Politecnico di Torino
Porto Institutional Repository
[Article] The exceptional set for the distribution of primes between
consecutive powers
Original Citation:
Bazzanella D. (2007). The exceptional set for the distribution of primes between consecutive
powers. In: ACTA MATHEMATICA HUNGARICA, vol. 116 n. 3, pp. 197-207. - ISSN 0236-5294
Availability:
This version is available at : http://porto.polito.it/1841396/ since: July 2008
Publisher:
Springer
Published version:
DOI:10.1007/s10474-007-5174-y
Terms of use:
This article is made available under terms and conditions applicable to Open Access Policy Article
("Public - All rights reserved") , as described at http://porto.polito.it/terms_and_conditions.
html
Porto, the institutional repository of the Politecnico di Torino, is provided by the University Library
and the IT-Services. The aim is to enable open access to all the world. Please share with us how
this access benefits you. Your story matters.
(Article begins on next page)
THE EXCEPTIONAL SET FOR THE
DISTRIBUTION OF PRIMES BETWEEN
CONSECUTIVE POWERS
by DANILO BAZZANELLA
Abstract
A well known conjecture about the distribution of primes asserts that
between two consecutive squares there is always at least one prime number. The proof of this conjecture is quite out of reach at present, even
under the assumption of the Riemann Hypothesis. This paper is concerned with the distribution of prime numbers between two consecutive
powers of integers, as a natural generalization of the afore-mentioned
conjecture.
Mathematical Subject Classification :
11NO5 - Distribution of Primes
This is the authors’ post-print version of an article published on
Acta Math. Hungar. 116 (3) (2007), 197–207,
DOI:10.1007/s10474-007-5174-y.
1
1
This version does not contain journal formatting and may contain minor changes
with respect to the published version.
The final publication is available at
http://dx.doi.org/10.1007/s10474-007-5174-y. The present version is accessible on PORTO,
the Open Access Repository of Politecnico di Torino (http://porto.polito.it), in compliance with the Publisher’s copyright policy as reported in the SHERPA-ROMEO website:
http://www.sherpa.ac.uk/romeo/issn/0236-5294/
1
2
1
D. BAZZANELLA
Introduction
A well known conjecture about the distribution of primes asserts that all intervals of type [n2 , (n + 1)2 ] contain at least one prime.
The proof of this conjecture is quite out of reach at present, even under
the assumption of the Riemann Hypothesis. To get a conditional proof of
the conjecture we need to assume a stronger hypothesis about the behavior of
Selberg’s integral in short intervals, see Bazzanella [2].
This paper is concerned with the distribution of prime numbers between
two consecutive powers of integers, as a natural generalization of the above
conjecture.
The well known result of Huxley about the distribution of prime in short
intervals, see [11], implies that, for n → ∞, all intervals [nα , (n + 1)α ] contain
and “almost all” contain the exthe expected numbers of prime for α > 12
5
pected numbers of prime for α > 56 . By “almost all” we mean that the number
of integers X ≤ n ≤ 2X for which the interval [nα , (n + 1)α ] does not contain
the expected number of primes is o(X). Our first result is an unconditional
estimate for the measure of the exceptional set for the distribution of primes
between two consecutive power of integers.
Theorem 1 Let ε > 0. Then for every [nα , (n+1)α ] ⊂ [N, 2N ] with O(N η(α)+ε )
exceptions we have the expected number of primes , where
η(α) =













