ − iy) + e 2 x1−σ−iy (1 − σ + iy)](σ 2 + y 2 )
≥ lim
x→∞ [(1 − σ)2 + y 2 ](σ 2 + y 2 )
iπ iπσ iπ
[e 4 xσ−iy (σ + iy) + e 2 + 4 xσ+iy (σ − iy)]{(1 − σ)2 + y 2 }
− lim
x→∞ [(1 − σ)2 + y 2 ](σ 2 + y 2 )
πi iπσ
[e 2 xiy (1 − σ − iy) + e 2 x−iy (1 − σ + iy)](σ 2 + y 2 ) 1−σ
∴ 0 ≥ lim x
x→∞ [(1 − σ)2 + y 2 ](σ 2 + y 2 )
iπ iπσ iπ
[e 4 x−iy (σ + iy) + e 2 + 4 xiy (σ − iy)]{(1 − σ)2 + y 2 } σ
− lim x
x→∞ [(1 − σ)2 + y 2 ](σ 2 + y 2 )

(24)

Now let’s consider the terms inside the two moduli of the R.H.S. We
can see that in the first modulus, even when y goes to infinity (since the
denominator is larger) the modulus doesn’t tend to infinity. The same is
also true for the second modulus. So the only term that we have to worry
about now is x±iy where x → ∞. Since |x±iy | = 1 we don’t have to worry
about it.
The reason why this is true is because when we let the limit x → ∞
for a function of the form Cxα±iy + Kxβ±iy (Where C and K are arbitrary
constants belonging to R) and take the absolute value of the limit, then we
find that
Cxα±iy + Kxβ±iy ≤ |Cxα±iy | + |Kxβ±iy |

⇒ Cxα±iy + Kxβ±iy ≤ |C|xα + |K|xβ

In this case we see that the R.H.S. of the above expression doesn’t depend
on |x±iy | and if α and β are less than or equal to 0 then the R.H.S. above

23
goes to zero and if and only if α and β are greater than 0 then the R.H.S.
goes to infinity as x → ∞.
So by the same reasoning, in inequality (24) x±iy doesn’t have any rel-
evant impact on the value of the R.H.S. and we can thus replace the two
absolute values with two positive (finite) constants that are fixed for specific
values of σ and y and don’t depend on x.
Now, letting the R.H.S. of inequality (24) be S, we can write

S = lim C1 x1−σ − C2 xσ
x→∞

Where C1 and C2 are two positive constants belonging to R. Now

lim C1 x1−σ − C2 xσ
x→∞

C1 x − C2 x2σ
= lim
x→∞ xσ
2σ
Let limx→∞ C1 x−C xσ
2x
= L. Then there are two cases.
Case − 1 : 0 < σ < 12
then 2σ < 1
2σ
So,L := limx→∞ C1 x−C xσ
2x
→ ∞ as C1 x grows much faster than C2 x2σ
and dividing by x doesn’t effect the result. Even if we only consider C1 x1−σ
σ

and C2 xσ then we can still see that C1 x1−σ grows much faster then C2 xσ for
all σ ∈ (0, 21 ) and the limit goes to infinity as x → ∞. Then inequality (24)
doesn’t hold true. Since in the interval 0 < σ < 12 inequality (24) doesn’t
hold true then we can clearly see a contradiction and thus prove that there
are no zeros of ξ(s) for <(s) ∈ (0, 12 ) and thus there are no zeros of ζ(s) for
<(s) ∈ (0, 12 ) . This applies for all =(s) ∈ R because we have already proven
that ζ(s̄) = 0 if ζ(s) = 0 and thus the positive or negative value of y doesn’t
effect the final result.
Case − 2 : 12 < σ < 1
then 2σ > 1
2σ
So, L := limx→∞ C1 x−C xσ
2x
→ −∞ because now C2 x2σ grows much
faster than C1 x and dividing by xσ doesn’t effect the result. Then inequality
(24) holds and we find no apparent contradiction. Then based on this one
might assume that ζ(s) can have zero in the interval <(s) ∈ ( 21 , 1). But we
know from the identity ξ(s) = ξ(1 − s) that if ζ(s) = 0 then ζ(1 − s) = 0
where s = σ + iy and 1 − s = 1 − σ − iy and as stated before, positive or
negative values of y doesn’t effect the result. Now σ ∈ ( 21 , 1) so 1−σ ∈ (0, 12 )
and from Case-1 we have already shown that ζ(s) 6= 0 for <(s) ∈ (0, 12 ). We

24
find another clear contradiction here and thus we can say that Case-2 is also
not true.
However, if we let <(s) = 12 then not only inequality (24) but also every
equation starting from equation(4) holds true. Thus we can conclude that
ζ(s) can have a non-trivial zero if and only if <(s) = 21 . And by using
Hardy’s theorem we can clearly state that all the non-trivial zeros of ζ(s)
,where s∈ C, lie on the line where <(s) = 12 and there are infinitely many of
them. 

