See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/335609455

On the twin prime conjecture

Preprint · October 2019

CITATIONS READS

0 1,273

2 authors:

Mbakiso Fix Mothebe Phúc Võ Đặng

University of Botswana American Mathematical Society
16 PUBLICATIONS   48 CITATIONS    23 PUBLICATIONS   85 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Number Theory View project

The conjecture of William Singer for the monomorphism of the algebraic transfer of rank 4 View project

All content following this page was uploaded by Phúc Võ Đặng on 25 August 2020.

The user has requested enhancement of the downloaded file.

 On the twin prime conjecture
- ă.ng Võ Phúc2
Mbakiso Fix Mothebe1 and D
1
Department of Mathematics, University of Botswana,
Pvt Bag 00704, Gaborone, Botswana
2
Faculty of Education Studies, University of Khanh Hoa,
01 Nguyen Chanh, Nha Trang, Khanh Hoa, Vietnam
arXiv:submit/3338755 [math.GM] 25 Aug 2020

Abstract
For n > 1 let pn denote the nth prime in the sequence of consecutive prime numbers in ascending
order. For a positive integer m, denote by π(m) the number of prime numbers less than or equal
to m. Let [ ] denote the floor or greatest integer function. In this paper, we show that:

p2n+1
" #
 
6 π p2n+1 .
n+1

As a consequence, we see that there are infinitely many primes (Euclid’s theorem). Then, we prove
that if we let π2 (m), denote the number of twin primes not exceeding m, then for all n > 2 :

p2n+3
" #
 
6 π2 p2n+3
3(n + 2)

and thereby prove the twin prime conjecture, namely that there are infinitely many prime numbers
p such that p + 2 is also prime.
Keywords:
Primes; Twin primes; Sieve methods.
2010 MSC: 11N05; 11N36.

Contents
1 Introduction and main results 2

2 Preliminary Result 3

3 Permutations of ordered finite sequences under the constraint of preserving

order 4

4 Proof of Theorem 1.2 4

5 A Sieve of Eratosthenes for Twin Primes 7

6 Proof of Theorem 1.5 9

- ă.ng Võ Phúc2 )
Email address: mothebemf@mopipi.ub.bw; dangvophuc@ukh.edu.vn ( Mbakiso Fix Mothebe1 and D

1
1. Introduction and main results
An integer p > 2 is called a prime if its only positive divisors are 1 and p. The prime numbers form
a sequence:

(1.1) 2, 3, 5, 7, 11, 13, 17, 19, 23, 29, 31, 37, 41, 43, . . . .

Euclid (300 B.C.) considered prime numbers and proved that there are infinitely many. Prime
numbers are odd except 2 and the only consecutive prime numbers are 2 and 3. Any two odd prime
numbers in the sequences (1.1) differ by at least 2. Pairs of prime numbers that differ by 2 as, for
example, in the sequence below

(1.2) (3, 5), (5, 7), (11, 13), (17, 19), (29, 31), (41, 43), . . .

are said to be twin primes.

Conjecture 1.1. There exist infinitely many twin primes.
The conjecture is still open. The first known published reference to this question was made by
Alphonse de Polignac in 1849, who conjectured that for every even number k, there are infinitely
many pairs of prime numbers p and p′ such that p′ − p = k (see [9]). The case k = 2 is the twin
prime conjecture. The conjecture has not yet been proven or disproven for a given value of k. In
2013, an important breakthrough was made by Yitang Zhang who proved the conjecture for some
value of k < 70 000 000 (see [13]). Later that same year, James Maynard announced a related
breakthrough which proved the conjecture for some k < 600 (see [6]). In 2014 D.H.J. Polymath
proved the conjecture for k 6 246. (see [10])
In this paper, we prove that Conjecture 1.1 is true. Let [ ] denote the floor or greatest integer
function. The following is our first main result.
Theorem 1.2. For n > 1 let pn denote the nth prime in the sequence of consecutive prime numbers
in ascending order. Then:
p2n+1
" #
 
6 π p2n+1 .
n+1
p2n+1
We see that the quotient n+1 is clearly much greater than n + 2. So, we get
Corollary 1.3. (Euclid’s theorem) There are infinitely many prime numbers.
For a positive integer x, let φ(x) denote the number of positive integers not exceeding x that
are relatively prime to x, which is said to be the Euler phi-function. As well known, if x > 1 has
the prime factorization x = pk11 pk22 . . . pkr r , then

φ(x) = (pk11 − pk11 −1 )(pk22 − pk22 −1 ) . . . (pkr r − pkr r −1 )

= x(1 − p11 )(1 − p12 ) . . . (1 − p1r ).

