Ilovepdf Merged (17) Removed 0

NUMBER THEORY
[MTH1C05]
STUDY MATERIAL
I SEMESTER
CORE COURSE
M.Sc. Mathematics
(2019 Admission onwards)
UNIVERSITY OF CALICUT
SCHOOL OF DISTANCE EDUCATION
CALICUT UNIVERSITY- P.O
MALAPPURAM- 673635, KERALA
190555
School of Distance Education
SCHOOL OF DISTANCE EDUCATION

UNIVERSITY OF CALICUT
STUDY MATERIAL FIRST

SEMESTER
M.Sc. Mathematics (2019 ADMISSION ONWARDS) CORE
COURSE:
MTH1C05-NUMBER THEORY
Prepared by:
Dr. Anil Kumar V

Professor
Department of Mathematics
University of Calicut
Scrutinized By:
Dr. Preethi Kuttipulackal

Associate Professor & Head
Department of Mathematics
MTH1C05-NUMBER THEORY
Number Theory
School of Distance Education
Anil Kumar V
Anil Kumar V. Number Theory
ii
Contents
1 Arithmetical Functions and its Averages 1

1.1 Arithmetical functions . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 The Möbius function µ(n) . . . . . . . . . . . . . . . . . . 1
1.1.2 Euler totient function ϕ(n) . . . . . . . . . . . . . . . . . 3
1.1.3 A relation connecting ϕ and µ. . . . . . . . . . . . . . . . 4
1.1.4 A product formula for ϕ(n) . . . . . . . . . . . . . . . . . 5
1.1.5 Dirichlet product of arithmetical functions . . . . . . . . . 9
1.1.6 Dirichlet inverses and Mobius inversion formula . . . . . . 12
1.1.7 The Mangoldt function Λ(n) . . . . . . . . . . . . . . . . 15
1.1.8 Multiplicative functions . . . . . . . . . . . . . . . . . . . 16
1.1.9 Examples of completely multiplicative functions . . . . . . 20
1.1.10 Examples of multiplicative functions . . . . . . . . . . . . 20
1.1.11 Multiplicative functions and Dirichlet multiplication . . . 21
1.1.12 The inverse of a completely multiplicative function . . . . 24
1.1.13 Liouville’s function λ(n) . . . . . . . . . . . . . . . . . . . 28
1.1.14 The divisor function σα (n) . . . . . . . . . . . . . . . . . 30
1.1.15 Generalised convolution . . . . . . . . . . . . . . . . . . . 32
1.1.16 Derivatives of arithmetical functions . . . . . . . . . . . . 34
1.1.17 The Selberg identity . . . . . . . . . . . . . . . . . . . . . 36
1.1.18 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
1.2 Averages of arithmetical functions . . . . . . . . . . . . . . . . . 38
1.2.1 The big oh notation. Asymptotic equality of functions . . 39
1.2.2 Partial sums of a Dirichlet product . . . . . . . . . . . . . 46
1.2.3 Applications to µ(n) and Λ(n) . . . . . . . . . . . . . . . 48
1.2.4 Another identity for the partial sums of a Dirichlet product 53
1.2.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
2 Some elementary theorems on distribution of prime numbers 57

2.1 Chebyshev’s functions ψ(x) and ϑ(x) . . . . . . . . . . . . . . . . 57
2.2 Relations connecting ϑ(x) and π(x) . . . . . . . . . . . . . . . . . 60
2.3 Some equivalent forms of the prime number theorem . . . . . . . 64
2.4 Inequalities for π(n) and pn . . . . . . . . . . . . . . . . . . . . . 72
2.5 Shapiro’s Tauberian theorem . . . . . . . . . . . . . . . . . . . . 79
2.6 Applications of Shapiro’s theorem . . . . . . . . . . . . . . . . . . 83
iii
2.7 An asymptotic formula for the partial sums . . . . . . . . . . . . 86

2.8 The partial sums of the Mobius function . . . . . . . . . . . . . . 88
3 Quadratic Residues 99
3.1 Definition and Examples . . . . . . . . . . . . . . . . . . . . . . . 99
3.2 Legendre’s symbol and its properties . . . . . . . . . . . . . . . . 101
3.3 Evalution of (−1 | p) and (2 | p) . . . . . . . . . . . . . . . . . . . 103
3.4 Quadratic reciprocity law . . . . . . . . . . . . . . . . . . . . . . 109
3.5 Jacobi symbol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
3.6 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
3.7 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
4 Cryptography 121
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
4.2 Terminologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
4.3 Types of Cryptography . . . . . . . . . . . . . . . . . . . . . . . 122
4.3.1 Symmetric Key Cryptography . . . . . . . . . . . . . . . . 122
4.3.2 Affine enciphering transformation . . . . . . . . . . . . . 123
4.3.3 Asymmetric Key Cryptography . . . . . . . . . . . . . . . 124
4.3.4 RSA cryptosystem. . . . . . . . . . . . . . . . . . . . . . 124
4.3.5 Hash Functions . . . . . . . . . . . . . . . . . . . . . . . . 126
4.3.6 Signatures . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
4.3.7 One way functions . . . . . . . . . . . . . . . . . . . . . . 126
4.3.8 Trapdoor One-Way Function . . . . . . . . . . . . . . . . 126
4.4 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
Bibliography 135
4.5 Syllabus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
4.6 Notations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
iv
Module 1
Arithmetical Functions and

its Averages
This module consists of two sections. Let N be the set of natural numbers.
Then f : N → C is called a sequence. In number theory such functions are
called arithmetical functions. In the first section introduces several arithmeti-
cal functions. The Dirichlet product of two arithmetical functions are included.
Moreover we will prove that the set of all arithmetical functions which do not
vanish at 1 form an abelian group under Dirichlet multiplication. Furthermore,
Möbius inversion is included. This section also includes generalized convolu-
tions and gives relation between associative property of Dirichlet multiplication
and convolution. Generalized inversion formula is also included. The second
section is devoted to averages of Arithmetical functions. The well known Euler’s
summation formula is derived. Several asymptotic formulas are also derived.
1.1 Arithmetical functions

Definition 1.1.1. A complex valued function defined on the set of natural num-
bers is called an arithmetical function or a number theoretic function. That is,
a function f : N → C is called an arithmetical function .
1.1.1 The Möbius function µ(n)

Definition 1.1.2. The Mobius function µ is defined as:

1
 if n = 1,
µ(n) = (−1) k
if n is a product k distinct primes,

0 otherwise.

From the definition of µ it clear that µ(n) = 0 if and only if n has a square
1
factor > 1.Obviously,

µ(1) = 1, µ(2) = −1, µ(3) = −1, µ(4) = 0, µ(5) = −1, µ(6) = 1, µ(8) = 0
and so on.
Theorem 1.1.1. If n ≥ 1 we have
(
X 1 1 if n = 1,
µ(d) = =
n 0 if n > 1.
d|n
Proof. We consider two cases.

Case 1: Suppose n = 1.
In this case X X
µ(d) = µ(d) = µ(1) = 1.
d|n d|1
Case 2: Suppose n > 1.

Then by fundamental theorem of arithmetic, n can be expressed in the following
form:
n = pa1 1 pa2 2 · · · pakk ,
where pi ’s are distinct primes and α1 , α2 , . . . , αk are integers greater than or
equal to 1. Now consider
X X
µ(d) = µ(d).
a a a
d|n d|p1 1 p2 2 ···pkk
Note that µ(d) = 0 if and only if d has a square factor. Therefore µ(d) 6= 0 if
and only if
d = 1; p1 , p2 , . . . , pk ; p1 p2 , . . . , pk−1 pk ; p1 p2 p3 , . . . pk−2 pk−1 pk ; · · · , p1 p2 · · · pk .
| {z } | {z } | {z } | {z }
(k1) (k2) (k3) (kk)
Hence
X
µ(d) = µ(1) + [µ(p1 ) + µ(p2 ) + · · · + µ(pk )] + [µ(p1 p2 ) + · · · + µ(pk−1 pk )] +
d|n
[µ(p1 p2 p3 ) + · · · + µ(pk−2 pk−1 pk )] + [µ(p1 p2 · · · pk )]

= 1 + [(−1) + (−1) + · · · + (−1)] + (−1)2 + (−1)2 + · · · + (−1)2 +

| {z } | {z }
(k1) (k2)
(−1)3 + (−1)3 + · · · + (−1)3 + · · · + (−1)k

| {z } | {z }
(k3) (kk)

k k k k
=1+ (−1) + (−1)2 + (−1)3 + · · · + (−1)k
1 2 3 k
= (1 − 1)n = 0.
This completes the proof.
2
1.1.2 Euler totient function ϕ(n)

Definition 1.1.3. If n ≥ 1 the Euler totient function ϕ(n) is defined as
ϕ(n) := #{a ∈ N : a ≤ n, (a, n) = 1}.
Thus
n 0
X
ϕ(n) = 1,
k=1
where dash denotes the sum is taken over those k which are relatively prime to
n. That is
n n
X X 1
ϕ(n) = 1= .
(n, k)
k=1 k=1
(k,n)=1

X
ϕ(d) = n.
d|n
Proof. Let
S = {1, 2, 3, . . . , n}
We partition the set S into mutually disjoint subsets as follows:
For each divisor d of n, let
A(d) := {k : (k, n) = d, 1 ≤ k ≤ n}.
Claim 1: If d1 and d2 are distinct divisors of n, then A(d1 ) ∩ A(d2 ) = ∅.
x ∈ A(d1 ) ∩ A(d2 ) ⇒ x ∈ A(d1 ) and x ∈ A(d2 )

⇒ (x, n) = d1 and (x, n) = d2
⇒ d1 = d2 , contradiction.
Claim 2: {1, 2, . . . , n} = ∪d|n A(d).

Obviously
∪d|n A(d) ⊆ S. (1.1)
Now
x ∈ S ⇒ (x, n) = d for some d, 1 ≤ d ≤ n

⇒ x ∈ A(d)
⇒ x ∈ ∪d|n A(d) for some d|n.
Thus
S ⊆ ∪d|n A(d). (1.2)
From equations (1.1) and (1.2) it follows that
S = ∪d|n A(d).
3
Therefore X
#(S) = #(A(d)).
d|n
This implies that

X
n= #(A(d)). (1.3)
d|n
Claim : #(A(d)) = ϕ(d).

Now
k ∈ A(d) ⇔ (k, n) = d
⇔ (k/d, n/d) = 1, 0 < k/d ≤ n/d
⇔ (q, n/d) = 1, 0 < q ≤ n/d
⇔ q ∈ {k : (k, n/d) = 1, 0 < k ≤ n/d}
| {z }
B
⇔q∈B
This implies that there is a one to one correspondence between the elements of
A(d) and B. Hence
#A(d) = #(B) = ϕ(n/d).
Hence equation (1.3) becomes:
X
n= ϕ(n/d) (1.4)
d|n
Observe that if d is a divisor of n, then n/d is also a divisor of n. The divisors

d and n/d are called conjugate divisors of n. Hence
X X
ϕ(n) = ϕ(n/d).
d|n d|n
Hence equation (1.4) can be written as

X
ϕ(d) = n.
d|n
This completes the proof of the theorem.
1.1.3 A relation connecting ϕ and µ.

X n
ϕ(n) = µ(d) .
d
d|n
4
Proof. By the definition of ϕ, we have

n
X 1
ϕ(n) =
(n, k)
k=1
Xn X
= µ(d) (by theorem 1.1.1)
k=1 d|(n,k)
n
X X
= µ(d)
k=1 d|n&d|k
 
X Xn X
=  µ(d) (1.5)
d|n k=1 d|k
Pn P
Note that for a fixed d, k=1 d|k µ(d) denotes the summation over all those
k in the set {1, 2, . . . , n} which are multiples of d.
Question: How many multiples of d are there in the set {1, 2, 3, . . . , n}?
Suppose 1 ≤ k ≤ n and k = qd. Now
1≤k≤n⇔0<k≤n
k n
⇔0< ≤
d d
k n
⇔1≤ ≤
d d
n
⇔1≤q≤
d
This implies that in the set {1, 2, . . . , n} there are n/d numbers which are mul-
tiples of d. These multiples of d are
{d, 2d, . . . , (n/d)d}
Hence equation (1.5) can be written as:

X n
ϕ(n) = µ(d) .
d
d|n
1.1.4 A product formula for ϕ(n)

Theorem 1.1.4. For n ≥ 1 we have
Y 1

ϕ(n) = n 1− ,
p
p|n
where p is a prime divisor of n.
5
Proof. Here we consider two cases.

Note that when n = 1, ϕ(n) = ϕ(1) = 1. Also when n = 1,
Y 1
Y
1

n 1− = 1−
p p
p|n p|1
Since there are no prime divide 1, we define

Y 1

1− = 1.
p
p|1
Hence the theorem is true for n = 1.

In this case n can be expressed in the form:
n = pα1 α2 αr
1 p2 · · · pr ,
where pi ’s are distinct primes and α1 , α2 , . . . , αr are integers ≥ 1. Consider

Y Y r
1 1
1− = 1−
p i=1
pi
p|n

1 1 1
= 1− 1− ··· 1 −
p1 p2 pr

1 1 1 1 1 1
=1− + + ··· + + + + ··· + −
p1 p2 pr p1 p2 p1 p3 pr−1 pr
(−1)r

1 1
+ ··· + + ··· +
p1 p2 p3 pr−2 pr−1pr p1 p2 · · · pr
r r 2 r
X (−1) X (−1) X (−1)3 (−1)r
=1+ + + + ··· +
i=1
pi i,j=1
pi pj pi pj pk p1 p2 · · · pr
i,j,k=1
i6=j i6=j6=k
r r r
X µ(pi ) X µ(pi pj ) X µ(pi pj pk ) µ(p1 p2 · · · pr )
=1+ + + + ··· +
i=1
pi i,j=1
pi pj pi pj pk p1 p2 · · · pr
i,j,k=1
i6=j i6=j6=k
X µ(d)
= .
d
d|n
Hence
Y 1
X µ(d)
n 1− =n
p d
p|n d|n
X n
= µ(d)
d
d|n
6
= ϕ(n) (by theorem 1.1.3 )
Theorem 1.1.5. Euler’s totient has the following properties:
(a) ϕ(pα ) = pα − pα−1 , for prime p and α ≥ 1.

d
(b) ϕ(mn) = ϕ(m)ϕ(n) ϕ(d) , where d = (m, n)
(c) ϕ(mn) = ϕ(m)ϕ(n) if (m, n) = 1.
(d) a|b implies ϕ(a)|ϕ(b).
(e) ϕ(n) is even for n ≥ 3. Moreover, if n has distinct r odd prime factors
then 2r |ϕ(n).
Proof.
(a) We have
Y 1

ϕ(n) = n 1− (theorem 1.1.24) (1.6)
p
p|n
Putting n = pα in equation (1.6), we obtain:

Y 1

ϕ(pα )n 1− (1.7)
p
p|pα

α 1
=p 1− (since p is the only prime divisor of pα ) (1.8)
p
= pα − pα−1 (1.9)
(b)We have
Y 1

ϕ(n) = n 1− (theorem 1.1.24)
p
p|n
Therefore

ϕ(n) Y 1
= 1− . (1.10)
n p
p|n
The above equation is true for all natural numbers n. Hence it is true for mn.
Hence
ϕ(mn) Y 1

= 1− . (1.11)
mn p
p|mn
7
Note that each prime divisor of mn is either a prime divisor of m or of n, and

those prime which divide both m and n also divide d = (m, n). Hence
ϕ(mn) Y 1

= 1−
mn p
p|mn
Q
1 1
Q
p|m 1 − p p|n 1 − p
= Q
1
p|d 1 − p
ϕ(m) ϕ(n)
m n
= ϕ(d)
(by theorem 1.1.24)
d

d
= ϕ(m)ϕ(n) .
ϕ(d)
(c)By part (b) we have

d
ϕ(mn) = ϕ(m)ϕ(n) , where d = (m, n).
ϕ(d)
Putting d = 1 in the above equation, we obtain
1
ϕ(mn) = ϕ(m)ϕ(n)
ϕ(1)
= ϕ(m)ϕ(n) (since ϕ(1) = 1).
(d)Given that a|b.
a|b ⇒ b = ac, 1 ≤ c ≤ b.
Here we consider 3 cases. Case 1: Suppose b = 1.

Since b = 1 and 1 ≤ c ≤ b, it follows that c = 1. This implies that a = 1. So
ϕ(a) = ϕ(b) = 1. Hence ϕ(a)|ϕ(b).
Case 2: Suppose b > 1, c = b.
In this case a = 1. This implies that ϕ(a) = 1. Note that 1 divides every
integer. In particular 1|ϕ(b). Hence ϕ(a)|ϕ(b).
Case 3: b > 1, c < b.
Now proof is by induction on b.
Step 1: b = 2.
Since 1 ≤ c < b and b = 2, it follows that c = 1. Hence a = 2. Therefore
ϕ(a)|ϕ(b).
Step 3: Assume that the result is true for all positive integers less than b(
That is, whenever λ|µ, µ < b, then ϕ(λ)|ϕ(µ)).
Step 2: Now to show that the result is true for b.
Let d = (a, c). Then d|c. Since c < b, by induction hypothesis, we have
ϕ(c)
ϕ(d)|ϕ(c) ⇒ ∈ N.
ϕ(d)
8
But now
ϕ(b) = ϕ(ac)
d
= ϕ(a)ϕ(c)
ϕ(d)

ϕ(c)
= ϕ(a)d
ϕ(d)
Since ϕ(c)/ϕ(d) is a natural number, it follows that ϕ(a)|ϕ(b).
(e)Let n ≥ 3. Then
n = 2k m, for some odd numberm.
If m = 1, then n = 2k , where k ≥ 2(since n ≥ 3). Therefore
ϕ(n) = ϕ(2k ) = 2k − 2k−1
= 2(2k−1 − 2k−2 ) = an even number.
Assume that m is an odd number greater than 1. Then m has an odd prime
divisor say p.
Now
p|m ⇒ ϕ(p)|ϕ(n)(∵ n = 2k m)
⇒ ϕ(p)|ϕ(n)
⇒ (p − 1)|ϕ(n)
⇒ 2|ϕ(n) (p is an odd prime)
⇒ ϕ(n) is even.
Next assume that
αk
n = pα1 α2
1 p2 · · · pk ,
where p1 , p2 , . . . , pr ∈ {2, 3, 5, 7, . . .}. Then
Y 1

ϕ(n) = n 1−
p
p|n

αk 1 1 1
= pα
1
1 α2
p 2 · · · p k 1 − 1 − · · · 1 −
p1 p2 pk
= pα
1
1 −1 α2 −1
p2 · · · pkαk −1 (p1 − 1)(p2 − 1) · · · (pn − 1)
Since each pi is odd it follows that each (pi − 1) is even. This implies 2r divides
ϕ(n). This completes the proof.
1.1.5 Dirichlet product of arithmetical functions

Definition 1.1.4. Let f and g be arithmetical functions. Then the Dirichlet
product of f and g is denoted by f ∗ g and is defined as
X n
(f ∗ g)(n) = f (d)g
d
d|n
9
X
= f (a)f (b), (1.12)
ab=n
where the sum on the right-hand side of (1.12) runs over all ordered pairs (a, b)
of positive integers satisfying ab = n.
Definition 1.1.5. If α is a complex number, the power function N α is an

arithmetical function is defined as
N α (n) = nα , ∀n ∈ N.
Definition 1.1.6. The unit function u is defined as
u(n) = 1 ∀n ∈ N.
Definition 1.1.7. The identity function I is defined as

(
1 1 if n = 1,
I(n) = =
n 0 if n > 1.
Remark 1. Theorem 1.1.1 can be stated in the form ϕ = µ ∗ u.
We have
X 1
µ(d) =
n
d|n
Now
X X n
µ(d) = µ(d)u
d
d|n d|n
= (µ ∗ u)(n)
Remark 2. Theorem 1.1.3 can be sated in the form µ ∗ N = ϕ.
We have
X n
ϕ(n) = µ(d)
d
d|n
X n
= µ(d)N
d
d|n
= (µ ∗ N )(n).
Theorem 1.1.6. Dirichlet multiplication is commutative and associative. That

is, for any multiplicative functions f, g, h we have
f ∗ g = g ∗ f (Commutative law)
f ∗ (g ∗ h) = (f ∗ g) ∗ h(Associative law)
10
Proof. Let f and g be two arithmetical functions. Then

X n
(f ∗ g)(n) = f (n)g
d
d|n
X
= f (a)g(b)
ab=n
X
= g(b)f (a)
ba=n
= (g ∗ f )(n).
Hence the Dirichlet product is commutative. That is f ∗ g = g ∗ f . Next, we
prove that Dirichlet multiplication is associative.
Let k = g ∗ h and ` = f ∗ g. Then
(f ∗ (g ∗ h))(n) = (f ∗ k)(n)
X
= f (a)k(d)
ad=n
X
= f (a)(g ∗ h)(d)
ad=n
X X
= f (a) g(b)h(c)
ad=n bc=d
X
= f (a)g(b)h(c).
abc=n
((f ∗ g) ∗ h)(n) = (` ∗ h)(n)

X
= `(d)h(c)
dc=n
X
= (f ∗ g)(d)f (c)
dc=n
X X
= f (a)g(b)h(c)
dc=n ab=d
X
= f (a)g(b)h(c).
abc=n
Hence (f ∗ g) ∗ h = f ∗ (g ∗ h).
Theorem 1.1.7. For any arithmetical function f we have
f ∗ I = I ∗ f = f,
where I is the identity function.
Proof. By the definition of Dirichlet product, we have
X n
(I ∗ f )(n) = I(d)f
d
d|n
11
X n
= I(1)f (n) + I(d)f
d
d|n
d>1
X 1 n
= f (n) + f
d d
d|n
d>1
= f (n) + 0 (∵ I(1) = 1I(d) = 0 if d > 1.)
= f (n).
Hence
I ∗ f = f.
Since Dirichlet product is commutative, we have f ∗ I = f .
1.1.6 Dirichlet inverses and Mobius inversion formula

Theorem 1.1.8. If f is an arithmetical function with f (1) 6= 0, then there is
a unique inverse f −1 of f called Dirichlet inverse of f such that
f ∗ f −1 = I = f −1 ∗ f
Moreover f −1 is given by

1
 f (1) if n = 1,
−1
f (n) = 1
P n
−1
− f (1) d|n f d f (d) if n > 1.
d<n
Proof. The proof is by induction on n. Given that f (1) 6= 0. Then

1
∈ C.
f (1)
Take f −1 (1) := 1
Then f −1 (1) is unique. Also
f (1) .