5 Remarks
The proof here has been done mostly based on methods from complex anal-
ysis.Though this approach to prove the Riemann hypothesis is very elemen-
tary but it should serve the cause.
According to the approach if any function has a Riemann hypothesis like
the zeta function then it must abide by an inequality similar to inequality
(24), which would force the function to only have nontrivial zeros on the
critical line, for this approach to apply for that function. So cases such as
the Epstein series [5] do not apply for this approach because even though
these functions have functional equations similar to the Riemann functional
equation, but these functions don’t give us any inequality similar to inequal-
ity (24) and thus our approach doesn’t apply. However, the approach does
not guarantee that all L-functions have a Riemann hypothesis, because for
that to hold true we need to prove that there are finitely or (more likely)
infinitely many nontrivial zeros that lie on the critical line and if some L-
function does not give us an inequality similar to inequality (24) then it
doesn’t automatically say that the function doesn’t have a Riemann hy-
pothesis but rather it would mean that our approach doesn’t apply for that
particular L-function. So this approach might not apply for the generalized
Riemann hypothesis but further investigation is needed for that assertion.
The approach we have created is so elementary that even the Euler prod-
uct was not explicitly used in our main theorem. However we see something
interesting when we examine the integral in equation (3). We know that the
Euler product the zeta function is the following
Y 1
ζ(s) =
p
1 − p−s

We know that the above expression leads to the usual series representation
of the zeta function. Now the usual series representation of the zeta function

25
is linear. Not only that but each term in the Euler product representation
is also linear. Because of this linearity we can derive the particular repre-
sentation of ξ(s) that we have utilized in our proof and from this particular
representation of the ξ(s) and the identity ξ(s) = ξ(1 − s) we derived equa-
tion (3) which holds true for <(s) > 1. This was because both the series
representation and product representation of the zeta function converges ab-
solutely in the region <(s) > 1. Then we analytically continued the integral
in equation (3) to also hold for the region <(s) ∈ (0, 1). This same logic
also holds for other L-functions that have a Euler product. However it is
not clear whether our new approach holds true for proving the Riemann
hypothesis in the general case (though it is most likely that if the approach
actually works for the ordinary Riemann hypothesis then a similar approach
will also work for other Dirichlet L-functions).
Finally even if it is the case that the given proof above is inadequate to
prove the Riemann hypothesis, the approach used in the proof is still worth
investigating further as such approaches may help uncover some properties
of L-functions which might not be apparent currently.
Data availability statement: The manuscript has no associated data

26
 Appendix

A Proof of the convergence of the integral in the

Riemann xi function
This theorem uses a modified version of the argument presented in [1] which
proves that the Riemann xi function is entire. Now let’s start by stating the
theorem

Theorem A.1. let´ ∞the definition of the Riemann xi function be the following
ξ(s) := s(s − 1) 0 xs−1 ∞ −n2 πx2 dx . Then ξ(s) is analytic for all
P
n=1 e
s ∈ C.

Proof. We have
ˆ ∞ ∞
2 πx2
X
ξ(s) = s(s − 1) xs−1 e−n dx
0 n=1
P∞ −n2 πx2
Lets only consider the integral first. For simplicity we will consider n=1 e =
ψ(x2 ). Then
ˆ ∞ ∞ ˆ ∞
−n2 πx2
X
s−1
x e dx = xs−1 ψ(x2 )dx
0 n=1 0

ˆ ∞ ˆ 1
= xs−1 ψ(x2 )dx + xs−1 ψ(x2 )dx
1 0
ˆ ∞ ˆ 1
= xs−1 ψ(x2 )dx + y s−1 ψ(y 2 )dy
1 0

Let y = x1 and dy = − x12 dx. Then the upper limit becomes 1 and the lower
limit becomes ∞. Then the latter integral becomes the following
ˆ 1
1 1
x1−s ψ( 2 )(− 2 )dx
∞ x x
ˆ ∞
1
= x−s−1 ψ( 2 )dx
1 x
We know that one of the Jacobi theta functions is
2 2
X
θ(x2 ) = e−n πx
n∈Z