Now, for each x ∈ N, let φ2 (x) denote the number of positive integers t, 1 6 t 6 x, for which
both 6t − 1and 6t + 1 are relatively prime to x. For example if x = 5, then φ2 (x) = 3 since the pairs
(11, 13) (17, 19) and (29, 31) are the only ones with components that are relatively prime to 30. In
[7], we showed that
Theorem 1.4. If the integer x > 1 has the prime factorization x = 2k1 3k2 pk33 . . . pkr r , then

φ2 (x) = 2k1 3k2 (pk33 − 2pk33 −1 ) . . . (pkr r − 2pkr r −1 ).

Using this, we shall prove the following theorem, which is our second main result.

2
Theorem 1.5. For n > 2 let pn denote the nth prime in the sequence of consecutive prime numbers
in ascending order. Then:
p2n+3
" #
 
6 π2 p2n+3 .
3(n + 2)
Here, denote byπ2 (m)the number of twin primes not exceeding the positive integer m. We see
p2n+3
that the quotient 3(n+2) is increasing and nonconstant. As a consequence, there are infinitely
many prime numbers p such that p + 2 is also prime. This confirms Conjecture 1.1.
Our work is organized as follows: In Section 2, we recall the definition of the well known sieve
of Eratosthenes. In Section 3, we give a necessary result for proving Theorems 1.2 and 1.5. In
Section 4, we prove Theorem 1.2. To make the paper self-contained, in Section 5, we recall a proof
of Theorem 1.4 as appears in [7] and extend the sieve of Eratosthenes to twin primes. Finally, the
proof of Theorem 1.5 is presented in Section 6.

2. Preliminary Result
The concepts required are elementary and can be obtained from introductory texts on number
theory, discrete mathematics and set theory ([3],[4], [12]).
Eratosthenes (276−194 B.C.) was a Greek mathematician whose work in number theory remains
significant. Consider the following lemma:
√
Lemma 2.1. Let a > 1 be an integer. If a is not divisible by a prime number p 6 a, then a is a
prime.
Eratosthenes used the above lemma as a basis of a technique called “Sieve of Eratosthenes" for
finding all the prime numbers less than a given integer x. The algorithm calls for writing down the
integers from 2 to x in their natural order. The composite numbers in the sequence are then sifted
out by crossing off from 2, every second number (all multiples of two) in the list, from the next
remaining number, 3, every third number, from the next remaining number, √ 5, every fifth number,
and so on for all the remaining prime numbers less than or equal to x. The integers that are left
on the list are primes. We shall refer to the set of integers left as the residue of the sieve. Thus
the order of the residue set is equal to π(x). √
Let x > 1 be an integer and let pn(x) be the largest prime number less than or equal to x. Let
p1 = 2, p2 = 3, p3 = 5, . . . , pn , . . . , pn(x) be the sequence of prime numbers less than or equal to
pn(x) in ascending order. For each n, 1 6 n 6 n(x), consider the following formula for finding all
integers less than or equal to x that are relatively prime to nr=1 pr :
Q

 #
n
"
X  X x 
(2.3) S(x, n) = x + (−1)j Qj .
j=1

16s1 <···<sj 6n i=1 ps i 

In particular S(x, n(x)) gives the order of the set consisting of the number 1 and all prime numbers
less than or equal x excluding the primes pn , 1 6 n 6 n(x). Thus π(x) = S(x, n(x)) + n(x) − 1 and
S(x, n) = φ(x) if pj divides x for all j.
To align S(x, n) to the sieve of Eratosthenes we may re-index the sum by the primes as follows:
h i h i h i
x x x
S(x, n) = x − 2 − 3 − 6 −...

h i   
x Pn−1
− + j=1 (−1)
j+1 Qxj
P
pn 16s1 <···<sj 6n−1 .
pn i=1
p si

In our applications we shall assume that S(x, n) is indexed as above by the primes.

3
3. Permutations of ordered finite sequences under the constraint of
preserving order
Let m, n with m > n be a pair of positive integers and let A = {1, 2, 3, . . . , m} and let B =
{1, 2, 3, . . . , n} be ordered subsets of N consisting of the first m and n elements respectively. We
shall say that a permutation a1 , a2 , a3 , . . . , am+n , of the sequence
(3.4) 1, 2, 3, . . . , n, 1, 2, 3, . . . , m
is order preserving if for all ai , aj ∈ A, i < j whenever ai < aj and for all ar , as ∈ B, r < s
whenever ar < as . Let C denote the family of all distinct order preserving permutations of the
sequence (3.4). We claim that:
Lemma 3.1. !
m+n n

|C| = 1− .
m m+1
Note that ! ! !
m+n n m+n m+n

1− = − .
m m+1 m m+1
Thus the orders of the sets C therefore yield elements of the Catalan triangle. The result of the
lemma is proved in [8]. Our application of the result of Lemma 3.1 only makes use of the well
known fact that
n
! !
m+n X m+n−i
=
m+1 i=1
m
and so its proof is omitted.
The result of Lemma 3.1 can be seen as a sieve that strikes out m+n


m+1 duplicate elements in a
m+n
m+n

set that would otherwise have m elements. As observed above m+1 is the sum total of the
entries on the diagonal containing the entry m+n