−1
X
−1 1 1
(f ∗ f )(1) = f (d)f = f (1)f −1 (1) = f (1) = 1.
d f (1)
d|1
Hence the result is true for n = 1.

Next assume that the result is true for all natural numbers less than n. That is
f −1 (r) is uniquely determined for all r < n, f ∗ f −1 (r) = I for all r < n and
1 X r −1
f −1 (r) = − f f (d) ∀ r < n.
f (1) d
d|r
d<r
We will prove that (f ∗ f −1 )(n) = I(n) and

1 X n −1
f −1 (n) = − f f (d).
f (1) d
d|n
d<n
12
Define
1 X n −1
f −1 (n) := − f f (d).
f (1) d
d|n
d<n
Consider
X n
(f ∗ f −1 )(n) = f f −1 (d),
d
d|n
n X n
=f f −1 (n) + f f −1 (d)
n d
d|n
d<n
X n
= f (1)f −1 (n) + f f −1 (d)
d
d|n
d<n
 
1 X n −1  X n
f −1 (d)

= f (1) −
 f (1) f f (d) + f
d  d
d|n d|n
d<n d<n
X n X n
=− f f −1 (d) + f f −1 (d) = 0.
d d
d|n d|n
d<n d<n
This implies that the result is true for n. Hence by mathematical induction the
result is true for all natural numbers.
Theorem 1.1.9. Let
A := {f : f is arithmetical and f (1) 6= 0}
Then A is an abelian group under Dirichlet product.
Proof.
Closure property: Let f, g ∈ A . Then f (1) 6= 0 and g(1) 6= 0. Also note

that f ∗ g is an arithmetical function. Also
(f ∗ g)(1) = f (1)g(1) 6== 0.
Hence f ∗ g ∈ A .
Associative property: By theorem 1.1.6 it follows that ∗ is associative.
Existence of identity: The existence of each element in A follows from the-

orem 1.1.8.
Existence of inverse: Note that I is the identity element (see theorem 1.1.7).
Commutative property: By theorem 1.1.6 it follows that ∗ is commutative.
13
Hence (A , ∗) is an abelian group.

Remark 3. The functions µ and u are inverses of each other.
Proof. From theorem 1.1.1 we have
X
µ(d) = I(n).
d|n
The above equation can be written as

X n
µ(d)u = I(n). (∵ u(n) = 1 ∀n)
d
d|n
That is
(µ ∗ u)(n) = I(n) ∀n
Theorem 1.1.10 (Mobius inversion formula). We have
X X n
f (n) = g(d) ⇔ g(n) = f (d)µ
d
d|n d|n
That is
f = g ∗ u ⇔ g = f ∗ µ.
Proof. Note that u−1 = µ. Hence
f = g ∗ u ⇔ f ∗ u−1 = (g ∗ u) ∗ u−1
⇔ f ∗ u−1 = (g ∗ u) ∗ u−1
⇔ f ∗ µ = g ∗ (u ∗ u−1 )
⇔ f ∗ µ = g ∗ I = g.
This completess the proof.
Corollary 1. If n ≥ 1 we have
X n
ϕ(n) = dµ .
d
d|n
Proof. We have X
ϕ(d) = n = N (n).
d|n
By Mobius inversion formula,
ϕ(n) = (N ∗ µ)(n)
X n
= µ(d)N (∵ N (n) = n ∀n)
d
d|n
X n
= µ(d) .
d
d|n
14
1.1.7 The Mangoldt function Λ(n)

Definition 1.1.8. For every integer n ≥ 1 the Mangoldt’s function Λ(n) is
defined as:
(
log p if n = pm for some prime p and some m ≥ 1
Λ(n) =
,0 otherwise.
A short table values of Λ(n) is given below:

n 1 2 3 4 5 6 7 8 9 10
Λ(n) 0 log 2 log 3 log 2 log 5 0 log 7 log 2 log 3 0
X
log n = Λ(n).
d|n
Proof. Here we consider two cases:

In this case X X
Λ(d) = Λ(d) = Λ(1) = 0
d|n d|1
Hence the theorem is true for n = 1.

In this case n can be expressed in the following form:
αk
n = pα 1 α2
1 p1 · · · p1 ,
where pi ’s are distinct primes and αi ≥ 1. Taking logarithms we have

k
X
log n = αi log pi . (1.13)
i=1
P
Observe that non zero terms in d|n Λ(d) occcurs only when
αk
d = p1 , p21 , . . . , pα 2 α2 2
k ; p2 , p2 , . . . , p2 ; . . . ; pk , pk , . . . , pk .
1
Hence
X
Λ(d) = Λ(p1 ) + Λ(p21 ) + · · · + Λ(pα 2 α2
1 ) + Λ(p2 ) + Λ(p2 ) + · · · + Λ(p2 ) + · · ·
1
d|n
Λ(pk ) + Λ(p2k ) + · · · + Λ(pα

k )
k
= (log p1 + log p1 + · · · + log p1 ) + (log p2 + log p2 + · · · + log p2 ) + · · ·

| {z } | {z }
α1 times α2 times
(log + log pk + · · · + log pk )

| k {z }
αk times
15
= α1 log p1 + α2 log p2 + · · · + αk log pk

k
X
= αi log pi . (1.14)
i=1
From the light of equations (1.13) and (1.14), we have

X
log n = Λ(d).
d|n

X n X
Λ(n) = µ(d) log =− µ(d) log d.
d
d|n d|n
Proof. By theorem 1.1.11, we have

X X n
log n = Λ(n) = Λ(n)u = (Λ ∗ u)(n).
d
d|n d|n
By Möbius inversion formula, we have

X n
Λ(n) = µ(d) log
d
d|n
X X
= µ(d) log n − µ(d) log d
d|n d|n
X X
= log n µ(d) − µ(d) log d
d|n d|n
X
1 X 1
= (log n) − µ(d) log d (∵ µ(d) = )
n n
d|n d|n
X
=− µ(d) log d.
d|n
1.1.8 Multiplicative functions

We have already proved that the set of all arithmetical function f with f (1) 6=
0(see theorem 1.1.9) forms an abelian group under Dirichlet multiplication. In
this section we discuss an important subgroup of this group, called multiplicative
functions.
Definition 1.1.9. An arithmetical function f is called multiplicative if
(a) f not identically zero,
16
(b) f (mn) = f (m)f (n), whenever (m, n) = 1.
Definition 1.1.10. An arithmetical function f is called completely multiplica-

tive if
(a) f not identically zero,
(b) f (mn) = f (m)f (n), for all m, n ∈ N.
Note :Consider the following sets:
A := {f : f is arithmetical and f (1) 6= 0},

M := {f ∈ A : f is multiplicative},
C := {f ∈ A : f is completely multiplicative}.
Then we have
M ⊂C ⊆A.
Moreover M is a subgroup of (A , ∗). Unlike M , which is a group with respect to
Dirichlet product, C is closed neither with respect to addition nor convolution.
Theorem 1.1.13. If f is multiplicative then f (1) = 1.
Proof. Since f is multiplicative, we have
f (mn) = f (m)f (n) whenever (m, n) = 1.
Note that for any n ∈ N, we have (n, 1) = 1. Hence by definition,
f (n) = f (n · 1) = f (n)f (1)
This implies that

f (n)[f (1) − 1] = 0.
Since f is not identically zero, we have f (1) − 1 = 0. That is f (1) = 1.
Theorem 1.1.14. Let f be an arithmetical function with f (1) = 1. Then
(a) f is multiplicative if and only if

αk αk
f (pα1 α2 α1 α2
1 p2 · · · pk ) = f (p1 ) f (p2 ) · · · f (pk )
for all primes pi and all integers αi ≥ 1.
(b) If f is multiplicative, then f is completely multiplicative if and only if
f (pα ) = f (p)α
for all primes p and all integers α ≥ 1.
17
Proof. (a) First assume that f is multiplicative. Then f (mn) = f (m)f (n)
whenever (m, n) = 1. We prove the the result by induction on k.
Step 1:Assume that k = 1. In this case we have
f (pα α1
1 ) = f (p1 )
1
Hence the result is true for n = 1.

Step 2:Assume that the result is true for all integers < k.
That is
f (pα 1 α2 αi α1 α2 αi
1 p2 · · · pi ) = f (p1 ) f (p2 ) · · · f (pi ) ∀ i < k.
Case 3:We will prove that the result is true for k.

Now
αk αk−1 αk
f (pα1 α2 α1 α2
1 p2 · · · pk ) = f p1 p2 · · · pk−1 pk
αk−1 αk
= f pα 1 α2
1 p2 · · · pk−1 f (pk ) (∵ f is multiplicative)
αk−1 αk
= f (pα α2
1 ) f (p2 ) · · · f pk−1 f (pk ) (by induction hypothesis)
1
Hence the result is true for k. Hence by mathematical induction the result is
true for all natural numbers.
Conversely assume that
αk αk
f (pα1 α2 α1 α2
1 p2 · · · pk ) = f (p1 ) f (p2 ) · · · f (pk ) ∀k ∈ N.
We will prove that f is multiplicative. Let m, n ∈ N such that (m, n) = 1. Let

αk
m = pα1 α2
1 p2 · · · pk ,
n = q1β1 q2β2 · · · qsβs
where pi ’s and qj ’s are distinct primes and αi , βi ≥ 1 are integers. Then

αk β1 β2
f (mn) = f pα 1
1 α2
p 2 · · · p q
k 1 2 q · · · q βs
s

= f (p1 ) f (p2 ) · · · f (pk ) f q1β1 f q2β2 · · · f qkβs (by assumption)
α1 α2 αk

αk β1 β2 βs
= f (pα 1
1 α2
p 2 · · · p k ) f q 1 2q · · · qk
= f (m)f (n)
(b) First assume that f is completely multiplicative. Then
f (mn) = f (m)f (n) ∀ m, n ∈ N
Claim : f (pα ) = f (p)α for all primes p and all integers αi ≥ 1. We will prove
the theorem by induction on α.
Step 1:Assume that α = 1. Note that f (p1 ) = f (p)1 . Hence the result is true
for n = 1.
18
Step 2: Assume that the result is true for all integers < k.
That is
f pi = f (p)i ∀i < k.

Step 3:We will prove that the result is true for k.

Now
f (pα ) = f pα−1 p

= f pα−1 f (p)(∵ f completely multiplicative)

α−1
= [f (p)] f (p) = [f (p)]α (by assumption).
Hence by mathematical induction f (pα ) = f (p)α for all α ∈ N.
Conversely assume that
f (pα ) = f (p)α for all k.
We will prove that f is completely multiplicative. Let m, n ∈ N. Let (m, n) = d.
Then m = ad and n = bd for some a, b ∈ N.Let
αk β1 β2 γ1 γ2 γt
a = pα 1 α2 βs
1 p2 . . . pk , b = q1 q2 . . . qs , d = d1 d2 . . . dt
Then
αk γ1 γ2 γt
m = pα1 α2
1 p2 . . . pk d1 d2 . . . dt
n = q1β1 q2β2 . . . qsβs dγ11 dγ22 . . . dγt t

αk β1 β2 βs 2γ1 2γ2 2γt
mn = pα1 α2
1 p2 . . . pk q1 q2 . . . qs d1 d2 . . . dt .
Now

αk β1 β2 βs 2γ1 2γ2 2γt
f (mn) = f pα 1 α2
1 p2 . . . pk q1 q2 . . . qs d1 d2 . . . dt

= f (pα α2 αk
q1β1 f q2β2 · · · f qsβs

1
1
) f (p2 ) · · · f (pk ) f

f d2γ1
1
f d22γ2 · · · f d2γ t
t
(∵ f is multiplicative)
α1 α2 αk β1 β2 βs
= f (p1 ) f (p2 ) · · · f (pk ) f (q1 ) f (q2 ) · · · f (qs )
2γ1 2γ2 2γt
f (d1 ) f (d2 ) · · · f (dt ) (by assumption)
Again

α k γ1 γ2 γt β1 β2 β s γ1 γ2 γt
f (m)f (n) = f (pα
1
1 α2
p2 . . . pk d d
1 2 . . . dt ) f q q
1 2 . . . qs d d
1 2 . . . dt
αk γ1 γ2 γt
= f (pα α2
1 ) f (p2 ) · · · f (pk ) f (d1 ) f (d2 ) · · · f (dt )
1

f q1β1 f q2β2 · · · f qsβs f (dγ11 ) f (dγ22 ) · · · f (dγt t )

α1 α2 αk β1 β2 βs
= f (p1 ) f (p2 ) · · · f (pk ) f (q1 ) f (q2 ) · · · f (qs )
2γ1 2γ2 2γt
f (d1 ) f (d2 ) · · · f (dt )
Hence f (mn) = f (m)f (n) for all m, n ∈ N. That is f is completely multi-
plicative.
19
1.1.9 Examples of completely multiplicative functions

1. The arithmetical function N α (n) = nα ∀ n ∈ N is completely multiplica-
tive.
For all m, n ∈ N, we have
N (mn) = (mn)α = N (m)N (m).
Hence N α is completely multiplicative. In particular N is completely

multiplicative.
2. The unit function u(n) = 1 for all n ∈ N is completely multiplicative.
For all m, n ∈ N, we have
u(mn) = 1 = 1.1 = u(m)u(n).
3. The identity function is completely multiplicative.

Here we consider several cases:
Case 1: Suppose m = 1, n = 1.
Then mn = 1. Therefore
I(m) = 1, I(n) = 1, I(mn) = 1
Hence
I(mn) = I(m)I(n)
Case 2: Suppose m = 1 and n > 1.
In this case mn > 1. So we have
I(m) = 1, I(n) = 0, I(mn) = 0
Thus
I(mn) = 0 = 1 × 0 = I(m)I(n).
Case 3: Suppose n = 1, m > 1.
In this case mn > 1. Moreover I(n) = 1, I(m) = 0 and I(mn) = 0. Hence
I(mn) = 0 = 0 × 1 = I(m)I(n).
Case 4:Suppose m, n > 1.

In this case mn > 1. Moreover I(m) = I(n) = I(mn) = 0. Hence
I(mn) = 0 = 0 × 0 = I(m)I(n).
1.1.10 Examples of multiplicative functions

1. The Euler function ϕ is multiplicative but not completely multiplicative.
Let m, n ∈ N such that (m, n) = 1. Then we have
ϕ(mn) = ϕ(m)ϕ(n) (see theorem 1.20(c))
20
Observe that ϕ is not completely multiplicative. Let m = 4 and n = 2.

Then (m, n) 6= 1. By the definition of ϕ we have
ϕ(8) = 4, ϕ(4) = 2, ϕ(2) = 1.
Hence
ϕ(mn) = ϕ(8) = 4 6= 2 × 1 = ϕ(4)ϕ(2) = ϕ(m)ϕ(n).
2. The Mobius function is multiplicative but not completely multiplicative.

Let m, n ∈ N such that (m, n) = 1.
Case 1: If either m = 1 or n = 1, then obviously
µ(mn) = µ(m)µ(n).
Case 2: m is square free and n is not square free.

Then clearly mn is not square free. Then by the definition of µ, we have
µ(mn) = 0 and µ(n) = 0. Hence
µ(mn) = 0 = µ(m)µ(n).
Case 3: m is not square free and n is square free.