27
 = 2ψ(x2 ) + 1
And it abides by the following identity
1 1
θ( ) = θ(x2 )
x x2
1
⇒ θ( ) = xθ(x2 )
x2
We can do this because x doesn’t approach 0 in the interval of the integral
we are concerned with. By using this identity we can write
1 1
2ψ( 2
) = θ( 2 ) − 1
x x
= 2xψ(x2 ) + x − 1
1 x 1
∴ ψ( 2
) = + xψ(x2 ) −
x 2 2
Then we can write
ˆ ∞ ˆ ∞
−s−1 1 x 1
x ψ( 2 )dx = x−s−1 ( + xψ(x2 ) − )dx
1 x 1 2 2
ˆ ∞ ˆ ∞ ˆ ∞
1 1
= x−s dx + x−s ψ(x2 )dx − x−s−1 dx
2 1 1 2 1
ˆ ˆ ˆ ∞
1 ∞ −s−1 1 ∞ −s
=− x dx + x dx + x−s ψ(x2 )dx
2 1 2 1 1
ˆ ∞
1 1
=− + + x−s ψ(x2 )dx
2s 2(s − 1) 1
ˆ ∞
1
= + x−s ψ(x2 )dx
2s(s − 1) 1
The above expression is always true for <(s) > 1. The integrand of the
last integral converges absolutely for all s ∈ C and x ∈ (1, ∞), because
−πx2
the integrand is of order O( e xs ). So this expression can be analytically
continued for all s ∈ C. Then if we simply include this dissected form of the
integral into ξ(s), we find
ˆ ∞ ˆ ∞ 
s−1 2 1 −s 2
ξ(s) = s(s − 1) x ψ(x )dx + + x ψ(x )dx
1 2s(s − 1) 1
ˆ ∞ ˆ ∞
1
⇒ ξ(s) = s(s − 1) s−1 2
x ψ(x )dx + + s(s − 1) x−s ψ(x2 )dx
1 2 1

28
 Here the integrand of the first integral is analytic and the last integral
converges absolutely as stated above. So the R.H.S.´ is analytic
P for −n all of s ∈
∞
C. Thus we can clearly state that ξ(s) = s(s − 1) 0 xs−1 ∞
2 πx2
n=1 e dx
is analytic for all s ∈ C. 
´ ∞ s−1 P∞ −n2 πx2
Corollary A.1.1. For <(s) ∈ (0, 1) the integral 0 x n=1 e dx
is analytic.
´∞
Proof. From the proof above we can write 0 xs−1 ∞ −n2 πx2 dx as the
P
n=1 e
following
ˆ ∞ ∞ ˆ ∞ ˆ ∞
X 2 2 1
xs−1 e−n πx dx = xs−1 ψ(x2 )dx + x−s−1 ψ( 2 )dx
0 1 1 x
n=1

and ˆ ∞ ˆ ∞
−s−1 1 1
x ψ( 2 )dx = + x−s ψ(x2 )dx
1 x 2s(s − 1) 1
P∞ −n2 πx2
Where ψ(x2 ) = n=1 e . Here we can clearly see that the integral
is
´ ∞analytic for s ∈ C − {0, 1} , so obviously for <(s) ∈ ´ (0, 1) the
P∞integral
x −s−1 ψ( 1 )dx is analytic and as a result the integral ∞ xs−1 −n2 πx2 dx
1 x2 0 n=1 e
is analytic. 

B An additional theorem
This additional
´ 0 P∞theorem will suffice to show that not including the integral
−n2 πx2
lim→0 ei 4 n=1 e
π (x−s̄ − xs−1 )dx in the expansion of the B segment
was only logical because this integral goes to zero. Let’s now state and prove
the theorem.
´0 P
Theorem B.1. Let s ∈ C and let lim→0 ei π4 ∞ n=1 e
−n2 πx2 (x−s̄ − xs−1 )dx
πi
lie on a line segment´ on the
P∞complex plane that goes from e 4 to 0 where
0 −n 2 πx2
 → 0. Then lim→0 ei π4 n=1 e (x−s̄ − xs−1 )dx = 0.
´∞
Proof. We recall from the corollary above that the integral 0 xs−1 ∞ −n2 πx2 dx
P
n=1 e
converges for <(s) ∈ (0, 1). So the integrand converges and thus holomorphic
in the complex plane if s ∈ C and <(s) ∈ (0, 1). So, if we take a semi-circle
within the region enclosed by the set U then we will find the contour inte-
gral along the boundary of the semicircle to be zero by Cauchy’s integral
theorem. Meaning
˛ ∞
2 2
X
−s̄
(x − x ) s−1
e−n πx dx = 0
Λ n=1

