Professional Documents
Culture Documents
Abstract
For the old question whether there is always a prime in the interval
[kn, (k+1)n] or not, the famous Bertrands postulate gave an affirmative
answer for k = 1. It was first proved by P.L. Chebyshev in 1850, and an
elegant elementary proof was given by P. Erdos in 1932 (reproduced in
[2, pp. 171-173]). M. El Bachraoui used elementary techniques to prove
the case k = 2 in 2006 [1]. This paper gives a proof of the case k = 3,
again without using the prime number theorem or any deep analytic
result. In addition we give a lower bound for the number of primes in
the interval [3n, 4n], which shows that as n tends to infinity, the number
of primes in the interval [3n, 4n] goes to infinity.
Notations
Throughout this paper, we let n run through the positive integers and p run
through the primes. We also let (n) be the prime counting function, which
counts the number of primes not exceeding n. Further define
1
1
f (x) = 2xx+ 2 ex e 12x
and
g(x) =
2xx+ 2 ex e 12x+1 .
Lemmas
Lemma 1.1. If n 8, then
(n)
n
.
2
Proof. This is trivial since 1, 9 and all even positive integers are not prime.
Andy Loo
1
, define the function
12
f (x + c)
.
g(c)g(x)
1
Then for x , h1 (x) is increasing.
2
Proof. It suffices to prove that the function
1
1
1
1
+ln(x+c)
2
2(x + c) 12(x + c)
2x
1
12 x +
1
12
!
,
2 +ln x
1
+ ln x.
2x
12(x + c)2
so that
1
2 , it suffices to prove that F1 (x) is increasing,
1
12 x +
12
1
1
+ ln(x + c)
+ ln x 0.
2(x + c)
2x
We actually have
1
1
2x 1
+ =
,
2
2x
x
2x2
1
which must be non-negative for all x . Therefore, the desired result follows.
2
F1 (x) =
Andy Loo
f (c)
.
g(x)g(c x)
1
c
c
Then when
x < , h2 (x) > 0; when x = , h2 (x) = 0 and when
2
2
2
1
c
< x c , h2 (x) < 0.
2
2
Proof. It suffices to prove that the function
1
1
12(cx)+1
where
F2 (x) =
Clearly, for
12
1
+
+ ln x.
2
(12(c x) + 1)
2x
1
x < c,
2
1
1
12(cx)+1
> 0.
2x 1
288
+
,
(12(c x) + 1)3
2x2
1
1
1
x < c. Thus whenever
x c ,
2
2
2
F2 (x) is increasing while F2 (c x) is decreasing, implying that there is at most
1
1
one value of x with x c satisfying F2 (x) = F2 (c x). It is clear
2
2
1
c
c
that x = is such a value. It follows that when x < , H2 (x) < 0 and
2
2
2
c
1
Andy Loo
Main Results
Now, suppose
n > e12 . The product of all primes p (3n, 4n], if any, must
divide 4n
. Let (p) be the power of p in the prime factorization of 4n
. Let
3n
3n
4n
= T1 T2 T3
3n
where
Y
T1 =
p(p) ,
T2 =
p 4n
p(p)
4n<p3n
and T3 =
p(p) .
3n<p4n
Bounding each multiplicand in T1 from above by 4n (see [3, p. 24]) and applying
Lemma 1.1,
4n
( 4n)
T1 < (4n)
(4n) 2 = (4n) n .
n
in [3, p. 24] manifests, for
Consider T2 . As the prime factorization of
j
where (r, s) = 1 if {s} {r} and (r, s) = [s r] + 1 if {s} < {r}. In both
cases, (r, s) s.
2n/7
4n/17
2n
4n/3
.
and D =
,C=
,B=
Now let A =
4n/15
3n/13
3n/2
n
We have the following observations:
Y
Y
n
p 4 6 (by Lemma 1.2)
p n
6
4n<p n
6
If
2n
n
<p
, then
6
11
2p <
Hence
n
<p 2n
6
11
n
3n
< 3p < 8p <
< 9p < 11p 2n.
2
2
p divides B.
