The Distribution of Prime Numbers

and its Applications

by
Samuel Budd

A project submitted to the Department of

Mathematical Sciences in conformity with the requirements
for Math 4301 (Honour’s Seminar)

Lakehead University
Thunder Bay, Ontario, Canada
copyright c (2015) Samuel Budd
 Abstract

This project will examine the distribution of prime numbers, as well as applications
of these results. We begin by approximating how many prime numbers exist that are
less than or equal to any given number N . This approximation is known as Tchebychev’s
Theorem. We then use this result to work through the proof of Mertens’ First and Second
Theorem. In the proof of Mertens’ first Theorem, we show that there exists a bound on
the difference between a series of fractions containing primes p, and log N , where p are
all the primes less than or equal to N . This difference is proved to be less than 4. We
also prove through Mertens’ Second Theorem that the difference between the sum of the
reciprocals of all the primes less than or equal to a given N and ln ln N is also less than a
relatively small constant. This constant is independent of N , and can be taken equal to
15.

i
 To my parents,
for their love and support.

ii
 Acknowledgements

I would like to thank my supervisor Dr. Razvan Anisca for his help and patience
throughout this project. This would not have been possible without his guidance. I am
also very grateful for the support that my family and friends have given me throughout
my education.

iii
 Contents

Abstract i
Acknowledgements iii
Chapter 1. Introduction 1
1. Preliminaries 2
Chapter 2. Tchebychev’s Theorem 4
Chapter 3. Mertens’ First Theorem 14
Chapter 4. Mertens’ Second Theorem 21
Chapter 5. Concluding Remarks 29
Bibliography 30

iv
 CHAPTER 1

Introduction

Prime numbers have been the focus of mathematicians for centuries. When Euclid’s
Elements were first published circa 300 B.C.E., there were already several important
results of prime numbers that had been discovered. In his ninth book, Euclid proved
that there are infinitely many prime numbers. In the same book, Euclid also included
a proof of the fundamental theorem of arithmetic, which allows us to break a number
into a distinct product of its prime factors: an immensely useful tool when working in
the area of number theory. For some time after that, the main focuses of prime number
theory were developing methods of determining whether a number was prime, and how
to find exceedingly larger prime numbers. During this time, the Greek mathematician
Eratosthenes developed one of the most effective methods for determining larger primes
up to a given limit. However, this method is only efficient for relatively small primes,
no higher than 10 million. After this, there was very little development in prime number
theory for over a millennium.
Around the beginning of the seventeenth century, Fermat, Mersenne, and Euler made
some significant contributions to the field of prime number theory, which included Fer-
mat’s Little Theorem as well as the idea of Mersenne Numbers. [3] A Mersenne number
is one of the form 2n ´ 1, where if n is prime, then the number 2n ´ 1 could potentially
be prime. If n was not prime, it was proved that 2n ´ 1 cannot be prime. Throughout
the century, Mersenne numbers that were determined to be prime were the largest prime
numbers that could be found. For the next few centuries, Cataldi, Euler, and Lucas
proved the existence of the Mersenne numbers with n “ 19, 31, 127, respectively, with
centuries between each discovery. To date, 45 Mersenne primes have been discovered, the
largest of which with n “ 43112609. The study of primes naturally led to the study of
their distribution, which will be the main focus of this project.
During the late 18th century, both Gauss and Legendre made similar conjectures
regarding πpnq, the number of prime numbers less than or equal to n. In particular,
x
[2] in 1798 Legendre estimated that πpxq could be estimated to log x´1.08366 . It is also
impressive that at the ageş of 15 or 16, Gauss conjectured a similar result to Legendre,
x
which was that πpxq « 2 log1 t dt. Now we focus our attention to the first person to
actually show that πpxq „ logx x . This result was obtained by Tchebychev in 1852, using
completely elementary methods in number theory. The work done by Gauss, Legendre,
and especially Tchebychev laid the ground work for solving what is now known as the
Prime Number Theorem. The Prime Number Theorem states that as x Ñ 8, then

1
Chapter 1. Introduction 2
x
πpxq “ ,
which was proved later by Jacques Hadamard and Charles-Jean-Gustave-
log x
Nicholas de la Vallée Poussin in 1896 using methods from complex analysis. [4] As a result
of the Prime Number Theorem, it has been proven that given any n ą 1, the probability
that a randomly selected number x is prime, with x ď n, is ln1n .

1. Preliminaries

For our study of Tchebychev’s theorem, we will require some basic knowledge of num-
ber theory. This chapter will include many important results that will, in effect, enable
us to work through the proof of Tchebychev’s theorem later on. It is interesting to note
that we can work through a relatively complicated and convoluted proof, with reasonably
elementary methods. We begin by formally defining what a prime number is:
Definition 1.1. A number p is prime if and only if its only divisors are 1 and itself.
Example 1.2. Given the number n “ 10, we can clearly break it up into it’s prime
factors so it will have the form 10 “ 2 ¨ 5. Because 10 is divisible by 1, 2, 5, and 10, we
say that 10 is not prime.
Example 1.3. Given the number n “ 23, we can see that 23 cannot be factored
further than what it already is, thus only 1 and 23 divide 23. Because only 1 and 23
divide 23, we say that 23 is prime
Definition 1.4. For n P Z we say that n is the integer part of a number a if
n ď a ă n ` 1,
and we denote it by n “ ras.
Example 1.5. We will look at several different cases: r0.2s “ 0, r0.9s “ 0, r3.8s “ 3,
r´3.2s “ ´4, r´5.9s “ ´6, r2s “ 2, and lastly r´7s “ ´7.
Corollary 1.6. Let a P R . Then a ´ 1 ă ras ď a for all a.

Proof. By the definition, we know that if n ď a ă n`1, then n “ ras. Setting n “ ras
in the inequality, we obtain n “ ras ď a ă n ` 1. For the other side of the inequality,
we simply subtract 1 from each side of the inequality. This gives us n ´ 1 ď a ´ 1 ă n,
and setting n “ ras again, we find that a ´ 1 ă ras. Combining our two inequalities, we
obtain a ´ 1 ă ras ď a, our desired result. 
Lemma 1.7. Given any real number a, then ra ´ ns “ ras ´ n.

Proof. Clearly, ras ´ n ď a ´ n because in general ras ď a. We also know that

a ´ 1 ă ras which implies that a ă ras ` 1. Using this, we find that a ´ n ă ras ` 1 ´ n.
Bringing these facts together, we obtain ras´n ď a´n ă ras´n`1. This is the definition
of the integer part of a ´ n, thus ras ´ n “ ra ´ ns, and therefore we obtain the result. 
Chapter 1. Introduction 3
Definition 1.8. We define an r-combination as the number of ways to choose a
particular number of elements r from a set of size n, with n P Z, n ě 0, and 0 ď r ď n.
In choosing our r elements, the order of the selected elements does not matter. We say n
choose r, and denote it by
ˆ ˙
n n!
Cpn, rq “ “ .
r r!pn ´ rq!
Example 1.9. Suppose we are asked to find the number of possible ways to choose 6
blocks from a group of 10 distinct blocks, regardless of their order. Then the number of
ways to do this is equal to
ˆ ˙
10 10! 10!
Cp10, 6q “ “ “ “ 210.
6 6!p10 ´ 6q! 6! ¨ 4!
Next we will introduce the Binomial Theorem, which helps us begin our proof of
Tchebychev’s theorem. The proof of the Binomial Theorem is irrelevant to this paper,
and is therefore omitted.
Theorem 1.10. (Binomial Theorem) Let a and b be variables and let n be a nonneg-
ative integer. Then
n ˆ ˙
ÿ n n n´k
a b “ pa ` bqn .
k“0
k

