A new criterion and a proof of the Riemann

hypothesis
Roupam Ghosh
arXiv:0912.4646v10 [[Link]] 16 Jun 2011

October 13, 2019

Abstract:
We introduce a new criterion which if satisfied implies the Riemann hypothesis.
Subsequently we verify the validity of the hypothesis itself.

1 Background
The Riemann Hypothesis is usually stated as, the non-trivial zeros of the Riemann
zeta function lie on the line ℜ(s) = 21 . Although, this is the standard formulation,
one of the exciting features of this problem is, it can be formulated in many dif-
ferent and unrelated ways. One of them was derived by Nyman and Beurling, and
now known as the Nyman-Beurling criterion for the Riemann Hypothesis.

Using a fundamental identity:

∞  
ζ(s) 1 s−1
Z
− =s ρ t dt (1)
s 0 t

where ρ(x) denotes the fractional part of x. Nyman and Beurling formulated the
Riemann Hypothesis in terms of the property of the function ρ. If C denotes the
linear manifold of functions,
n  
X θk
f (x) = ck ρ 0 < θk ≤ 1
k=1
x
Pn
where ck are constants such that k=1 ck θk = 0. Beurling showed that :

The Riemann zeta-function is free from zeros in the half-plane σ > 1/p where
1 < p < ∞ if and only if C is dense in the space Lp (0, 1)

1
which is a really wonderful statement that links two separate branches of mathe-
matics in a very elegant manner. This beautiful result prompted further study of
Nyman-Beurling’s approach. Báez-Duarte was among those whose work on this
criterion deserves mention.

The approach I take in this paper is influenced by Beurling’s 1955 paper: A clo-
sure problem related to the Riemann zeta function and Báez-Duarte’s 2001 paper:
New versions of the Nyman-Beurling criterion for the Riemann Hypothesis. In
this paper I would be studying the function ν corresponding to the Dirichlet eta
function η just like ρ corresponds with the ζ function.

2 Dirichlet eta function η(s) and ν(x)

Theorem 2.1: For all ℜ(s) > 0 and ν(t) = {t/2} + 1/2 − {t/2 + 1/2} we have
Z ∞
η(s) = s ν(t)t−s−1 dt (2)
1

Proof: The Dirichlet eta function is defined as, for all ℜ(s) > 0,
∞
X 1 1
η(s) = s
−
k=1
(2k − 1) (2k)s
Now we have for all t ∈ R+ ν(t) = 0 or 1. It is not hard to see that ν(t) = 0
whenever t ∈ [2k, 2k + 1) and 1 whenever t ∈ [2k − 1, 2k) for all positive integers
k. Hence, we can write the integral (2) as
Z ∞ ∞ Z 2k
X
−s−1
s ν(t)t dt = st−s−1 dt
1 k=1 2k−1

Giving us the sum

∞
X 1 1
s
−
k=1
(2k − 1) (2k)s
which is nothing but η(s). Since we already know that this sum converges for
ℜ(s) > 0, we get our result.

Alternatively, we can write equation (2) as

Z 1  
η(s) 1 s−1
= ν t dt (3)
s 0 t

2
3 Technical Lemmas
Remarks and Definitions:
(x) = ∞
P
Before we proceed further, we set fP k=1 µ(k)ν(x/k). We shall denote it
m
simply by f . In other cases, f (x) = k=1 µ(k)ν(x/k) will be denoted by fm . Here,
µ is the usual Möbius-mu function.

Also for Dirichlet’s eta function we have η(s) = (1 − 21−s )ζ(s) for ℜ(s) > 0.

Let us also keep in mind the following properties of ν(x) for x ∈ R:

1. ν(x) takes only the values 0 and 1.
2. ν(x) is constant in any interval [n, n + 1) where n ∈ N
3. ν(x/y) = 0 if x < y for x, y ∈ R and x > 0 and y > 0
R1 θ
xs−1 dx = θs η(s)


Lemma 3.1: For 0 < θ ≤ 1 and ℜ(s) > 0 we have 0
ν x s

Proof: We have for 0 < θ ≤ 1

Z θ  
θ θs η(s)
ν xs−1 dx =
0 x s
If θ = 1 we have our result. Hence suppose that 0 < θ < 1. Since ν(θ/x) = 0
whenever θ < x therefore Z 1  
θ
ν xs−1 dx = 0
θ x
Summing up the above two integrals we get the following equation
Z 1  
θ θs η(s)
ν xs−1 dx = (4)
0 x s

