A REMARK ON THE RIEMANN ZETA FUNCTION

YUKITAKA ABE

Abstract. We prove that if the number of nontrivial zeros of the

Riemann zeta function which are not on the critical line is finite,
then every nontrivial zero is on the critical line.

1. Introduction
The Riemann zeta function ζ(s) is defined on σ > 1 by
∞
X 1
ζ(s) = ,
n=1
ns
where s = σ+it. It is analytically continued to a meromorphic function
on the whole plane with a pole at s = 1. It is well-known that negative
even integers are zeros of ζ(s). Other zeros of ζ(s) are called complex
zeros or nontrivial zeros.
Riemann stated the following statement, the so-called Riemann hy-
pothesis, in [6] in 1859.

The Riemann hypothesis. All nontrivial zeros of ζ(s) lie on the

critical line σ = 12 .

Hilbert listed it as the eighth problem of his 23 problems in his 1900

address to the Paris International Congress of Mathematicians. This
is one of the most important unsolved problems in the twenty-first
century.
Nobody has succeeded to prove it up to the present, but many com-
putational results are known. In the early part of the twentieth century,
they were obtained by hand computation ([1], [4], [5] and [7]). Numer-
ical computations by computers have permitted us to check the truth
of the Riemann hypothesis to extremely large t. We refer to [3] for a
history of numerical verifications.
In this paper, we give a remark that if the number of nontrivial zeros
of the Riemann zeta function which are not on the critical line is finite,
2020 Mathematics Subject Classification. 11M26.
Key words and phrases. Riemann hypothesis.
1
2 YUKITAKA ABE

then every nontrivial zero is on the critical line. First we prepare a

lemma on Dirichlet series. Combining this lemma and the theorem of
de la Vallée Poussin, we obtain our result.

2. Lemma on Dirichlet series

This section is devoted to the following lemma which is a variant of
Lemma 3.12 in [8].
Lemma 1. We assume that a Dirichlet series
∞
X an
f (s) =
n=1
ns
is absolutely convergent for σ > 1. Take σ0 < 1 and c > 0 with
σ0 + c > 1. Let g(s) be an entire function with finite number of zeros
Zg = {α1 , . . . , αN } such that σ0 < Re(αj ) < σ0 + c (j = 1, . . . , N ) and
|g(s)| → ∞ as s → ∞. Then, there exist meromorphic functions hn (s)
(1 ≤ n < x) on C whose poles are at most α1 , . . . , αN and all simple
such that
Z c+iT
X an  1 f (s + w) xw

1
(2.1) + h n (s) = dw
n<x
ns g(s) 2πi c−iT g(s + w) w
X an X an
+ R
s n
(c, x, T )(s) + Q (c, x, T )(s)
s n
n<x
n x<n
n

for any s ∈ D(σ0 , T ) \ Zg , where a positive number x is not an integer,

T is a positive number with |Im(αj )| < T4 (j = 1, . . . , N ), D(σ0 , T ) =
{σ + it; σ0 ≤ σ, |t| < T4 }, and Rn (c, x, T )(s) and Qn (c, x, T )(s) are
holomorphic functions depending only on g(s), n, c, x, T and satisfying
(2.2)
M  x c M  x c
|Rn (c, x, T )(s)| < and |Qn (c, x, T )(s)| <
AπT n AπT n
on D(σ0 , T ) for some constants A and M . Therefore, the right side of
(2.1) converges absolutely and uniformly on D(σ0 , T ).
1 x w 1


Proof. Take any s ∈ D(σ0 , T ). A function g(s+w) n w
of w has poles
1

α −s
at w = 0, α1 −s, . . . , αN −s. It has the residues g(s) and aj nx j αj1−s
1
at w = 0 and αj − s respectively, where aj is the residue of g(s) at αj .
We define a meromorphic function
N  x αj −s
X 1
hn (s) := aj .
j=1
n αj − s
 A REMARK ON THE RIEMANN ZETA FUNCTION 3

If n < x, then we obtain

Z c−iT Z c+iT Z −∞+iT 
1 1  x w dw
+ +
2πi −∞−iT c−iT c+iT g(s + w) n w
1
= + hn (s)
g(s)
by the residue theorem. Let
  