This case is exactly similar to case 2.
Case 4: Both m and n are square free.
Since m and n are relatively prime, mn is square free. Let
m = p1 p2 · · · pm ,
where pi ’s are distinct primes. Similarly, let
q = q1 q2 . . . qn ,
where qi ’s are distinct primes. Then
µ(m) = (−1)m , µ(n) = (−1)n and µ(mn) = (−1)m+n .
Hence
µ(mn) = (−1)m+n = (−1)m (−1)n = µ(m)µ(n).
Hence µ is completely multiplicative. Next we will show that µ is not
completely multiplicative. Let m = 2 and n = 2. Then µ(4) = 0 and
µ(2) = −1. Note that µ(4) 6= µ(2)µ(2). Hence µ is not completely
multiplicative.
1.1.11 Multiplicative functions and Dirichlet multiplica-

tion
Theorem 1.1.15. If f and g are multiplicative, so is their Dirichlet product
f ∗ g.
21
Proof. Since f and g are multiplicative functions, we have f (1) = 1 and g(1) = 1.
Now X
(f ∗ g)(1) = f (d)g(1/d) = f (1)g(1) = 1
d|1
Hence (f ∗ g)(1) 6= 0. Let m, n ∈ N such that (m, n) = 1. Note that
d|mn ⇔ d = ab, where a|m, b|m and(a, b) = 1.
Now
X mn
(f ∗ g)(mn) = f (d)g
d
d|mn
X mn
= f (ab)g
ab
a|m
b|n
X m n
= f (a)f (b)g g
a b
a|m
b|n
m n
(∵ f and g are multiplicative, (a, b) = 1& , = 1)
m X n a b
X
= f (a)g f (b)
a b
a|m b|n
= (f ∗ g)(m)(f ∗ g)(n).
Remark 4. The Dirichlet product of two completely multiplicative functions

need not be completely multiplicative.
Proof. Observe that the arithmetical functions u and N are completely multi-
plicative. We will prove that the function f = u ∗ N is not completely multi-
plicative. To prove that f is not completely multiplicative it suffices to prove
that for any prime number p, we have f (pα ) 6= f (p)α .
Now
f (pα ) = (u ∗ N )(pα )
X
= u(d)N (pα /d)
d|pα
X pα
=
α
d
d|p
α
p pα pα pα
= + + 2 + ··· + α
1 p p p
α α−1
=p +p + · · · + 1.
22
Again
f (p) = (u ∗ N )(p)
X
= u(d)N (p/d) = u(1)N (p) = p
d|1
Therefore
f (p)α = pα 6= pα + pα−1 + · · · + 1 = f (pα ) .
Hence f is not completely multiplicative.
Theorem 1.1.16. Let f, g be arithmetic functions. If both g and f ∗ g are
multiplicative, then f is also multiplicative.
Proof. Since g and f ∗ g are multiplicative, we have g(1) = 1 and (f ∗ g)(1) = 1.
Now

X 1
1 = (f ∗ g)(1) = f (d)g = f (1)g(1) = f (1)(1) = f (1).
d
d|1
Hence f 6= 0.
Claim :f is multiplicative.That is, f (mn) = f (m)f (n) whenever (m, n) = 1.
We prove the claim by induction on mn.
Step 1: Assume that mn = 1.
In this case m = n = 1. Note that
f (mn) = f (1) = 1 = f (m)f (n).
Hence the result is true for mn = 1.
Step 2: Assume that the result is true for all products ab less than mn.
Step 3: We prove the claim for mn.
Note that
X mn
(f ∗ g)(mn) = f (d)g
d
d|mn
X mn
= f (ab)g
ab
a|m
b|n
X mn
= f (mn)g(1) + f (ab)g (∵ (a, b) = 1)
ab
a|m
b|n
ab<mn
X m n m n
= f (mn) + f (a)f (b)g g ∵ , =1
a b a b
a|m
b|n
ab<mn
X m n
= f (mn) + f (a)g f (b)g + f (m)f (n) − f (m)f (n)
a b
a|m
b|n
ab<mn
23
X m n m n

= f (mn) + f (a)g f (b)g + f (m)g f (n)g − f (m)f (n)
a b m n
a|m
b|n
ab<mn
X m n
= f (mn) + f (a)g f (b)g − f (m)f (n)
a b
a|m
b|n
X m X n
= f (mn) + f (a)g f (b)g − f (m)f (n)
a b
a|m b|n
= f (mn) + (f ∗ g)(m)(f ∗ g)(n) − f (m)f (n)

= f (mn) + (f ∗ g)(mn) − f (m)f (n) (∵ f ∗ g is multiplicative)
This implies that f (mn) − f (m)f (n) = 0. That is f (mn) = f (m)f (n).
Theorem 1.1.17. If g is multiplicative, then its Dirichlet multiplicative g −1 is
also multiplicative.
Proof. First observe that I = g ∗ g −1 . We have already seen that I is completely
multiplicative and hence multiplicative. Hence g and g ∗ g −1 are multiplicative.
So by theorem 1.1.16, g −1 is multiplicative.
1.1.12 The inverse of a completely multiplicative function

Theorem 1.1.18. Let f be a multiplicative function. Then f is multiplicative
if and only if
f −1 (n) = µ(n)f (n) for all n ≥ 1.
Proof. Assume that f is completely multiplicative. Let g(n) = µ(n)f (n). We
want to show that f −1 (n) = g(n). Now
X n
(g ∗ f )(n) = g(d)f
d
d|n
X n
= µ(d)f (d)f
d
d|n
X n
= µ(d)f d (∵ f is completely multiplicative )
d
d|n
X
= µ(d)f (n)
d|n
X
= f (n) µ(d)
d|n
= f (n)I(n)
(
1 if n = 1
=
0 if n > 1.
24
Remark 5. The Dirichlet inverse of the Euler totient function is given by

X
ϕ−1 (n) = µ(d)d.
d|n
Proof. Since ϕ = µ ∗ N we have
ϕ−1 = µ−1 ∗ N −1
We have already proved that N is completely multiplicative. Therefore by

previous theorem,
N −1 (n) = µ(n)N (n).
Hence
ϕ−1 = µ−1 ∗ N −1 = µ−1 ∗ µN = u ∗ µN
Therefore
ϕ−1 (n) = (u ∗ µN )(n)

X n
= µN (d)u
d
d|n
X
= (µN )(d)
d|n
X X
= µ(d)N (d) = µ(d)d.
d|n d|n
Theorem 1.1.19. If f is multiplicative we have

X Y
µ(d)f (d) = (1 − f (p)).
d|n p|n
Proof. Let
X
g(n) = f (d)µ(d)
d|n
X
= (f µ)(d)
d|n
X n
= (f µ)(d)u
d
d|n
= (f µ ∗ u)(n)
Hence
g = f µ ∗ u.
26
Since f and µ are multiplicative, so is f µ. Also observe that u is multiplicative.

This implies that f µ ∗ u is multiplicative.
Now
g(pα ) = (f µ ∗ u)(pα )
α
X p
= (f µ)(d)u
d
d|pα
X
= (f µ)(d)
d|pα
X
= f (d)µ(d)
d|pα
= f (1)µ(1) + f (p)µ(p) + f (p2 )µ(p2 ) + · · · + f (pα )µ(pα )

= 1 − f (p) (∵ f (1) = 1, µ(p2 ) = µ(p3 ) = · · · = 0)
Let
αk
n = pα1 α2
1 p2 · · · pk ,
where pi ’s are distinct primes and α ≥ 1.. Since g is multiplicative we have

αk
g(n) = g(pα α2
1 )g(p2 ) · · · g(pk )
1
= (1 − f (p1 )(1 − f (p2 )) · · · (1 − f (pk ))

Y
= (1 − f (p))
p|n
Corollary 2. The Dirichlet inverse of ϕ is given by

Y
ϕ−1 = (1 − f (p)).
p|n
Proof. From remark 5, we have

X
ϕ−1 (n) = µ(d)d
d|n
X
= µ(d)N (d)
d|n
Y
= (1 − N (p)) (by theorem 1.1.19)
p|n
Y
= (1 − p)
p|n
27
1.1.13 Liouville’s function λ(n)

Definition 1.1.11. The Liouville’s function λ is defined as:
(
1 if n = 1,
λ(n) = αk
(−1)α1 +α2 +···+αk if n = pα 1 α2
1 p2 · · · pk ,
where pi ’s distinct primes and αi ≥ 1 are integers.
Remark 6. The Liouville’s function is a completely multiplicative function.
Proof. From the definition of λ we have
λ(pα ) = (−1)α
= [λ(p)]α .
This implies that λ is completely multiplicative.
Theorem 1.1.20. For every n ≥ 1, we have

(
X 1 if n is a square,
λ(d) =
d|n
0 otherwise.
Moreover
λ−1 (n) = |µ(n)| for all n.
Proof. Let
X
g(n) = λ(d)
d|n
X n
= λ(d)u
d
d|n
Hence
g(n) = (λ ∗ u)(n)
Since λ and µ are multiplicative, λ ∗ µ is multiplicative. Putting n = pα in the
equation X
g(n) = λ(d)
d|n
we obtain:
X
g(pα ) = λ(d)
d|n
= λ(1) + λ(p) + · · · + λ(pα )

= 1 + (−1) + (−1)2 + (−1)3 + · · · + (−1)α
28
(
1 if α is even,
=
0 if α is odd.
Let
αk
n = pα1 α2
1 p2 · · · pk ,
where pi ’ s are distinct primes and i ≥ 1 are integers. Therefore

αk
g(n) = g(pα α2
1 )g(p2 ) · · · g(pk )
1
(
1 if each αi is even,
=
0 if at least one αi is odd.
Assume that each αi is even. Then αi = 2βi for βi ,i = 1, 2, . . . , k. This implies

that
n = p2β 1 2β2
1 p2 · · · p2β
k
k
2 2 2
= pβ1 1 pβ2 2 · · · pβkk ,
= a square
Hence (
X 1 if n is a square,
g(n) = λ(d) =
d|n
0 otherwise.
Note that λ is completely multiplicative. Therefore
λ−1 (n) = λ(n)µ(n) (1.15)
Claim : λ(n)µ(n) = |µ(n)| for all n.

We have

1
 if n = 1,
k
µ(n) = (−1) if n = p1 p2 · · · pk , for distinct primes,

0 otherwise.


1
 if n = 1,
αk
λ(n) = (−1)α1 +α2 +···+αk if n = pα 1 α2
1 p2 · · · pk ,

0 otherwise.

Here we consider 3 cases:

Case 1:Suppose n = 1.
In this case λ(n)µ(n) = 1 = |µ(1)|.
Case 2:Suppose n = p1 p2 · · · pk , where pi ’s are distinct primes.
Then
µ(n) = (−1)k , λ(n) = (−1)k .
Therefore
λ(n)µ(n) = (−1)k (−1)k = 1 = |µ(n)|.
29
αk
Case 3: Suppose n = pα1 α2
1 p2 · · · pk , where pi ’s are distinct primes and αi ≥ 1.
In this case
µ(n) = 0, λ(n) = (−1)α1 +α2 +···+αn
Therefore
λ(n)µ(n) = 0 = |µ(n)|.
Hence
λ−1 (n) = µ(n)λ(n) = |µ(n)| for all n.
1.1.14 The divisor function σα (n)

Definition 1.1.12. For any real or complex α and any integer n ≥ 1 the divisor
function is defined by
X
σα (n) := dα ,
d|n
= sum of the αth powers of the divisors of n

Remark 7. σα (n) is multiplicative.
Proof. Note that N α and u are multiplicative functions. Now
X
σα (n) = dα
d|n
X n
= N α (d)u
d
d|n
= (N α ∗ u)(n).
Hence σα is the Dirichlet product of two multiplicative functions N α and u.
Thus σα is multiplicative.
Remark 8. If α = 0, then the divisor function σα can be written as:
X
σ0 (n) = d0
d|n
X
= 1
d|n
= Number of the divisors of n.

The function σ0 is usually denoted by d(n).
Remark 9. If α = 1, then the divisor function σα can be written as:
X
σ1 (n) = d
d|n
= sum of the divisors of n.

The function σ1 is denoted by σ(n).
30
Theorem 1.1.21. For n ≥ 1, we have

X n
σα−1 (n) = dα µ(d)µ .
d
d|n
Proof. We have
X
σα (n) = dα
d|n
X n
= N α (d)u
d
d|n
= (N α ∗ u)(n)
Hence σα = N α ∗ u. Therefore
σα−1 = (Nα )−1 ∗ u−1 = (Nα )−1 ∗ µ. (1.16)
Since N α is completely multiplicative, we have
(N α )−1 (n) = N α (n)µ(n) = (N α µ)(n).
Inserting the value of (N α )−1 in equation (1.16) we obtain
σα−1 (n) = (N α µ) ∗ µ(n)

X n
= (N α µ)(d)µ
d
d|n
X n
= N α (d)µ(d)µ
d
d|n
X n
= dα µ(d)µ .
d
d|n
αk
Theorem 1.1.22. If n > 1 and n = pα 1 α2
1 p2 · · · pk , where pi ’s distinct primes
and αi ≥ 1., then
k
Y
d(n) = (αi + 1).
i=1
Proof. Since d is a multiplicative function we have

αk αk
d (pα1 α2 α1 α2
1 p2 · · · pk ) = d (p1 ) d (p2 ) · · · d (pk ) (1.17)
To compute d (pα αi
i ) we note that the divisors of pi are
i
1, pi , p2i , . . . , pα
i
i
31
Hence the number of divisors of pα

i are (αi + 1). Therefore,
i
d (pα
i ) = (αi + 1)
i
Hence
k
Y
d(n) = (αi + 1)
i=1
αk
Theorem 1.1.23. If n > 1 and n = pα 1 α2
1 p2 · · · pk , where pi ’s distinct primes
and αi ≥ 1., then
k
Y pα
i
i +1
−1
σ(n) = αi
i=1
p i − 1
Proof. Since σ is a multiplicative function we have

αk αk
σ (p1α1 pα α1 α2
2 · · · pk ) = σ (p1 ) σ (p2 ) · · · σ (pk )
2
(1.18)
To compute σ (pα αi
i ) we note that the divisors of pi are
i
1, pi , p2i , . . . , pα
i
i
Hence
(αi +1)
pi −1
σ (pα 2 αi
i ) = 1 + pi + pi + . . . + pi =
i
.
pi − 1
Hence
k
Y pαi +1 − 1
i
σ(n) =
i=1
pα
i −1
i
1.1.15 Generalised convolution

Definition 1.1.13. Let F be a real or complex valued function defined on the
positive real axis (0, ∞) such that F (x) = 0 for 0 < x < 1. Let α be an
arithmetical function. Then the sum
X x
α(n)F
n
n≤x
is called the generalized convolution of G and α and is denoted by α ◦ F . Thus

X x
(α ◦ F )(x) := α(n)F .
n
n≤x
Theorem 1.1.24 (Associative property relating α and ∗). For any arithmetical
functions α and β, we have
α ◦ (β ◦ F ) = (α ∗ β) ◦ F.
32
Proof. For x > 0 we have

X x
(α ◦ (β ◦ F ))(x) = α(n)(β ◦ F )
n
n≤x
X X x
n
= α(n) β(m)F
x m
n≤x m≤ n
X X x
= α(n) β(m)F
mn
n≤x mn≤x
X x
= α(n)β(m)F
mn
mn≤x

X k x
= α(n)β F , k = mn
n k
k≤x
 

X X k  x
=  α(n)β F
n k
k≤x n|k
X x
= (α ∗ β)(k)F
k
k≤x
= ((α ∗ β) ◦ F )(x).
Therefore α ◦ (β ◦ F ) = (α ∗ β) ◦ F . This completes the proof.
Theorem 1.1.25 (Generalized inversion formula). Let α has Dirichlet inverse

α−1 . Then
X x X x
G(x) = α(n)F ⇔ F (x) = α−1 (n)G .
n n
n≤x n≤x
That is
G = α ◦ F ⇔ F = α−1 ◦ G.
Proof. First assume that G = α ◦ F . Then
α−1 ◦ G = α−1 ◦ (α ◦ F )
= (α−1 ∗ α) ◦ F (by theorem 1.1.24)
= I ◦ F = F.
Conversely assume that F = α−1 ◦ G. Then
α ◦ F = α ◦ (α−1 ◦ G)
= (α ∗ α−1 ) ◦ G(by theorem 1.1.24)
= I ◦ G = G.
33
Theorem 1.1.26 (Generalized Mobius inversion formula). Let α be completely

multiplicative. Then
X x X x
G(x) = α(n)F ⇔ F (x) = α(n)µ(n)G
n n
n≤x n≤x
That is
G = α ◦ F ⇔ F = α−1 ◦ G = (αµ) ◦ G.
Proof. First assume that G = α ◦ F . Then
α−1 ◦ G = α−1 ◦ (α ◦ F )
= (α−1 ∗ α) ◦ F
= I ◦ F = F.
That is
F = α−1 ◦ G = (αµ) ◦ G (∵ α is completely multiplicative)
Conversely assume that F = α−1 ◦ G. Then
α ◦ F = α ◦ (α−1 ◦ G)
= (α ∗ α−1 ) ◦ G
= I ◦ G = G.
1.1.16 Derivatives of arithmetical functions

Definition 1.1.14. Let f be an arithmetical function. Then the derivative of
f is defined as
f 0 (n) := f (n) log n, n ≥ 1.
Example 1. The derivative of the function I is 0.
By definition, the derivative of I is
I 0 (n) = I(n) log(n)

1
= log n
n
(
1 × log 1 if n = 1
=
0 × log n if n > 1.
= 0.
Remark 10. We have Λ ∗ u = u0 .
34
Proof. By the definition of derivative, we have

u0 (n) = u(n) log n
= log n (u(n) = 1)
X
= Λ(d) (by theorem 1.1.11)
d|n
X n
= Λ(d)u
d
d|n
= (Λ ∗ u)(n).
Theorem 1.1.27. If f and g are arithmetical functions we have:
(a) (f + g)0 = f 0 + g 0 .
(b) (f ∗ g)0 = f 0 ∗ g + f ∗ g 0 .
(c) (f −1 )0 = −f −1 ∗ (f ∗ f )−1 , provided that f (1) 6= 0.
Proof. (a) By the definition of derivative, we have
(f + g)0 (n) = (f + g)(n) log n
= (f (n) + g(n)) log n
= f (n) log n + g(n) log n
= f 0 (n) + g 0 (n).
(b)Note that
(f ∗ g)0 (n) = (f ∗ g)(n) log n
X n
= f (d)g log n
d
d|n
X n n
= f (d)g log(d )
d d
d|n
X n X n n
= f (d)g log d + f (d)g log
d d d
d|n d|n
X n X n n
= f (d) log dg + f (d)g log )
d d d
d|n d|n
X n X n
= f 0 (d)g + f (d)g 0
d d
n|d d|n
= (f ∗ g)(n) + (f ∗ g 0 )(n).
0
(c) Note that I 0 = 0. This implies that (f ∗ f −1 )0 = 0. Then by part (b) we

have
0 = f 0 ∗ f −1 + f ∗ (f −1 )0
35
This implies that

f ∗ (f −1 )0 = −(f 0 ∗ f −1 )
This implies that
f −1 ∗ f ∗ (f −1 )0 = −f −1 ∗ (f 0 ∗ f −1 )

That is,
(f −1 ∗ f ) ∗ (f −1 )0 = −f 0 ∗ (f −1 ∗ f −1 )
That is
I ∗ (f −1 )0 = −f 0 ∗ (f ∗ f )−1
That is
(f −1 )0 = −f 0 ∗ (f ∗ f )−1 .
1.1.17 The Selberg identity

X n X
Λ(n) log n + Λ(d)Λ = µ(d) log2 (n/d).
d
d|n d|n
Proof. By Remark 10 we have:
Λ ∗ u = u0 . (1.19)
Differentiation of this equation gives:
Λ0 ∗ u + Λ ∗ u0 = u00 .
That is
Λ0 ∗ u + Λ ∗ (Λ ∗ u) = u00 .
Now multiply both sides by µ = u−1 to obtain:
(Λ0 ∗ u) ∗ u−1 + (Λ ∗ (Λ ∗ u) ∗ u−1 ) = u00 ∗ µ.
That is
Λ0 + Λ ∗ Λ = u00 ∗ µ
This implies that
(Λ0 + Λ ∗ Λ)(n) = (u00 ∗ µ)(n)
That is
X n X n
Λ(n) log n + Λ(d)Λ = µ(d)u00
d d
d|n d|n
36
X n
= µ(d)u0 log(n/d)
d
d|n
X n
= µ(d)u (log(n/d))2
d
d|n
X
= µ(d)(log(n/d))2 .
d|n
1.1.18 Exercises
1. Find all integers n such that
(a) ϕ(n) = n/2,

(b) ϕ(n) = ϕ(2n)
(c) ϕ(n) = 12.
2. For each of the following statements either give a proof or exhibit a counter
example.
(a) If (m, n) = 1 the (ϕ(m), ϕ(n)) = 1,

(b) If n is composite, then (n, ϕ(n)) > 1,
(c) If the same prime divide m and n, then nϕ(m) = mϕ(n).
3. Prove that
n X µ2 (d)
=
ϕ(n) ϕ(d)
d|n
4. Prove that ϕ(n) > 6 for all n with at most 8 distinct prime factors.
5. Prove that d2 |n µ(d) = µ2 (n).