Lemma 3.2: For all x ≥ 2 , f (x) = −1

Proof: For ℜ(s) > 0 we get from Lemma 3.1

Z 1   ∞
1 s−1 η(s) X µ(k)
f x dx = (5)
0 x s k=1 k s

Fix s with ℜ(s) > 1. For ℜ(s) > 1 we know that ∞ s

P
k=1 µ(k)/k = 1/ζ(s).
Hence, Z 1  
1 1 − 21−s
f xs−1 dx =
0 x s

3
Giving us,
1
1 1 − 21−s
  
1
Z
1+f xs−1 dx = +
0 x s s

1 Z 1
1 1 − 21−s
    
1 1
Z
2
s−1
1+f x dx + 1+f xs−1 dx = +
0 x 1
2
x s s

Now, since f (x) = 1 whenever x ∈ [1, 2) giving us for all ℜ(s) > 1 Hence, we get
1 Z 1
1 1 − 21−s
  
1
Z
2
s−1
1+f x dx + 2xs−1 dx = + (6)
0 x 1
2
s s

ie.,
1   
1
Z
2
1+f xs−1 dx = 0 (7)
0 x
Then we have from equation (7) that for all ℜ(s) > 1
1    Z 1  
1 1
Z
2 3
s−1
1+f x dx + 1+f xs−1 dx + ... = 0
1
3
x 1
4
x

Since, f (r) is always constant in any given [r, r + 1) where r ∈ N, we get

   
1 1 1 1
(1 + f (2)) s − s + (1 + f (3)) s − s + ... = 0
2 3 3 4
It follows from the uniqueness property of a Dirichlet series that each of the terms
1 + f (r) = 0. Therefore for all x ≥ 2 , f (x) = −1.

3.1 The Criterion, two proofs

Theorem 3.1: The Riemann Hypothesis is true if
Z ∞  2
n
X 
lim 1 + µ(k)ν(x/k) x−2 dx = 0 (8)
 
n→∞ n  
k=1

Proof 1: We get from (5) for ℜ(s) > 0

Z 1   ∞
1 s−1 η(s) X µ(k)
f x dx =
0 x s k=1 k s

4
Giving us,
1   Z 1   ∞
1 1 η(s) X µ(k)
Z
2
s−1 s−1
f x dx + f x dx =
0 x 1
2
x s k=1 k s

Since f (x) = 1 whenever x ∈ [1, 2), we get

1   Z 1 ∞
1 η(s) X µ(k)
Z
2
s−1 s−1
f x dx + x dx =
0 x 1
2
s k=1 k s

Hence for ℜ(s) > 0

1 ∞
η(s) X µ(k) 1 − 21−s
  
1
Z
2
1+f xs−1 dx = − (9)
0 x s k=1 k s s

Define f ∗ (x) = f (1/x) for 0 < x ≤ 1/2. Using Holders inequality we have for
L2 (0, 12 )  
∞
 η(s) X µ(k) 1 − 21−s 
−  ≤ ||1 + f ∗ ||2 .||xs−1||2 (10)

 s ks s


k=1

If ℜ(s) > 21 , we get for s = σ + it

!1/2
1/2
1
Z
||xs−1 ||2 = |xs−1 |2 dx = σ− 12
√
0 2 2σ − 1

Now, in equation P (10) if ||1 + f ∗ ||2 = 0 then the Riemann Hypothesis follows,
because it implies ∞ s 1
n=1 µ(n)/n converges in the half-plane ℜ(s) > 2 and where
∗ 2 1
it equals 1/ζ(s). To show ||1 + f ||2 = 0 in L (0, 2 ) we have to prove that

1/2
  2 !1/2
1 + fn 1  dx
Z
 
lim =0
n→∞ 0
 x 

But since by Lemma 3.2, fn (x) = nk=1 µ(k)ν(x/k) = −1 for 2 ≤ x ≤ n, hence we

P
have Z 1/2   2 Z ∞
1
|1 + fn (x)|2 x−2 dx
 
lim 1 + fn  dx = lim (11)
n→∞ 
0 x  n→∞ n

which completes our proof for the criterion.

5
 P∞ s
Proof 2: In this second proof we show that k=1 µ(k)/k forms a Cauchy se-
quence for ℜ(s) > 12 if
Z ∞ 1/2
2 −2
lim |1 + fn (x)| x dx =0 (12)
n→∞ n

This
Pn simply follows from the following. From lemma 3.1 and using the fact
that k=1 µ(k)ν(x/k) = 1 for n > 2 and x ∈ [1, 2) we get for n > m > 2
n Z 1 n  !
η(s) X µ(k) 2 X 1
= 1 + µ(k)ν xs−1 dx (13)
s ks 0 kx
k=m+1 k=m+1