 1  3 5
M := sup    ; s = u + it, −∞ < u < ∞, T ≤ |t| ≤ T .
g(s)  4 4
Then we have 0 < M < ∞ by the assumption of g(s). Therefore, the
following estimation holds
 Z c−iT  x w dw  c
 1 1
  < M (x/n) .
 2πi g(s + w) n w  2πT log(x/n)
−∞−iT

Similarly we have
 Z −∞+iT  x w dw  c
 1 1
  < M (x/n) .
 2πi g(s + w) n w  2πT log(x/n)
c+iT

We define
Z c−iT Z −∞+iT 
1 1  x w dw
Rn (c, x, T )(s) := + .
2πi −∞−iT c+iT g(s + w) n w
Then we obtain
Z c+iT
1 1  x w dw 1
(2.3) = + hn (s) − Rn (c, x, T )(s)
2πi c−iT g(s + w) n w g(s)
and
M (x/n)c
|Rn (c, x, T )(s)| <
πT log(x/n)
for s ∈ D(σ0 , T ).
For x < n, we similarly obtain
Z c+iT Z c−iT Z ∞−iT 
1 1  x w dw
+ + = 0,
2πi ∞+iT c+iT c−iT g(s + w) n w
because there is no residue term. If we set
Z ∞+iT Z c−iT 
1 1  x w dw
Qn (c, x, T )(s) := + ,
2πi c+iT ∞−iT g(s + w) n w
then we have
Z c+iT
1 1  x w dw
(2.4) = −Qn (c, x, T )(s).
2πi c−iT g(s + w) n w
4 YUKITAKA ABE

We also obtain
M (x/n)c
|Qn (c, x, T )(s)| <
πT | log(x/n)|
for s ∈ D(σ0 , T ) by the same way as above.
From (2.3) and (2.4), it follows that
Z c+iT X an 1 Z c+iT
1 f (s + w) xw 1  x w dw
dw =
2πi c−iT g(s + w) w n<x
ns 2πi c−iT g(s + w) n w
X an 1 Z c+iT 1  x w dw
+
x<n
ns 2πi c−iT g(s + w) n w
X an  
1
= s
+ hn (s)
n<x
n g(s)
X an
− R (c, x, T )(s)
s n
n<x
n
X an
− Qn (c, x, T )(s).
x<n
ns
Then we obtain (2.1). Furthemore, we can take A > 0 such that
| log(x/n)| > A for any n ∈ N, by the assumption of x. Hence, we also
obtain (2.2). □

3. Result
Assume that the number of nontrivial zeros of ζ(s) which are not on
the critical line is finite. Then, the following Proposition is immediate.
Proposition 1. Under the above assumption, there exists T0 > 0 such
that any s = σ + it with 21 < σ < 1 and t > T0 is not a zero of ζ(s).
Let
B := sup{β; ζ(β + iγ) = 0, γ ̸= 0}.
The Riemann hypothesis states B = 21 . We recall the theorem of de
la Vallée Poussin ([2]) which says that there is a constant A > 0 such
that ζ(s) is not zero for
A
σ ≥1− (t > t0 ),
log t
where t0 is some positive constant. We may restate it as follows:
if s = σ + it (t > t0 ) satisfies
 
A
(3.1) t ≤ exp ,
1−σ
 A REMARK ON THE RIEMANN ZETA FUNCTION 5

then ζ(s) ̸= 0.
1
Proposition 2. It holds that 2
≤ B < 1.
Proof. Since
 
A
exp −→ ∞ as σ −→ 1 − 0,
1−σ
A


there exists σ0 < 1 such that T0 < exp 1−σ for σ0 < σ < 1, where
T0 is the constant in Proposition 1. Then, there is no zero of ζ(s)
in a region σ0 < σ by Proposition 1 and the theorem of de la Vallée
Poussin. □
1
Proposition 3. If 2
< B < 1, then there is no zero of ζ(s) on the line
σ = B.
Proof. If the function ζ(s) has a zero on σ = B, then the number
of zeros of ζ(s) on σ = B is finite by Proposition 1. Let ZB =
{ρ1 , ρ1 , . . . , ρN , ρN } be the set of zeros of ζ(s) on the line σ = B. By
Proposition 1, there exists δ0 > 0 such that there are no zeros of ζ(s)
except ZB in the set {σ + it; B − δ0 ≤ σ, −∞ < t < ∞}.
It is well-known that
∞
1 X µ(n)
=
ζ(s) n=1 ns
for σ > 1. We define an entire function g(s) by
N
Y
g(s) := (s − ρj )(s − ρj ).
j=1