m in Pascal’s triangle. Thus given any two positive
integers m, n with m > n we write the integers from 1 to m+n


m in their natural order. Since we
m+n−1

shall be taking ratios, we may assume that m numbers in the sequence are then sifted out by
being crossed out in an evenly distributed fashion. From the remaining integers, m+n−2


m numbers
in the sequence are crossed out in an
 evenly distributed
 fashion and so on for each of the integers
m+n
n
j, 3 6 j 6 n. In any case the m 1 − m+1 elements of the residue set are assumed to be evenly
m+n

distributed over the interval 1 to m . Thus in our application any two consecutive elements in
m+1
the residue of the sieve may be assumed to be separated by an interval of m−n+1 . We shall refer to
this quantity as the average density of the residue of the sieve of Lemma 3.1. Thus the larger the
average density, then the more sparse are the elements of the residue set.
The above sieve is compared to the sieve of Eratosthenes or that of Equation (2.3) and thereby
generates our main results for the primes and twin primes.

4. Proof of Theorem 1.2

In Lemma 3.1, if we put m = n, then the order of the residue set
! !
2n n 1 2n

|C| = 1− = .
n n+1 n+1 n
Now let x be a positive rational number and suppose that x has n consecutive prime number divisors
p1 = 2, p2 = 3, p3 = 5, . . . , pn such that
 !
n
!
X
j
 X x  1 2n
(4.5) x+ (−1) Qj = .
j=1

16s1 <···<sj 6n i=1 ps i  n+1 n
4
Our wish is to align the above expression with
n
! ! !
2n X 2n − i 1 2n
− =
n i=1
n n+1 n

and make a comparison between x and 2n

n
Then since pj divide x for all j we have
n
! 
1 2n  Y pj 
x= .
n+1 n j=1
(pj − 1)

We claim that:
! n
! 
2n 1 2n  Y pj 
(4.6) >
n n+1 n j=1
(pj − 1)

for all n > 2. This is easily seen to be the case when n = 2. Proceeding by induction assume that
the statement is true for some integer n > 2. Then to obtain the expression for n + 1 we multiply
the left hand side of (4.6) by 2(2n+1)
n+1 and the right hand side by

2(2n + 1) (n + 1) pn+1
.
(n + 1) (n + 2) (pn+1 − 1)
(n+1) pn+1
But (n+2) (pn+1 −1) < 1 for all n > 2 and this proves our claim.
It follows from Inequality (4.6) that
! n

n
! 
2n  Y 1 2n  Y
(pj − 1) > pj 
n j=1
n + 1 n j=1

for all n > 2 (with strict inequality if n > 2). Now

 ! n  ! n 
1 2n  Y 1 2n  Y
(4.7) φ pj   = (pj − 1)
n+1 n j=1
n+1 n j=1

while
Q  
n 2n
Pn 2n−i

j=1 (pj − 1) n − i=1 n

Q 
1 2n
n
(4.8) = n+1 n j=1 (pj − 1) .
Q 
2n
n
Let yn = n j=1 (pj − 1) ,
 #
n
"
X  X yn 
T (yn ) = yn + (−1)j Qj
j=1

16s1 <···<sj 6n i=1 ps i 

and  
n n
! !!
Y 2n X 2n − i
R(yn ) =  (pj − 1) − .
j=1
n i=1
n
Then T (yn ) > R(yn ).
2n

For all n > 3 let xn+1 = p2n+1 − 1, Cn2n = n ,
 
n
" #
X  X xn+1 
P (xn+1 ) = xn+1 + (−1)j Qj ,
i=1 psi
 
j=1 16s1 <···<sj 6n

5
and let  !
n
xn+1 2n − j  xn+1
R(yn ) = Cn2n −
X
.
yn j=1
n Cn2n
The residue of the sieve T consisting of integers below xn+1 is
 
n
"" # #
xn+1 ∗  yn xn+1 
(−1)j
X X
T (yn ) = xn+1 + Qj .
yn j=1

16s1 <···<sj 6n i=1 ps i yn 

We claim that for each n > 3:

p2n+1
" # !!
xn+1 1 2n
6
n+1 (n + 1) Cn2n n

xn+1
= R(yn )
yn
xn+1 ∗
6 T (yn )
yn
 
6 π p2n+1 .
xn+1 ∗
6 π p2n+1 . It suffices to show that


We first show that yn T (yn )

xn+1 ∗
T (yn ) = P (xn+1 ).
yn
xn+1 ∗
Comparing the terms of yn T (yn ) with those of P (xn+1 ) first note that for all j and all si , 1 6
s1 < · · · < sj 6 n,
" # & '
yn xn+1 yn xn+1 yn xn+1
Qj 6 Qj 6 Qj .
i=1 ps i yn i=1 psi
yn i=1 psi
yn