P
√ √
6. Let f (n) = [ n] − n − 1 . Prove that f is multiplicative but not com-
pletely multiplicative.
7. Assume that f is multiplicative. Prove that
(a) f −1 (n) = µ(n)f (n) for every square free n.

(b) f −1 (p2 ) = f (p)2 − f (p2 ) for every prime p.
(a) Assume that f is multiplicative. Prove that f is completely multi-
plicative if, and only if, f −1 (pα ) = 0 for all primes p and all integers
a ≥ 2.
(b) Prove that the following statement
P or exhibit a counter example. If
f is multiplicative, then F (n) = d|n f (d) is multiplicative.
37
1.2 Averages of arithmetical functions

An arithmetical function is defined to be a function f (n), defined for n ∈ N,
which maps to a complex number such that f : N → C. Examples of arithmetic
functions include: the number of primes less than a given number n, the number
of divisors of n, etc. While the behavior of values of arithmetic functions are
hard to predict, it is easier to analyze the behavior of the averages of arithmetic
functions which we define as
f (1) + f (2) + · · · + f (n)
lim =L
n→∞ n
where L the average value of f (n).
1. Let r(n) be the number of representations of n as a sum of two squares
n = x2 + y 2
Then average number of representations of a natural number as a sum of

two squares is π. That is,
r(1) + r(2) + · · · + r(n)
lim =π
n→∞ n
2. Let f (n) be the number of decompositions of a natural number n into a

sum of one or more consecutive prime numbers. For example, f (395) = 2
because
395 = 127 + 131 + 137 = 71 + 73 + 79 + 83 + 89.
Then average number of decompositions of a natural number into a sum

of one or more consecutive prime numbers is log 2. That is
f (1) + f (2) + · · · + f (n)

lim = log 2
n→∞ n
3. Consider the divisor function d(n) defined by
X
d(n) = 1.
d|n
Note that d(p) = 2 for all primes numbers and d(n) is large when n has a
large number of divisors. Thus the values of d(n) fluctuate as n increases.
In this case it is difficult to determine their behavior as n increases. We
can prove that
d(1) + d(2) + · · · + d(n)
=1
n log n
In chapter paper, we will examine averages of several different arithmetic func-
tions.
38
1.2.1 The big oh notation. Asymptotic equality of func-

tions
Definition 1.2.1. Let f and g be real valued functions such that g(x) > 0 for
all x ≥ a. We write f (x) = O(x)(and say that f (x) is big oh of g(x)) if there
exists a constant M > 0 such that
|f (x)| ≤ M g(x) for all x ≥ a.
• f (x) = O(g(x)) means that |f (x)/g(x)| is bounded for all x ≥ a.

• The equation of the form
f (x) = h(x) + O(g(x))
means that
f (x) − h(x) = O(g(x))
Rx Rx
• If f (t) = O(g(t)) for some t ≥ a, then a f (t) dt = O( a g(t) dt)
Proof. Note that f (t) = O(g(t)) means that there exists a constant M > 0 such
that
|f (t)| ≤ M g(t), for t ≥ a
This implies that
Z x Z x
|f (t)|dt ≤ M g(t)dt, for x ≥ a
a a
That is, Z
x

Z x

f (t)dt ≤ M g(x), for x ≥ a
a a
That is Z x Z x
f (t) dt = O g(t) dt .
a a
Example 2. Note that
x = O(x2 ), x2 + 10x + 7 = O(x2 ), xn = O(ex ) for all n ∈ N.
Theorem 1.2.1.
1. If f1 is O(g1 ) and f2 is O(g2 ), then (f1 + f2 ) is O(max{g1 , g2 }).
2. If f1 and f2 are both O(g), then (f1 + f2 ) is O(g).
3. If f1 is O(g1 ) and f2 is O(g2 ), then (f1 f2 ) is O(g1 g2 ).
4. If f1 is O(f2 ) and f2 is O(f3 ), then f1 is O(f3 ).
39
5. If f is O(g), then (af ) is O(g) for any constant

Definition 1.2.2. We say f (x) is asymptotic to g(x) as x → ∞ if
f (x)
lim =1
x→∞ g(x)
and write f (x) ∼ g(x), x → ∞.
Here we need the following:
Definition 1.2.3.
P Let f be an arithmetical function and x is any positive real
number. Then n≤x f (n) is defined as:
[x]
X X
f (n) = f (n)
n≤x n=1
Definition 1.2.4. Let f be an arithmetical function. Then the average of f is

defined as
1 X
f ∼ (n) := f (n)
n
n≤x
Theorem 1.2.2. (Euler summation formula) If f has a continuous deriva-

tive f 0 on the interval [y, x], where 0 < y < x, then
X Z x Z x
f (n) = f (t)dt + (t − [t]) f 0 (t)dt + f (x) ([x] − x) − f (y) ([y] − y) .
y<n≤x y y
Equivalently,
X Z x Z x
f (n) = f (t)dt + {t}f 0 (t)dt + f (y){y} − f (x){x},
y<n≤x y y
where {t} = t − {t}.

Let m = [y] and k = [x]. If n, n − 1 ∈ [y, x], then:
Z k k
X Z n
0
[t]f (t) dt = [t]f 0 (t) dt
m n=m+1 n−1
k Z
X n
= (n − 1)f 0 (t)
m+1 n−1
k
X
= (n − 1)[f (n) − f (n − 1)]
m+1
k
X
= ([nf (n) − (n − 1)f (n − 1)] − f (n))
m+1
40
k
X k
X
= ([nf (n) − (n − 1)f (n − 1)]) − f (n)
m+1 m+1
X
= kf (k) − mf (m) − f (n)
y<n≤x
X Z k
f (n) = − [t]f 0 (t) dt + kf (k) − mf (m)
y<n≤x m
Z x Z y Z x
=− [t]f 0 (t)dt − [t]f 0 (t) + [t]f 0 (t)dt + kf (k) − mf (m)
y m k
Z x Z k Z x Z k Z y Z ! x
∵ = + = − +
y y k m m k
Z x Z y Z x
=− [t]f 0 (t)dt − mf 0 (t) + kf 0 (t)dt + kf (k) − mf (m)
y m k
Z x
=− [t]f 0 (t)dt − m[f (y) − f (m)] + k[f (x) − f (k)] + kf (k) − mf (m)
y
Z x
=− [t]f 0 (t)dt − mf (y) + kf (x)
y
Z x
=− [t]f 0 (t)dt − [y]f (y) + [x]f (x)
y
Z x
=− [t]f 0 (t)dt + (y − [y]) f (y) − (x − [x]) f (x) + xf (x) − yf (y)
y
Z x
=− [t]f 0 (t)dt + {y}f (y) − {x}f (x) + xf (x) − yf (y)
y
Z x Z x Z x
=− [t]f 0 (t)dt + {y}f (y) − {x}f (x) + tf 0 (t)dt + f (t)dt
y y y
Z x Z x
= (t − [t])f 0 (t)dt + {y}f (y) − {x}f (x) + f (t)dt
y y
Z x Z x
0
= {t}f (t)dt + f (t)dt + {y}f (y) − {x}f (x).
y y
Z x Z x
tf 0 (t)dt =tf (t)]xy − f (t)dt
y y
Z x
=xf (x) − yf (y) − f (t)dt.
y
Definition 1.2.5. The Euler’s constant is defined by the equation

1 1 1
C = lim 1 + + + · · · + − log n .
n→∞ 2 3 n
41
Definition 1.2.6. The Riemann zeta function is defined by the equation

(P∞ 1
n=1 ns if s > 1
ζ(s) = P 1 x1−s

limx→∞ n≤x ns − 1−s if 0 < s < 1.
Theorem 1.2.3. If x ≥ 1 we have

1 1
P
(a) n≤x n = log x + C + O x , where C is Eulers constant.
x1−s
1
+ ζ(s) + O(x−s ) if s > 0, s 6= 1.
P
(b) n≤x ns = 1−s
1
= O(x1−s ) if s > 1.
P
(c) n>x ns
xα+1
ns = + O(xs ) if α ≥ 0.
P
(d) n≤x α+1
Proof. (a)By Euler’s Summation formula, we have

X Z x Z x
f (n) = f (t)dt + (t − [t]) f 0 (t)dt + f (x) ([x] − x) − f (y) ([y] − y) .
y<n≤x y y
(1.20)
Putting y = 1 and f (t) = 1/t in the above formula, we obtain:

X 1 Z x Z x
1
= dt − (1 − [t])(−1/t2 )dt + (1/x)([x] − x)
n 1 t 1
1<n≤x
That is
Z x
X 1 t − [t] x − [x]
= log x − 2
dt −
n 1 t x
1<n≤x
Equivalently,
Z x
X 1 t − [t] x − [x]
= log x − 2
dt + 1 − (1.21)
n 1 t x
1≤n≤x
Consider the functions

x − [x] 1
f (x) = − , g(x) =
x x
Note that
|f (x)|
= x − [x] < 1.
g(x)
Thus
f (x) = O(g(x)).
42

Z x
X 1 t − [t] 1
= log x − 2
dt + 1 + O
n 1 t x
1≤n≤x
Z ∞ Z ∞
t − [t] t − [t] 1
= log x − 2
dt − 2
dt + 1 + O
1 t x t x
Z ∞ Z ∞
t − [t] t − [t] 1
= log x − 2
dt + 2
dt + 1 + O (1.22)
1 t x t x
Note that ∞ ∞
t − [t]
Z Z
1
0≤ dt < dt = 1 < ∞
1 t2 1 t2
Thus ∞
t − [t]
Z
< ∞.
1 t2
Again note that
t−[t] ∞ Z ∞
t − [t] t − [t]
Z
t2 1 1 1
1 = t−[t] < 1 ⇒ =O ⇒ dt = O dt = O
t2
t2 t2 x t2 x t 2 x

Z ∞ Z ∞
X 1 t − [t] t − [t] 1
= log x − 2
dt + 2
dt + 1 + O
n 1 t x t x
1≤n≤x
Z ∞
t − [t] 1 1
= log x + 1 − 2
dt +O + O
1 t x x
| {z }
B

1
= log x + B + O (1.23)
x
Taking x → ∞ in the above equation, we obtain:
 
X 1
lim  − log x = B
x→∞ n
1≤n≤x
This implies that B = C. Hence equation (1.23) becomes:

X 1
1
= log x + C + O
n x
n≤x
(b) Putting y = 1 and f (t) = t−s , s > 0, s 6= 1 in the Euler’s summation

formula, we obtain:
Z x Z x
X 1 1 x − [x]
s
= s
dt + (t − [t])(−st−s−1 )dt −
n 1 t 1 xs
1<n≤x
43
Z x
x−s+1

1 t − [t] x − [x]
= − −s dt −
1−s 1−s 1 ts+1 xs
−s+1 Z x
x 1 t − [t] 1
= − −s dt + O
1−s 1−s 1 ts+1 xs
Therefore,
Z x
x−s+1

X 1 1 t − [t] 1
s
= + 1 − − s s+1
dt + O
n 1−s 1−s 1 t xs
1≤n≤x
Z ∞ Z ∞
x1−s

1 t − [t] t − [t] 1
= − +1−s s+1
dt − s+1
dt + O s
1−s 1−s 1 t x t x
∞
x1−s

t − [t]
Z
1 1 1
= − +1−s s+1
dt + O s
+O s
1−s 1−s 1 t x x
∞
x1−s

t − [t]
Z
1 1
= + 1− −s s+1
dt +O s
. (1.24)
1−s 1−s 1 t x
| {z }
B(s)
t − [t] s
s =O
ts+1 tZs+1
∞ Z ∞
t − [t] s
⇒ s s+1 dt = O dt
x t x ts+1

1
=O
xs
Case 1: Suppose s > 1.

Letting x → ∞ in the above equation, we obtain:
∞
X 1
= B(s)
n=1
ns
x1−s
1
+ ζ(s) + O(x−s ).
P
That is ζ(s) = B(s). Hence if s > 1, n≤x ns = 1−s
Thus equation (2.31) becomes:
X 1 x1−s
s
= + ζ(s) + O(x−s ).
n 1−s
n≤x
Case 2: Suppose 0 < s < 1.

Letting x → ∞ in equation (2.31), we obtain
 
X 1 1−s
x
lim  −  = B(s)
x→∞ ns 1−s
n≤x
44
1.2.2 Partial sums of a Dirichlet product

Theorem 1.2.4. If h = f ∗ g, let
X X X
H(x) = h(n), F (x) = f (n) and G(x) = g(n).
n≤x n≤x n≤x
Then we have
X x X x
H(x) = f (n)G = g(n)F .
n n
n≤x n≤x
Proof. Let (
0 if 0 < x < 1,
U (x) =
1 if x ≥ 1.
Claim 1:F = f ◦ U .
Now
X x
(f ◦ U )(x) = f (n)U
n
n≤x
X x
= f (n) ∵ 0 < x ≤ n ⇒ > 1 ⇒ U (x/n) = 1.
n
n≤x
= F (x).
Claim 2: G = g ◦ U.
Now
X x
(g ◦ U )(x) = g(n)U
n
n≤x
X x
= g(n) ∵ 0 < x ≤ n ⇒ > 1 ⇒ U (x/n) = 1.
n
n≤x
= G(x).
Claim 3: h ◦ U = H.
X x
(h ◦ U )(x) = h(n)U
n
n≤x
X
= h(n) = H(x).
n≤x
Claim 4: H = f ◦ G.
Now
f ◦ G = f ◦ (g ◦ U ) = (f ∗ g) ◦ U = h ◦ U = H.
Claim 5: H = g ◦ F.
Now
g ◦ F = g ◦ (f ◦ U ) = (g ∗ f ) ◦ U = h ◦ U = H.
46
P
Theorem 1.2.5. thm33 If F (x) = f (n) we have
n≤x
XX X hxi X x
f (d) = f (n) = F .
n n
n≤x d|n n≤x n≤x
Proof. From the above theorem, we have
H = f ◦ G, and
H = g ◦ F.
If g = 1, then we have:
X
G(x) = 1 = [x].
n≤x
Now
H(x) = (f ◦ G)(x)
X x
= f (n)G
n
n≤x
X hxi
= f (n) (1.25)
n
n≤x
Again
H(x) = (g ◦ F )(x)
X x
= g(n)F
n
n≤x
X x
= F (1.26)
n
n≤x
Also
X
H(x) = h(n)
n≤x
X
= (f ∗ g)(n)
n≤x
XX x
= f (n)g . (1.27)
n
n≤x d|n
From equations (1.25), (1.26) and (1.27), we have

XX X hxi X x
f (d) = f (n) = F .
n n
n≤x d|n n≤x n≤x
47
1.2.3 Applications to µ(n) and Λ(n)

X hxi
µ(n) =1
n
n≤x
and X hxi
Λ(n) = log[x]!.
n
n≤x
Proof. Putting f (n) = µ(n) theorem we obtain:

X hxi XX X 1
µ(n) = µ(d) = = 1.
n n
n≤x n≤x d|n n≤x
Again putting f (n) = Λ(n) theorem we obtain:
X hxi XX X
Λ(n) = Λ(d) = log n = log[x]!.
n
n≤x n≤x d|n n≤x

Theorem 1.2.7. For all x ≥ 1 we have

X

µ(n)
≤1
n≤x n

with equality holding only if x < 2.

Proof. Here we consider two cases:
Case1: Suppose x < 2.
In this case
X µ(n) µ(1)
= = 1.
n 1
n≤x
Note that in this case equality occurs.

Case 2: Suppose x ≥ 2.
For each real number y, let
{y} := y − [y].
From theorem 1.2.3 we have:
X hxi
1= µ(n)
n
n≤x
X x n x o
= µ(n) −
n n
n≤x
48
X µ(n) X nxo
=x − µ(n)
n n
n≤x n≤x
This implies that

X µ(n) X nxo
x =1+ µ(n)
n n
n≤x n≤x
Therefore

n x o
X µ(n) X
x = 1 + µ(n)
n≤x n n

n≤x
X nxo
≤1+
n
n≤x
X nxo
= 1 + {x} +
n
2≤n≤x
X
< 1 + {x} + 1 (∵ 0 ≤ {y} < 1)
2≤n≤x
 
X
= 1 + {x} + [x] − 1 ∵ 1 = [x]
n≤x
= 1 + x − [x] + [x] − 1 = x.
Dividing by x we obtain:
X
µ(n)
<1

n≤x n

Theorem 1.2.8 (Legendre’s identity). For every x ≥ 1 we have

Y
[x]! = pα(p) ,
p≤x
where the product is extended over all primes ≤ x, and

∞
X x
α(p) =
m=1
pm
Proof. We have
X hxi
Λ(n) = log[x]! (see theorem 1.2.3)
n
n≤x
49
Putting n = pm in the above equation, we obtain:

X
m x
log[x]! = Λ(p ) m
p
pm ≤x
X h x i X h x i
= Λ(2m ) m + Λ(3m ) m + · · ·
2 3
2,22 ,...2m ≤x 3,32 ,...3m ≤x
 
∞
X X
m x 
= Λ(p ) m 
p

p≤x m=1
pm ≤x
 
∞
X X x 
= log p m 
p

m=1
p≤x
pm ≤x
 
∞
X X x 
= log p 

pm 
p≤x m=1
pm ≤x
∞
X X x
= α(p) log p, α(p) =
m=1
pm
p≤x
pm ≤x
X
α(p)
= log p
p≤x
 
Y
= log  pα(p) 
p≤x
pα(p) . This completes the proof.