8
1
−
5α 2
5
−1
 2α











72 − 9α − 8α2
3α(α + 12)
6
6
<α≤ +c
5
5
27
53
<α≤
16
26
53
12
≤α<
26
5
and c positive suitable constant.
For the sake of simplicity we consider the function c(α) only for the extremes
and more interesting values of the parameter α. Arguing in the same way we
can obtain the explicit values of the function c(α) for every α.
A corollary of this theorem is Theorem 1 of Bazzanella [2], which states that for
every [n2 , (n + 1)2 ] ⊂ [N, 2N ] with O(N 1/4+ε ) exceptions we have the expected
number of primes.
PRIMES BETWEEN CONSECUTIVE POWERS
3
Assuming some hypotheses we can have strong result about the regularity of
distribution of primes and then a better estimate for the number of exceptions
for the distribution of primes between two consecutive powers of integers.
First we assume the Lindelöf hypothesis, which states that, for every η > 0,
the Riemann Zeta-function satisfies
1
ζ(σ + it) tη (σ ≥ , t ≥ 2).
2
Assuming this hypothesis we can prove that all intervals [nα , (n + 1)α ] contain
the expected numbers of prime for α > 2 and “almost all” contain the expected numbers of prime for α > 1, see Bazzanella [1]. Our second result is a
conditional estimate for the measure of the exceptional set for the distribution
of primes between two consecutive powers of integers under the assumption of
the Lindelöf hypothesis.
Theorem 2 Assume the Lindelöf hypothesis, let ε > 0 and 1 < α ≤ 2. Then
for every [nα , (n + 1)α ] ⊂ [N, 2N ] with O(N 2/α−1+ε ) exceptions we have the
expected number of primes.
A corollary of this theorem is Theorem 2 of Bazzanella [1], which states that
almost all intervals [nα , (n + 1)α ] contain the expected number of primes for
α > 1.
Assuming The Riemann Hypothesis we can set our last result.
Theorem 3 Assume the Riemann Hypothesis, let 1 < α ≤ 2 and g(x) → ∞
arbitrarily slowly.
Then for every [nα , (n + 1)α ] ⊂ [N, 2N ] with O(N 2/α−1 log2 N g(N )) exceptions we have the expected number of primes.
A corollary of this theorem is Theorem 2 of Bazzanella [2], which states that
for every [n2 , (n + 1)2 ] ⊂ [N, 2N ] with O(g(N ) log2 N ) exceptions we have the
expected number of primes.
2
The necessary lemmas
First lemma is a result about the structure of the exceptional set for the
asymptotic formula
ψ(x + h(x)) − ψ(x) ∼ h(x) as x → ∞.
(1)
4
D. BAZZANELLA
Let X be a large positive number, δ > 0 and let | | denote the modulus of
a complex number or the Lebesgue measure of a set. Let h(x) be an increasing
function such that xε ≤ h(x) ≤ x for some ε > 0,
∆(x, h) = ψ(x + h(x)) − ψ(x) − h(x)
and
Eδ (X, h) = {X ≤ x ≤ 2X : |∆(x, h)| ≥ δh(x)}.
It is clear that (1) holds if and only if for every δ > 0 there exists X0 (δ) such
that Eδ (X, h) = ∅ for X ≥ X0 (δ). Hence for small δ > 0, X tending to ∞ and
h(x) suitably small with respect to x, the set Eδ (X, h) contains the exceptions,
if any, to the expected asymptotic formula for the number of primes in short
intervals. Moreover, we observe that
Eδ (X, h) ⊂ Eδ0 (X, h) if 0 < δ 0 < δ.
We will consider increasing functions h(x) of the form h(x) = xθ+ε(x) , with
some 0 < θ < 1 and a function ε(x) such that |ε(x)| is decreasing,
!
|y|
.
ε(x) = o(1) and ε(x + y) = ε(x) + O
x
A function satisfying these requirements will be called of type θ.
First lemma provides the basic structure of the exceptional set Eδ (X, h).
Lemma 1 Let 0 < θ < 1, h(x) be of type θ, X be sufficiently
√ large depending
on the function h(x) and 0 < δ 0 < δ with δ − δ 0 ≥ exp(− log X). If x0 ∈
Eδ (X, h) then Eδ0 (X, h) contains the interval [x0 −ch(X), x0 +ch(X)]∩[X, 2X],
where c = (δ − δ 0 )θ/5. In particular, if Eδ (X, h) 6= ∅ then
|Eδ0 (X, h)| θ (δ − δ 0 )h(X).
Lemma 1 is part (i) of Theorem 1 of Bazzanella and Perelli, see [3], and
essentially says that if we have a single exception in Eδ (X, h), with a fixed δ,
then we necessarily have an interval of exceptions in Eδ0 (X, h), with δ 0 little
smaller than δ. The interesting consequence of this lemma is that we can use
an average estimate to prove the non-existence of the exceptions.
PRIMES BETWEEN CONSECUTIVE POWERS
5
Remainder lemmas provide conditional and unconditional bounds for the Selberg’s integral and the fourth-power integral linked to the distribution of
primes in short intervals.
Lemma 2
Let ε > 0 then exists c > 0 such that
Z 2X
|ψ(x + δx) − ψ(x) − δx + B(x, θ)|2 dx X (11+14θ)/10+ε
X
with
uniformly for X θ−1
1
1
< θ < + c,
6
6
Xθ
,
B(x, θ) log X
δ X θ−1 .
Proof.
In order to prove the lemma we use the classical explicit formula, see ch.
17 of Davenport [5], to write
ψ(x + δx) − ψ(x) − δx =
X
x% c% (δ) + O(
|γ|≤T
X log2 X
),
T
(2)
uniformly for X ≤ x ≤ 2X, where 10 ≤ T ≤ X, % = β + iγ runs over the
non-trivial zeros of ζ(s),
c% (δ) =
(1 + δ)% − 1
%
c% (δ) min(X θ−1 ,
and
1
).
|γ|
(3)
Choose
T = X 1−θ log3 X,
(4)
We use Theorem 3 of Halász-Turán [6], which asserts that there exists a
constant c1 > 0 such that
N (σ, T ) T (1−σ)
3/2
log3
1
1−σ
(5)
for 1 − c1 ≤ σ ≤ 1. From (3) - (5) and Vinogradov’s zero-free region, see ch. 6
of Titchmarsh [14], by a standard argument we see that there exists a constant
c2 > 0 such that
X
|γ|≤T
1−c2 ≤β≤1
xρ cρ (δ) X θ−1 log2 X
max X σ N (σ, T ) 1−c2 ≤σ≤1
Xθ
log X
(6)
6
D. BAZZANELLA
uniformly for X ≤ x ≤ 2X.
Again by a standard argument, from (3), (4) and Ingham-Huxley density
estimate, we obtain
Z 2X
2
xρ cρ (δ) dx X 2θ−1+ε
X
X
max
1/2≤σ≤1−c2
X 2σ N (σ, T ) X (11+14θ)/10+ε ,
|γ|≤T
1−c2 ≤β≤1
for θ > 1/6 and sufficiently small. This completes the proof of the Lemma 2.
Lemma 3
Let ε > 0 then
Z 2X
|ψ(x + ∆x) − ψ(x) − ∆x + B(x, θ)|4 dx X µ(θ)+ε
X
where
µ(θ) =