If we let n → ∞ then
∞ Z 1 ∞  !
η(s) X µ(k) 2 X 1
s
= 1+ µ(k)ν xs−1 dx (14)
s k=m+1 k 0 k=m+1
kx

But using the result of Lemma 3.2 in the RHS of (14) equals
Z 1 ∞  !
2 X 1
1+ µ(k)ν xs−1 dx
0 k=m+1
kx
Z 1 ∞  ! Z 1 m  !
2 X 1 2 X 1
= 1+ µ(k)ν xs−1 dx − 1+ µ(k)ν xs−1 dx
0 k=1
kx 0 k=1
kx
Z 1  
2 1
=− 1 + fm xs−1 dx
0 x
(15)
Using a similar treatment as in proof 1, we get by applying Holder’s inequality for
L2 (0, 12 ) to equation (14) and replace the RHS by (15) to get,
 
∞
 η(s) X µ(k) 
∗
 ≤ ||1 + fm ||2 .||xs−1 ||2 (16)
 
 s ks 

k=m+1

Where   X m    
1 1 ∗ 1
fm = µ(k)ν and fm (x) = fm
x k=1
kx x
∗
||2 = 0 in L2 (0, 1/2) as m → ∞ then we can say that nk=1 µ(k)/k s
P
Here, if ||1 + fm
forms a Cauchy sequence and hence converges for ℜ(s) > 1/2 and this implies the
Riemann Hypothesis. The criterion is then derived in a similar fashion as in the
first proof.

6
3.2 The Proof of the Riemann hypothesis
Theorem 3.2: The Riemann Hypothesis is true.
Pn 2
Proof: Set gn (x) = |1 + k=1 µ(k)ν(x/k)|

Since g(x) is always constant in the interval [k, k + 1) Rwhere k ∈ N giving us

∞
gn′ (x) = 0 whenever x ∈ [k, k + 1). Treating the integral n gn (x)x−2 dx as a sum
of its constituents we get
∞ ∞  k+1 X ∞ Z k+1  
gn (x) 1
Z X
−2
gn (x)x dx = − − − gn′ (x)dx (17)
n k=n
x k k=n k x

Since, gn′ (x) = 0 we have to only evaluate

∞  k+1
X gn (x)
− (18)
k=n
x k

For x ∈ [k, k + 1) we have gn (x) = gn (k) and hence

∞  k+1 X∞
X gn (x) gn (k)
− = (19)
x k k(k + 1)
k=n k=n

Now since ν(x) takes the values 1 or 0, ideally | nk=1 µ(k)ν(x/k)| can take a
P
maximum value where all its negative µ(k) have ν(x/k) = 1 and all the positive
µ(k) have ν(x/k) = 0. Hence we have,
 
n
√
 
X  X X n
µ(k)ν(x/k) ≤ 1+ 1 + ... = π(n) + O( n) = O
 
log n

 
k=1 p1 ≤n p1 p2 p3 ≤n

where p1 ...p2r+1 ≤ n denotes the primes less than n taken 2r + 1 at a time where
r ∈ N such that their product is ≤ n, and π(n) denotes the number of primes ≤ n.
Therefore, gn (x) = O (n2 /(log n)2 ). Hence as n → ∞ from equation (17) and (19)
we get, Z ∞
lim g(x)x−2 dx = 0 (20)
n→∞ n
Using Theorem 3.1 we have a proof of Riemann Hypothesis.

7
References
[1] Peter Borwein, Stephen Choi, Brendan Rooney, Andrea Weirathmueller ”The
Riemann Hypothesis: A Resource for the Afficionado and Virtuoso Alike”

[2] E.C. Titchmarsh, ”The Theory of the Riemann Zeta-Function”

[3] H. M. Edwards, ”Riemann’s Zeta Function”

[4] Sondow, Jonathan and Weisstein, Eric W. ”Riemann Zeta Function.” From
MathWorld – A Wolfram Web Resource.
[Link]

[5] T.M. Apostol, ”Introduction to Analytic Number Theory”

[6] L. Baez-Duarte, A new necessary and sufficient condition for the Riemann
Hypothesis, 2003, [Link]/0307215

[7] A Beurling, A closure problem related to the Riemann Zeta function, 1955

[8] J. F. Burnol, A note on Nyman’s equivalent formulation of the Riemann hy-

pothesis, 2001

[9] B. Bagchi, On Nyman, Beurling and Baez-Duarte’s Hilbert space reformula-

tion of the Rieamnn hypothesis, 2006,[Link]/0607733v1

[10] L. Báez-Duarte, New versions of the Nyman-Beurling criterion for the Rie-
mann Hypothesis, 2001

[11] W W L Chen, Distribution of prime numbers

[Link]