You are on page 1of 9

Proof of the Riemann hypothesis

Werner Raab
Professor, Dr. phil., retired member of the
Mathematical Institute of the University of Bonn, Germany
Residence: Anton-Klieber-Str. 14, 6410 Telfs, Austria
E-mail: werner.raab@hotmail.com

Abstract
It is shown that the Mellin transform
v(s) =

=
sin(s)(1/2 s)(3/2 s)

ts1 w(t) dt

of the function

2
()
w(t) =
arctan
t =1

t
2
()
=
arctan

t =1

with the Mobius numbers () is holomorphic within the complex


strip 0 < s < 1 , since w(t) = O(1) when t 0 and w(t) = O(1/t)
when t .

Keyword: Riemann hypothesis


Mathematics Subject Classification: 11M26

Abels test of convergence

We consider the series

2
()
w(t) =
arctan
t =1

(1)

for positive values of the real variable t. The Mobius numbers () may be
dened by the Dirichlet series

()
1
=
(s) =1 s

(2)

for the reciprocal of the Riemann zeta-function (s). This series converges
absolutely in the complex half-plane s > 1 and vanishes at s = 1:

()
=1

The identity

= 0.

(3)

arctan
= arctan

2
t
and the vanishing of the series (3) lead us to the expression

()
2
w(t) =
arctan
t =1

(4)

for the function (1). Now we can conclude by Abels test of convergence that
both the series (1) and (4) converge for each positive real
number t, since
the series (3) converges and since the numbers arctan t/ as well as the

numbers arctan /t form each a monotone and bounded sequence.

Dirichlets test of convergence

According to Dirichlets test of convergence, we conclude from the estimate


n

()



1

(5)

=1

for the partial sums of the series (3), that is to say from their boundedness,
that the series (1)
converges for each positive real number t, since the positive
numbers arctan t/ decrease monotonically to the limit zero.
Dirichlets test is not t for a proof of the convergence of the series (4).
For more information about the Mobius function we refer to Landaus

classical Handbuch [2], especially to the chapter 40, Uber


die Summen, welche
(n) enthalten. There a proof of the estimate (5) is given.
2

Mellin transformations

1. If 0 < s < 1/2 , then we have the Mellin transform

t/


t
dx
s3/2
s3/2
t
arctan
dt =
t
dt

1 + x2
0
0
0
dx
ts1
1 s1 1
1 1

t
=
dt
=
dt dx
0
0 0 1 + x2 t/
0 1 + x2 t/

ts1

dt =
1+t
sin(s)(1 2s) 1/2s
0
0
for each positive real number . Therefore the partial sums
=

s1/2

2s

dx

n
2
()
wn (t) =
arctan
t =1

of the function (1) are transformed into the partial sums

vn (s) =

s1

wn (t) dt = 2

()
=1

s3/2

arctan

t
dt

()

sin(s)(1/2 s) =1 3/2s

of the function
v(s) =

.
sin(s)(1/2 s)(3/2 s)

(6)

2. If 1/2 < s < 1 , then we have the Mellin transform



/t

dx
s3/2
s3/2
t
arctan
dt =
t
dt
t
1 + x2
0
0
0

=

s2

1
dx
ts2

dt
=
dt dx
1 + x2 /t
1 + x2 /t
0
0

ts1

dt =
.
1+t
sin(s)(2s 1) 1/2s
0
0
These results can also easily be derived from formulas that are contained
in the chapter XI, Integral transforms, of the reference work Formulas and
Theorems [3] as examples for the Mellin transformations of the functions
arctan x and arccot x.
= s1/2

x2s2 dx

Dominated convergence

In order to show that the function (6) is the Mellin transform

v(s) =

ts1 w(t) dt

(7)

of the function (1) at least within the complex strip 0 < = s < 1/2 , we
can also use the estimate (5) and Abels inequality





()

n ()

t



max
arctan
arctan t arctan t


1n =1
=1

for each positive integer n and for each positive real number t. Hence we
obtain the estimate

s1
|t wn (t)| dt 2
t3/2 arctan t dt =
.
sin()(1/2 )
0
0
This shows that the proposition (7) is true according to Lebesgues principle
of dominated convergence.

Power series expansion

The series

arctan

t
(1)k
=
k=0 1 + 2k

( )1/2+k

converges when 0 t < 1 and 1 . Hence we have

()
2
(1)k
w(t) =
t =1 k=0 1 + 2k

(t)k
()
k=0

1/2 + k

=1

3/2+k

( )1/2+k

(t)k
.
k=0 (1/2 + k)(3/2 + k)

We introduce the function


u(s) =
in order to write
w(t) =

1
(s 1)(s)

u(3/2 + k)(t)k .

(8)

(9)

k=0

This power series has the initial value w(0) = u(3/2). Thus we have shown
that w(t) = O(1) when t 0. This concludes the trivial part of our proof.
4

Mellin inversion

We remark at this place that the series (9) may be considered as a sum of
residues that emerges from the Mellin inversion formula
1 1/2+i s
t v(s) ds
2i 1/2i

w(t) =

(10)

1
1
tiy
=

dy .
t 0 cosh(y)y (1 + iy)
We dispense with a justication of this formula.