Q
Hence [x]! = p≤x
Theorem 1.2.9. If x ≥ 2 we have
log[x]! = x log x − x + O(log x),
and hence X hxi

Λ(n) = x log x − x + O(logx).
n
n≤x
Proof. Putting f (t) = log t in Euler’s summation formula with y = 1 we obtain:

Z x Z x
X 1
log n = log tdt + (t − [t]) dt + log x([x] − x) − 0
1 1 t
1<n≤x
Z x
1
= x log x − x + 1 + O dt + O(log x) (1.28)
1 t
50
Note that
t − [t]1/t
= t − [t] < 1 ⇒ (t − [t])1/t = O(1/t)
1/t
Therefore
x Z x
t − [t]1/t
Z
1
dt = O dt
1 1/t 1 t
Also

log x([x] − x)
= |[x] − x| < 1 ⇒ log x([x] − x) = O(log x)
log x
Adding 1 on both sides of equation (1.28) we obtain:

X
log n = x log x − x + 2 + O(log x) + O(log x)
n≤x
= x log x − x + O(log x)
That is
log[x]! = x log x − x + O(log x).
From the light of theorem 1.2.3 it follows that

X hxi
Λ(n) = x log x − x + O(log x)..
n
n≤x
Theorem 1.2.10. For x ≥ 2 we have

X x
log p = x log x + O(x),
p
p≤x
where the sum is extended over all primes ≤ x.
Proof. We have hxi

X
log[x]! = Λ(n) .
n
n≤x
Therefore,

X x m
log[x]! = Λ(p ) m
p
pm ≤x
X h x i X h x i
= Λ(2m ) m + Λ(3m ) + ···
2 m
2 3m
2,2 ,...2 ≤x 3,32 ,...3m ≤x
51
 
∞
X X x 
= Λ(pm ) m 
p

m=1
p≤x
pm ≤x
 
∞
X X x 
= log p
pm
 
p≤x m=1
pm ≤x
 
∞
X X x 
= log p 

pm

p≤x m=1
pm ≤x
XX ∞
X x x
= Λ(p) + Λ(pm ) m
p p
p≤x p≤x m=2
∞
X x XX x
= Λ(p) + log p m .
p m=2
p
p≤x p≤x
Therefore
∞
X x XX x
Λ(p) = log[x]! − log p m
p m=2
p
p≤x p≤x
∞
XX x
= x log x − x + O(log x) − log p (by theorem 1.2.9)
pm
p≤x m=2
(1.29)
P∞ h i
x
P
Claim : p≤x m=2 log p pm = O(x).
∞ XX∞
XX x x
log p ≤ log p m
pm p
p≤x m=2 m=2 p≤x
∞
X X x
= log p
m=2
pm
p≤x
∞
XX 1
=x log p
pm
p≤x m=2

X 1 1
=x log p 2 + 3 + · · ·
p p
p≤x

X 1 1
=x log p 2 1 + + · · ·
p p
p≤x

X 1 1
=x log p 2
p 1 − 1/p
p≤x
52
X 1 p
=x log p
p2 p − 1
p≤x
X log n
≤x
n(n − 1)
n≤x
| {z }
<∞
This implies that

∞
XX x
log p = O(x).
pm
p≤x m=2

X x
Λ(p) = x log x − x + O(log x) − O(x)
p
p≤x
= x log x + O(x).
1.2.4 Another identity for the partial sums of a Dirichlet

product
Theorem 1.2.11. If a and b are positive real numbers such that ab = x, then
X X x X x
f (d)g(q) = f (n)G + g(n)F − F (a)G(b).
n n
q,d n≤a n≤b
qd≤x
Proof. Let
X
F (x) = f (n),
n≤x
X
G(x) = g(n) and
n≤x
X
H(x) = h(x), h = f ∗ g. (1.30)
n≤x
Then
XX n
H(x) = f (d)g
d
n≤x d|n
X
= f (d)g(q). (1.31)
q,d
qd≤x
Since a and b are positive real numbers such that ab = x it follows that (a, b) is
a point on the rectangular hyperbola qd = x.
53
The sum H(x) in equation (1.31) is extended over all lattice points in the first
quadrant of (q, d)- plane, below the rectangular hyperbola qd = x between the
two lines q = 1 and d = 1. Let
D = {(d, q) : q ≥ 1, d ≥ 1, qd ≤ x}
A = {(d, q) : d ≥ 1, q > b, qd ≤ x}
B = {(d, q) : q ∈ [1, a], d ∈ [1, b]}
C = {(d, q) : d > a, q ≥ 1, qd ≤ x}.
Then
D = A ∪ B ∪ C.
Now
X X X
= f (d)g(q)
(d,q)∈A∪B d≤a q≤x/d
X X
= f (d) g(q)
d≤a q≤x/d
X x
= f (d)G
d
d≤a
54
X x
= f (n)G
n
n≤a
X X X
f (d)g(q) = f (d)g(q)
(d,q)∈B∪C q≤b d≤x/q
X X
= g(q) f (d)
q≤b d≤x/q

X x
= g(q)F
q
q≤b
X x
= g(n)F .
n
n≤b
X XX
= f (d)g(q)
(d,q)∈B d≤a q≤b
X X
= f (d) g(q)
d≤a q≤b
= F (a)G(b).
Hence
X X
f (d)g(q) = f (d)g(q)
d,q (d,q)∈D
qd≤x
X X X
= f (d)g(q) + −
(d,q)∈A∪B (d,q)∈B∪C (d,q)∈B
X x X x
= f (n)G + g(n)F − F (a)G(b).
n n
n≤a n≤b
1.2.5 Exercises
1. Use Euler’s summation formula to deduce the following for x ≥ 2:

P log n 1 2 log x
(a) n≤x n = 2 log x + A + O x , where A is constant.

1 1
P
(b) 2≤n≤x n log n = log(log x) + B + O x log x , where B is a constant.
(a) Prove that [2x] − [x] is either 0 or 1.
55
56
Module 2
Some elementary theorems

on distribution of prime
numbers
2.1 Chebyshev’s functions ψ(x) and ϑ(x)

Definition 2.1.1. For x > 0 the Chebyshev’s ψ-function is defined as
X
ψ(x) = Λ(n).
n≤x
Recall the Mangoldt Λ function:

(
log p if n = pm for some prime p and some m ≥ 1,
Λ(n) =
0 otherwise.
Using this function we can write the definition of ψ as follows:

X
ψ(x) = Λ(n)
n≤x
∞
X X
= Λ(pm )
p≤x m=1
pm ≤x
∞
X X
= Λ(pm )
m=1 p
pm ≤x
X∞ X
= log p (2.1)
m=1 p≤x1/m
57
Note that the sum in (2.1) is a finite sum. For, if x1/m < 2 then
P
p≤x1/m log p =
0 ( that is the summation is empty). Now
1 log x
x1/m < 2 ⇒ log x < log 2 ⇒ m > = log2 x.
m log2
Hence if m > log2 x, the sum in (2.1) is zero. Hence ψ(x) can be written as:
X X
ψ(x) = log p.
m≤log2 x p≤x1/m
Definition 2.1.2. If x > 0 the Chebyshev’s ϑ function is defined as

X
ϑ(x) = log p,
p≤x
where p runs over all primes ≤ x.
Using Chebyshev’s ϑ function, we can write Chebyshev’s ψ function as fol-

lows: X
ψ(x) = ϑ x1/m .
m≤log2 x
Theorem 2.1.1. For x > 0 we have

ψ(x) ϑ(x) (log x)2
0≤ − ≤ √ .
x x 2 x log 2
Proof. Note that

X X X
ψ(x) − ϑ(x) = log p − log p
m≤log2 x p≤x1/m p≤x
X X X
= log p + log p − log p
2≤m≤log2 x p≤x p≤x
X X
= log p ≥ 0.
2≤m≤log2 x p≤x1/m
Hence
X X
0 ≤ ψ(x) − ϑ(x) = log p
2≤m≤log2 x p≤x1/m
X
= ϑ(x1/m ) (2.2)
2≤m≤log2 x
We have
X
ϑ(x) = log p
p≤x
58
X
< log x
p≤x
< x log x.
Therefore
ϑ(x1/m ) < x1/m log x1/m
Hence (2.2) can be written as
X
0 ≤ ψ(x) − ϑ(x) < x1/m log(x1/m )
2≤m≤log2 x

= x1/2 log x1/2 + x1/3 log x1/3 + · · · + x1/ log2 x log x1/ log2 x
1 1/2 1 x1/ log2 x
= x log x + x1/3 log x + · · · + log x
2 3 log2 x
1 1/2 1 1
≤ x log x + x1/2 log x + · · · + x1/2 log x (∵ log is increasing)
2 3 log2 x
1 1/2 1 1
≤ x log x + x1/2 log x + · · · + x1/2 log x
2 2 2
1 1/2 1 1/2 1
≤ x log x log2 x = x log x log x
2 2 log 2
√
x (log x)2
=
2 log 2
Dividing by x we obtain:
0≤ − ≤ √
x x 2 x log 2
Corollary 3. We have

ψ(x) ϑ(x)
lim − = 0.
n→∞ x x
Proof. We have
0≤ − ≤ √
x x 2 x log 2
To prove the result it suffices to show that
(log x)2
lim √ = 0.
x→∞ 2 x log 2
By L Hospital rule we have
(log x)2 2(log x) x1

lim √ = lim 1
n→∞ 2 x log 2 n→∞ 2 √ log 2
2 x
59
2(log x) 2( x1 ) 4
= lim √ = lim 1 = lim √ = 0.
n→∞ x log 2 n→∞ √
2 x
log 2 n→∞ x log 2
2.2 Relations connecting ϑ(x) and π(x)

Theorem 2.2.1 (Abel’s identity). For any arithmetical function a(n) let
X
A(x) = a(n),
n≤x
where A(x) = 0 if x < 1. Assume that f has a continuous derivative on the

interval [y, x], where 0 < y < x. Then we have
X Z x
a(n)f (n) = A(x)f (x) − A(y)f (y) − A(t)f 0 (t)dt.
y<n≤x y
Proof. Let k = [x] and m = [y]. Then

X
A(x) = a(n)
n≤x
[x] k
X X
= a(n) = a(n) = A(k).
n=1 n=1
Similarly A(y) = A(m). Now

[x]
X X
a(n)f (n) = a(n)f (n)
y<n≤x n=[y]+1
k
X
= a(n)f (n)
n=m+1
k
X
= [A(n) − A(n − 1)]f (n)
n=m+1
k
X k
X
= A(n)f (n) − A(n − 1)f (n)
n=m+1 n=m+1
k
X k−1
X
= A(n)f (n) − A(n)f (n + 1)
n=m+1 n=m
k−1
X k−1
X
= A(n)f (n) + A(k)f (k) − A(n)f (n + 1)
n=m+1 n=m
60
k−1
X k−1
X
= A(n)f (n) + A(k)f (k) − A(m)f (m + 1) − A(n)f (n + 1)
n=m+1 n=m+1
k−1
X
= A(n)[f (n) − f (n + 1)] + A(k)f (k) − A(m)f (m + 1)
n=m+1
k−1
X Z n+1
=− A(n) f 0 (t)dt + A(k)f (k) − A(m)f (m + 1)
n=m+1 n
Note that if t ∈ [n, n + 1],we have

X n
X
A(t) = a(n) = a(i) = A(n)
n≤t i=1
. Therefore the above equation can be written as:
X k−1
X Z n+1
a(n)f (n) = − A(t)f 0 (t)dt + A(k)f (k) − A(m)f (m + 1)
y<n≤x n=m+1 n
Z k
=− A(t)f 0 (t) + A(k)f (k) − A(m)f (m + 1) (2.3)
m+1
Now
Z m+1 Z m+1
0
A(t)f (t) = A(m)f 0 (t)dt
y y
= A(m)[f (m + 1) − f (y)]
= A(m)f (m + 1) − A(m)f (y)
= A(m)f (m + 1) − A(y)f (y)
Therefore
Z m+1
A(m)f (m + 1) = A(y)f (y) + A(t)f 0 (t)dt (2.4)
y
Again
Z x Z x
A(t)f 0 (t) = A(k)f 0 (t)dt
k k
= A(k)[f (x) − f (k)]
= A(k)f (x) − A(k)f (k)
= A(x)f (x) − A(k)f (k)
Therefore
Z x
A(k)f (k) = − A(t)f 0 (t)dt + A(x)f (x) (2.5)
k
61
Putting the values A(m)f (m + 1) and A(k)f (k) in equation (2.3), we obtain:
X Z k Z x
a(n)f (n) = − A(t)f 0 (t)dt − A(t)f 0 (t)dt + A(x)f (x) − A(y)f (y)
y<n≤x m+1 k
Z m+1
− A(t)f 0 (t)dt
y
"Z #
m+1 Z k Z x
0 0 0
=− A(t)f (t)dt + A(t)f (t)dt + A(t)f (t)dt
y m+1 k
+ A(x)f (x) − A(y)f (y)

Z x
=− A(t)f 0 (t)dt + A(x)f (x) − A(y)f (y).
y

Corollary 4. (Euler summation formula) If f has a continuous derivative f 0
on the interval [y, x], where 0 < y < x, then
X Z x Z x
f (n) = f (t)dt + (t − [t])f 0 (t)dt + f (x)([x] − x) − f (y)([y] − y).
y<n≤x y y
Proof. Note that when a(n) = 1 for all n ∈ N, then

X X
A(x) = a(n) = 1 = [x].
n≤xx n≤x
Puttting a(n) = 1 for all n ∈ N in the Abel’s formula we obtain:

X Z x
f (n) = [x]f (x) − [y]f (y) − [t]f 0 (t)dt
y<n≤x y
Z x Z x
= [x]f (x) − [y]f (y) − ([t] − t)f 0 (t)dt − tf 0 (t)dt
y y
Z x Z x
= [x]f (x) − [y]f (y) − ([t] − t)f 0 (t)dt − [tf (t)]xy − f (t)dt
y y
Z x Z x
= [x]f (x) − [y]f (y) − ([t] − t)f 0 (t)dt − xf (x) + yf (y) + f (t)dt
y y
Z x Z x
= f (t)dt + ([x] − x)f (x) − ([y] − y)f (y) + (t − [t])f 0 (t)dt.
y y
Alternate proof of Abel’s Identity

Note that every finite sum can be written as a Riemann Stieltjes integral. That
is,
X Z x
a(n)f (n) = f (t)dA(t)
y<n≤x y
62
Integration by parts gives us:

X Z x
a(n)f (n) = [f (t)A(t)]xy − f 0 (t)A(t)dt
y<n≤x y
Z x
= f (x)A(x) − f (y)A(y) − f 0 (t)A(t)dt.
y
Theorem 2.2.2. For x ≥ 2 we have

Z x
π(t)
ϑ(x) = π(x) log x − dt (2.6)
t
Z x 2
ϑ(x) ϑ(t)
π(x) = + 2 dt (2.7)
log x 2 t log t
Proof. Define an arithmetical function a(n) by

(
1 if n is a prime
a(n) =
0 otherwise.
Then we have
X
π(x) = 1
p≤x
X
= a(n) = A(x).
n≤x
and
X X
ϑ(x) = log p = a(n) log n
p≤x n≤x
We have
X Z x
y<n≤x y
Putting f (t) = log t and y = 1 in the above equation, we obtain:

Z x
X 1
a(n) log n = A(x) log x − A(1) log 1 − A(t) dt.
1 t
1<n≤x
This implies that

Z x
π(t)
ϑ(x) = π(x) log x − dt
1 t
63
Z 2 Z x
π(t) π(t)
= π(x) log x − dt − dt
1 t 2 t
Z x
π(t)
= π(x) log x − dt (π(t) = 0 if t < 2)
2 t
Let b(n) = a(n) log n. Then

X X
B(x) = b(n) = a(n) log n
n≤x n≤x
Therefore we have
X
ϑ(x) = log p
p≤x
X
= a(n) log n = B(x)
n≤x
and
X X b(n)
π(x) = a(n) =
log n
n≤x n≤x
1
Putting f (t) = log t and y = 3/2 in the Abel’s identity, we obtain:
x
−1
Z
X b(n) 1 1
= B(x) − B(3/2) − B(t) dt.
log n log x log(3/2) 3/2 (log t)2
3/2<n≤x
This implies that

Z x
ϑ(x) ϑ(3/2) ϑ(t)
π(x) = − + 2 dt.
log x log(3/2) 3/2 t log t
Z 2 Z x
ϑ(x) ϑ(t) ϑ(t)
= + 2 dt + 2 dt
log x 3/2 t log t 2 t log t
Z x
ϑ(x) ϑ(t)
= + 2 dt.
log x 2 t log t
2.3 Some equivalent forms of the prime number

theorem
Theorem 2.3.1. The following relations are logically equivalent
π(x) log x
lim = 1. (1)
x→∞ x
64
ϑ(x)
lim = 1. (2)
x→∞ x
ψ(x)
lim = 1. (3)
x→∞ x
Proof. (1) ⇒ (2):

Assume that
π(x) log x
lim = 1.
x→∞ x
We want to show that
θ(x)
lim = 1.
x→∞ x
From theorem 2.2.2, we have
Z x
π(t)
ϑ(x) = π(x) log x − dt (2.8)
2 t
The above equation can be written as:

Z x
ϑ(x) π(x) log x 1 π(t)
= − dt
x x x 2 t
Therefore, we have
Z x
ϑ(x) π(x) log x 1 π(t)
lim = lim − lim dt
x→∞ x x→∞ x x→∞ x 2 t
1 x π(t)
Z
= 1 − lim dt (2.9)
x→∞ x 2 t
To prove that (1) ⇒ (2), we need only show that
1 x π(t)
Z
lim dt = 0.
x→∞ x 2 t
Note that
π(x) log x
lim = 1.
x→∞ x
The above limit can be written as:

π(x)
x
lim = 1.
x→∞ 1
log x
This implies that

π(t) 1
=O
t log t
65
This implies that

Z ∞ Z ∞
1 π(t) 1 1
dt = O dt
x 2 t x 2 log t
This implies that there exists a constant M > 0 such that
1 ∞ π(t)
Z Z ∞
1 1
dt ≤ M dt (2.10)
x 2 t x 2 log t
Now,
√
Z x Z x Z x
1 1 1
dt = dt + √
dt
2 log t 2 log t x log t
√
Z x Z x
1 1
≤ dt + √ dt
2 log 2 x log x
1 √ 1 √
≤ ( x − 2) + √ (x − x)
log 2 log x
1 √ 1 √
≤ x+ √ (x − x)
log 2 log x
Therefore, we have
Z x √
1 1 1 1
0≤ dt ≤ √ + √ (x − x)
x 2 log t x log 2 x log x
This implies that

Z x
1 1
lim dt = 0.
x→∞ x 2 log t
In the light of equation (2.10), we have
1 x π(t)
Z
lim dt = 0.
x→∞ x 2 t
This completes the proof of (1) ⇒ (2).

(2) ⇒ (1):
Assume that limx→∞ ϑ(x)
x = 1. To prove that
π(x) log x
lim = 1.
x→∞ x
From theorem 2.2.2, we have
Z x
ϑ(x) ϑ(t)
π(x) = + dt
log x 2 t log2 t
66

Z x
π(x) log x θ(x) log x ϑ(t)
= + dt
x x x 2 t log2 t
Therefore
log x x ϑ(t)
Z
π(x) log x θ(x)
lim = lim + lim 2 dt
x→∞ x x→∞ x x→∞ x 2 t log t
Z x
log x ϑ(t)
= 1 + lim 2 dt
x→∞ x 2 t log t
To prove the result, we need only show that
log x x ϑ(t)
Z
lim 2 dt = 0.
x→∞ x 2 t log t
By hypothesis,
ϑ(x)
lim =1
x→∞ x
This implies that
θ(t) = O(t).
Therefore
Z x Z x
log x ϑ(t) log x ϑ(t)dt
dt = O (2.11)
x 2 t log2 t x 2 log2 t
Now
√
Z x Z x Z x
ϑ(t)dt 1 1
= dt + dt
2 log2 t 2 log2 t √
x log2 t
√
Z x Z x
1 1
≤ 2 dt + √ 2 √ dt
2 log 2 x log x
1 √ 1 √
= 2 [ x − 2] + 2 √ (x − x)
log 2 log x
√
x 4 √
≤ 2 + 2 (x − x)
log 2 log x
Therefore
x √
√
Z
log x ϑ(t)dt log x x log x 4
2 ≤ 2 + 2 (x − x)
x 2 log t x log 2 x log x
√
1 log x 4(x − x)
= √ +
log2 2 x x log x
67
This implies that

√
log x x ϑ(t)

(x − x)
Z
log x
lim 2 dt = 0 ∵ lim = 0, lim = 0
x→∞ x 2 t log t x x log x
This completes the proof of (2) ⇒ (1).
(2) ⇒ (3).
Assume that limx→∞ ϑ(x)x = 1. We want to show that limx→∞ ψ(x)
x = 0. By
corollary 3, we have
ψ(x) ϑ(x)
lim − = 0.
n→∞ x x
This implies that limx→∞ ψ(x)
x = 0.
(3) ⇒ (1).
Assume that limx→∞ ψ(x)x = 0. We want to prove that limx→∞ ϑ(x)
x = 0. By
corollary 3, we have
ψ(x) ϑ(x)
lim − = 0.
n→∞ x x
ϑ(x)
This implies that limx→∞ x = 0.
Theorem 2.3.2. Let pn denote the n th prime. Then the following asymptotic
relations are logically equivalent:
π(x) log x
lim = 1. (1)
x→∞ x
π(x) log π(x)
lim = 1. (2)
x→∞ x
pn
lim = 1. (2)
n→∞ n log n
Proof. (1) ⇒ (2):

Asuume that (1) holds. That is
π(x) log x
lim =1
x→∞ x
Taking logarithms we obatin:
lim [log π(x) + log log x − log x] = 0.
x→∞
This implies that

log π(x) log log x
lim log x + − 1 = 0.
x→∞ log x log x
This implies that
" log π(x) log log x
#
log x + log x −1
lim = 0.
x→∞ 1/ log x
68
This implies that

log π(x) log log x
+ − 1 = O(1/ log x)
log x log x
This implies that there exists a constant M > 0 such that

log π(x) log log x 1
+ −1≤M .
log x log x log x
This implies that

log π(x) log log x
lim + − 1 = 0.
x→∞ log x log x
This implies that
log π(x) log log x
lim + lim −1 = 0.
x→∞ log x x→∞ log x
| {z }
=0
This implies that

log π(x)
lim = 1.
x→∞ log x
Now
π(x) log π(x) log π(x) π(x) log x
lim = lim ×
x→∞ x x→∞ log x x
log π(x) π(x) log x
= lim × lim = (1)(1) = 1.
x→∞ log x x→∞ x
(2) ⇒ (3):
Assume that (2) holds. That is
π(x) log π(x)
lim = 1.
x→∞ x
Putting x = pn in the above equation we obtain:
π(pn ) log π(pn )
lim = 1.
n→∞ pn
This implies that
n log n
lim = 1. (∵ π(pn ) = n)
n→∞ pn
This implies that
pn
lim = 1.
n→∞ n log n
69
Thus (2) ⇒ (3).