Definition 1.11. We define the factorial function as f : N Ñ Z` , denoted by

f pnq “ n!. The value of f pnq “ n! is the product of the first n positive integers, so
f pnq “ 1 ¨ 2 ¨ 3 ¨ . . . ¨ pn ´ 1q ¨ n, and f p0q “ 0! “ 1.
Example 1.12. Take 5!. By the definition, 5! “ 5 ¨ 4 ¨ 3 ¨ 2 ¨ 1 “ 120. Similarly,
13! “ 13 ¨ 12 ¨ 11 ¨ 10 ¨ 9 ¨ 8 ¨ 7 ¨ 6 ¨ 5 ¨ 4 ¨ 3 ¨ 2 ¨ 1 “ 6, 227, 020, 800.
Theorem 1.13. (Fundamental Theorem of Arithmetic) Every positive integer greater
than 1 can be written uniquely as a prime or as the product of two or more primes where
the prime factors are written in order of nondecreasing size.
Example 1.14. Take for instance n “ 21. Then by the Fundamental Theorem of
Arithmetic, 21 “ 3 ¨ 7. Similarly, for n “ 420, we see that 420 “ 22 ¨ 3 ¨ 5 ¨ 7.
 CHAPTER 2

Tchebychev’s Theorem

We begin this section with some necessary definitions and results that will prove to
be useful in the proof of Tchebychev’s Theorem. Our first definition plays a central role
throughout the duration of this paper.
Definition 2.1. Given any number N ą 1, we define the number of prime numbers
less than or equal to N as πpN q.
Lemma 2.2. Let n, m P Z. Then πpn ` mq ă πpnq ` m.

Proof. We know that the number of prime numbers less than or equal to n is πpnq.
Similarly, the number of prime numbers less than or equal to n`m is πpn`mq. If we take
the difference between the number of prime numbers less than n ` m, and the number of
prime numbers less than n, this difference is clearly less than or equal to m. We can see
this with the following inequality:
πpn ` mq ´ πpnq ď m.
Rearranging, we obtain πpn ` mq ď πpnq ` m, the required result. 
Lemma 2.3. Let b be any real number. Then
#
1
r2bs if b ´ rbs ă 2
(0.1) 2 rbs “ 1
r2bs ´ 1 if b ´ rbs ě 2

Proof. To prove this, we will have to consider two cases, the first of which being if
b ´ rbs ă 12 . Rearranging, we obtain b ă rbs ` 12 . From here, clearly 2b ă 2 rbs ` 1, and
since in general rbs ď b, then 2 rbs ď 2b ă 2 rbs ` 1. This inequality implies that 2 rbs is the
integer part of 2b, and hence 2 rbs “ r2bs. For the second case, we have that b ´ rbs ě 21 .
Rearranging, we obtain rbs ď b ´ 12 . From here, we can see that 2 rbs ď 2b ´ 1, where
2b ´ 1 is also equal to 2b ´ 2 ` 1. Factoring, we obtain 2pb ´ 1q ` 1, and since in general
we have that b ´ 1 ă rbs, then clearly 2b ´ 1 “ 2pb ´ 1q ` 1 ă 2 rbs ` 1. Simplifying the
inequality, we find that 2 rbs ď 2b ´ 1 ă 2 rbs ` 1. This by definition implies that 2 rbs is
the integer part of 2b ´ 1, and hence 2 rbs “ r2b ´ 1s. Using Lemma (1.7), we have that
2 rbs “ r2b ´ 1s “ r2bs ´ 1, which proves the second case. 
Lemma 2.4. Given any positive integer N,
N
N N N´ 2k
N` ` . . . ` k´1 “ 1 .
2 2 1´ 2

4
Chapter 2. Tchebychev’s Theorem 5
N
Proof. We begin by setting sk “ N ` ` . . . ` 2k´1 . Multiplying sk by 12 , we
2
N

obtain s2k “ N2 ` N4 ``. . . ` 2Nk . Now, subtracting

˘
sk
2
from sk , we find that sk ´ s2k “
N ` N2 ` . . . ` 2k´1
N
´ N2 ` N4 ` . . . ` 2Nk . We can see that all the middle terms of the
right hand side cancel, and we are left with sk ´ s2k “ N ´ 2Nk . Factoring out sk , we have
` ˘ N´ N
that sk 1 ´ 12 “ N ´ 2Nk , which after further rearrangement gives us sk “ 1´ 21k . Now,
2
N N
` N N
˘ N ´ Nk
since sk “ N ` 2
` . . . ` 2k´1 , we obtain N ` 2
` ... ` 2k´1
“ 2
1´ 12
, the result we are
looking for. 

With the above information, we now have the necessary tools to prove Tchebychev’s
Theorem. We will now formally introduce this very important Theorem, and its proof
will immediately follow.

Theorem 2.5. (Tchebychev’s Theorem) Let N be any positive integer greater than or
equal to 1. Then πpN q has the following upper and lower bounds:

N N
(0.2) 0.1 ă πpN q ă 4 .
log N log N

Proof. Given two numbers a and b, the binomial theorem states that

n ˆ ˙
ÿ n
an bn´k “ pa ` bqn .
k“0
k
Substituting 2n for n, and setting both a and b equal to 1, then we have
2n ˆ ˙ ˆ ˙ ˆ ˙ ˆ ˙ ˆ ˙
ÿ 2n 2n´k k 2n 2n 2n 2n
p1q p1q “ 1` ` `. . .` `. . .` `1 “ p1`1q2n “ 22n .
k“0
k 2 3 n 2n ´ 1
ˆ ˙
2n
It is clear that ą 0 for any k, thus by rearranging, we can see that
k
ˆ ˙
2n
(0.3) ă 22n .
n

We now claim that 2n´k

n´k
ą 2 for all k ă n. Indeed, multiplying both sides by pn ´ kq,
which is positive when k ă n, we obtain
p2n ´ kqpn ´ kq
ą 2 ¨ pn ´ kq.
pn ´ kq
Simplifying on the left, and expanding on the right, we obtain 2n ´ k ą 2n ´ 2k. After
further simplification, we reach our result as it is clear that 2k ą k.
Chapter 2. Tchebychev’s Theorem 6
ˆ ˙
2n
Next, we expand the combination .
n
ˆ ˙
2n p2nq!
“
n pn!q2
2n ¨ p2n ´ 1q ¨ p2n ´ 2q ¨ p2n ´ 3q . . . p2n ´ pn ´ 1qq ¨ p2n ´ nq!
“
pn!q2
2n ¨ p2n ´ 1q ¨ p2n ´ 2q ¨ p2n ´ 3q . . . pn ` 1q ¨ pn!q
“
pn!q2
2n ¨ p2n ´ 1q ¨ p2n ´ 2q ¨ p2n ´ 3q . . . pn ` 1q
“
n!
2n 2n ´ 1 2n ´ 2 n`1
“ ¨ ¨ ¨ ... ¨
n n´1 n´2 1
ě 2looooomooooon
¨ 2 ¨ 2...2
n times
n
“2 .

From above, we obtain the following expression

ˆ ˙
n 2n
(0.4) 2 ď .
n
Now, from (0.3) and (0.4), we can see that

ˆ ˙
n 2n
(0.5) 2 ď ă 22n .
n
˙ ˆ
2n
We will now look at the prime factorization of . First, we will look at the prime
n
factorization of k! for any positive integer k. Letting p be a prime number less than or
equal to k, we can write k “ npr ` `, where p has order r, n P N, and ` ă pr . Clearly,
since ` ă pr , then k “ npr ` ` ă pr n ` pr “ pr pn ` 1q. From here, we obtain the following
inequality
k´` k
n“ r
ď r ă n ` 1,
p p
” ı
therefore, by definition, n is the integer part of pkr , thus n “ pkr
Now, expanding k! we obtain the following expression,
k! “ 1 ¨ 2 ¨ . . . ¨ p ¨ . . . ¨ 2p ¨ . . . ¨ pp ¨ . . . ¨ 2pp ¨ . . . ¨ ppp ¨ . . . ¨ 2ppp ¨ . . . ¨ k,
” ı
where p is any prime less than or equal to k. Therefore we obtain n1 “ kp multiples of
” ı
p which are less than k, n2 “ pk2 multiples of p2 which are less than k, and in general
Chapter 2. Tchebychev’s Theorem 7
” ı
k
nr “ pr
multiples of pr which are less than k. This allows us to count all the multiples
of p of order 1, as well as all the multiples of p2 with order 2, and so on, up to all multiples
of pr of order r. For this reason, the power of p in the prime factorization of k! is
„  „  „ 
k k k
α“ ` 2 ` ... ` q ,
p p p
where q is the largest integer such that pq ă k
ˆ ˙
2n p2nq!
If we now apply this knowledge to the expansion of “ pn!q2
, we find that it will
n
contain the following,

