The truth of the Riemann hypothesis Werner Raab Abstract We shall show that the meromorphic function v(s) =
π sin(πs)(1/2 − s)ζ(3/2 − s)
is holomorphic within the complex strip: 0 < <s < 1, since it is the Mellin transform Z ∞ v(s) = ts−1 w(t) dt 0
of the function
1 w(t) = 2πi
Z 1/2+i∞ 1/2−i∞
t−s v(s) ds
with the properties w(t) = O(1)
when t → 0
and w(t) = O(1/t) when t → ∞.
Keyword: Riemann hypothesis Mathematics Subject Classification: 11M26
1
A Mellin transformation
We consider the reciprocal u(s) =
1 (s − 1)ζ(s)
of the regularized zeta-function (s − 1)ζ(s) with the complex variable s. 1
(1)
The famous Riemann hypothesis may be expressed by the claim that the meromorphic function π v(s) = u(3/2 − s) (2) sin(πs) is holomorphic within the complex strip: 0 < <s < 1. Theorem 1 Within the complex strip: 0 < <s < 1/2, the function v(s) is equal to the Mellin transform v(s) =
Z ∞ 0
ts−1 w(t) dt
of the series
(3) s
∞ 2 X t µ(ν) w(t) = √ arctan ν t ν=1 ν with the values µ(ν) of the M¨obius arithmetic function.
(4)
Proof. The M¨obius numbers µ(ν) can be defined by the Dirichlet series ∞ X 1 µ(ν) = . ζ(s) ν=1 ν s
This converges in the half plane: <s > 1 and vanishes at s = 1. If 0 < <s < 1/2, then it follows from the Mellin transformation s Z ∞ Z √t/ν Z ∞ t dx dt = ts−3/2 dt ts−3/2 arctan ν 1 + x2 0 0 0 √ Z = ν
√ Z 1 Z ∞ ts−1 dx t dt = ν dt dx ν + x2 t 0 0 ν + x2 t 0 0 Z ∞ s−1 Z 1 t dx ν s−1/2 s−1/2 =ν dt = π 1+t sin(πs)(1 − 2s) 0 0 x2s
that
∞
s−1
Z 1
∞ X π µ(ν) v(s) = sin(πs)(1/2 − s) ν=1 ν 3/2−s
∞ X µ(ν) Z ∞ s−3/2 t arctan =2 ν=1
ν
0
s
Z ∞ ∞ X µ(ν) t dt = 2 ts−3/2 arctan ν 0 ν=1 ν
s
t dt. ν
In addition, we remark that µ
∞ 2 X µ(ν) π w(t) = √ − arctan 2 t ν=1 ν
r
ν t
2
¶
∞ 2 X µ(ν) = −√ arctan t ν=1 ν
r
ν . t
(5)
2
Power series expansions
Theorem 2 If 0 ≤ t < 1, then w(t) is represented by the power series w(t) =
∞ X
u(3/2 + k)(−t)k .
(6)
k=0
Proof. According to the definition (4) we have ∞ ∞ 2 X µ(ν) X (−1)k w(t) = √ t ν=1 ν k=0 1 + 2k
=
∞ ∞ X (−t)k X µ(ν) k=0
=
1/2 + k ν=1 ν 3/2+k
µ ¶1/2+k
t ν
∞ X
(−t)k . k=0 (1/2 + k)ζ(3/2 + k)
Another way to derive this result is the Mellin inversion ∞ X 1 Z 1/2+i∞ −s t−s u(3/2 − s) w(t) = . t v(s) ds = π Ress=−k 2πi 1/2−i∞ sin(πs) k=0
Theorem 3 If 0 ≤ t < ∞, then w(t) is represented by the series µ
∞ 1 X t w(t) = ∆m u(3/2) 1 + t m=0 1+t
¶m
(7)
with the differences m
∆ u(3/2) =
m X k=0
Ã
!
(−1)k m . k (1/2 + k)ζ(3/2 + k)
Proof. The identities Ã
∞ 1 X m t = 1 + t m=k k
!µ
k
t 1+t
¶m
hold for t > −1/2. According to Theorem 2 we obtain the power series (1 + t)w(t) = =
∞ X
u(3/2 + k)(−1)
k=0 ∞ X m=0
µ
∞ X
k
m=k
t 1+t
à ! ¶m X m m k=0
3
k
Ã
!µ
m k
t 1+t
¶m
(−1)k u(3/2 + k).