X (3+5θ)/2





X (55+60θ
11
27
<θ≤
27
53
2 −108θ)/(39−36θ)
and
B(x, θ) 27
7
≤θ< .
53
12
Xθ
log X
uniformly for X θ−1 ∆ X θ−1 .
Proof.
We proceed along similar lines of the Lemma 2, using fourth power moments
instead of mean square estimates. Using The Ingham-Huxley density estimate
we can prove that for every η > 0 we have
X
xρ cρ (∆) X θ−1 log2 X
|γ|≤T, β6∈I
max
1/2≤σ≤1, σ6∈I
X σ N (σ, T ) with
I = [a, b],
a = max
1
, 3θ − 1 − η
2
Xθ
log X
(7)
PRIMES BETWEEN CONSECUTIVE POWERS
and
b=
(10 − 9θ)
+η
7






7
11
27
<θ≤
27
53
,
4
27
7
−θ+η
≤θ<
3
53
12
uniformly for X ≤ x ≤ 2X and 11/27 < θ < 7/12.
We bound the contribution of the zeros such that β ∈ I by a fourth power
moment and we get





Z 2X
X
|
xρ cρ (ξ)|4 dx X 4θ−3+ε max X 4σ N ∗ (σ, T ).
X
σ∈I
|γ|≤T, β∈I
(8)
From Theorem 2 of Heath-Brown [8] we have
T (10−11σ)/(2−σ) logk T








N ∗ (σ, T )  T (18−19σ)/(4−2σ) logk T






T 12(1−σ)/(4σ−1) logk T
if
1
2
≤σ≤
2
3
if
2
3
≤σ≤
3
4
if
3
4
≤ σ ≤ 1,
(9)
where k is an absolute constant. Hence from (8) and (9) we obtain
Z 2X
X
|
X
xρ cρ (ξ)|4 dx X µ(θ)+ε ,
|γ|≤T, β∈I
with
µ(θ) =