(3) ⇒ (2) :
pn
lim = 1.
n→∞ n log n
We want to show that
π(x) log x
lim = 1.
x→∞ x
Given x ∈ [2, ∞). Then we can find consecutive prime numbers pn and pn+1
such that
pn ≤ x ≤ pn+1 . (2.12)
Then by the definition of π, we have π(x) = n. Equation (2.12) can be written

as
pn x pn+1
≤ ≤ . (2.13)
n log n n log n n log n
Taking limits as n → ∞, we obtain:
pn x pn+1
lim ≤ lim ≤ lim . (2.14)
n→∞ n log n n→∞ n log n n→∞ n log n
This implies that
x pn+1
1 ≤ lim ≤ lim . (2.15)
n→∞ n log n n→∞ n log n
pn+1
We will show that limn→∞ n log n = 1. Now
pn+1 pn+1 (n + 1) log(n + 1)

= ×
n log n (n + 1) log(n + 1) n log n
Therefore
pn+1 pn+1 (n + 1) log(n + 1)
lim = lim × lim
n→∞ n log n n→∞ (n + 1) log(n + 1) n→∞ n log n
= (1)(1) = 1.
Thus
pn+1
lim = 1.
n→∞ n log n
pn+1
Inserting limn→∞ n log n = 1 in equation (2.13), we obtain:
x
lim = 1.
n→∞ n log n
70
Equivalently,
x
lim = 1.
n→∞ π(x) log π(x)
(2) ⇒ (1) :
π(x) log π(x)

lim =1
x→∞ x
Taking log on both sides, we obtain:
lim log π(x) + log log π(x) − log(x) = 0.

x→∞
Equivalently,

log log π(x) log x
lim log π(x) 1 + − = 0.
x→∞ log π(x) log π(x)
That is
 
log log π(x) log x
1+ log π(x) − log π(x)
lim   = 0.
x→∞ 1/ log π(x)
This implies that

log log π(x) log x
1+ − = O(1/ log π(x))
log π(x) log π(x)
This implies that there is a constant M > 0 such that
log log π(x) log x
1+ − ≤ M (1/ log π(x))
log π(x) log π(x)
This implies that

log log π(x) log x
lim 1+ − = 0.
x→∞ log π(x) log π(x)
This implies that
log log π(x) log x
1 + lim − lim = 0.
x→∞ log π(x) x→∞ log π(x)
| {z }
=0
This implies that

log x
lim = 0.
x→∞ log π(x)
71
Now

π(x) log x log x π(x) log π(x)
lim = lim ×
x→∞ x x→∞ log π(x) x
log x π(x) log π(x)
= lim × lim
x→∞ log π(x) x→∞ x
= (1)(1) = 1.
2.4 Inequalities for π(n) and pn

Lemma 1. For n ≥ 1 we have

2n
2n < < 4n .
n
Proof. Note that
2n
X 2n 2n
4n = (1 + 1)2n = > .
k n
k=0
Next we show that

n 2n
2 <
n
We will prove the inequality by induction on n.
Step 1: Suppose n = 1.
In this case
2 2×1
21 = =
1 1
Thus the result is true when n = 1.
Step 2:Assume that the result is true for n = m. That is

m 2m
2 < .
m
Step 3: To prove that the result is true for n = m + 1. That is we have to
prove that
m 2m + 2
2 <
m
Note that

2(m + 1) 2m + 2 2m + 1 2m + 1
= = +
m+1 m+1 m m+1

2m 2m 2m 2m
= + + +
m m−1 m m+1
72

2m 2m 2m
=2 + +
m m−1 m+1

2m
>2 > 2 × 2m = 2m+1 .
m

1 n n
< π(n) < 6 .
6 log n log n
Proof. By lemma 1, we have

2n
2n < < 4n .
n
That is,
2n!
2n < < 4n .
(n!)2
Taking logarithms we obtain
n log 2 ≤ log(2n)! − 2 log n! < n log 4. (2.16)
But theorem 1.2.8 implies that

Y
[x]! = pα(p) ,
p≤x
P∞ h i
where α(p) = m=1 pxm .
Taking logarithms we obtain:
X
log[x]! = α(p) log p,
p≤n
That is
∞
XX x
log[x]! = log p (2.17)
m=1
pm
p≤x
h i
x
Note that if x < pm , we have x/pm < 1 and hence pm = 0. Now
x < pm ⇒ log x < m log p

log x
< m.
log p
73
This implies that

x log x
=0 if m > .
pm log p
log x
log p ]
X [X x

log[x]! = log p (2.18)
m=1
pm
p≤x
Putting x = n in (2.18) equation, we obtain:

log n
X [X
log p ]
n

log n! = log p
m=1
pm
p≤n
Again putting x = 2n in equation (2.18), we obtain

log 2n
X [Xlog p ]
2n

log 2n! = log p
pm
p≤2n m=1
Hence
log 2n log n
X [Xlog p ]
2n
X [X
log p ]
n

log 2n! − 2 log n! = log p − 2 log p
pm pm
p≤2n m=1 m=1
p≤n
log 2n log n
X [ X ] 2n
log p
X [X
log p ]
n

= log p − 2 log p
pm pm
p≤2n m=1 m=1
p≤n
[ log 2n
log p ]
X X n
−2 log p
p>n m=[ log n ]+1
pm
log p
| {z }
=0
 
log 2n
[ log p ]
X X  2n
n 
= 
 pm − 2 m 
 log p (2.19)
p
p≤2n m=1 | {z }
0 or 1
log 2n
X [X] log p
X log 2n
≤ log p = log p
log p
p≤2n m=1 p≤2n
X log 2n X
≤ log p = log 2n 1 = log(2n)π(2n).
log p
p≤2n p≤2n
Thus
log 2n! − 2 log n! ≤ π(2n) log(2n). (2.20)
74
Consider the leftmost inequality in (2.16):
n log 2 ≤ log(2n)! − 2 log n!. (2.21)
In the light of equation (2.20), we have
n log 2 ≤ π(2n) log(2n).
This implies that

n log 2 2n log 2
π(2n) ≥ = ×
log(2n) log 2n 2
2n 1 1
> × (∵ log 2 > 1/2)
log 2n 2 2
1 2n
=
4 log 2n
1 2n
>
6 log 2n
Thus
1 2n
< π(2n) (2.22)
6 log 2n
Since π is an increasing function, we have
1 2n
π(2n + 1) ≥ π(2n) >
4 log 2n
1 2n 2n + 1
> ×
4 log(2n + 1) log(2n + 1)
1 2 2n + 1 1 2n + 1
> =
4 3 log(2n + 1) 6 log(2n + 1)
That is
1 2n + 1
< π(2n + 1). (2.23)
6 log(2n + 1)
From equations (2.22) and (2.23), we have:

1 n
< π(n) for all n ≥ 2. (2.24)
6 log n
This proves the leftmost inequality in the theorem. To prove the other inequality
we consider equation (2.19):
 
log 2n
[ log p ]
X X  2n n 
log 2n! − 2 log n! =  pm − 2 pm  log p
 
p≤2n m=1 | {z }
0 or 1
75
log 2n
X 2n
n X [X log p ]
2n

n
= −2 log p + −2 log p
p p p p
p≤2n p≤2n m=2
X 2n
n
≥ −2 log p
p p
p≤2n
 
X  2n
n 
≥  p −2 p  log p
 
n<p≤2n | {z } |{z}
=1 =0
X
= log p
n<p≤2n
X X
= log p − log p
p≤2n p≤n
= ϑ(2n) − ϑ(n).
Hence
ϑ(2n) − ϑ(n) ≤ log(2n)! − log(n)! < n log 4. (by Eqn(2.16))
Taking n = 2r in the above inequality we obtain:
ϑ(2r+1 ) − ϑ(2r ) < log(2n)! − log(n)! < 2r log 4 = 2r+1 log 2.
Summing the above inequality from r = 0 to r = k, we obtain:

k
X k
r+1 X
ϑ(2 ) − ϑ(2r ) < 2r+1 log 2.
r=0 r=0
The above inequality can be written as:
ϑ(2r+1 ) < (2k+1 − 1) log 4 < 2k+1 log 2.
Hence
ϑ(2r+1 ) < 2k+1 log 2. (2.25)
Now we choose k so that 2k ≤ n ≤ 2k+1 . Since ϑ is an increasing function we

have
ϑ(n) ≤ ϑ(2r+1 ) < 2k+2 log 2 ≤ 4n log 2. ∵ 2k ≤ n

(2.26)
But if 0 < α < 1, we have

 
X
(π(n) − π(nα )) =  1 log nα
nα <p≤n
76
X
< log p (∵ log nα < log p)
nα <p≤n
≤ ϑ(n) < 4n log 2 (by Eqn (2.26))
Hence
4n log 2
π(n) < + π(nα )
α log n
4n log 2
< + nα
α log n

n 4 log 2 log n
= + 1−α (2.27)
log n α n
Consider the function
f (x) = x−c log x, c > 0, x ≥ 1
Note that the function f attains its maximum at x = e1/c and the maximum
value of f is 1/ce.
Putting α = 2/3 in equation (2.27), we obtain:

n 4 log 2 log n
π(n) < + 1/3
log n 2/3 n
n −1/3
= [6 log 2 + n log n]
log n
n
≤ [6 log 2 + 3/e] < 6n/ log n
log n
Theorem 2.4.2. For n ≥ 1 the nth prime pn satisfies the inequalities

1 12
n log n < pn < 12 n log n + n log
6 e
Proof. For k ≥ 2 we have(Theorem 2.4.1)
1 k k
< π(k) < 6 . (2.28)
6 log k log k
If k = pn then k ≥ 2 and n = π(k). Putting k = pn in the righttmost inequality

in 2.28 we obtain:
pn
n = π(pn ) < 6
log pn
This implies that
pn < 6n log pn (2.29)
77
Again consider the leftmost inequality in (2.4.1):

1 k
< π(k)
6 log k
Putting k = pn in the above inequality we obtain
1 pn
< π(pn ) = n
6 log pn
This implies that
pn < 6n log pn
pn 6n log pn
√ < √ (2.30)
pn pn
Now consider the function
log x
f (x) = √
x
Then we have
2 − log x
f 0 (x) =
2x3/2
Also f 0 (x) = 0 implies that x = e2 . One can easily verify that f 00 (e2 ) < 0.
Hence the function f has a maximum at x = e2 and the maximum value is
given by 2/e. Hence equation (2.31) can be written as:

√ 6n log pn 2 12n
pn < √ < 6n = . (2.31)
pn e e
Taking logarithms we obtain:

1 12
log pn < log + log n
2 e
This implies that

12
log pn < 2 log + 2 log n (2.32)
e
Hence equations (2.29) and (2.32) we have:

12
pn < 6n 2 log + 2 log n
e

12
= 12 n log n + n log
e
78
P∞ 1
Corollary 5. The series k=1 pn is divergent.
Proof. From the leftmost inequality in theorem 2.4.2 we have

1
n log n < pn .
6
This implies that
1 6
<
pn n log n
P∞ 1
Note that the series n=2 n log n is a divergent series. So by comparison test,
P∞ 1
the series n=1 pn is divergent.
2.5 Shapiro’s Tauberian theorem

Theorem 2.5.1. Let {an } be a sequence such that
X hxi
a(n) = x log x + O(x) for all x ≥ 1.
n
n≤x
Then
(a) For x ≥ 1 we have

X a(n)
= log x + O(1).
n
n≤x
(b) There is a constant B > 0 such that

X
a(n) ≤ Bx for all x ≥ 1.
n≤x
(c) There is a constant A > 0 and an x0 > 0 such that

X
a(n) ≥ Ax for all x ≥ x0 .
n≤x
Proof. Let
X
S(x) = a(n),
n≤x
X hx i
T (x) = a(n) .
n
n≤x
79
It is given that
X hxi
a(n) = x log x + O(x) for all x ≥ 1.
n
n≤x
That is,
T (x) = x log x + O(x) for all x ≥ 1. (2.33)
First we prove (b). For this first we establish the following inequality:
x x
S(x) − S ≤ T (x) − T .
2 2
Now
x X hxi hxi
X
T (x) − 2T = a(n) −2 a(n)
2 n 2n
n≤x n≤x/2
X hxi X hxi X hxi
= a(n) + a(n) −2 a(n)
n n 2n
n≤x/2 x/2<n≤x n≤x/2
 
X h x i h x i X hxi
= a(n) 
 n − 2 + a(n)
{z 2n } n

n≤x/2 | x/2n≤x
0 or 1
X hxi
≥ a(n)
n
x/2<n≤x
X hxi X hxi
= a(n) − a(n)
n n
n≤x n≤x/2
x
= S(x) − S
2
Hence
x x
S(x) − S ≤ T (x) − T
2 2
Now
x hx
x x i
T (x) − T = [x log x + O(x)] − 2 +Olog
2 2 2 2
x
= x log x + O(x) − x log x + x log 2 + O
2
= x log 2 + O(x) = O(x) + O(x) = O(x).
Hence
x
T (x) − T = O(x)
2
80
This implies that

x
S(x) − S ≤ O(x).
2
This means that there is some constant K > 0 such that
x
S(x) − S ≤ Kx for all x ≥ 1.
2
Replace x successively by x/2, x/4, . . . to get
x x x
S −S ≤K ,
x2 4
x 2
x
S −S ≤K ,
4 8 4
..
.
Adding the above inequalities we get
S(x) ≤ K x + (x/2) + (x/22 ) + · · · = |{z}

2K x
B
That is
S(x) ≤ Bx
That is,
X
a(n) ≤ Bx.
n≤x
This completes the proof of (b). Next we prove (a). Note that
x hxi
− <1
n n
This implies that
hxi x
− = O(1).
n n
Hence
X hxi
T (x) = a(n)
n
n≤x
X x

= + O(1) a(n)
n
n≤x
 
X a(n) X
=x +O a(n)
n
n≤x n≤x
81
X a(n)
=x + O(S(x))
n
n≤x
X a(n)
=x + O(x) (∵ S(x) ≤ Bx)
n
n≤x
Hence
X a(n) 1
= T (x) + O(1)
n x
n≤x
1
= [x log x + O(x)]
x
= log x + O(1).
This completes the proof of (a).
Finally, we prove (c). Let
X a(n)
A(x) =
n
n≤x
Then by part (a) of the theorem, we have

A(x) = log x + O(1)
Equivalently,
A(x) = log x + R(x),
where R(x) = O(1). Since R(x) = O(1) we have |R(x)| ≤ M for some M > 0.
Choose α such that 0 < α < 1. Consider
X a(n) X a(n)
A(x) − A(αx) = −
n n
n≤x n≤αx
= log x + R(x) − [log αx + R(αx)] (if x ≥ 1 and αx ≥ 1)

= log x + R(x) − log α − log x − R(αx)
= − log α + R(x) − R(αx)
≥ − log α − |R(x)| − |R(αx)|
≥ − log α − 2M
Now choose α so that − log α − 2M = 1. This implies that α = e−(2M +1) .

Obviously 0 < α < 1. For this α we have the inequality
1
A(x) − A(αx) ≥ 1 if x ≥ .
α
But
X a(n) X a(n) X a(n)
A(x) − A(αx) = − =
n n n
n≤x n≤αx αx<n≤x
82
X a(n) 1 X 1
≤ < a(n) = S(x).
n αx αx
n≤x n≤x
Thus
S(x) 1
> A(x) − A(αx) ≥ 1 if x ≥ .
αx α
Thus
1
S(x) ≥ αx if x ≥ .
α
Taking A = α and x0 = 1/α we have
S(x) ≥ Ax, x > x0 .
2.6 Applications of Shapiro’s theorem

X Λ(n)
= log x + O(1).
n
n≤x
Also, there exist positive constants c1 and c2 such that
ψ(x) ≥ c1 x for all x ≥ 1
and
ψ(x) ≥ c2 x for all sufficiently large x.
Proof. We have
X hxi
Λ(n) = log[x]!
n
n≤x
= x log x − x + O(log x)
= x log x + O(x) + O(log x)
= x log x + O(x).
Hence X hxi
Λ(n) = x log x + O(x)
| {z } n
n≤x
a(n)
So by Shapiros theorem we have

X Λ(n)
(a) = log x + O(1).
n
n≤x
83
Also there exists c1 > 0 and c2 > 0 such that

X
(b) Λ(n) ≤ c1 x for x ≥ 1.
n≤x
| {z }
ψ(x)
X
(c) Λ(n) ≥ c2 x for sufficiently large x.
n≤x
| {z }
ψ(x)

X log p
= log x + O(1).
p
p≤x
Also there exist positive constants c1 and c2 such that
ϑ(x) ≤ c1 x for all x ≥ 1.
and
ϑ(x) ≥ c2 x for sufficiently large x.
Proof. We know that for x ≥ 2,

X x
log p = x log x + O(x). (2.34)
p
p≤x
Define
(
log p if n is a prime
Λ1 (n) =
0 otherwise
Then equation (2.34) can be written as:

X hxi
Λ1 (n) = x log x + O(x).
| {z } n
n≤x
a(n)≥0
So by Shapiros theorem, we have the following

X Λ1 (n)
(a) = log x + O(1)
n
n≤x
This implies that

X log p
= log x + O(1).
p
p≤x
84
X
(b) Λ1 (n) ≤ C1 x, for all x ≥ 1.
n≤x
This implies that

X
log p ≤ c1 x
p≤x
| {z }
ϑ(x)
for all x ≥ 1.
X
(c) Λ1 (n) ≥ C2 x,
n≤x
for sufficiently large x. This implies that
ϑ(x) ≥ c2 x,
for sufficiently large x.