A monotony

If [t] denotes the integral part of t, then the integrals

[t]
ts+1

dt =
(

=1

+1

dt
ts+1

1
1
1
=

s
s =1

( + 1)s

( 1)
1
1+
=
s
s
=2

1
(s)
1
=
s
s =1
s

and

+1

[t]2
1
1
1
dt
2
dt =
=

2
s+2
s+2
s+1
t
t
s + 1 =1

( + 1)s+1

=1
(

1
2 ( 1)2
=
1+
s+1
s+1
=2

1
2 1
2(s) (s + 1)
=
=
s+1
s + 1 =1
s+1

are holomorphic functions in the half-plane s > 1.


Combining these equations we obtain the relation
(s + 1)s (s)(s 1) = s(s + 1)

t [t]
[t] dt
ts+2

(11)

between (s) and (s + 1), which is now valid in the half-plane s > 0 and
shows that the dierence
u(s) = u(s) u(s + 1) = u(s)u(s + 1)((s + 1)s (s)(s 1))
takes on only positive values for positive real values of s, since u(s) is a
positve function of the positive real variable s.
5

Euler series transformation

We need an estimate of the function w(t) when the positive real variable t
tends to inniity.
To this purpose we consider the generalization
f (s, t) =

u(s + k)(t)k

(12)

k=0

of the function w(t) = f (3/2, t) for positive values of the real variable s. Now
we replace each power tk in the series (12) by the series
(

1
m
tk =
1 + t m=k k

)(

t
1+t

)m

(13)

which emerges from the relations

m=k

m m zk
z =
k!
k

d
dz

)k

1
1
=
1z
1z

z
1z

)k

with

z
t
or t =
.
1+t
1z
We suppose that |t| < 1 and t > 1/2, which means that |t| < |1 + t|.
Hence we obtain the absolutely convergent double series
z=

1
m
f (s, t) =
u(s + k)(1)k
1 + t k=0
k
m=k

1
t
=
1 + t m=0 1 + t

)(

t
1+t

)m

( )
)m
m
m
k=0

u(s + k)(1)k .

We introduce the nite dierences


m

u(s) =

k=0

m
u(s + k)(1)k
k

(14)

of the function u(s) in order to write


(

f (s, t) =

1
t
m u(s)
1 + t m=0
1+t

)m

(15)

A boarderline case

The Euler transformation (15) yields the holomorphic continuation of the


power series (12) from the unit disk |t| < 1 into the half-plane t > 1/2.
Now we shall prove the very important fact that the series (15) converges
still at the point t = 1/2.
We consider the entire function
g(s, t) =

u(s + k)

k!

k=0

tk

(16)

of the variable t for any xed positive real number s. The Cauchy product

u(s + k)

g(s, t)et =

k!

k=0

m!

m=0

(t)m

m!

m=0

m
m u(s)
u(s + k)(1)k =
(t)m
m!
k
m=0

(t)m

tk

k=0

is also an entire function of t. We have

u(s + k)

k!

k=0

tk e2t dt =

0
T

=
0

T
u(s + k) k 2t
t e dt =
g(s, t)e2t dt
k!
0
k=0

m u(s)

m!

m=0

(t)m et dt =

m u(s)

m!

m=0

(t)m et dt ,

because the term-by-term integrations are apparently allowed. The estimate

k 2t

t e

T /2

dt < e

k 3t/2

t e

T /2

dt = e

( )k+1

2
3

tk et dt

3T /2

shows that the remainder

u(s + k)
k=0

k!

tk e2t dt < eT /2

( )k+1

u(s + k)

k=0

2
3

vanishes when T tends to innity. Therefore we obtain the result

u(s + k)
k=0

2k+1

u(s + k)
k=0

k!

tk e2t dt =

m=0

m u(s)(1)m .

(17)

10

Abels theorem of continuity

We conclude from the convergence of the series (17) that


lim n u(s) = 0

(18)

for each positive real number s. The recursions


m+1 u(s) = m u(s) m u(s + 1)

(19)

let recognize that the partial sums


n1

m u(s + 1) =

m=0

n1

m u(s) m+1 u(s) = u(s) n u(s)

m=0

tend to the limit

m u(s + 1) = u(s) lim n u(s) = u(s) .


n

m=0

In particular we obtain the limit


u(1/2) =

u(3/2) =

m=0

(
m

u(3/2) lim

m=0

t
= lim
u(3/2)
t
1+t
m=0

t
1+t

)m+1

)m+1

= lim f (3/2, t)t = lim w(t)t


t

due to Abels theorem on the continuity of power series.


Thus we have derived the estimate
w(t) = O(1/t) when t ,

(20)

which completes our proof of the Riemann hypothesis, because it shows that
the function (6), the Mellin transform (7) of w(t), is holomorphic within the
complex strip 0 < s < 1 , as Riemann had conjectured in 1859.
For more details concerning series transformations we refer to Hardys
great book Divergent Series [1], especially to the sections on Eulers and
Borels methods of summation.
Although we took into account that there are always billions of rational
reasons not to look at a problem which has been unsuccessfully looked at by
generations of mathematicians (Alain Connes), we ventured on the dicult
task.
A.M.D.G.
8

References
[1] G. H. Hardy, Divergent Series. Oxford University Press, 1949.
[2] 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, by Teubner, Leipzig, 1909.)
[3] W. Magnus, F. Oberhettinger, R. P. Soni, Formulas and Theorems for
the Special Functions of Mathematical Physics. Third enlarged edition,
Springer-Verlag, Berlin, Heidelberg, New York, 1966.