Therefore " # " #

xn+1 yn xn+1
Qj 6 Qj
i=1 psi i=1 psi
yn
   
xn+1
since yn < 1 if n > 3 and so Qxjn+1 is the integer part of Qj yn xn+1
yn . If n > 3 we may
p
i=1 si i=1
p si
assume that xyn+1
n
< 1 since, by Bertrand’s postulate, there is a prime p such that n < p 6 2n, while
2(n+1)
Cn+1 > 2Cn for all n > 1. Thus xn+1
2n ∗
yn T (yn ) = P (xn+1 ) as required.
Finally we show that
xn+1
R(yn ) 6 P (xn+1 ).
yn
But as explained in the proof of Theorem 1.5 in Section 6 it is sufficient to show that:
 
n
1  pj 
Y
lim = 0.
n→∞ n + 1 j=1 (pj − 1)

The limit may be verified using similar methods as in the proof of Theorem 1.5 in Section 6 so the
details are omitted.
The result of Theorem 1.2 has however been confirmed independently by an anonymous reviewer
as follows: That
p2n+1
" #
 
< π p2n+1
n+1
6
follows from the results of Rosser and Schoenfeld (see [11]). More precisely they show that π(x) >
x
logx for x > 17. Thus, it is enough that

p2n+1 p2
> n+1
(2log(pn+1 ) (n + 1)

for n > 7, which is equivalent to logpn+1 < (n+1)

2 . But in the same Rosser, Schoenfeld paper it is
shown that
pn+1 < (n + 1)(log(n + 1) + loglog(n + 1))
for n > 6. . Thus, it is enough that
(n + 1)
log((n + 1)(log(n + 1) + loglog(n + 1))) < .
2
Regarding log((n + 1)(log(n + 1) + loglog(n + 1))) as a function of n, we get its derivative to be
3
less than (n+1) which is smaller than 0.5 for n > 6. Thus for values of n large enough log((n +
(n+1)
1)(log(n + 1) + loglog(n + 1))) is less than 2 . For smaller values of n one can check the inequality
individually.

5. A Sieve of Eratosthenes for Twin Primes

For purposes of completeness we, in this section, recall a proof of Theorem 1.4 as appears in [7]
and extend the sieve of Eratosthenes to twin primes.
Proof of Theorem 1.4. We first show that the Euler phi-function can be extended to pairs of integers
(6t − 1, 6t + 1) : t > 0. In particular the extension may be viewed as a sieve for the twin primes.
For each x ∈ N, let φ2 (x) denote the number of positive integers t, 1 6 t 6 x, for which both
6t − 1and 6t + 1 are relatively prime to x. We show that we can evaluate φ2 (x) from the prime
factorization of n. Our arguments are based on those used by Burton [3] to show that the Euler
phi-function is multiplicative. We have the following.
Proposition 5.1. Let k and s be nonnegative numbers and let p > 5 be a prime number. Then:
(i) φ2 (2k ) = 2k .
(ii) φ2 (3s ) = 3s .  
(iii) φ2 (pk ) = pk − 2pk−1 = pk 1 − 2p .

Proof. (i) and (ii). For all nonnegative integers k and s and t :

gcd(6t − 1, 2k ) = gcd(6t + 1, 2k ) = 1

and
gcd(6t − 1, 3s ) = gcd(6t + 1, 3s ) = 1.
(iii) Clearly gcd(6t − 1, pk ) = 1 if and only if p does not divide 6t − 1 and gcd(6t + 1, pk ) = 1
if and only if p does not divide 6t + 1. There is one integer between 1 and p that satisfies the
congruence relation 6t ≡ 1 (mod p). Hence there are pk−1 integers between 1 and pk that satisfy
6t ≡ 1 (mod p). Similarly there are pk−1 integers of the form 6t + 1 between 1 and pk divisible by
p. Thus the set
{(6t − 1, 6t + 1) | 1 6 t 6 pk }
contains exactly pk − 2pk−1 pairs corresponding to integers t for which both gcd(6t − 1, pk ) = 1 and
gcd(6t + 1, pk ) = 1. Thus φ2 (pk ) = pk − 2pk−1 .

7
 We recall that a number theoretic function f is said to be multiplicative if f (mn) = f (m)f (n)
whenever gcd(m, n) = 1. From the proof of Theorem 5.1 it is clear that for all integers k s, we have
φ2 (2k 3s ) = φ2 (2k )φ2 (3s ).
We now show that the function φ2 is multiplicative. This will be our basis for the proof of
Theorem 1.4. We need some lemmas.
Lemma 5.2. Given integers m, n. Then, gcd(6t − 1, mn) = 1 if and only if gcd(6t − 1, m) = 1 and
gcd(6t − 1, n) = 1.
Similarly given integers m, n, gcd(6t + 1, mn) = 1 if and only if gcd(6t + 1, m) = 1 and gcd(6t +
1, n) = 1. This is an immediate consequence of the following standard result.
Lemma 5.3. Given integers m, n, k. Then, gcd(k, mn) = 1 if and only if gcd(k, m) = 1 and
gcd(k, n) = 1.
We note also the following standard result.
Lemma 5.4. If a = bq + r, then gcd(a, b) = gcd(b, r).
Proposition 5.5. The function φ2 is multiplicative, that is, if gcd(m, n) = 1, then
φ2 (mn) = φ2 (m)φ2 (n).
Proof. The result holds if either m or n equals 1. We shall therefore assume neither m nor n equals
1. Arrange the integer pairs (6t − 1, 6t + 1), 1 6 t 6 mn, in an n × m array as follows:

 
(5,7) ... (6m − 1, 6m + 1)

 (6(m + 1) − 1, 6(m + 1) + 1) ... (6(2m) − 1, 6(2m) + 1) 

 .. .. .. 
. . .
 
 

.. .. .. 
. . .
 
 
(6((n − 1)m + 1) − 1, 6((n − 1)m + 1) + 1) ... (6(mn) − 1, 6(mn) + 1)
We know that φ2 (mn) is equal to the number of pairs (6t − 1, 6t + 1) in this matrix for which both
6t − 1 and 6t + 1 are relatively prime to mn. By virtue of Lemma 5.2 this is the same as the number
of pairs (6t − 1, 6t + 1) in the same matrix for which both 6t − 1 and 6t + 1 are relatively prime to
each of m and n.
We first note, by virtue of Lemma 5.4, that gcd(6(qm + t) − 1, m) = gcd(6t − 1, m) and likewise
gcd(6(qm + t) + 1, m) = gcd(6t + 1, m). Therefore the pairs (6(qm + t) − 1, 6(qm + t) + 1) in the tth
column are both relatively prime to m if and only if both 6t − 1 and 6t + 1 are relatively prime to
m. Therefore only φ2 (m) columns contain pairs (6t − 1, 6t + 1) both relatively prime to m and every
other pair in the column will constitute of integers both relatively prime to m. The problem now is
to show that in each of these φ2 (m) columns there are exactly φ2 (n) integer pairs (6t − 1, 6t + 1)
that are both relatively prime to n, for then altogether there would be φ2 (m)φ2 (n) pairs in the
table that are relatively prime to both m and n.
The entries that are in the tth column (where it is assumed gcd(6t − 1, m) = gcd(6t + 1, m) = 1)
are:
(6t − 1, 6t + 1), (6(m + t) − 1, 6(m + t) + 1), . . . , (6((n − 1)m + t) − 1, 6((n − 1)m + t) + 1)
There are n pairs in this sequence and for no two pairs
(6(qm + t) − 1, 6(qm + t) + 1), (6(jm + t) − 1, 6(jm + t) + 1)
in the sequence do we have
6(qm + t) − 1 ≡ 6(jm + t) − 1 (mod n)
8
and
6(qm + t) + 1 ≡ 6(jm + t) + 1 (mod n),
since otherwise we would arrive at a contradiction q ≡ j (mod n).
Thus the terms of the sequence,

t, m + t, 2m + t, . . . , (n − 1)m + t

are congruent modulo n to 0, 1, 2, . . . n − 1 in some order.

Now suppose r is congruent modulo n to qm+t. Then the integers 6(qm+t)−1 and 6(qm+t)+1
are both relatively prime to n if and only if both 6r − 1 and 6r + 1 are relatively prime to n. The
implication is that the tth column contains as many pairs of integers that are relatively prime to n
as does the set {(5, 7), (11, 13), . . . , (6n − 1, 6n + 1)}, namely φ2 (n) pairs. Thus the number of pairs
of integers (6t − 1, 6t + 1) in the matrix that are relatively prime to m and n is φ2 (m)φ2 (n). This
completes the proof of the proposition.
Now, Theorem 1.4 follows from Propositions 5.1 and 5.5.
From the statement of Theorem 1.4 we note that if n > 3 and p3 = 5, p4 = 7, p5 = 11, . . . , pn
is an ordered sequence of prime numbers in ascending order and k = x = 5.7.11. . . . .pn , then
2 2 2
φ2 (k) = k(1 − p3 )(1 − p4 ) . . . (1 − pn )
  
j−2 k
Q2j
Pn j
=k+ j=3 (−1)
P
36s1 <···<sj 6n .
i=1
p si

But by definition φ2 enumerates the pairs (6t − 1, 6t + 1) less than 6k both of which are relatively
prime to k. By Lemma 2.1 the pairs (6t − 1, 6t + 1) in the residue of the sieve φ2 which are both
less than p2n+1 are therefore twin primes. Thus if we denote this set by T2 (p2n+1 ) then we must have
|T2 (p2n+1 )| 6 π2 p2n+1 .