Then, its zeros are ZB , and |g(s)| −→ ∞ as s → ∞. We take T > 0

such that |Im(ρj )| < T4 for j = 1, . . . , N . Putting σ0 = B − δ0 and
c = 2, we apply Lemma 1. We take x > 1 which is not an integer, and
fix it. Then we have
Z 2+iT
X µ(n)  1 xw

1 1
s
+ hn (s) = dw
n<x
n g(s) 2πi 2−iT ζ(s + w)g(s + w) w
X µ(n)
(3.2) + s
Rn (2, x, T )(s)
n<x
n
X µ(n)
+ Qn (2, x, T )(s)
x<n
ns
for s ∈ D(σ0 , T ) \ ZB , where hn (s) is a meromorphic function on C
whose poles are at most ZB and all simple.
6 YUKITAKA ABE

1 x w
We consider the integral of ζ(s+w)g(s+w) w
along C = C0 + C1 + C2 +
C3 , where C0 , C1 , C2 and C3 are segments from 2 − iT to 2 + iT , from
2 + iT to − δ20 + iT , from − δ20 + iT to − δ20 − iT and from − δ20 − iT to
2 − iT , respectively. We set
 
1 1 T
D0 := σ + it; B − δ0 < σ < B + δ0 , |t| < .
4 4 4
Then D0 ⊂ D(σ0 , T ) and ZB ⊂ D0 . For any s ∈ D0 \ ZB , the poles
1 xw
of ζ(s+w)g(s+w) w
in a domain surrounded by C are w = 0, ρ1 − s, ρ1 −
1 xw
s, . . . , ρN − s and ρN − s. The residue of ζ(s+w)g(s+w) w
at w = 0 is
1 1
ζ(s)g(s)
. Let aj and bj be the residues of ζ(s)g(s) at ρj and ρj respectively.
ρj −s
1 xw
Then, the residues of ζ(s+w)g(s+w) w
at ρj − s and ρj − s are aj xρj −s and
ρj −s
bj xρj −s respectively. We define
N 
xρj −s xρj −s
X 
Q(s) := aj + bj .
j=1
ρ j − s ρ j − s

Then, Q(s) is a meromorphic function on C whose poles are at most

ZB and all simple. We note that Re(s + w) > B − 34 δ0 if s ∈ D0 and w
is on C. If s ∈ D0 and w is on C2 , then Re(s + w) < B − 41 δ0 . Then,
1 xw
ζ(s+w)g(s+w) w
is holomorphic on C as a function of w for any s ∈ D0 .
By the residue theorem, we obtain
xw
Z
1 1 1
dw = + Q(s)
2πi C ζ(s + w)g(s + w) w ζ(s)g(s)
for s ∈ D0 \ ZB . Therefore we have
Z 2+iT
1 1 xw 1
(3.3) dw = + Q(s) + P (s)
2πi 2−iT ζ(s + w)g(s + w) w ζ(s)g(s)
for s ∈ D0 \ ZB , where
xw
Z
1 1
P (s) = − dw
2πi C1 +C2 +C3 ζ(s + w)g(s + w) w
is a holomorphic function on D0 . We set
 
3 1 3 5
E := σ + it; B − δ0 ≤ σ ≤ B + 2 + δ0 , T ≤ |t| ≤ T
4 4 4 4
[ 3 1 5

σ + it; B − δ0 ≤ σ ≤ B − δ0 , |t| ≤ T .
4 4 4
 A REMARK ON THE RIEMANN ZETA FUNCTION 7

Then we have {s + w; s ∈ D0 , w ∈ C1 ∪ C2 ∪ C3 } ⊂ E. Since there is

1
no pole of ζ(s)g(s) on E, we can take M0 > 0 such that
 
 1 
 ζ(s)g(s)  < M0
 

on E. Then we obtain
xw  M0 x2
Z 
 1
 dw <
C1 ζ(s + w)g(s + w) w
  T log x
and
xw  M0 x2
Z 
 1
 dw <