X (3+5θ)/2





X (55+60θ
11
27
<θ≤
27
53
2 −108θ)/(39−36θ)
27
7
≤θ< ,
53
12
and then Lemma 3 follows.
Lemma 4 Assume the Lindelöf hypothesis and let ε > 0. Then there exists
a function E(y, T ) such that
Z 2X
X
ψ y+
y
T
− ψ(y) −
4
y
+ B(y, T ) dy X 4+ε T −3
T
8
D. BAZZANELLA
and
B(y, T ) y
T ln y
uniformly for X ≥ 2, 1 ≤ T ≤ X and X ≤ y ≤ 2X.
Lemma 4 is Lemma B of Yu, see [15].
Lemma 5
Assume the Riemann hypothesis, then for x ≥ 4
Z 2X
2
|ψ(x + θx) − ψ(x) − θx| dx θX
2
X
2
log
θ
2
,
uniformly in 0 < θ ≤ 1.
This last lemma is Lemma 5 of Saffari and Vaughan, see [13].
3
Proof of the Theorems
We define H = (n + 1)α − nα and
Aδ (N, α) = {N 1/α ≤ n ≤ (2N )1/α : |ψ((n + 1)α ) − ψ(nα ) − H| ≥ δH}.
This set contains the exceptions, if any, to the expected asymptotic formula
for the number of primes in intervals of the type [nα , (n + 1)α ] in [N, 2N ].
The main step of the proofs is to connect the exceptional set Aδ (N, α) with
the exceptional set for the distribution of primes in short intervals and show
that
Eδ/2 (N, h)
+1
(10)
|Aδ (N, α)| N 1−1/α
for every δ > 0, α > 1 and h(x) = (x1/α + 1)α − x.
In order to prove (10) we let n ∈ Aδ (N, α) and x = nα ∈ [N, 2N ]. From
the definition of the set Aδ (N, α) we get
|ψ((n + 1)α ) − ψ(nα ) − H| ≥ δH,
and then
|ψ(x + h(x)) − ψ(x) − h(x)| ≥ δh(x),
PRIMES BETWEEN CONSECUTIVE POWERS
9
which implies x ∈ Eδ (N, h). Using Lemma 1, with δ 0 = δ/2, we obtain that
there exists an effective constant c such that
[x, x + ch(x)] ∩ [N, 2N ] ⊂ Eδ/2 (N, h).
Let m ∈ Aδ (N, α), m > n. As before we can define y = mα ∈ [N, 2N ] such
that
[y, y + ch(y)] ∩ [N, 2N ] ⊂ Eδ/2 (N, h).
Choosing c < 1 we find
y − x = mα − nα ≥ (n + 1)α − nα > ch(x),
and then
[x, x + ch(x)] ∩ [y, y + ch(y)] = ∅.
Hence (10) is proved, since for every n ∈ Aδ (N, α) and x = nα , with at most
one exception, we have
[x, x + ch(x)] ⊂ [N, 2N ].
Now we can conclude the proof of the theorems providing a suitable bound
for the measure of the exceptional set Eδ/2 (N, h). Let g(x) a function such
that g(x) → ∞ arbitrarily slowly, for x → ∞. Then we subdivide [N, 2N ] into
g(N ) intervals of type Ij = [Nj , Nj + Y ] with
N ≤ Nj < 2N
and
Y N
.
g(N )
For every x ∈ Eδ/2 (N, h) we have
|ψ(x + h(x)) − ψ(x) − h(x)| N 1−1/α
and then
|Eδ/2 (N, h)|N 2−2/α Z
|ψ(x + h(x)) − ψ(x) − h(x)|2 dx
Eδ/2 (N,h)
X
j
Z
j
Eδ/2
(N,h)
|ψ(x + h(x)) − ψ(x) − h(x)|2 dx ,
10
D. BAZZANELLA
1/α
j
where Eδ/2
(N, h) = Eδ/2 (N, h) ∩ [Nj , Nj + Y ]. Choosing Tj = Nj
deduce
|Eδ/2 (N, h)|N 2−2/α x
x
ψ(x + ) − ψ(x) −
Tj
Tj
Z
X
j
/α we can
2
dx
j
Eδ/2
(N,h)
≤
X Z 2N
ψ(x +
N
j
x
x
) − ψ(x) −
Tj
Tj
2
dx.
Hence
|Eδ/2 (N, h)| N
−2+2/α
X Z 2N
j
N
x
x
ψ(x + ) − ψ(x) −
Tj
Tj
2
dx.
(11)
Follow the same line, but choosing the function g(x) N ε , for every ε > 0,
we can also obtain
|Eδ/2 (N, h)| N
−2+2/α
X Z 2N
N
j
2
x
x
ψ(x + ) − ψ(x) −
+ B(x, Tj ) dx,
Tj
Tj
(12)
and
4
X Z 2N
x
x
|Eδ/2 (N, h)| N
ψ(x + ) − ψ(x) −
+ B(x, Tj ) dx,
Tj
Tj
N
j
(13)
for every function B(x, T ) such that
−4+4/α
B(x, T ) x
T log N
uniformly for 1 ≤ T ≤ N and N ≤ x ≤ 2N .
Lemma 2 and (12) imply that
|Eδ/2 (N, h)| N 1/2+3/(5α)+ε .
(14)
Lemma 3 and (13) lead to
|Eδ/2 (N, h)| N τ (α)+ε ,
(15)
PRIMES BETWEEN CONSECUTIVE POWERS
with
τ (α) =