We can consider φ2 (k) as defining a sieve on the set of integers 1, 2, 3, 4, . . . , k or, equivalently,
on the set
S = {(6t − 1, 6t + 1) | 1 6 t 6 k}
which for each pj , 3 6 j 6 n, strikes out t whenever 6t − 1 or 6t + 1 is divisible by pj . Since the
primes pj are unevenly distributed they strike out the values t in an unevenly distributed manner.
However this is achieved in a cyclic pattern in the sense that for j < n, pj strikes out the same
number of values of t, with the same irregularity, in each interval
j
Y j
Y
s pi 6 t < (s + 1) pi .
i=3 i=3

The average density or sparsity of elements of the residue set is therefore

n
Y pj
.
j=3
(pj − 2)

6. Proof of Theorem 1.5

Some of the arguments in this section parallel those of Section 4 so some details are omitted. In
Lemma 3.1 if we put m = n + 1, then the order of the residue set
! !
2n + 1 n 2 2n + 1

(6.9) |C| = 1− = .
n+1 n+2 n+2 n+1

9
 Now let x = 6k + 1 be a positive rational number and suppose that k has n consecutive prime
number divisors p3 = 5, p4 = 7, p5 = 11, . . . , pn+2 such that
 
n+2
! !
X  X 2j−2 k  2 2n + 1
(6.10) k+ (−1)j Qj = .
j=3

36s1 <···<sj 6n+2 i=1 psi
 n+2 n+1

Our aim is to align the above expression with

n
! ! !
2n + 1 X 2n − i + 1 2 2n + 1
− =
n+1 i=1
n+1 n+2 n+1

and then compare k with 2n+1

n+1 .
Then since pj divide k for all j we have
! 
2 2n + 1 n+2
Y pj 
k= .
n+2 n+1 j=3
(pj − 2)

We claim that if n > 9, then:

! 
2n + 1 n+2
!
2n + 1 2 Y pj 
(6.11) >
n+1 n+2 n+1 j=3
(pj − 2)

or, equivalently,
 
2 n+2
Y pj 
(6.12) 1>
n + 2 j=3 (pj − 2)

In general we have the following result which compares the average density of the elements of
the residue of the sieve of Equation (6.9) to the average density of the elements of the residue of
the sieve φ2 .
Lemma 6.1. Let a > 1 be an integer. Then there exists an integer n0 such that for all integers
n > n0 :
n+2
Y pj n+2
a < .
j=3
(p j − 2) 2

Proof. It suffices to show that

 
2 n+2
Y pj 
(6.13) lim =0
n→∞ n + 2 (pj − 2)
j=3

since in this event we can always find an integer n0 such that

 
2 n+2
Y pj  1
<
n + 2 j=3 (pj − 2) a
 Qn+2 pj 
2
for all n > n0 . Given n+2 j=3 (pj −2) we may obtain the next product inductively by multiplying
(n+2) pn+3 (n+2) pn+3
the given product by (n+3) (pn+3 −2) . For
sufficiently large values of n (n > 9) we have (n+3) (pn+3 −2) <
1. We prove a statement logically equivalent to (6.13), namely that
 
n + 2 n+2
Y (pj − 2)
(6.14) lim  = ∞.
n→∞ 2 j=3
p j

10
 (n+2) pn+3
But the reciprocal product of (n+3) (pn+3 −2) may be expressed in the form
! !
n+3 pn+3 − 2 1 2
   
= 1+ 1−
n+2 pn+3 n+2 pn+3
1 2 1
   
= 1+ − 1+
n+2 pn+3 n+2
Now if n > 7, then pn+3 > 3(n + 2) and (n + 2)2 > 6(n + 2) and so
1 2 1 1 2 1
       
1+ − 1+ > 1+ − 1+
n+2 pn+3 n+2 n+2 3(n + 2) n+2
!
1 2 1 1
 
= 1+ − +
n+2 3 n + 2 (n + 2)2
!
1 2 1 1
 
> 1+ − +
n+2 3 n + 2 6(n + 2)
!
2 9n + 20
= 1+ =
9(n + 2) 9n + 18

Thus for all k > 7 :

k  k 
!
9n + 20 n+3 pn+3 − 2
Y  Y 
(6.15) 6 .
n=7
9n + 18 n=7
n+2 pn+3

It is therefore sufficient to show that:

∞  ∞ 
9n + 20 2
Y  Y 
= 1+ = ∞.
n=7
9n + 18 n=7
9n + 18

But if {an } is a sequence of positive real numbers, then the series ∞

n=1 an and the product
∞
n=1 (1+
P Q
P∞ 2
an ) converge or diverge together [5]. The required result therefore follows since n=7 ( 9n+18 ) di-
verges. This proves the result of the lemma.
Now, it follows from Inequality (6.11) that
!  ! 
2n + 1 n+2
Y 2 2n + 1 n+2
Y
(pj − 2) > pj 
n+1 j=3
n+2 n+1 j=3

for all n > 9. Now

 !  ! 
2 2n + 1 n+2
Y 2 2n + 1 n+2
Y
(6.16) φ2  pj   = (pj − 2)
n+2 n+1 j=3
n+2 n+1 j=3

while
Q  
n+2 2n+1
Pn 2n−i+1

j=3 (pj − 2) n+1 − i=1 n+1

Q 
2 2n+1
n+2
(6.17) = n+2 n+1 j=3 (pj − 2) .