29
Here I have taken the semicircle over the set U 0 ⊂ U and I have taken the
contour in such a manner that the path along the contour goes clockwise
instead of counter-clockwise. Meaning
˛ X
∞  X
∞
−n2 πx2 −s̄ 2 πx2
e (x −x s−1
)dx = e−n (x−s̄ − xs−1 )dx
Λ n=1 Λ n=1

Let’s dissect the contour integral

 X
∞ ˆ 0 ∞ ˆ X
∞
−n2 πx2 −n2 πx2 2 2
X
−s̄ −s̄
e (x −x s−1
)dx = lim e (x −x s−1
)dx+ e−n πx (x−s̄ −xs−1 )dx
Λ n=1 →0 ei π4 λ n=1
n=1

ˆ 0 ∞  X
∞
−n2 πx2 2 πx2
X
−s̄
⇒ lim π
e (x −x s−1
)dx = e−n (x−s̄ − xs−1 )dx−
→0 ei 4 Λ n=1
n=1

ˆ X
∞
2 πx2
e−n (x−s̄ − xs−1 )dx
λ n=1
ˆ 0 ∞ ˆ X
∞
−n2 πx2 2 πx2
X
−s̄
⇒ lim π
e (x −x s−1
)dx = 0 − e−n (x−s̄ − xs−1 )dx
→0 ei 4 λ n=1
n=1
ˆ 0 ∞ ˆ X
∞
−n2 πx2 2 πx2
X
−s̄
∴ lim e (x −x s−1
)dx = − e−n (x−s̄ − xs−1 )dx
→0 e iπ
4
n=1 λ n=1

. Now let’s use the ML inequality on the integral in the R.H.S.

ˆ X
∞
2 πx2
e−n (x−s̄ − xs−1 )dx ≤ Maximum value of the integrand in the interval
λ n=1

times the length of the semi-circle

ˆ X
∞ ∞
2 πx2
X 2 πx2 π
⇒ e−n (x−s̄ − xs−1 )dx ≤ lim max0 e−n (x−s̄ − xs−1 )
λ n=1 →0 x∈U 2
n=1

Here the length of the semi-circle is such because the full length of the
interval is . So the radius of the semi-circle will be 2 . The max value of
2
the integrand is of order O(e−π −σ ) where σ ∈ (0, 1). If we multiply this
with the length of the semi-circle then the function in the R.H.S. becomes

30
 2
of order O(e−π 1−σ ). Here 1 − σ is positive for σ ∈ (0, 1). So we can write
2
the R.H.S. as O(e−π a ), where a is a positive real number.
ˆ X
∞
2 πx2 2
∴ e−n (x−s̄ − xs−1 )dx ≤ lim O(e−π a )
λ n=1 →0

Here the power of  is always positive. So as  → 0 the R.H.S. tends to zero.

Then ˆ X ∞
2 2
e−n πx (x−s̄ − xs−1 )dx ≤ 0
λ n=1
ˆ X
∞
2 πx2
∴ e−n (x−s̄ − xs−1 )dx = 0
λ n=1

And thus ˆ 0 ∞
2 πx2
X
lim π
e−n (x−s̄ − xs−1 )dx = 0
→0 ei 4 n=1


31
References
[1] E.C. Titchmarsh. ‘The theory of the Riemann Zeta-function’ Second edi-
tion. Edited with a preface by D.R. Heath-Brown, Clarendon Press Ox-
ford University Press, New York ,1986.

[2] J. B. Conrey, ‘More than two fifths of the zeros of the zeros of the Rie-
mann zeta function are on the critical line Journal für die reine und
angewandte Mathematik, 1989.

[3] B. Riemann, ‘Üeber die Anzahl der Primzahlen unter einer gegebenen
Grösse ’.(German)[On the Number of Prime Numbers less than a given
Quantity. ] Monatsberichte der Berliner Akademie, November 1859. En-
glish translation by D.R. Wilkins 1998.

[4] E. Bombieri, ‘Problems of the millennium: The Riemann Hypothesis,’

Clay, 2000.

[5] H. Ki, ‘All but infinitely many non-trivial zeros of the approximations
of Epstein zeta function are simple and on the critical line’ London
Mathematical Society,2005.

e-mail: [email protected]

32