Andy Loo
If
4n
2n
<p
, then
11
21
p<
Y
Hence
n
4n
< 2p < 5p < n < 6p < 7p
.
3
3
p divides A.
2n
<p 4n
11
21
If
4n
n
< p , then
21
5
5p n < 6p < 15p 3n < 16p < 20p 4n < 21p.
Hence (p) = 0.
If
2n
n
<p
, then
5
9
p<
Y
Hence
n
4n
< 2p < 4p < n < 5p < 6p
.
3
3
p divides A.
n
<p 2n
5
9
If
3n
2n
<p
, then
9
13
4p < n < 5p < 13p < 3n < 14p < 17p < 4n < 18p.
Hence (p) = 0.
Y
p divides C.
3n
<p 4n
13
17
If
4n
n
< p , then
17
4
4p n < 5p < 12p 3n < 13p < 16p 4n < 17p.
Hence (p) = 0.
If
4n
n
<p
, then
4
15
p<
Hence
n
<p 4n
4
15
n
4n
< 2p < 3p < n < 4p < 5p
.
3
3
p divides A.
Andy Loo
p divides D.
4n
<p 2n
15
7
If
2n
3n
<p
, then
7
10
3p < n < 4p < 10p 3n < 11p < 13p < 4n < 14p.
Hence (p) = 0.
If
n
3n
< p , then
10
3
p<
Hence
3n
n
< 2p < 4p <
< 5p < 6p 2n.
2
2
p divides B.
3n
<p n
10
3
If
4n
n
<p
, then
3
9
4n
n
< p < 2p < n < 3p
.
3
3
Hence
p divides A.
n
<p 4n
3
9
If
4n
n
< p , then
9
2
2p n < 3p < 6p 3n < 7p < 8p 4n < 9p.
Hence (p) = 0.
If
2n
n
<p
, then
2
3
n
3n
< p < 2p <
< 3p 2n.
2
2
Hence
p divides B.
n
<p 2n
2
3
If
3n
2n
<p
, then
3
4
p < n < 2p < 4p 3n < 5p < 4n < 6p.
Hence (p) = 0.
Andy Loo
If
4n
3n
<p
, then
4
5
3n
n
<p<
< 2p 2n.
2
2
Hence
p divides B.
3n
<p 4n
4
5
If
4n
< p n, then
5
p n < 2p < 3p 3n < 4p 4n < 5p.
Hence (p) = 0.
Y
p divides A.
n<p 4n
3
If
4n
<p
3
3n
,
2
then
n < p < 2p 3n < 4n < 3p.
Hence (p) = 0.
Y
p divides B.
3n
<p2n
2
T2 4 6 ABCD.
Note that by Lemma 1.3,
(4n)!
4n
=
3n
(3n)!n!
g(4n)
>
f (3n)f (n)
n
1
1
1
2
256
=
e 48n+1 36n 12n
,
27
6n
Andy Loo
and similarly,
4n
!
4n [4n/3]
4n
3
4n/3
A=
=
n
4n
n
3
3
n !
n!
3
4n
f
4n
3
<
4n
3
n
g(n)g
3
4n
f
4n
3
(by Lemma 1.4)
n
3
g(n)g
3
r
4 n
4n
43
2 1 1 1
=
e 16n 12n+1 4n+1
,
3 n
3
3n
+1
2n
2n
2
2n
= 2n
B=
2n
3n/2
3n [3n/2] + 1
[3n/2]
2n
2
3n
+1
f (2n)
< 2n 2
3n
3n
3n
2n
g
+ 1 g 2n
+1
2
2
2
f (2n)
(by Lemma 1.5)
< (6n + 4)
3n
3n
g 2n
g
2
2
n
16
12n + 8 1 1 1
=
e 24n 18n+1 6n+1
,
3
3n
32
3n
+1
[4n/17]
4n [4n/17]
4n
13
4n/17
C=
=
3n/13
3n [3n/13] + 1
4n
17 [3n/13]
17
17
13
4n
3n
f
+
1
4n
17
13
3n
3n
4n
3n
17 4n
1
g
+1 g
+1
17
13
13
17
13
Andy Loo
4n
f
4n 51n + 221
17
<
4n 3n
3n
17 n 221
g
g
13
17
13
133 174 n
13
221
17
1
4
4n 51n + 221
13
26 48n
36n+13
12n+221
=
,
221 221
e
17 n 221
3
17
6n
and
4n
+1
[2n/7]
2n [2n/7]
2n
15
2n/7
=
D=
4n/15
4n [4n/15] + 1
2n
7 [4n/15]
7
7
15
2n
4n
f
+
1
2n
7
15
4n
2n
4n
4n
7 2n
1
+1 g
+1
g
7
15
15
7
15
2n
f
2n 28n + 105
7
(by Lemmas 1.4 and 1.5)
<
2n 4n
4n
7 2n 105
g
g
15
7
15
2 4 2 n
5
35
7
4n2 + 15n
105 105 15 15 2 7
15
=
.