” ı
2n 2n
`...`r 2n
pr p s pq s
`
p2
´ n ”nı ¯2
r ps
` 2 `...`r pnq s
p p

” ı
2n 2n
`...`r 2n
pr p s pq s
`
p2
“ ´ ” ı ¯
2 n ` n `...`r pnq s
p r p s p2
´ ” ı ¯ ´ ´ ” ı ¯¯
2n
r p s
` 2n
2 `...`r 2n
q s ´ 2 rn s` n2 `...`r pnq s
“p p p p p

´” ı ” ı¯
2n
´2 n ` 2n n
`...`pr 2n ´2r pnq sq
“ ppr p s r p sq pq s
´2
p2 p2 ,

with p a prime less than or equal to 2n and q the largest integer such that pq ď 2n.
Next, using Lemma 2.3, by setting b “ a2 , we obtain

# “ ‰
a
”aı ras if 2
´ a2 ă 1
2
(0.6) 2 “ a
“ ‰
2 ras ´ 1 if 2
´ a2 ě 1
2
“a‰ “a‰
Now,“a‰we will examine the expression ras ´ 2 2
. From
“a‰ (0.8), if 2 2
“ ras,“ then
‰
ras ´ 2 2 “ ras ´ ras “ 0. On the other hand, if 2 2 “ ras ´ 1 then ras ´ 2 a2 “
ras ´ pras ´ 1q “ 1. From this, it is clear that

”aı
(0.7) ras ´ 2 ď 1 for all a.
2
Returning to the following expression,
´” ı ” ı¯
2n
´2 n ` 2n n
`...`pr 2n ´2r pnq sq
ppr p s r p sq pq s
´2
p2 p2 ,
2n a
if we set a “ p
from (0.7), then 2
“ np . This can therefore be used to show that
Chapter 2. Tchebychev’s Theorem 8

ˆ„  „ ˙ ˆ„  „ ˙ ˆ„  „ ˙
2n n 2n n 2n n
´2 ` 2
´2 2 ` ... ` q
´2 q ď 1 ` 1 ` ... ` 1 “ q
p p p p p p
ˆ ˙
2n
Now, if we denote by pi the prime numbers in the prime factorization of , then
n
we can see that each pi will have a power less than or equal to qi , where pqi i ď 2n. This
is shown below:
ˆ ˙
2n
ď pq11 ¨ pq22 ¨ . . . ¨ pqrr ď 2n ¨ 2n ¨ . . . ¨ 2n “ p2nqr .
n
loooooooomoooooooon
r times

Since there
ˆ are˙ r many primes less than or equal to 2n, we can set r “ πp2nq. This
2n
means that ď 2nπp2nq .
n
Now, we will take a look back at the following expression:
ˆ ˙
2n ¨ p2n ´ 1q ¨ . . . ¨ pn ` 1q
2n
“ .
nn ¨ pn ´ 1q ¨ . . . ¨ 1
ˆ ˙
2n
We will take all the primes that divide , and put them into two groups. First of all,
ˆ ˙ n
2n
we denote the primes that divide by p1 , p2 , . . . , ps , ps`1 , . . . , pr . We will make our
n
first group of primes all the primes less than or equal to n, and they will be the primes
p1 , p2 , . . . , ps . Our second group of primes will consist of all the primes greater than n
but less than or equal to 2n, and they will be the remaining primes;ˆ ps`1 ˙, ps`2 , . . . , pr .
2n
Clearly, the product of all of the primes in our second group divides “ p2nq!
pn!q2
since
n
they appear in the numerator only. Thus:
ˆ ˙
2n
ě ps`1 ¨ ps`2 ¨ . . . ¨ pr ą n ¨ n ¨ . . . ¨ n “ nr´s .
n
Now since p1 up to ps consist of s many primes, and p1 to pr consist of r many primes,
then it is clear that the amount of primes from ps`1 to pr is exactly equal to r ´ s. We
may also note that the primes p1 up to ps are all less than or equal to n as defined above,
which by definition means that there are exactly πpnq of them. By setting s “ πpnq, we
obtain the following expression:

ˆ ˙ ˆ ˙
r´s 2n πp2nq πp2nq´πpnq 2n
(0.8) n ă ď 2n which implies n ă ď 2nπp2nq .
n n
Chapter 2. Tchebychev’s Theorem 9
˙ˆ
n 2n
By comparing (0.5) with the above result, we see that 2 ď ď 2nπp2nq which
n
implies 2n ď 2nπp2nq . Taking the logarithm of both sides, we find that

logp2n q ď logp2nπp2nq q
n ¨ logp2q ď πp2nq ¨ logp2nq
n ¨ logp2q
ď πp2nq
logp2nq
2n logp2q
¨ ď πp2nq
logp2nq 2
2n
¨ p0.1505149...q ď πp2nq.
logp2nq

Therefore for any even number N “ 2n, we have shown the first half of the desired
inequality. For N odd, we can use a similar argument to satisfy the inequality. We first
claim that 2n ě 23 ¨ p2n ` 1q for n ě 1. Clearly, 6n ě 2p2n ` 1q “ 4n ` 2. Collecting like
terms we find that 2n ě 2, which is equivalent to n ě 1. Now, using this fact, we take
any odd N “ 2n ` 1. We know from above that n ¨ logp2q ď πp2nq ¨ logp2nq. We can use
this to show the following:

n ¨ logp2q ď πp2nq ¨ logp2nq ď πp2n ` 1q ¨ logp2n ` 1q

2 logp2q logp2q
¨ p2n ` 1q ¨ ď 2n ¨ ď πp2n ` 1q ¨ logp2n ` 1q
3 2 2
p2n ` 1q 2 logp2q
¨ ¨ ď πp2n ` 1q
logp2n ` 1q 3 2
p2n ` 1q
¨ p0.1003433...q ď πp2n ` 1q.
logp2n ` 1q

Since for any even N we have logNN ¨ p0.1505149...q ď πpN q, and for any odd N we have
N
log N
¨ p0.1003433...q ď πpN q, then we can conclude that for all N ą 1,

N
(0.9) ¨ p0.1003433...q ď πpN q.
log N
Chapter 2. Tchebychev’s Theorem 10
We now begin the proof of the second half of the desired inequality. From (0.5) and
(0.8), we find that nπp2nq´πpnq ă 22n . Taking the logarithm of both sides, we find that
logpnπp2nq´πpnq q ă logp22n q
pπp2nq ´ πpnqq ¨ log n ă 2n ¨ log 2
n
πp2nq ´ πpnq ă 2 ¨ log 2 ¨
log n
n
“ p0.60206q ¨ .
log n
“ ‰
We now suppose that x ą 1, for x P R. Setting n “ x2 and letting x “ a from
our previous“ x ‰ result (0.8), we find that
“ x ‰either rxs “ 2n or rxs “ 2n ` 1. For the first
case,“ if‰ 2 2 “ rxs, then since n ““ ‰2 , we have that rxs “ 2n. For the second case,
if 2 x2 “ rxs ´ 1, then with n “ x2 , we find that 2n “ rxs ´ 1 which implies that
rxs “ 2n ` 1.
We now observe two results. The first result is that πpxq “ πprxsq “ πp2n ` 1q ď
πp2nq ` 1, using Lemma“(2.2). ‰ This is true because
“ x ‰ rxs is either 2n ` 1 or 2n. The second
x x
result is that πp 2 q “ πp 2 q “ πpnq, as n “ 2 . From these two results, we obtain the
following expression:

x
πpxq ´ πp q ď πp2nq ` 1 ´ πpnq “ πp2nq ´ πpnq ` 1
2
n
ă 2 ¨ log 2 ¨ `1
log n
n n
ă 2 ¨ log 2 ¨ `
log n log n
n
“ p2 ¨ logp2q ` 1q ¨
log n
n
“ p1.60206q ¨ .
log n