X x
ψ = x log x − x + O(log x)
n
n≤x
and
X x
ϑ = x log x + O(x).
n
n≤x
Proof. Note that if f (n) is an arithmetical function, then

X hxi X x
f (n) = F ,
n n
n≤x n≤x
P
where F (x) = n≤x f (n).
We have X
ψ(x) = Λ(n).
n≤x
Then
X hxi X x
Λ(n) = ψ (2.35)
n n
n≤x n≤x
We have already proved that

X hxi
Λ(n) = x log x − x + O(log x) (2.36)
n
n≤x
85
In the light of equations (2.35) and (2.36) we have

X x
ψ = x log x − x + O(log x)
n
n≤x
Again we have
X X
ϑ(x) = log p = Λ1 (n).
n≤x n≤x
Then
X hxi X x
Λ1 (n) = ϑ
n n
n≤x n≤x
But
X hxi
Λ1 (n) = x log x + O(x).
n
n≤x
Hence
X x
ϑ = x log x + O(x).
n
n≤x
2.7 An asymptotic formula for the partial sums
Theorem 2.7.1. There is a constant A such that

X1
1
= log log x + A + O for all x ≥ 2.
p log x
p≤x
Proof. Let
X log p
A(x) =
p
n≤x
and let
(
1 if n is prime,
a(n) =
0 otherwise
Then
X1 X a(n) X a(n)
= and A(x) = log n.
p n n
n≤x n≤x n≤x
86
Recall Abel’s identity:

X Z x
y<n≤x y
Putting y = 3/2, f (t) = log1 t and an = a(n)

n log n in the Abel’s identity, we
obtain:
X a(n) Z x
1 1 1 A(t)
log n = A(x) − A(3/2) + 2 dt.
n log n log x log(3/2) 3/2 log t
t
3/2<n≤x
That is
Z x
X a(n) A(x) A(t)
= + dt.
3/2<n≤x
n log x 3/2 t log2 t
Thta is
X1 Z 2 Z x
A(x) A(t) A(t)
= + 2 dt + 2 dt
p log x 3/2 t log t 2 t log t
p≤x
Z x
A(x) A(t)
= + 2 dt (2.37)
log x 2 t log t
Now
X log n
A(x) = a(n)
n
n≤x
X log p
=
p
p≤x
= log x + O(1)
Inserting the value of A(x) in equation (2.37) we obtain:
X1 Z x
log x + O(1) log t + R(t)
= + dt, R(t) = O(1).
p≤x
p log x 2 t log2 t
Z x Z x
1 1 R(t)
=1+O + dt + 2 dt (2.38)
log x 2 t log t 2 t log t
Now
Z x Z x
1 1/t
dt = dt
2 t log t 2 log t
x
= (log log t)2 = log(log x) − log(log 2).
Also
Z x Z ∞ Z ∞
R(t) R(t) R(t)
dt = dt − dt (2.39)
2 t log2 t 2 t log2 t x t log2 t
87
Note that
R(t) = O(1)
Therefore
Z ∞ Z ∞ Z ∞
R(t) 1 d −1 1
dt = O dt = O = O
x t log2 t x t log2 t x dt log t log x

Z ∞ Z ∞
1 R(t) 1
dt = dt +O
x t log t t log2 t log x
|2 {z }
=B

1
=B+O .
log x
Hence equation (2.38) becomes:

X1
1 1
=1+O + + log(log x) − log(log 2) + B + O .
p log x log x
p≤x

1
= log log x + O + A.
log x
2.8 The partial sums of the Mobius function

Definition 2.8.1. If x ≥ 1 we define
X
M (x) = µ(n).
n≤x
The exact bounds of M (x) is not known. Numerical evidence suggests that
√
|M (x)| < x, if x > 1.
Definition 2.8.2. If x ≥ 1 we define

X
H(x) = µ(n) log n.
n≤x
Theorem 2.8.1. We have

M (x) H(x)
lim − = 0.
x→∞ x x log x
88
Proof. Recall Abel’s identity:

X Z x
y<n≤x y
Putting y = 1, a(n) = µ(n) and f (t) = log t in the Abel’s identity we obtain
Z x
X 1
µ(n) log(n) = M (x) log x − M (1) log 1 − M (t) dt.
y t
1<n≤x
That is,
Z x
X M (t)
µ(n) log(n) = M (x) log x − dt.
1 t
1<n≤x
| {z }
H(x)
That is
Z x
M (t)
H(x) = M (x) log x − dt.
1 t
Dividing both sides by x log x we obtain:
Z x
H(x) M (x) 1 M (t)
= − dt.
x log x x x log x 1 t
That is
Z x
M (x) H(x) 1 M (t)
− = dt
x x log x x log x 1 t
Therefore
Z x
M (x) H(x) 1 M (t)
lim − = lim dt.
x→∞ x x log x x→∞ x log x 1 t
R x M (t)
1
Claim: limx→∞ x log x 1 t dt = 0. But we have the trivial estimate M (x) =
O(x) so
Z x Z x
M (t)
dt = O dt = O(x),
1 t 1
Therefore
Z x
1
M (t) 1
dt = O(x)
1x log x
t x log x

1
=O
log x
R x M (t)
1
This implies that limx→∞ x log x 1 t dt = 0.
89
Theorem 2.8.2. The prime number theorem implies that
M (x)
lim = 0.
x→∞ x
Proof. By theorem 2.8.1 we have

M (x) H(x)
lim − = 0.
x→∞ x x log x
To prove the theorem it suffices to show that
H(x)
lim = 0.
x→∞ x log x
By definition of H(x) we have

X
−H(x) = − µ(n) log n
n≤x
x
P P
Claim:− n≤x µ(n) log n = n≤x µ(n)ψ n . We have
X
log n = Λ(d).
d|n
So by Mobius inversion formula, we have

X n
Λ(n) = µ(d) log
d
d|n
X X
= µ(d) log n − µ(d) log(d)
d|n d|n
X
1
= log n − µ(d) log d
n
d|n
X
=− µ(d) log d.
d|n
Hence
X
Λ(n) = − µ(d) log d
d|n
Again by Mobius inversion formula,

X n
−µ(n) log n = µ(d)Λ
d
d|n
90
Hence
X XX n
− µ(n) log n = µ(d)Λ
d
n≤x n≤x d|n
X
= (µ ∗ Λ)(n)
n≤x
X X
= µ(n) Λ(n)
n≤x n≤x/n
X x
= µ(n)ψ
n
n≤x
Hence
X X x
−H(x) = − µ(n) log n = µ(n)ψ
n
n≤x n≤x
We know that
π(x) log x ψ(x)
lim = 1 ⇔ lim = 1.
x→∞ x x→∞ x
ψ(x)
Hence prime number theorem is equivalent to ψ(x) ∼ x. Since limx→∞ x = 1,
given > 0 there exists A > 0 such that

ψ(x)

x − 1 < for all x ≥ A.

This implies that
|ψ(x) − x| < for all x ≥ A. (2.40)

x
Choose x ≥ A. Let y = A . We have
X x
−H(x) = µ(n)ψ
| {z } n
n≤x
I
X x X x
= µ(n)ψ + µ(n)ψ (2.41)
n n
n≤y y<n≤x
| {z } | {z }
I1 I2
Now
hxi x
n≤y⇒n≤ <
A A
x
⇒A≤ .
n
Hence from equation (2.40) we have
x x x
− < for all n ≤ y.

n n n
ψ
91
Consider
X x
I1 = µ(n)ψ
n
n≤y
X hx
xi x
= µ(n) +ψ −
n n n
n≤y
X µ(n) X h x xi
=x + µ(n) ψ −
n n n
n≤y n≤y
Therefore

X µ(n) X x x
|I1 | ≤ y + |µ(n)| −

n ψ
n n
n≤y n≤y
Xx
≤ x(1) +
n
n≤y
X1 Z y
1
=x+x < x + x dx
n 1 x
n≤y
= x + x(log y − 1) < x + x(log y + 1)

< x + x(log x + 1)
Consider the integral

X x
I2 = µ(n)ψ
n
y<n≤x
Now
hxi
y<n≤x⇒ < n ≤ [x]
hA
xi
⇒ +1≤n
A
⇒y+1≤n
1 1
⇒ ≥
y+1 n
x x
⇒ ≥ (2.42)
y+1 n
Also
x A 1
y≤ <y+1⇒ > (2.43)
A x y+1
From equations (2.42) and (2.43), we have

x x
≤ < A.
n y+1
92
Since ψ is an increasing function, we have

x
ψ < ψ(A).
n
Therefore

x
X
|I2 | < µ(n)ψ
y<n≤x n
< xψ(A).
Hence
|H(x)| ≤ |I1 | + |I2 |

< x + x(log x + 1) + xψ(A) if x ≥ A
= x log x + (2 + ψ(A))x.
Dividing throughout by x log x, we obtain:
|H(x)| 2 + ψ(A)
<+ for all x ≥ A. (2.44)
x log x log x
Note that
2 + ψ(A)
lim = 0.
x→∞ x log x
This implies that there exists B > 0 such that
2 + ψ(A)
< for all x ≥ B. (2.45)
x log x
Hence
|H(x)|
< + for all x ≥ max{A, B}. (2.46)
x log x
This implies that

H(x)
lim = 0.
x→∞ x log x
Definition 2.8.3. We say that f is little oh of g as x → ∞( f = o(g) as

x → ∞) if
f (x)
lim = 0.
x→∞ g(x)
Remark 11. lim Mx(x) = 0 implies that M (x) = o(x) as x → ∞.
93
Theorem 2.8.3. The relation
M (x) = o(x)asx → ∞
implies that ψ(x) ∼ x as x → ∞.

Proof. Consider
[x] − ψ(x) − 2C,
where C is the Euler constant. now
X X X 1
[x] − ψ(x) − 2C = 1− Λ(n) − 2c
n
n≤x n≤x n≤x
X
1
= 1 − Λ(n) − 2c (2.47)
n
n≤x
We have
X
σ0 (n) = 1
n≤x
Therefore by Mobius inversion formula

X n
1= µ(d)σ0 (2.48)
d
d|n
Also we have
X n
Λ(n) = µ(d) log (2.49)
d
d|n
X
1
= µ(d) (2.50)
n
d|n
Using equations (2.48), (2.49) and (2.50) in equation (2.47) we obtain

XX n n
[x] − ψ(x) − 2C = {µ(d)σ0 − µ(d) log − 2Cµ(d)}
d d
n≤x d|n
XX n n
= µ(d){σ0 − log − 2C}
d d
n≤x d|n
X n n
= µ(d) {σ0 − log − 2C}
d {z d
q,d | }
qd≤x f (q)
X
= µ(d)f (q). (2.51)
q,d
qd≤x
94
P
Claim: µ(d)f (q) = o(x) as x → ∞.
q,d
qd≤x P
By theorem 1.2.11 we can write q,d µ(d)f (q) in the following form:
qd≤x
X X x X x
µ(d)f (q) = µ(n)F + f (n)M − F (a)M (b),
n n | {z }
q,d n≤b n≤a
qd≤x | {z } | {z } I3
| {z } I1 I2
I
where a and b are any positive numbers such that ab = x and

X
F (x) = f (n)
n≤x
Now
X
F (x) = f (n)
n≤x
X
= [σ0 (n) − log n − 2C]
n≤x
X X X
= σ0 (n) − log n − 2C 1
n≤x n≤x n≤x
√
= [x log x + (2C − 1)x + O( x)] − [x log x − x + O(log x)] − 2C[x + O(1)]
√
= O( x) + O(log x) + O(1)
√ √ √
= O( x) + O( x) + O( x)
√
= O( x).
Hence √
F (x) = O( x).
This implies that there exists B > 0 such that
√
|F (x)| ≤ B x for x ≥ 1. (2.52)
Now

X x X x

|I1 | = µ(n)F ≤
n n
F
n≤b n≤x
√ X 1
r
X x
≤B =B x √
n n
n≤b n≤x
√
Z b
1 √ √ √
≤B x √ dt = 2B[ b − 1] < 2B x b
1 t
√ √ Ax A
= A x b = √ < x if √ <
a a
95
Hence
|I1 | < x.
By hypothesis
M (x)
lim = 0.
x→∞ x
Then corresponding to the same > 0 there exits c > 0 such that
|M (x)|
≤ for x ≥ c,
x K
P |f (n)|
where K = n≤x n . Now
X x
|I2 | = | f (n)M |
n
n≤a
X
≤ |f (n)||M (x/n)|
n≤x
X x x
≤ |f (n)| for ≥c
n K n
n≤x
x
= for ≥ c.
n
Hence
|I2 | < for x ≥ ac.
Now
|I3 | = |F (a)||M (b)|

√ √ √ √
≤ B a b (∵ |F (x)| ≤ B x, |M (x)| < x)
√ √ √ √ √ √
≤ 2B a b = A ab < A ab < ( a) ab = ab = x
Hence
|I3 | < x
Thus
|I| < 3x.
provided x > a and x > ac.

Theorem 2.8.4. If
X µ(n)
A(x) =
n
n≤x
96
the relation
A(x) = o(1) as x → ∞
implies prime number theorem. In other words, the prime number theorem is a
consequence of the statement that the series
∞
X µ(n)
n=1
n
converges and has sum 0.

Proof. It suffices to show that
M (x)
lim = 0.
x→∞ x
Recall the Abel’s identity:
X Z x
y<n≤x y
µ(n)
Taking y = 1, f (t) = t and a(n) = in Abel’s identity we obtain:
n
X µ(n) Z x
n = A(x)x − A(t)dt
n 1
1<n≤x
That is,
X Z x
µ(n) = A(x)x − A(t)dt
n≤x 1
That is,
Z x
M (x) = A(x)x − A(t)dt
1
Therefore
Z x
M (x) 1
= A(x) − A(t)dt (2.53)
x x 1
Since A(x) = o(1),
lim A(x) = 0. (2.54)

x→∞
Rx
Claim: limx→∞ x1 1 A(t)dt = 0. Since limx→∞ A(x) = 0, given > 0 there
exists c > 0 such that
|A(x)| < ∀x ≥ c.
97
Now
Z x Z c
1 x
Z
1 1

x A(t)dt=
x A(t)dt + A(t)dt
1 1 x c
Z c Z x
1 1
≤ A(t)dt + A(t)dt
x 1 x c
Z c
1 x
Z
1
= A(t)dt + A(t)dt
x 1 x c
1 c 1 x
Z Z
≤ |A(t)| dt + |A(t)| dt
x 1 | {z } x c | {z }
<1 <
1 c 1
< (c − 1) + (x − c) = − (1 + c + )
x x x x
Therefore
Z
1 x

lim A(t)dt = 0.
x→∞ x 1
Hence from equation (2.53) it follows that
M (x)
lim = 0.
x→∞ x
98
Module 3
Quadratic Residues
In first section of this chapter, we will introduce quadratic residues of an odd

prime p. In the second section we will discuss Euler’s criterion, which specifies
when an integer is a quadratic residue modulo p. Whether an integer n is a
quadratic residue of p is indicated by a symbol called Legendre’s symbol. We
will also discuss properties of Legendre Symbol. In the fourth chapter we prove
quadratic reciprocity law. In the last section we discuss a generalization of
Legendre symbol viz Jacobi symbol.
3.1 Definition and Examples

Definition 3.1.1. Let p be an odd prime and (n, p) = 1 . We say that n is a
quadratic residue modulo p if the equation
x2 ≡ n(mod p)
has a solution. If n is not a quadratic residue of p, we call it a quadratic non-

residue modulo p.
Remark 12. Let x be a solution of x2 ≡ n(mod p). Then −x is also a solution

of x2 ≡ n(mod p).
Note that
(−x)2 = (p − x)2 = p2 − 2px + x2 ≡ x2 (mod p)
Example 3. Find the quadratic residues modulo 11.
12 ≡ 1(mod 11) 102 ≡ (−1)2 ≡ 3(mod 11)

22 ≡ 4(mod 11) 92 ≡ (−2)2 ≡ 4(mod 11)
32 ≡ 9(mod 11) 82 ≡ (−3)2 ≡ 9(mod 11)
42 ≡ 5(mod 11) 72 ≡ (−4)2 ≡ 5(mod 11)
52 ≡ 3(mod 11) 62 ≡ (−5)2 ≡ 3(mod 11)
99
Hence the quadratic residues mod 11 are 1, 3, 4, 5, 9, and the non residues
are 2, 6, 7, 8, 10.
Remark 13. Since (−x)2 ≡ x2 (mod p), to find the quadratic residues mod p
we need only consider x2 , 0 < x < [p/2].
Remark 14. The nonzero residues (mod p) can be listed as:
Z×
p = {1, 2, 3, . . . , p − 1}.
Note that Z× ×
p is a group under multiplication. The elements of Zp can be written
as:
Z×
p = {1, 2, 3, . . . , (p − 1)/2, (p + 1)/2, . . . , p − 1}
| {z } | {z }
<p/2 >p/2
= {1, 2, 3, . . . , (p − 1)/2, p − (p − 1)/2, p − (p − 3)/2 . . . , p − 1}

= {j, p − j : j = 1, 2, 3, . . . , (p − 1)/2}
Moreover, p − j ≡ −j(mod p). Thus,

p−1
Z×
p = ±j : 1 ≤ j ≤
2

p−1 p−2 p−1
.= − ,− , . . . , −1, 1, 2, . . . ,
2 2 2
Hence
p−1
|aj | : aj ∈ Z×

p = 1, 2, 3, . . . ,
2
Observe that
p+1 p−1
≤x≤p−1⇔1≤p−x≤
2 2
Theorem 3.1.1. If a and b are quadratic residues mod p, then so is ab.
Proof. Suppose
a ≡ r2 (mod p), b ≡ r2 (mod p)
Then
ab ≡ (rs)2 (mod p).
Theorem 3.1.2. Let p be a prime. Then every reduced residue system mod p:
Z×
p = {1, 2, . . . , (p − 1)}
contains exactly p−1 p−1

2 quadratic residues and exactly 2 quadratic nonresidues.
Moreover the quadratic residues are
( 2 )
p − 1
12 , 22 , 32 , . . . ,
2
100
Proof. Note that Z×p is a field. Let Q denote the set of quadratic residues mod
p. Define a map ϕ : Z× ×
p → Zp by
ϕ(r) = r2 (mod p)
Then
ker(ϕ) = {x ∈ Z×
p : ϕ(x) = 1}
= {x ∈ Z× 2
p : r ≡ 1(mod p)}
= {1, −1}.
Also
im(ϕ) = {r ∈ Z×
p : ϕ(x) = r}
= {r : r ≡ x2 (mod p)}
= Q.
By the first isomorphism theorem of group theory,
|ker(ϕ)||im(ϕ)| = |Z×
p|
That is
p−1
2|Q| =
2
Threfore |Q| = p−1
2 . This implies that the number of quadratic nonresidues in
p−1
Zp is also 2 . Since Zp is a group, x (mod p) ∈ Z×
× × 2
p for x = 1, 2, · · · , (p − 1)/2.
Next, we will show that 12 , 22 , 32 , . . . , ((p − 1)/2)2 are all distinct.
Assume that
p−1
x2 ≡ y 2 (mod p), 1 ≤ x, y ≤
2
⇒(x − y)(x + y) ≡ (mod p)
⇒(x − y) ≡ (mod p)(∵ 1 < x + y < p)
⇒x ≡ y(mod p).
3.2 Legendre’s symbol and its properties

The Legendre symbol is a convenient notation to indicate whether an integer
n is a quadratic residue modulo of an odd prime p. The Legendre symbol of n
modulo p is denoted by (a | p).
Definition 3.2.1. Let p be an odd prime. We define Legendre’s symbol (n|p)
as follows:

 0 if p divides n;

(n | p) = 1 if n is a quadratic residue modulo p;

−1 if n is a quadratic nonresidue modulo p.

101
Theorem 3.2.1. (Euler’s criterion)Let p be an odd prime. Then for all n

we have
(n | p) ≡ n(p−1)/2 (mod p).
Proof.
Case 1: Suppose n is a multiple of p. In this case (n|p) = 0 and n(p−1)/2 ≡
0(mod p). Hence the result is obvious if n is a multiple of p.
Case 2: Assume that (n | p) = 1.
Since n is a quadratic residue modulo p, we have
n ≡ m2 (mod p), for some integer m
This implies that
n(p−1)/2 ≡ mp−1 (mod p)

≡ 1(mod p) (By Fermat’s Theorem)
Hence
n(p−1)/2 ≡ (n|p)(mod p)
Case 3: Suppose that (n | p) = −1.