3
Finite differences
In order to estimate the fuction w(t) when the positive real variable t tends to infinity, we glance at the general theory of the finite differences m
∆ u(s) =
m X
Ã
!
m (−1)k u(s + k) = ∆m+1 u(s) + ∆m u(s + 1). k
k=0
(8)
The sequence of the partial sums n−1 X
n−1 X
∆m u(s + 1) =
m=0
³
´
∆m u(s) − ∆m+1 u(s) = u(s) − ∆n u(s)
m=0
converges to the limit ∞ X
∆m u(s + 1) = u(s) − lim ∆n u(s), n→∞
m=0
if limn→∞ ∆n u(s) exists. Theorem 4 If s is a positive real number, then lim ∆n u(s) = 0.
(9)
n→∞
Proof. For each fixed positive value of the real variable s the series f (s, t) =
∞ X u(s + k) k t k=0
k!
represents an entire function of the variable t. The Cauchy product −t
f (s, t)e
∞ ∞ X u(s + k) k X (−1)m m = t t
k!
k=0
=
à ! ∞ m X (−t)m X m m=0
m!
k=0
k
m!
m=0
(−1)k u(s + k) =
∞ X ∆m u(s) m=0
m!
(−t)m
may be considered as a generating function of the differences ∆m u(s). The series Z ∞ ∞ ∞ X u(s + k) Z ∞ k −2t u(s + k) X = t e dt = f (s, t)e−2t dt k+1 2 k! 0 0 k=0 k=0 =
∞ X ∆m u(s) m=0
m!
m
(−1)
Z ∞ 0
tm e−t dt =
∞ X
(−1)m ∆m u(s)
m=0 m m
converges apparently. Therefore the terms (−1) ∆ u(s) tend to zero. 4
4
Conclusion
Theorem 5 The function µ
∞ X
t (1 + t)w(t) = ∆ u(3/2) 1+t m=0 m
¶m
of the positive real variable t is bounded, i. e. w(t) = O(1/t)
(10)
when t tends to infinity. Proof. From Theorem 4 it is plain that the limit ∞ X m=0
∆m u(3/2) = u(1/2) − n→∞ lim ∆n u(1/2) = u(1/2)
(11)
exists. According to Abel’s theorem of continuity for power series, we have u(1/2) =
∞ X m=0
µ m
∆ u(3/2) lim
t→∞
t 1+t
¶m
∞ X
µ
t = lim ∆ u(3/2) t→∞ 1+t m=0 m
¶m
.
Now we have arrived at our goal: the Mellin integral Z ∞ π v(s) = = ts−1 w(t) dt sin(πs)(1/2 − s)ζ(3/2 − s) 0 represents a holomorphic function not only within the strip: 0 < <s < 1/2 (Theorem 1), but moreover within the entire strip: 0 < <s < 1, as Riemann conjectured. Finally, we note an interesting consequence of our derivation: the formulas (4) and (5) on the one hand, and (6) and (11) on the other hand, show that ∞ 2 X µ(ν) √ u(3/2) = w(0) = lim arctan t→0 t ν=1 ν
and
s
∞ X t µ(ν) √ =2 ν ν=1 ν ν
r ∞ ∞ √ X X µ(ν) µ(ν) ν √ . arctan = −2 u(1/2) = lim w(t)t = −2 lim t t→∞ t→∞ t ν ν=1 ν=1 ν
The idea of our proof is due in substance to Marcel Riesz [2]. For details concerning the fundamentals of the Riemann zeta-function we refer to Landau’s classical Handbuch [1]. A.M.D.G. 5
References [1] E. Landau, Handbuch der Lehre von der Verteilung der Primzahlen. Third (corrected) edition, two volumes in one, Chelsea Publishing, New York, 1974. (First edition, in two volumes, 1909.) [2] M. Riesz, Sur l’hypoth`ese de Riemann. Acta Math. 40 (1916), 185-190. Collected Papers, Springer-Verlag, Berlin, Heidelberg, New York, 1988, 165-170.
Werner Raab Dr. phil., Professor, retired member of the Mathematical Institute of the University of Bonn Residence: Anton-Klieber-Str. 14, 6410 Telfs, Austria E-mail:
[email protected]
6