The density of the residue of the sieve of Equation (6.17) is n+2 2 which, when n > 9, is greater
Qn+2 pj
than j=3 (pj −2) , the average density of the residue of the sieve of Equation (6.16). Thus the sieve
of Equation (6.17) is more dominant or sifts out more elements than φ2 or the sieve of Equation
(6.16).

11
 Q 
2 2n+1
n+2
Let sn = n+2 n+1 j=3 pj and and consider the sieve
Q  
2 n+2 2n+1
Pn 2n−j+1

(6.18) R2 (sn ) = n+2 j=3 pj n+1 − j=1 n+1

 2 Q 
2 2n+1
n+2
= n+2 n+1 j=3 pj .

We shall apply the result of the following lemma to show that the sieve 6.18 is dominant over the
sieve (6.16) in certain subintervals of the interval 1, . . . , sn .
Lemma 6.2. Let n > 12 be a fixed integer and let pn denote the nth prime in the sequence of
consecutive prime numbers in ascending order. Then for all integers r > 1 :

r(n − 1)n < pn pn+r−1 .

Proof. Let t be the largest integer such that ps > ts for all s > n. Since n > 12, then t > 3. Then

r(n − 1)n < t2 n2 < p2n < pn pn+r−1

for all r, 1 6 r 6 t2 . If r > t2 , then we must have

pn pn+r−1
  
n<
n−1 r
 
since in general pn+r−1
r increases and cannot decrease by a significant amount. This proves the
result of the lemma.
In the result of the above lemma we note that each value of r is a count of the number of
multiples of (n − 1)n as well as that of the number of terms in the sequence of products {pn pn+r−1 }.
We note therefore that if k is an integer greater than p2n , then the number of multiples of n(n − 1)
less than k is more than the number of terms of the sequence {pn pn+r−1 } that terminates with the
largest product less or equal to k.
p2 −1 2n+1
= 2n+1


For n > 9, let kn+3 = n+36 , Cn+1 n+1 . Then
!
kn+3 2 2n + 1 kn+3
R2 (sn ) = 2n+1
sn n + 2 n + 1 Cn+1

is a lower bound for the order of the residue set of the sieve of Equation (6.18) consisting of integers
less than kn+3 .  
2 2n+1
Qn+2 2
With k(n) = sn = n+2 n+1 j=3 j , let T2 (pn+3 ) denote the residue set of the sieve φ2
p
consisting of pairs of integers (6t − 1, 6t + 1) less than p2n+3 (so that t < kn+3 .) Clearly R2 (sn ) 6
φ2 (k(n)) if n > 9.
We claim that if n > 9, then:
p2n+3
" # !
2(kn+3 ) 1 2n + 1
6 2n+1
3(n + 2) n + 2 Cn+1 n+1
kn+3
= R2 (sn )
sn
6 |T2 (p2n+3 )|
 
6 π2 p2n+3 .

We must show that

kn+3
R2 (sn ) 6 |T2 (p2n+3 )|.
sn
12
Let T2d (n + 3) denote the average density of the elements of T2 (p2n+3 ). Since the sieve φ2 is fairly
even over the entire interval 1, . . . , k(n), T2d (n + 3) cannot vary by a large amount from the overall
pj
average density n+2 j=3 (pj −2) of the residue of φ2 over the entire interval. Further since the average
Q

density of the residue of the sieve of Equation (6.18), n+2 2 , is even over the entire interval, the above
inequality follows, for sufficiently large values of n, from the result of Lemma 6.1. We all the same
show independently that the sieve of Equation (6.18) is indeed dominant over φ2 on a subinterval
1, . . . , kn+3 of 1, . . . , k(n).
pj
Since R2 (sn ) 6 φ2 (k(n)), the above inequality follows if we assume that T2d (n+3) 6 n+2
Q
j=3 (pj −2) .
pj
We may therefore assume that T2d (n + 3) is greater than n+2 n+2
j=3 (pj −2) but less than 2 .
Q

We proceed by induction on n, the index of the primes and show that kn+3 2
sn R2 (sn ) 6 |T2 (pn+3 )|
implies ksn+1
n+4
R2 (sn+1 ) 6 |T2 (p2n+4 )|. By direct computation it is easy to show that the above in-
kn+3
equality holds for 9 6 n 6 3400. Assume that sn R2 (sn ) 6 |T2 (p2n+3 )| for some n > 3400, that
is,
 !
n
!
2n + 1 X 2n − j + 1  kn+3 2
(6.19)  − 2n+1 6 |T2 (pn+3 )|
n+1 j=1
n + 1 C n+1

for some n > 3400. We show that we must have

 !
n+1
!
2(n + 1) + 1 X 2(n + 1) + 1 − j kn+4
 − 
2(n+1)+1
6 |T2 (p2n+4 )|.
n+2 j=1
n + 2 C n+2