We now claim that logn n ă logx x for n ě 3 and n ă x. We begin by looking at the nth
root of the first few integers:
?
2
?
3
?
4
?5
2 “ 1.41 3 “ 1.44 4 “ 1.41 5 “ 1.37 . . .
? ?
We can see that for n ě 3, n`1 n ` 1 ă n n. Taking the logarithm of both sides, we find
1
that n`1 ¨ logpn ` 1q ă n1 ¨ logpnq. This implies that for x ą n, logx x ă logn n . Rearranging,
“ ‰ “ ‰
we obtain logn n ă logx x . Now, since n “ x2 , then for x2 ě 3,
x x
πpxq ´ πp q ă p2 ¨ logp2q ` 1q ¨ .
2 log x
“ ‰ “ ‰
We can also show that the above inequality holds for x2 ă 3. First, if x2 ă 3, this
is equivalent to saying x ă 6. But we know for any x ă 10, then log x ă 1, so clearly
Chapter 2. Tchebychev’s Theorem 11
1
ą 1. Using what we have just shown, we now look at the same expression as before
log x
πpxq ´ πp x2 q. Clearly, for all x ą 1,

´x¯ x x
πpxq ´ π ă πpxq ă x ă ă p2 ¨ logp2q ` 1q ¨ .
2 log x log x

We now look at the following inequality:

´x¯ ´x¯
πpxq logpxq ´ π log
2¯
´x 2 ´x¯ ´x¯ ´x¯
“ πpxq logpxq ´ π logpxq ` π logpxq ´ π log
” ´ x ¯ı 2 ´ x ¯2 ” ´2x ¯ı 2
“ πpxq ´ π ¨ logpxq ` π ¨ logpxq ´ log
2 2 2
x ´x¯
ă p2 ¨ logp2q ` 1q ¨ ¨ logpxq ` π ¨ rlogpxq ´ plogpxq ´ logp2qqs
logpxq 2
´x¯
ă p2 ¨ logp2q ` 1q ¨ x ` ¨ rlogp2qs
2
logp2q
“ p2 ¨ logp2q ` 1 ` q¨x
2
“ p1.75257 . . .q ¨ x.

Now, taking N to be any arbitrary positive integer, from above we obtain the following
set of inequalities:

˙ ˆ
ˆ ˙
N N
πpN q logpN q ´ π log ă p1.75257 . . .q ¨ N,
2 2
ˆ ˙ ˆ ˙ ˆ ˙ ˆ ˙
N N N N N
π log ´π log ă p1.75257 . . .q ¨ ,
2 2 4 4 2
ˆ ˙ ˆ ˙ ˆ ˙ ˆ ˙
N N N N N
π log ´π log ă p1.75257 . . .q ¨ ,
4 4 8 8 4
..
.
ˆ ˙ ˆ ˙ ˆ ˙ ˆ ˙
N N N N N
π k´1
log k´1
´π k
log k
ă p1.75257 . . .q ¨ k´1 .
2 2 2 2 2
Chapter 2. Tchebychev’s Theorem 12
k
We now choose k so that 2 ą N . Adding all the inequalities from the previous step, we
obtain the following:
ˆ ˙ ˆ ˙ ˆ ˙ ˆ ˙
N N N N
πpN q logpN q ´ π log `π log
2 2 2 2
ˆ ˙ ˆ ˙ ˆ ˙ ˆ ˙
N N N N
´π log ` ... ` π k´2
log
4 4 2 2k´2
ˆ ˙ ˆ ˙ ˆ ˙ ˆ ˙ ˆ ˙ ˆ ˙
N N N N N N
´π k´1
log k´1
`π k´1
log k´1
´π k
log
2 2 2 2 2 2k
N N
ă p1.75257 . . .q ¨ N ` p1.75257 . . .q ¨ ` . . . ` p1.75257 . . .q ¨ k´1
ˆ 2˙ 2
N N
“ p1.75257 . . .q N ` ` . . . ` k´1 .
2 2

The above expression can simplify to

ˆ ˙ ˆ ˙ ˆ ˙
N N N N
πpN q logpN q ´ π log ă p1.75257 . . .q N ` ` . . . ` k´1 .
2k 2k 2 2
Now, using Lemma 2.4, we find that
˜ ¸
N ´ 2Nk
ˆ ˙ ˆ ˙
N N
πpN q logpN q ´ π log ă p1.75257 . . .q
2k 2k 1 ´ 12
ˆ ˙
N
“ p3.50514q N ´ k
2
ă p3.50514q ¨ N ă 4N,
N
using the fact that 0 ă 2k
ă 1. Setting πp 2Nk q “ 0 due to the fact that N
2k
ă 1, we can
therefore conclude that
ˆ ˙ ˙ˆ
N N
πpN q logpN q ´ π log ă 4N
2k 2k
ˆ ˙
N
πpN q logpN q ´ 0 ¨ log ă 4N
2k
πpN q logpN q ă 4N
N
πpN q ă 4 ¨ .
log N
Comparing the above inequality with (0.9), we find that
N N
p0.1003433...q ¨ ď πpN q ă 4 ¨ ,
log N log N
thus completing the proof.

Chapter 2. Tchebychev’s Theorem 13
We can also complete this proof using the natural logarithm, as opposed to the loga-
rithm of base 10. The proof remains the exact same, except for when computing values
in certain parts of the proof. Instead of computing values using log base 10, we can use
log base e, which is equivalent to ln. Computing these values, we find that (0.9) can be
changed to
N
¨ p0.2310406q ď πpN q,
ln N
and the other side of the inequality can be changed to
N
πpN q ă 6 ¨
ln N
This implies that in terms of the natural logarithm, πpN q can be approximated in the
following way:
N N
¨ p0.23q ď πpN q ă 6 ¨ .
ln N ln N
 CHAPTER 3

Mertens’ First Theorem

In this section, we use what we have proved in the previous chapter to examine an
application of the distribution of prime numbers. We begin by stating some useful tools
which will be helpful in the proof Mertens’ First Theorem.
Lemma 3.1. Let N be an integer and pi a prime number. Then
N N
N N N N pi
´ q `1
pi i
` 2 ` 3 ` . . . ` qi “ .
pi pi pi pi 1 ´ p1i
N N N N 1
Proof. We begin by setting sk “ pi
` p2i
` p3i
` ... ` q .
pi i
Multiplying sk by pi
we
sk
obtain pi
“ N
p2i
` N
p3i
` ... `Subtracting sk from spki we obtain
N
q `1 .
pi i
ˆ ˙
sk N N N N N N N
sk ´ “ ` 2 ` 3 ` . . . ` qi ´ ` 3 ` . . . ` qi `1
pi pi pi pi pi p2i pi pi
N N N N N N N N
“ ` 2 ` 3 ` . . . ` qi ´ 2 ´ 3 ´ . . . ´ qi ´ qi `1
pi pi pi pi pi pi pi pi
N N
“ ´ qi `1 .
pi pi
Factoring out sk from the left side, we obtain sk p1 ´ p1i q “ N
pi
N
q `1 . Dividing both sides
´
pi i
by p1´ p1i q, and substituting N
pi
` pN2 ` pN3 `. . .` pNqi for sk , we obtain N pi
` pN2 ` pN3 `. . .` pNqi “
i i i i i i
N
pi
´ qN`1
p i
i
1´ p1
, the result we are looking for. 
i