3
2α
27
53
≤α≤
16
26






36 + 24α − 5α2
3α(α + 12)
53
12
≤α<
26
5
11
Theorem 1 is consequence of (10), (14) and (15).
We now assume the Lindelöf hypothesis. Lemma 4 and (13) yields
|Eδ/2 (N, h)| N −4+4/α
X
N 4+ε Tj−3 N 1/α+2ε .
(16)
j
From (10) and (16) we can conclude
|Aδ (N, α)| Eδ/2 (N, h)
N 1/α+2ε
+
1
+ 1 N 2/α−1+2ε ,
N 1−1/α
N 1−1/α
for every 1 < α ≤ 2 and δ > 0. This completes the proof of Theorem 2.
Lastly we assume the Riemann hypothesis. Lemma 5 and (11) lead to
|Eδ/2 (N, h)| N −2+2/α
X
Tj−1 N 2 log2 (2Tj ) N 1/α log2 N g(N ).
(17)
j
From (10) and (17) we can conclude
|Aδ (N, α)| Eδ/2 (N, h)
N 1/α log2 N g(N )
+
1
+ 1 N 2/α−1 log2 N g(N ),
N 1−1/α
N 1−1/α
for every 1 < α ≤ 2 and δ > 0. Thus Theorem 3 is proved.
References
[1] D. Bazzanella, Primes between consecutive powers, preprint 2005.
[2] D. Bazzanella, Primes between consecutive square, Arch. Math. 75
(2000), 29-34.
12
D. BAZZANELLA
[3] D. Bazzanella and A. Perelli, The exceptional set for the number of primes
in short intervals, J. Number Theory 80 (2000), 109-124 .
[4] R. C. Baker, G. Harman and J. Pintz, The difference between consecutive
primes II, Proc. London Math. Soc. (3) 83 (2001), no. 3, 532–562.
[5] H. Davenport, Multiplicative Number Theory, volume GTM 74, Springer
- Verlag, 1980. second edition.
[6] G.Halász and P.Turán, On the distribution of roots of Riemann zeta and
allied functions, J. Number Theory 1 (1969), 121-137.
[7] D.R. Heath-Brown, The difference between consecutive primes II, J.
London Math. Soc. (2), 19 (1979), 207–220.
[8] D.R. Heath-Brown, Zero density estimates for the Riemann zeta-function
and Dirichlet L-function, J. London Math. Soc. (2), 19 (1979), 221–232.
[9] D.R. Heath-Brown, The difference between consecutive primes IV, A tribute to Paul Erdös (ed. A Baker, B. Bollobás, and A. Hajnal, Cambridge
University Press, 1990), 277–278.
[10] D.R. Heath-Brown, The difference between consecutive primes IV, A tribute to Paul Erdös (ed. A Baker, B. Bollobás, and A. Hajnal, Cambridge
University Press, 1990), 277–278.
[11] M.N.Huxley, On the difference between consecutive primes, Invent. Math.
15 (1972), 164–170.
[12] C. Jia, Almost all short intervals containing prime numbers, Acta Arith.
76 (1996), no. 1, 21–84.
[13] B. Saffari and R. C. Vaughan, On the fractional parts of x/n and related
sequences II, Ann. Inst. Fourier, Grenoble, 27 (1977), 1-30.
[14] E.C.Titchmarsh, The Theory of the Riemann Zeta-function, second ed.,
Oxford U.P., 1986.
[15] G. Yu, The differences between consecutive primes, Bull. London Math.
Soc. 28 (1996), no. 3, 242–248.
PRIMES BETWEEN CONSECUTIVE POWERS
Danilo Bazzanella
Dipartimento di Matematica
Politecnico di Torino
Corso Duca degli Abruzzi, 24
10129 Torino, Italy
e-mail : danilo.bazzanella@polito.it
web: http://staff.polito.it/danilo.bazzanella
13