But the average density of the elements in the residue of the sieve
 !
n+1
!
2(n + 1) + 1 X 2(n + 1) + 1 − j kn+4
 − 
2(n+1)+1
n+2 j=1
n+2 Cn+2
n+3 n+3
is equal to 2 which is an increase by n+2 from the average density of the residue of the sieve
 
2(p2n+4)
of Equation (6.18). This is equivalent to sifting out (n+2)(n+3) of the elements of the residue
of the sieve of Equation 6.18 which are less than p2n+4 . On the other hand let Let φ2 (k(n)) be
computed iteratively as follows: for each j, 3 6 j 6 n + 2, let φ2 (k(n), 3) = k(n)(1 − p23 ) and for
j > 4 let φ2 (k(n), j) = φ2 (k(n), j − 1)(1 − p2j ). Then each pj sifts out pairs (6t − 1, 6t + 1) where
6t − 1 or 6t + 1 is equal to pj or a product of pj and a prime or a product of primes not less than
pj . Therefore, by Lemma 6.2, the sieve of Equation 6.18 remains dominant over φ2 in the interval
1, . . . , kn+4 . Now T2 (p2n+4 ) is obtained from the residue of the sieve φ2 by sifting out, over the interval
1, . . . , kn+4 , a pair (6t − 1, 6t + 1) one of whose components is equal to pn+3 as well as those pairs
(6t − 1, 6t + 1), kn+3 6 t 6 kn+4 , one of whose components is a multiple of pn+3 . By the result of
2(p2n+4 )
Lemma 6.2, (n+2)(n+3) is greater than the number of those pairs (6t − 1, 6t + 1), kn+3 6 t 6 kn+4 ,
one of whose components is a multiple of pn+3 by a prime greater than pn+3 . Thus we must have
kn+4 2
sn+1 R2 (sn+1 ) 6 |T2 (pn+4 )| and this completes the induction.
Finally we note that the result of Theorem 1.5 has been shown to be true over a large range of
consecutive integers n > 2 starting with the case n = 2. This completes the proof of Theorem 1.5.
 
p2n+1
is a weak lower bound for π p2n+1 . For large values of n,


Remark. We have seen that n+1
 
p2n+1
is also a lower bound for π2 p2n+1 , that is, for sufficiently large values of n,


n+1

p2n+1
" #
 
6 π2 p2n+1 .
n+1
For instance this is the case when n = 3400.
13
 Acknowledgments. We would like to thank Professor Stephan Wagner and Berndt Gensel for their
helpful comments on earlier versions of this paper. We are also thankful for the useful conversations
we had with Shalin Singh and Dr. Abebe Tufa.

References
[1] J.A. Beachy and W.D. Blair, Abstract Algebra, Waveland Press, Inco. 2006.
[2] R. Betti, The twin prime conjecture and other curiosities regarding prime numbers, Lettera Matematica, 5 (2017), 297-303.
[3] D.M. Burton, Elementary Number Theory, McGraw-Hill, International Editions 2002.
[4] E.G. Goodaire and M.M. Parmenter, Discrete Mathematics with Graph Theory, Prentice Hall, 2002.
[5] Mattias Flygare, Some Properties of Infinite Series , Bachelor Mathematics Degree Project, Karlstads University, 2012.
[6] J. Maynard, Small gaps between primes, Annals of Mathematics, 181(2) (2015), 383-413.
[7] M. F. Mothebe and S. Samuel, Properties of the Euler phi-function on pairs of positive integers (6x − 1, 6x + 1), African Jour.
Math. and Comput. Sci. Res. Vol. 9(2):12-14. (2016). DOI: 10.5897/AJMCSR2016.0648.
[8] M.F. Mothebe, A note on admissible monomials of degree 2λ − 1, Jour. of Math. Sciences and Appl. Vol. 7 No. 1 10-14. (2019).
DOI:10.12691/jmsa-7 -1 -2.
[9] Alphonse de Polignac, Recherches nouvelles sur les nombres premiers, Comptes Rendus des Séances de l’Académie des Sciences,
1849.
[10] D.H.J. Polymath, Variants of the Sielberg sieve, and bounded intervals containing many primes, Res. Math. Sci. 1 (2014) Art.
12.
[11] J.H. Rosser and L. Schoenfeld, Approximate formulas for some fuctions of prime numbers, Illinois Jour. Math. 64-94. (1962).
[12] J.K. Strayer, Elementary Number Theory, PWS Publishing Company, 1994.
[13] Yitang Zhang, Bounded gaps between primes, Annals of Mathematics, 179(3) (2014), 1121-1174.

14

View publication stats