C3 ζ(s + w)g(s + w) w T log x
δ0
for s ∈ D0 . Since we have w = − δ20 + it and xw = eit log x x− 2 on C2 ,
we obtain
xw 
Z 
1 4 δ0
dw < M0 T x− 2



C2ζ(s + w)g(s + w) w δ0
for s ∈ D0 . It follows from the above estimates that
1 2M0 x2
 
4 δ
− 20
(3.4) |P (s)| < + M0 T x
2π T log x δ0
on D0 .
By (3.2) and (3.3), we obtain
X µ(n)  1 
1
+ hn (s) = + Q(s) + P (s)
n<x
ns g(s) ζ(s)g(s)
X µ(n)
(3.5) + Rn (2, x, T )(s)
n<x
ns
X µ(n)
+ Qn (2, x, T )(s)
x<n
ns
for s ∈ D0 \ ZB . We see that a function
X µ(n) X µ(n)
P (s) + R n (2, x, T )(s) + Qn (2, x, T )(s)
n<x
ns x<n
ns
is bounded on D0 by the properties of Rn (2, x, T )(s) and Qn (2, x, T )(s),
1
and (3.4). The functions g(s) , hn (s) and Q(s) are meromorphic func-
tions on C whose poles are at most ZB and all simple. On the other
1
hand, the function ζ(s)g(s) has poles of order at least 2 at every point
in ZB . This contradicts to the equation (3.5). Hence, there is no zero
of ζ(s) on the line σ = B. □
8 YUKITAKA ABE

Theorem 1. Assume that the number of nontrivial zeros of ζ(s) which

are not on the critical line is finite. Then, the Riemann hypothesis is
true.
Proof. We may assume 21 ≤ B < 1 by Proposition 2. Suppose that
1
2
< B < 1. Then there is no zero of ζ(s) on the line σ = B by
Proposition 3. Hence, we can take B ′ with 21 < B ′ < B such that
ζ(s) ̸= 0 for B ′ < σ by Proposition 1. This contradicts to the definition
of B. Thus we conclude B = 21 . □
References
[1] R. Backlund, Sur les zéros de la fonction ζ(s) de Riemann, C. R. Acad. Sci.
Paris, V., 158 (1914), 1979–1982.
[2] C. J. de la Vallée Poussin, Sur la fonction ζ(s) de Riemann et le nombre des
nombres premiers inférieurs à une limite donée, Mém. Couronnés et Autres
Mém. Publ. Acad. Roy. Sci. des lettres Beaux-Arts Belg., 59 (1899–1900), 1–
74.
[3] X. Gourdon, The 1013 first zeros of the Riemann zeta
function, and zeros computation at very large height,
http://numbers.computation.free.fr/Constants/Miscellaneouszetazeros1e13-
1e24.pdf.
[4] J. Gram, Note sur les zéros de la fonction ζ(s) de Riemann, Acta Math., 27
(1903), 289–304.
[5] J. I. Hutchinson, On the roots of the Riemann zeta-function, Trans. Amer.
Math. Soc., 27 (1925), 49–60.
[6] B. Riemann, Über die Anzahl der Primzahlen unter einer gegebenen Grösse,
Monat. der Königl. Preuss. Akad. der Wissen. zu Berlin aus der Jahre 1859,
(1860), 671–680.
[7] E. C. Titchmarsh, The zeros of the Riemann zeta-function, Proc. Roy. Soc.
London, 151 (1935), 234–255; also ibid., 157 (1936), 261–263.
[8] E. C. Titchmarsh, The Theory of the Riemann Zeta Function, second edition,
edited and with a preface by D. R. Heath-Brown, The Clarendon Press, Oxford
University Press, New York, 1986.

Department of Mathematics, University of Toyama, Toyama 930-

8555, Japan