Therefore
1
4n
1
4n
>
T3 =
n
n
3n (4n) 4 6 ABCD
3n T1 T2
3
3
n 2 (n 221)(2n 105)
3 2 E n
n
>
e M (4n)
4160
(3n + 2)(3n + 13)(4n + 15)
3
3
n 2 n2
3 2 E n
n
>
e M (4n)
4160
(4n)(4n)(5n)
3
5
3 2 E n
e M (4n) n n 2
=
332800
10
Andy Loo
where
E=
1
1
1
1
1
1
1
1
+
+
+
48n + 1 36n 12n 16n 12n + 1 4n + 1 24n 18n + 1
1
17
13
221
7
5
35
+
+
+
+
+
6n + 1 48n 36n + 13 12n + 221 24n 16n + 5 8n + 35
and
4
1 3 4
2 4 2
3
1 221 3 13 17 17 2 105 4 15 7 7 1
256 1 3 3 2
(3)
4 6 > 1.
M=
7 4
16 221
13
4
105
15
2
Obviously
lim eE = 1.
Moreover, we have
5
n
n 52
ln M (4n)
n
= n ln M n ln(4n) ln n.
2
5
5
o(n). Thus, ln M n (4n) n n 2 goes to infinity and so does M n (4n) n n 2 .
It follows that
lim T3 = +,
which means that there exists some n0 such that for all n n0 , T3 > 1.
In fact, it
is routine to check (using WolframAlpha for instance) that when
3
3 2 E n
5
12
n > e ,
e M (4n) n n 2 is always greater than 1 and so T3 > 1.
332800
Direct verification, on the other hand, ensures that there is always a prime in
the interval [3n, 4n] for all positive integers n < e12 . Therefore, our desired
result ensues:
Theorem 2.1. For every positive integer n, there is a prime in the interval
[3n, 4n]. Plainly, it follows that when n 2, there is always a prime in the
interval (3n, 4n).
4(n + 2)
.
Corollary 2.2. If n 3, then there is a prime in the interval n,
3
Proof. If n 0 (mod 3), then the result follows directly from Theorem2.1. If n 1(mod 3), then by Theorem 2.1 there exists a prime
4(n + 2)
. If n 2 (mod 3), then by Theorem 2.1 there exp n + 2,
3
4(n + 1)
ists a prime p n + 1,
.
3
11
Andy Loo
Next, we establish a lower bound for the number of primes in the interval
[3n, 4n]. Bounding each prime in the interval from above by 4n, we have the
following
Theorem 2.3. For n 4, the number of primes in the interval (3n, 4n) is
at least
3
3 2 E n
n 52
e M (4n)
n
log4n
.
332800
Note that
log4n
5
3 2 E n
e M (4n) n n 2
332800
3
3 2
5
ln n + n ln M n ln(4n) + E + ln
2
332800
=
ln n + ln 4
3
5
3 2
+ ln 4
n ln M n ln(4n) + E + ln
5
332800
2
=
2
ln n + ln 4
ln(4n)
n ln M
5
n
.
>
2 ln n
2
ln(4n)
lim
n
= +.
ln n
References
[1] M. El Bachraoui, Prime in the Interval [2n, 3n], International Journal of
Contemporary Mathematical Sciences, Vol.1 (2006), no. 13, 617-621.
12
Andy Loo