Lemma 3.2. Take N P N, and let s ą 1 for s P R. Then

1 1 1 1
1 ` s ` s ` s ` ... ` s
2 3 4 N
1 s
tends to a limit lying between s´1 and s´1 .
Lemma 3.3. Let n be any positive integer. Then n! satisfies the following inequality:
c
4 ? ´ n ¯n ? ´ n ¯n
¨e¨ n¨ ă n! ă e ¨ n ¨ .
5 e e
Lemma 3.4. Let a P R. Then ?
2 log a ă a,
for a ą 0.
14
Chapter 3. Mertens’ First Theorem 15
?
Proof. To show this, we rearrange our expression to look like a ´ 2 log a ą 0.
Treating this expression like a function, it is sufficient to show that this function is always
?
positive for a ą 0. Taking the derivative of the function f paq “ a ´ 2 log a, we obtain
? ?
1 1 2 log e a ´ 4 ap0.43429q a ´ 1.73716
f paq “ ? ´ “ ? “ .
2 a a 2a a 2a
Setting our derivative equal to 0, we find that
?
a ´ 1.73716 ?
0“ ñ a “ 1.73716 ñ a “ 3.01772,
2a
therefore we must have a maximum or minimum at a=3.01772, and an asymptote at
a “ 0. Substituting values that are in neighborhood of a=3.01772 into our derivative, we
find that f 1 p3q is negative, and f 1 p3.1q is positive, meaning that our function decreases to
a “ 3.01772 and then increases indefinitely. This is illustrated by substituting values into
our original function:

f p3q “ 0.777808 f p3.01772q “ 0.777801 f p3.1q “ 0.77795

Because f p3.01772q is our local minimum, and it is positive, we can therefore conclude
? ?
that for a ą 0, f paq “ a ´ 2 log a is positive. This implies that a ´ 2 log a ą 0 and
?
therefore 2 log a ă a. 

Theorem 3.5. (Mertens’ First Theorem) Let 2, 3, 5, 7, 11, 13, . . . , p be the primes not
exceeding a given integer N . Then there exists a constant R, that can be taken equal to
4, such that
ˇ ˇ
ˇ log 2 log 3 log 5 log p ˇ
ˇ
ˇ 2 ` ` ` . . . ` ´ log N ˇ ă R.
3 5 p ˇ

Proof. We begin by decomposing a given integer N ! into its prime factors. As we

have seen in the proof of Tchebychev’s Theorem, N ! has the following form:

(0.10) N ! “ pα1 1 ¨ pα2 2 ¨ . . . ¨ pαr r ,

with p1 , p2 , . . . , pr the primes less than or equal to N , and

„  „  „  „ 
N N N N
αi “ ` 2 ` 3 ` . . . ` qi ,
pi pi pi pi

where pqi i ď N . Taking logarithms of both sides of (0.10), we obtain

log N ! “ logppα1 1 ¨ pα2 2 ¨ . . . ¨ pαr r q

“ log pα1 1 ` log pα2 2 ` . . . ` log pαr r
“ α1 log p1 ` α2 log p2 ` . . . ` αr log pr .
Chapter 3. Mertens’ First Theorem 16
Now, we can estimate log N ! in two ways. By the definition of αi , we can see that the
right hand side of the above equation has the form
ˆ„  „  „  „ ˙
N N N N
(0.11) ` 2 ` 3 ` . . . ` q1 log p1
p1 p1 p1 p1
ˆ„  „  „  „ ˙
N N N N
(0.12) ` ` 2 ` 3 ` . . . ` q2 log p2
p2 p2 p2 p2
ˆ„  „  „  „ ˙
N N N N
(0.13) ` ... ` ` 2 ` 3 ` . . . ` qr log pr .
pr pr pr pr

We now wish to simplify the above expression by changing

„  „  „  „ 
N N N N N N N N
` 2 ` 3 ` . . . ` qi to ` 2 ` 3 ` . . . ` qi ,
pi pi pi pi pi pi pi pi

where pi remains less than N for 1 ď i ď r, and with pqi i ď N . By removing the integer
part notation, we introduce an error of no more than 1 per term in the estimation of
log N !. Indeed, by definition
„  „ 
N N N
qi ď qi ď qi ` 1,
pi pi pi

thus, for each sum of (0.11), (0.12), and (0.13), we have an error of q1 , q2 , . . . , qr respec-
tively, as there are exactly qi many terms in the sum of
„  „  „  „ 
N N N N
` 2 ` 3 ` . . . ` qi .
pi pi pi pi

We have now found a total error on our estimation of log N !, namely

q1 ¨ log p1 ` q2 ¨ log p2 ` q3 ¨ log p3 ` . . . ` qr ¨ log pr

“ log pq11 ` log pq22 ` log pq33 ` . . . ` log pqrr
ď log N ` log N ` log N ` . . . ` log N “ r log N.
loooooooooooooooooooooomoooooooooooooooooooooon
r times

In the above expression, we can see that r is equal to the number of primes less than or
equal to N , thus by Tchebychev’s Theorem,
N
r “ πpN q ă B ,
log N
where B can be taken equal to 4. We can now see that our estimate of log N ! is bounded
by
N
r log N ă B ¨ log N “ BN.
log N
Chapter 3. Mertens’ First Theorem 17
This allows us to construct an upper and lower bound for log N !:
ˆ ˙ ˆ ˙
N N N N N N N N
` ` ` . . . ` q1 log p1 ` ` ` ` . . . ` q2 log p2 ` . . .
p1 p21 p31 p1 p2 p22 p32 p2
ˆ ˙ ˆ ˙
N N N N N N N N
... ` ` ` ` . . . ` qr log pr ě log N ! ą ` ` ` . . . ` q1 log p1
pr p2r p3r pr p1 p21 p31 p
ˆ ˙ ˆ ˙ 1
N N N N N N N N
` ... ` ` 2 ` 3 ` . . . ` q2 log p2 ` ` 2 ` 3 ` . . . ` qr log pr ´ BN.
p2 p2 p2 p2 pr p r pr pr

We can now modify Lemma 3.1 in the following way:

N N N
N N N N pi
´ q `1
N
pi i pi
` 2 ` 3 ` . . . ` qi “ ă “ ,
pi pi pi pi 1 ´ p1i 1´ 1
pi
pi ´ 1

and we also note that

N N N N N
ď ` 2 ` 3 ` . . . ` qi .
pi pi pi pi pi
Using these two facts, we find that
N N N
p˚q log p1 ` logp2 ` . . . ` log pr ě log N !
p1 ´ 1 p2 ´ 1 pr ´ 1
N N N
ą log p1 ` log p2 ` . . . ` log pr ´ BN.
p1 p2 pr
We wish to simplify the above inequality even further, so as it has the following form:
ˆ ˙
log p1 log p2 log pr
p˚˚q N ` ` ... ` ` K ą log N !
p1 p2 pr
ˆ ˙
log p1 log p2 log pr
ąN ` ` ... ` ´B ,
p1 p2 pr
where K is a positive constant. To do this, we must find the difference between the left
and right hand sides of p˚q and p˚˚q. We can begin with the obvious case, which is the
right side. Clearly, N has simply been factored from all the terms on the right hand side
of p˚q, thus the right hand sides of p˚q and p˚˚q are equivalent. For the former case, we
first find how much the left hand side of p˚q differs from
ˆ ˙
log p1 log p2 log pr
(0.14) N ` ` ... ` .
p1 p2 pr
This difference will ultimately be our K value. After factoring out N from p˚q, we find
that the difference for each term is equal to
log pi log pi pi log pi ´ ppi ´ 1q log pi log pi
´ “ “ ,
pi ´ 1 pi pi ppi ´ 1q pi ppi ´ 1q
Chapter 3. Mertens’ First Theorem 18
for i “ 1, 2, 3, . . . , r, thus the total difference between the left hand side of p˚q and (0.14)
is
ˆ ˙
log p1 log p2 log pr
(0.15) N ` ` ... ` .
p1 pp1 ´ 1q p2 pp2 ´ 1q pr ppr ´ 1q

We now claim that for any a ě 2

log a 1
ă ? .
apa ´ 1q a a

It is clear that the above inequality implies log a ă a´1

? , which is equivalent to 2 log a ă
a
´ ¯
a´1
2 ?a . Working with the right hand side of the inequality, we find that for a ě 2,
ˆ ˙ ˆ? ˙ ˆ ? ˙ ˆ ? ˙
a´1 apa ´ 1q ? a ? a ?
2 ? “2 “2 a´ ě2 a´ “ a.
a a a 2

Now, using Lemma 3.4, since

ˆ ˙
? a´1 ?
2 log a ă a and 2 ? ě a,
a
´ ¯
a´1
we can therefore conclude that 2 log a ă 2 ?
a
.
We can now find an upper bound for the sum inside the brackets of (0.15):

log p1 log p2 log pr

` ` ... `
p1 pp1 ´ 1q p2 pp2 ´ 1q pr ppr ´ 1q
1 1 1
ă ? ` ? ` ... ` ?
p1 p1 p2 p2 p r pr
1 1 1 1
ă 1 ` ? ` ? ` ? ` ... ` ? .
2 2 3 3 4 4 N N

Using Lemma 3.2, taking s “ 23 , we find that the last sum of our previous inequality is
bounded:
3
1 1 1 1 2
1 ` ? ` ? ` ? ` ... ` ? ă 3 “ 3.
2 2 3 3 4 4 N N 2
´1

We can therefore conclude that the difference between the left hand side of p˚q and (0.14),
namely K, can be taken equal to 3.
We now find another estimate for log N ! using Lemma 3.3. Recall that
ˆ ˙N ˆ ˙N
? N ? N
Φ1 ¨ N ¨ ă N ! ă Φ2 ¨ N ¨ .
e e
Chapter 3. Mertens’ First Theorem 19
b
4
where Φ1 “ e and Φ2 “ e. Taking logarithms of each part of the inequality, we obtain
5
˜ ˆ ˙N ¸ ˜ ˆ ˙N ¸
? N ? N
log Φ1 ¨ N ¨ ă log N ! ă log Φ2 ¨ N ¨
e e
ˆ ˙N ˆ ˙N
? N ? N
log Φ1 ` log N ` log ă log N ! ă log Φ2 ` log N ` log
e e
1 1
p§q log Φ1 ` log N ` N plog N ´ log eq ă log N ! ă log Φ2 ` log N ` N plog N ´ log eq.
2 2
n x
We proved previously that log n
ă log x
for n ě 3 and n ă x. Rearranging this
inequality, we find that
log x log n
ă .
x n
Thus, it is clear that for N ě 3,
1 log N 1 log 3
ă ă 0.08
2 N 2 3
and
log Φ2 log e
ă ă 0.15.
N 3
Using these two inequalities, we can modify p§q in the following way
1 1
log Φ1 ` log N ` N plog N ´ log eq ă log N ! ă log Φ2 ` log N ` N plog N ´ log eq
2 2
log Φ2 log N
N plog N ´ log eq ă log N ! ă N `N ` N plog N ´ log eq
ˆN 2N ˙
log Φ2 1 log N
N plog N ´ log eq ă log N ! ă N ` ` log N ´ log e
N 2 N
N plog N ´ log eq ă log N ! ă N p0.15 ` 0.08 ` log N ´ log eq
N plog N ´ log eq ă log N ! ă N plog N ´ log e ` 0.23q .

Now, we recall from (˚˚) that

ˆ ˙
log p1 log p2 log pr
N ` ` ... ` ` K ą log N !
p1 p2 pr
ˆ ˙
log p1 log p2 log pr
ąN ` ` ... ` ´B .
p1 p2 pr
Comparing this inequality with the previous inequality, we come to two conclusions. First,
log p1 log p2 log pr
` ` ... ` ` K ą log N ´ log e,
p1 p2 pr
and second,
log p1 log p2 log pr
` ` ... ` ´ B ă log N ´ log e ` 0.23.
p1 p2 pr
Chapter 3. Mertens’ First Theorem 20
Rearranging the two inequalities, we obtain
log p1 log p2 log pr
` ` ... ` ´ log N ą ´K ´ log e
p1 p2 pr
and
log p1 log p2 log pr
` ` ... ` ´ log N ă B ´ log e ` 0.23.
p1 p2 pr
Using the fact that K and B can be taken equal to 3 and 4 respectively, we can
calculate the right hand sides of the two expressions above. Thus, ´K ´ log e “ ´3 ´
0.43429 “ ´3.43429 and B ´ log e ` 0.23 “ 4 ´ 0.43429 ` 0.23 “ 3.79571. Taking the
larger of the absolute value of these two numbers, namely | ´ 3.43429| ă |3.79571| ă 4,
we obtain
log p1 log p2 log pr
´4 ă ` ` ... ` ´ log N ă 4.
p1 p2 pr
This was the result we were looking for, which therefore completes this proof.

 CHAPTER 4

Mertens’ Second Theorem

We begin this chapter by introducing an extremely useful formula which, with the
help of Mertens’ First Theorem, will enable us to prove Mertens’ Second Theorem.

Theorem 4.1. (Abel’s Formula) Consider the sum

S “ a1 b1 ` a2 b2 ` a3 b3 ` . . . ` an bn ,

where a1 , a2 , a3 , . . . , an and b1 , b2 , b3 , . . . , bn are any two sequences of numbers, and denote

the sums b1 , b1 ` b2 , b1 ` b2 ` b3 , . . . , b1 ` b2 ` b3 ` . . . ` bn by B1 , B2 , B3 , . . . , Bn
respectively. Then

S “ pa1 ´ a2 qB1 ` pa2 ´ a3 qB2 ` pa3 ´ a4 qB3 ` . . . ` pan´1 ´ an qBn´1 ` an Bn .

Theorem 4.2. (Mertens’ Second Theorem) Given any N ą 1, for all p ă N with p
prime, the expression

1 1 1 1 1
(0.16) ` ` ` ` ¨ ¨ ¨ ` ´ ln ln N
2 3 5 7 p

is less than some relatively small constant T, where T is independent of T.

Proof. To begin this proof, we first examine the series

1 1 1 1 1
S“ ` ` ` ` ¨¨¨ ` ,
p1 p2 p3 p4 pr
where p1 , p2 , . . . , pr are all the primes less than or equal to any integer N . We now set

1 1 1 1
a1 “ , a2 “ , a3 “ , . . . , ar “
log p1 log p2 log p3 log pr

log p1 log p2 log p3 log pr

b1 “ , b2 “ , b3 “ , . . . , br “ .
p1 p2 p3 pr

21
Chapter 4. Mertens’ Second Theorem 22
Using Theorem 4.1, denoting the sums b1 , b1 ` b2 , b1 ` b2 ` b3 , . . . , b1 ` b2 ` b3 ` . . . ` br
by B1 , B2 , B3 , . . . , Br respectively, we obtain
log p1
B1 “ ,
p1
log p1 log p2
B2 “ ` ,
p1 p2
..
.
log p1 log p2 log pr
Br “ ` ` ... ` .
p1 p2 pr

We can now write our series S in terms of a1 , a2 , . . . , ar , and b1 , b2 , . . . , br . Clearly

a1 b 1 ` a2 b 2 ` a3 b 3 ` . . . ` ar b r
1 log p1 1 log p2 1 log pr
“ ¨ ` ¨ ` ... ` ¨
log p1 p1 log p2 p2 log pr pr
1 1 1 1 1
“ ` ` ` ` ¨¨¨ ` ,
p1 p 2 p3 p4 pr

but by Theorem 4.1,

a1 b 1 ` a2 b 2 ` a3 b 3 ` . . . ` ar b r
“ pa1 ´ a2 qB1 ` pa2 ´ a3 qB2 ` pa3 ´ a4 qB3 ` . . . ` par´1 ´ ar qBr´1 ` ar Br ,

thus
1 1 1 1 1
` ` ` ` ¨¨¨ `
p p 2 p3 p4 p
ˆ1 ˙ ˆ r ˙
1 1 1 1
“ ´ B1 ` ´ B2
log p1 log p2 log p2 log p3
ˆ ˙ ˆ ˙
1 1 1 1 1
` ´ B3 ` . . . ` ´ Br´1 ` ¨ Br
log p3 log p4 log pr´1 log pr log pr
ˆ ˙ ˆ ˙
1 1 1 1
p˚q “ ´ B1 ` ´ B2
log p1 log p2 log p2 log p3
ˆ ˙ ˆ ˙
1 1 1 1
` ´ B3 ` . . . ` ´ Br´1
log p3 log p4 log pr´1 log pr
ˆ ˙
1 1 1
` ´ Br ` ¨ Br .
log pr log N log N
Chapter 4. Mertens’ Second Theorem 23
Now by Theorem 3.5, we can deduce that:

ˇ ˇ
ˇ log p1 ˇ
ˇ
ˇ p1 ´ log p 1
ˇ ă K ñ |B1 ´ log p1 | ă K ñ B1 ă log p1 ` K,
ˇ
ˇ ˇ
ˇ log p1 log p2 ˇ
ˇ
ˇ p1 ` ´ log p 2
ˇ ă K ñ |B2 ´ log p2 | ă K ñ B2 ă log p2 ` K,
p2 ˇ
ˇ ˇ
ˇ log p1 log p2 log p3 ˇ
ˇ p1 ` p2 ` p3 ´ log p3 ˇ ă K ñ B3 ă log p3 ` K,
ˇ ˇ

..
.
ˇ ˇ
ˇ log p1 log p2 log p r ˇ
ˇ
ˇ p1 ` ` . . . ` ´ log p r
ˇ ă K ñ Br ă log pr ` K,
p2 pr ˇ
ˇ ˇ
ˇ log p1 log p2 log pr ˇ
ˇ p1 ` p2 ` . . . ` pr ´ log N ˇ ă K ñ Br ă log N ` K,
ˇ ˇ

where K can be taken equal to 4. Thus we can conclude that

B1 ă log p1 ` K, B2 ă log p2 ` K, B3 ă log p3 ` K, . . .

. . . , Br ă log pr ` K, and Br ă log N ` K.

From above, it follows that

1 1 1 1 1
` ` ` ` ¨¨¨ `
p p2 p 3 p4 pr
ˆ1 ˙ ˆ ˙
1 1 1 1
ă ´ plog p1 ` Kq ` ´ plog p2 ` Kq
log p1 log p2 log p2 log p3
ˆ ˙ ˆ ˙
1 1 1 1
` ´ plog p3 ` Kq ` . . . ` ´ log pr´1 ` K
log p3 log p4 log pr´1 log pr
ˆ ˙
1 1 1
` ´ plog pr ` Kq ` ¨ plog N ` Kq
log pr log N log N
ˆ ˙ ˆ ˙
1 1 K K 1 1 K K
“ ´ log p1 ` ´ ` ´ log p2 ` ´
log p1 log p2 log p1 log p2 log p2 log p3 log p2 log p3
ˆ ˙ ˆ ˙
1 1 K K 1 1
` ´ log p3 ` ´ ` ... ` ´ log pr´1
log p3 log p4 log p3 log p4 log pr´1 log pr
ˆ ˙
K K 1 1 K K 1 K
` ´ ` ´ log pr ` ´ ` log N ` .
log pr´1 log pr log pr log N log pr log N log N log N
Chapter 4. Mertens’ Second Theorem 24
1 1 1 1 1
The bound on our series S “ p1
` p2
` p3
` p4
` ¨¨¨ ` pr
can simplify to

ˆ ˙ ˆ ˙
´ log p1 K 1 1 1 1
1` ` ` ´ log p2 ` ´ log p3
log p2 log p1 log p2 log p3 log p3 log p4
ˆ ˙ ˆ ˙
1 1 1 1 1
` ... ` ´ log pr´1 ` ´ log pr ` log N,
log pr´1 log pr log pr log N log N

which after further rearrangement yields

„ ˆ ˙ ˆ ˙
´ log p1 1 1 1 1
p˚˚q 1 ` ` ´ log p2 ` ´ log p3 ` . . .
log p2 log p2 log p3 log p3 log p4
ˆ ˙ ˆ ˙ 
1 1 1 1 1 K
` ´ log pr´1 ` ´ log pr ` log N ` .
log pr´1 log pr log pr log N log N log p1

Now we will focus on the expression found inside the square brackets. We can expand the
expression to look like

´ log p1 log p2 log p2 log p3 log p3

` ´ ` ´ ` ...
log p2 log p2 log p3 log p3 log p4
log pr´1 log pr´1 log pr log pr log N
... ` ´ ` ´ ` ,
log pr´1 log pr log pr log N log N

then factor the common denominators, which will yield the following

1 1 1
plog p2 ´ log p1 q ` plog p3 ´ log p2 q ` plog p4 ´ log p3 q
log p2 log p3 log p4
1 1
` . . . ` plog pr ´ log pr´1 q ` plog N ´ log pr q .
log pr log N

We can now estimate the above expression geometrically by constructing a graph to

model the series. We can depict the expression by the sum of N ´ 1 rectangles. The area
of the first rectangle will have a base of length plog p2 ´log p1 q and a height of length log1p2 ,
as per the first term in our expression. The second rectangle will have a base of length
plog p3 ´ log p2 q, and a height of length log1p3 . This pattern continues up to our pN ´ 1qth
rectangle, which will have a base of length plog N ´ log pr q and a height of length log1N .
These rectangles are plotted in Figure (4.1). We now observe that when x “ log p2 , that
y “ log1p2 . Similarly, when x “ log p3 , y “ log1p3 , and so on, up to x “ log N , and y “ log1N .
In general we find that y “ x1 , which is a hyperbola that intersects the top right corner of
each rectangle. This hyperbola is also shown in Figure 4.1.
Chapter 4. Mertens’ Second Theorem 25

Figure 4.1

Now that we have defined the hyperbola such that the shaded rectangles are all below
its graph, we have thus found an upper bound on all the rectangles, and therefore an
upper bound on the series S. We simply need to calculate the area below the hyperbola
y “ x1 between the points x “ log p1 and x “ log N . This is done easily using calculus.
Clearly
ż log N
1 ˇlog N
(0.17) dx “ pln xqˇ “ lnplog N q ´ lnplog p1 q.
ˇ
log p1 x log p1

We can therefore conclude that the expression in the square brackets from p˚˚q, is
bounded by lnplog N q ´ lnplog p1 q, and thus
1 1 1 1 1 K
(0.18) ` ` ` ` ¨¨¨ ` ă 1 ` lnplog N q ´ lnplog p1 q ` .
p1 p2 p3 p4 pr log p1
1 1 1
We now focus our attention on finding a lower bound for our series S “ p1
` p2
` p3
`
1
p4
` ¨ ¨ ¨ ` p1r . Recall that
log p1 log p2 log pi
Bi “ ` ` ... ` ,
p1 p2 pi
where i “ p1, 2, . . . , rq, and pr ď N . Adding logpi`1
pi`1
and subtracting a, where a ą log pi`1
pi`1
,
we find that
log p1 log p2 log pi log pi`1
(0.19) Bi ą ` ` ... ` ` ´ a.
p1 p2 pi pi`1
Here, we can let a “ log3 3 , as for n ě 3, log 3
3
ă log n
n
. Now, as a result of Mertens’ First
Theorem, we find that
log p1 log p2 log p3 log pi log pi`1
´K ă ` ` ` ... ` ` ´ log pi`1 ,
p1 p2 p3 pi pi`1
Chapter 4. Mertens’ Second Theorem 26
which implies

log p1 log p2 log p3 log pi log pi`1

log pi`1 ´ K ă ` ` ` ... ` ` .
p1 p2 p3 pi pi`1

Returning to equation (0.19), we can conclude that

ˆ ˙
log p1 log p2 log p3 log pi log pi`1
Bi ą ` ` ` ... ` ` ´a
p1 p2 p3 pi pi`1
ą plog pi`1 ´ Kq ´ a,

for 1 ď i ă r. Using the same argument, we find that

log p1 log p2 log p3 log pr

` ` ` ... ` ą log N ´ K ´ a,
p1 p2 p3 pr

with the ´a added in for convenience. From above, we can conclude that

B1 ą log p2 ´ K ´ a, B2 ą log p3 ´ K ´ a, . . .
. . . , Br´1 ą log pr ´ K ´ a, Br ą log N ´ K ´ a.

Now, referring back to p˚q, we find that

1 1 1 1 1
` ` ` ` ¨¨¨ `
p1 p 2 p3 p4 pn
ˆ ˙ ˆ ˙
1 1 1 1
ą ´ plog p2 ´ K ´ aq ` ´ plog p3 ´ K ´ aq
log p1 log p2 log p2 log p3
ˆ ˙ ˆ ˙
1 1 1 1
` ´ plog p4 ´ K ´ aq ` . . . ` ´ plog pr ´ K ´ aq
log p3 log p4 log pr´1 log pr
ˆ ˙
1 1 1
` ´ plog N ´ K ´ aq ` ¨ plog N ´ K ´ aq
log pr log N log N
ˆ ˙ ˆ ˙
1 1 K `a K `a 1 1
“ ´ log p2 ´ ` ` ´ log p3
log p1 log p2 log p1 log p2 log p2 log p3
ˆ ˙
K `a K `a 1 1 K `a K `a
´ ` ` ´ log p4 ´ ` ` ...
log p2 log p3 log p3 log p4 log p3 log p4
ˆ ˙ ˆ ˙
1 1 K `a K `a 1 1
... ` ´ log pr ´ ` ` ´ log N
log pr´1 log pr log pr´1 log pr log pr log N
K `a K `a 1 K `a
´ ` ` log N ´ .
log pr log N log N log N
Chapter 4. Mertens’ Second Theorem 27
With further rearrangement, we obtain
ˆ ˙ ˆ ˙
1 1 K `a 1 1
“ ´ log p2 ´ ` ´ log p3
log p1 log p2 log p1 log p2 log p3
ˆ ˙ ˆ ˙
1 1 1 1
` ´ log p4 ` . . . ` ´ log pr
log p3 log p4 log pr´1 log pr
ˆ ˙
1 1 1
` ´ log N ` log N
log pr log N log N
log p2 log p2 log p3 log p3 log p4 log p4 log pr log pr
“ ´ ` ´ ` ´ ` ... ` ´
log p1 log p2 log p2 log p3 log p3 log p4 log pr´1 log pr
log N log N 1 K `a
` ´ ` log N ´
log pr log N log N log p1
„
1 1 1
“ plog p2 ´ log p1 q ` plog p3 ´ log p2 q ` plog p4 ´ log p3 q ` ...
log p1 log p2 log p3

1 K `a
. . . ` plog N ´ log pr q `1´ .
log pr log p1
log N log p1
In this last step, we omitted the ´ log N
term, and replaced it with ´ log p1
. Now, as we
have previously done with the upper bound on S, we can see that the sum inside the
square brackets above is simply the sum of differently sized rectangles. This time, the
first rectangle will have a base of length plog p2 ´ log p1 q and a height of length log1p1 . The
second rectangle will have a base of length plog p3 ´ log p2 q and a height of length log1p2 ,
and so on, up to a rectangle with a base of length plog N ´ log pr q and height of length
1
log pr
. These rectangles are shown in Figure 4.2.

Figure 4.2

We can see that when x “ log p1 , y “ log1p1 . This pattern holds for all pr , with r ď N .
We also notice that, as before, these coordinates define the function y “ x1 , only this
Chapter 4. Mertens’ Second Theorem 28
time, the curve intersects the top left hand corner of the rectangles, as shown in Figure
4.2. This implies that the area bounded by the curve between log p1 and log N is less
than the area of the shaded rectangles. We simply need to calculate this area, which will
in turn give us a lower bound for our series, but we have already calculated this to be
lnplog N q ´ lnplog p1 q from (0.17).
Now that we have a lower bound for the sum of the rectangles, we find that
1 1 1 1 1 K `a
(0.20) ` ` ` ` ¨¨¨ ` ą lnplog N q ´ lnplog p1 q ` 1 ´ .
p1 p2 p 3 p4 pr log p1
1
Next, combining equations (0.18) and (0.20), we find that for S “ p1
` p12 ` p13 ` p14 `¨ ¨ ¨` p1r
K `a K
(0.21) lnplog N q ´ lnplog p1 q ` 1 ´ ă S ă 1 ` lnplog N q ´ lnplog p1 q ` .
log p1 log p1
Now, as opposed to using logarithms of two different bases in our expression, we will use
the fact that
log N “ M ln N,
where M “ log e. This changes (0.21) to
K `a K
lnpM ln N q ´ lnplog p1 q ` 1 ´ ă S ă 1 ` lnpM ln N q ´ lnplog p1 q `
log p1 log p1
K `a K
ln ln N ` ln M ´ lnplog p1 q ` 1 ´ ă S ă 1 ` ln ln N ` ln M ´ lnplog p1 q ` .
log p1 log p1
From here we can simply evaluate our expression using the fact that M “ log e “
0.43429..., p1 “ 2, K “ 4, and a “ log3 3 “ 0.15904.... Clearly,
K `a
ln ln N ` ln M ´ lnplog p1 q ` 1 ´
log p1
4 ` 0.15904
“ ln ln N ` lnp0.43429q ´ lnplog 2q ` 1 ´
log 2
“ ln ln N ` p´0.83404q ´ p´1.20054q ` 1 ´ p13.81603q
“ ln ln N ´ 12.44953 ą ln ln N ´ 13.
Similarly,
K
1 ` ln ln N ` ln M ´ lnplog p1 q `
log p1
4
“ ln ln N ` 1 ` p´0.83404q ´ p´1.20054q `
0.30102
“ ln ln N ` 14.65465 ă ln ln N ` 15.
Comparing the two inequalities above, we finally obtain
1 1 1 1 1
ln ln N ´ T ă ` ` ` ` ¨¨¨ ` ă ln ln N ` T,
p1 p2 p3 p4 pr
where T “ 15, the larger of the upper and lower bounds. This completes the proof.

 CHAPTER 5

Concluding Remarks

This project explored how prime numbers are distributed throughout the set of real
numbers. We proved that there exists an upper and lower bound on πpN q, which is
dependent on N . It is fascinating to see how such a complicated yet relatively elementary
proof can yield such an elegant result. We used this result to explore further problems in
the study of prime numbers, specifically, Mertens’ First and Second Theorem.
In Mertens’ First Theorem, we used similar arguments as in the proof of Tchebychev’s
Theorem, and simply found the magnitude of errors between different series. The proof
of Mertens’ Second Theorem was more geometrically intuitive. We used Mertens’ First
Theorem as well as Abel’s Formula to transform our series into another more convenient
series, which could be interpreted geometrically. This allowed us to find an upper and
lower bound using integration, which enabled us to ultimately complete the proof.
From this point, the next step in research could be exploring the proof the the Prime
Number Theorem. This would involve a much more complicated proof, using methods
from Complex Analysis.

29
 Bibliography

[1] A.M. Yaglom, I.M. Yaglom, Challenging Mathematical Problems with Elementary Solutions. New
York, Dover Publication, (1987).
[2] D. Goldfeld, The Elementary Proof of the Prime Number Theorem: An Historical Perspective.
Columbia University, New York. 1
[3] J.J. O’Connor, E.F. Robertson, Prime Numbers, The MacTutor History of Mathematics Archive.
May 2009. Web. http://www-history.mcs.st-and.ac.uk/HistTopics/Prime_numbers.html 1
[4] K.H. Rosen, Discrete Mathematics and Its Applications, New York, McGraw-Hill, (2007). 2

30