In this case n is not a quadratic residue mod p. Consider the polynomial
f (x) ≡ x(p−1)/2 − 1(mod p). (3.1)

2
Claim 1:12 , 22 , . . . , p−1
2 are roots of equation(3.1)
For x = 1, 2, . . . , (p − 1)/2, we have
(p−1)/2
x2 = xp−1 ≡ 1(mod p) (by Fermats theorem)
Thus every quadratic residues in Z× p is a root of equation(3.1). Since

× (p−1)/2
Zp is a field, the equation f (x) ≡ x − 1(mod p) has at most (p −
1)/2 solutions. Note that Z× p contains (p − 1)/2 quadratic residues, viz,
12 , 22 , . . . , ((p−1)/2)2 . Hence the equation f (x) = x(p−1)/2 −1 has exactly
(p − 1)/2 solutions in Z× p . So quadratic nonresidue is not a root of the
(p−1)/2
polynomial f (x) = x − 1.
Claim 2: x(p−1)/2 ≡ ±1(mod p).
By Fermats theorem we have
xp−1 ≡ 1(mod p)

x(p−1)/2 − 1 x(p−1)/2 + 1 ≡ 0(mod p)
This implies that

x(p−1)/2 ≡ ±1(mod p).
102
If n is a quadratic residue modulo p, then by claim 1, n(p−1)/2 ≡ 1(mod p).

Therefore, if n is not a quadratic residue modulo p, we must have x(p−1)/2 ≡
−1(mod p). Hence
n(p−1)/2 ≡ (n | p)(mod p)

Theorem 3.2.2. Legendre’s symbol (n|p) is a completely multiplicative func-
tion.
Proof. Let f (n) = (n | p), for all n ∈ N.
Case 1: Suppose p | m or p | n.
Then p | mn. In this case (m | p)(n | p) = 0 and (mn | p) = 0. Hence
(mn | p) = (m | p)(n | p).
Case 2: Suppose p - m and p - n.

Note that
(mn|p) ≡ (mn)(p−1)/2 (mod p)(by Eulers criteria)

≡ m(p−1)/2 n(p−1)/2 (mod p)
≡ (m | p)(n | p)(mod p)(by Eulers criteria)
Therefore
(mn | p) − (m | p)(n | p) ≡ 0(mod p). (3.2)
Note that (m | p), (n | p), (mn | p) ∈ {0, 1, −1}. So
(mn | p) − (m | p)(n | p) ∈ {0, 2, −2}.
Since p is an odd prime and (mn | p) − (m | p)(n | p) is divislble by p(see

(3.2)) it follows that (mn | p) − (m | p)(n | p) = 0. Hence
(mn | p) = (m|p)(n | p).
That is
f (mn) = f (m)f (n) for all m, n ∈ N.
3.3 Evalution of (−1 | p) and (2 | p)

Theorem 3.3.1. For every odd prime p we have
(
1 if p ≡ 1(mod 4)
(−1 | p) = (−1)(p−1)/2 =
−1 if p ≡ 3(mod 4)
103
Proof. By Euler’s criterion,
(−1 | p) ≡ (−1)(p−1)/2 (mod p)
Equivalently,
(−1 | p) − (−1)(p−1)/2 ≡ 0(mod p)
Note that (−1 | p) ∈ {1, −1} and (−1)(p−1)/2 ∈ {1, −1}. So
(−1 | p) − (−1)(p−1)/2 ∈ {0, 2, −2}
Since p is an odd prime, we have (−1 | p) − (−1)(p−1)/2 = 0. Equivalently
(−1 | p) = (−1)(p−1)/2
Now
(−1 | p) = 1 ⇔ (−1)(p−1)/2 = 1
⇔ (p − 1)/2 = 2m for some m ∈ Z
⇔ p = 1 + 4m
⇔ p ≡ 1(mod 4)
Also,
(−1 | p) = −1 ⇔ (−1)(p−1)/2 = −1
⇔ (p − 1)/2 = 2m + 1 for some m ∈ Z
⇔ p = 3 + 4m
⇔ p ≡ 3(mod 4)

Theorem 3.3.2. For every odd prime p we have
(
2 1 if p ≡ ±1(mod 8)
(2 | p) = (−1)(p −1)/8 =
−1 if p ≡ ±3(mod 8)
Proof. Consider the following (p − 1)/2 congruences:
p − 1 ≡1(−1)1 (mod p)
2 ≡2(−1)2 (mod p)
p − 3 ≡3(−1)3 (mod p)
4 ≡4(−1)4 (mod p)
..
.
p−1
r≡ (−1)(p−1)/2 (mod p)
2
104
where r = p + 1/2 or (p − 1)/2.

Multiply the above congruences, we obtain:

p−1
2 · 4 · 6 · · · (p − 1) ≡ !(−1)1+2+···+(p−1)/2 (mod p)
2
This gives us:

p−1 p−1 2
2 (p−1)/2
!≡ !(−1)(p −1)/8 (mod p)
2 2
This implies that
2
−1)/8
2(p−1)/2 ≡ (−1)(p (mod p)
By Eulers criterion,
2
−1)/8
(2 | p) ≡ 2(p−1)/2 ≡ (−1)(p (mod p) (3.3)
Case 1 Suppose p ≡ ±1(mod 8).

Then p ± 1 = 8m for some m ∈ Z. So p = 8m ± 1. Hence
p2 = (8m ± 1)2 = 64m2 ± 16m + 1

2
−1)/8
Consequently, (p2 − 1)/8 = 0. Thus (−1)(p = (−1)0 = 1.
Case 2 Suppose p ≡ ±3(mod 8).
Then p ± 3 = 8m for some m ∈ Z. So p = 8m ± 3. Therefore
p2 = (8m ± 3)2 = 64m2 ± 48m + 9
Hence
p2 − 1
≡ 1(mod 8)
8
2
−1)/8
Thus (−1)(p = (−1)1 = −1.
Theorem 3.3.3. (Gauss’ lemma)Let p be an odd prime and let n ∈ Z with

(n, p) = 1. Consider the set:

p−1
S = n, 2n, 3n, . . . , n
2
and let
p−1
jn ≡ aj (mod p), 1 ≤ j ≤
2
Let m be the number of aj ’s which are negative. In other words, let m denote
the number of integers in the set S whose residues (mod p) are greater than p/2.
Then
(n | p) = (−1)m .
105
Proof. Note that

aj ∈ Z×
p , j = 1, 2, . . . , (p − 1)/2.
First we prove that a1 , a2 , . . . , a(p−1)/2 are all distinct.

For 1 ≤ j ≤ p−1
2 ,
aj ≡ ak (mod p) ⇔ jn ≡ kn(mod p)
⇔ (jn − kn) ≡ 0(mod p)
⇔ (j − k)n ≡ 0(mod p)
⇔ (j − k) ≡ 0(mod p)(∵ (p, n) = 1)
⇔ j = k.
Hence a1 , a2 , . . . , a(p−1)/2 are all distinct.

Also note that

× p−1 p+1
Zp = 1, 2, 3, . . . , , ,...,p − 1
2 2

p−1 p−1 p−3
= 1, 2, 3, . . . , ,p − ,p − ...,p − 1
2 2 2

p−1
= j, p − j : j = 1, 2, 3, . . . ,
2

p−1
= ±j : 1 ≤ j ≤ (∵ p − j ≡ −j(mod p))
2

p−1 p−2 p−1
= − ,− , . . . , −1, 1, 2, . . . ,
2 2 2
Since {a1 , a2 , . . . , a(p−1)/2 } ⊂ Z× p , and ai ’s are distinct we have

p−1 p−1
|aj | : 1 ≤ j ≤ = 1, 2, 3, . . . ,
2 2
Now
p−1
n(2n)(3n) · · · n ≡ a1 a2 · · · a(p−1)/2 (mod p)
2
This implies

p−1 p−1
n(p−1)/2 1 · 2 · · · ≡ (−1)m 1 · 2 · · · (mod p)
2 2
That is,
p−1 p−1
n(p−1)/2 ! ≡ (−1)m !(mod p)
2 2
This implies that
n(p−1)/2 ≡ (−1)m (mod p)
106
Example 4. Let us consider p = 13 and n = 5. Then

p−1
S = n, 2n, 3n, . . . , n (mod 13)
2
= {5, 2 × 5, 3 × 5, 4 × 5, 5 × 5, 6 × 5} (mod 13)
= {5, 10, 2, 7, 12, 4} = {2, 4, 5, 7, 10, 12}
| {z } | {z }
<p/2 >p/2
= { 2, 4, 5 , −6, −3, −1}

| {z } | {z }
0<r<p/2 −p/2<r<0
Therefore (5|13) = (−1)3 = −1.
Theorem 3.3.4. Let m be the number defined in Gauss’ lemma. Then

(p−1)/2
p2 − 1

X tn
m≡ + (n − 1) (mod 2).
t=1
p 8
If n is odd,
(p−1)/2
X tn
m≡ (mod 2).
t=1
p
Proof. Let
p−1
S= n, 2n, 3n, . . . , n
2
For t = 1, 2, . . . , (p − 1)/2, write

tn tn tn
= + ,
p p p
n o
tn
where 0 < p < 1.[Note that {x} = x − [x] for all x ∈ R with 0 ≤ {x} < 1
and {x} = 0 if and only if x is an integer]
So for t = 1, 2, . . . , (p − 1)/2

tn tn
tn = p +p
p p

tn
=p + rt , (3.4)
p
where rt ∈ Z× p . Note that rt is nothing but the remainder in the Euclidean

division of tn by p. Let
A = {rt : rt < p/2} = {a1 , a2 , . . . , ak }

B = {rt : rt > p/2} = {b1 , b2 , . . . , bm }.
107
p−1
Then k + m = 2 and A ∪ B = {r1 , r2 , . . . , r(p−1)/2 }. That is,
{r1 , r2 , . . . , r(p−1)/2 } = {a1 , a2 , . . . , ak , b1 , b2 , . . . , bm } (3.5)
Note that
p/2 < bi < p ⇒ −p/2 > −bi > −p

⇒ p − p/2 > p − bi > −p + p
⇒ 0 < p − bi < p/2.
Therefore

p−1
1, 2, 3, . . . , = {a1 , a2 , . . . , ak , p − b1 , p − b2 , . . . , p − bm }
2 | {z } | {z } | {z }
c1 c2 cm
= {a1 , a2 , . . . , ak , c1 , c2 , . . . , cm } (3.6)
| {z } | {z }
0<rt <p/2 0<rt <p/2
Now, summing Equation (3.4) from 1 to (p − 1)/2, we get:

(p−1)/2 (p−1)/2 (p−1)/2
X X tn X
n t=p + rt
t=1 t=1
p t=1
| {z }
S(n,p)
k m
p2 − 1 X X
n = pS(n, p) + ai + bi
8 i=1 i=1
k
X m
X
= pS(n, p) + ai + (p − ci )
i=1 i=1
k
X m
X
= pS(n, p) + ai + mp − ci
i=1 i=1
k m
!
X X
≡ pS(n, p) + ai + ci + mp(mod 2)(∵ −1 ≡ 1(mod 2))
i=1 i=1
(p−1)/2
X
≡ pS(n, p) + t + mp(mod 2)
t=1
2
p −1
≡ pS(n, p) + + mp(mod 2).
8
This implies that
p2 − 1
mp ≡ −pS(n, p) + (n − 1) (mod 2)
8
108
Hence
p2 − 1
m ≡ S(n, p) + (n − 1) (mod 2) (∵ p ≡ 1(mod 2), −1 ≡ 1(mod 2)
8
If n is odd, then n − 1 ≡ 0(mod 2). So
(p−1)/2
X tn
m≡ (mod 2).
t=1
p
3.4 Quadratic reciprocity law

Let p and q be distinct odd primes. The question is whether the the following
pair of quadratic congruence are solvable.
x2 ≡ p(mod q), y 2 ≡ q(mod p)
Using Legendre’s symbol the question is how (p | q) is related to (q|p). Consider
the following bivariate table:
q/p 3 5 7 11 13 17 19 -
3 0 -1 1 -1 +1 -1 +1 -
5 -1 0 -1 +1 -1 -1 +1 -
7 -1 -1 0 + -1 -1 -1 -
11 +1 +1 -1 0 -1 -1 -1 -
13 +1 -1 -1 -1 0 +1 -1 -
17 -1 -1 -1 -1 +1 0 +1 -
19 -1 +1 +1 + -1 +1 0 -
- - - - - - - - -
We observe that row-5 and column-5 are identical in the above table. Similar is
the case for row-13, column-13, and row-17, column-17. There is slightly more
complicated regularity in cases of 3, 7, 11, and 19. There is a flip in sign when
both
p, q ≡ 3(mod 4).
If one of p or q is of the form 4k + 1, then it seems that (p | q)(q | p) = 1. and if
both are of the form 4k + 3, then it seems that (p | q)(q | p) = −1. The relation
was guessed by Euler and an incomplete proof was given by Legendre. Gauss
rediscovered it at the age of 18 and obtained the first complete proof. In fact
he gave several proofs. It is well known as Quadratic Reciprocity Law,
(p|q)(q|p) = (−1)((p−1)/2)((q−1)/2)
The law states that if one of p or q has remainder 1 when divided by 4, then
either both are solvable or both are unsolvable. If the remainders of both p and
q are 3, then if one is solvable then the other is not.
109
Theorem 3.4.1. (Quadratic Reciprocity Law) Let p and q are distinct odd
primes. Then
(
((p−1)/2)((q−1)/2) −1 if p ≡ q ≡ 3(mod 4)
(p|q)(q|p) = (−1) =
1 p ≡ 1(mod 4) or q ≡ 1(mod 4)
Proof. (Ferdinand Eisenstine’s proof) By Gauss’ lemma we have

(q|p) = (−1)m ,
where
(p−1)/2
X xq
m≡ (mod 2)..
x=1
p
Similarly,
(p|q) = (−1)n ,
where
(q−1)/2
X yp
n≡ (mod 2).
y=1
q
Hence
(p|p)(q|p) = (−1)m (−1)n = (−1)m+n .
Claim: m + m = p−1 2
q−1
2
Consider the set:

p−1 q−1
S = (x, y) : 1 ≤ x ≤ ,1 ≤ y ≤
2 2

p−1 p−1
= the set of lattice points in the rectangle 1, × 1,
2 2
Clearly the number of elements in S is equal to
p−1q−1
#(S) = . (3.7)
2 2
Consider the line joining (0, 0) and (p/2, q/2). Its equation is y = mx, where
m = q/p i.e. py = qx. We claim that there is no lattice point on the line py = qx
within S. If a lattice point (u, v) is on the line py = qx within S, then pv = qu,
implies that q | pv. But gcd(p, q) = 1 and 0 < v < q- so it is impossible. Let S1
and S2 be the regions of S above and below the line py = qx. Therefore
S1 = {(x, y) : qx < py} ,
S2 = {(x, y) : qx > py} .
Number of lattice points within S are the number of lattice points of S1 and
S2 . That is
S = S1 ∪ S2 .
We know that the number of integers in the interval (0, py/q) are [py/q].
110
(i) Note that S1 can be written as:
S1 = {(x, y) : qx < py}

q−1
2
[
py
= (x, y) : x <
y=1
q
q−1
2
[
= Sy ,
y=1
n o
py
where Sy = (x, y) : x < q . Note that, number of lattice points in
Sy is equal to number of integers in the interval (0, py/q). Therefore
#(Sy ) = [py/q]. Thus
q−1 q−1
2 2
X X py
#(S1 ) = #(Sy ) =
y=1 y=1
q
(ii) The number of elements in S2 is

p−1
2
X qx
#(S2 ) =
x=1
p
Hence the number of elements in S is

q−1 p−1
2 2
X py X qx
#(S1 ) + #(S2 ) = + =m+n (3.8)
x=1
q x=1
p
From eqquations (3.7) and (3.8), we have:

p−1q−1
m+n= .
2 2
3.5 Jacobi symbol

Definition 3.5.1. Let P be an odd integer. Then Jacobi symbol (n|P ) is defined
as: (
1 if P = 1
(n|P ) = Qr Qr
i=1 (n|p i )ai
if P = i=1 pai i
111
Note that (n|P ) ∈ {0, 1, −1}.

Theorem 3.5.1. If P and Q are odd positive integers, we have
(a) (m | P )(n | P ) = (mn | P )
(b) (n|P )(n|Q) = (n|P Q)
(c) (m|P ) = (n|P ) whenever m ≡ n(mod P )
(d) (a2 n|P ) = (n|P ) whenever (a, P ) = 1.
Proof. (a) Let P = p1 p2 · · · pk , where pi ’s are prime not necessarily distinct.
Then
k
Y k
Y
(m | P )(n | P ) = (m | pi ) (n | pi )
i=1 i=1
k
Y
= (m | pi )(n | pi )
i=1
k
Y
= (mn | pi ) = (mn | P )
i=1
(b) Let P = p1 p2 · · · pk and Q = q1 q2 · · · qs , where pi ’s and qi ’s are primes not

necessarily distinct.
k
Y s
Y
(n|P )(n|Q) = (n | pi ) (n | qi )
i=1 i=1
= (n | P Q)
(c) Let P = p1 p2 · · · pk . Assume that m ≡ n(mod P ). Then m ≡ n(mod pi )

for all i. Then (m | pi ) = (n | pi ) for all i. Therefore
k
Y k
Y
(m | pi ) = (m | pi )
i=1 i=1
That is (m | P ) = (n|P ).
(d)
(a2 n | P ) = (a2 | P )(n | P )

= (a | P )(a | P )(n | P )
= (a | P )2 (n | P )
= (n | P )
112
Theorem 3.5.2. If P is an odd positive integer we have
(−1|P ) = (−1)(P −1)/2
Proof. Let
P = p1 p2 . . . pm
where pi ’s are not necessarily distinct primes. This can also be written as
m
Y m
Y m
X X
P = pi = (1 + pi − 1) = 1 + (pi − 1) + (pi − 1)(pj − 1) + · · ·
i=1 i=1 i=1 i6=j
| {z }
divisible by 4
Hence
m
X
P −1≡ (pi − 1)(mod 4)
i=1
or
m
P − 1 X (pi − 1)
≡ (mod 2)
2 i=1
2
Therefore
m
Y
(−1|P ) = (−1|pi )
i=1
Ym
= (−1)(pi −1)/2
i=1
Pm
= (−1) i=1 (pi −1)/2
= (−1)(P −1)/2
Theorem 3.5.3. If P is an odd positive integer we have

2
−1)/8
(2|P ) = (−1)(P
Proof. Let
P = p1 p2 . . . pm
where pi ’s are not necessarily distinct primes. This can also be written as
m
Y m
Y m
X X
P2 = p2i = (1 + p2i − 1) = 1 + (p2i − 1) + (p2i − 1) (p2j − 1) + · · ·
| {z } | {z }
i=1 i=1 i=1 i6=j
≡0(mod 8) ≡0(mod 8)
Hence
m
X
2
P −1≡ (p2i − 1)(mod 64)
i=1
113
or
m
P 2 − 1 X (p2i − 1)
≡ (mod 8)
8 i=1
8
This implies that
m
X (p2 − 1)i P2 − 1
= + 8m for some m ∈ Z
i=1
8 8
Therefore
m
Y
(2|P ) = (2|pi )
i=1
m
Y 2
= (−1)(pi −1)/8
i=1
Pm 2
= (−1) i=1 (pi −1)/8
2 2
−1)/8+8m −1)/8
= (−1)(P = (−1)(P
Theorem 3.5.4. (Reciprocity law for Jacobi symbols)If P and Q are positive
odd integers with (P, Q) = 1, then
(P |Q)(Q|P ) = (−1)(P −1)(Q−1)/4
Proof. Let P = p1 p2 · · · pm and Q = q1 q2 · · · qn , where the pi and qi are primes
not necessarily distinct. Then
  !
Yn Ym
(P |Q)(Q|P ) =  (P |qj ) (Q|pi ) (By the definition of Legendre symbol)
j=1 i=1
    !
m
Y n
Y n
Y m
Y
=  (pi |qj )  (qj |pi ) 
i=1 j=1 j=1 i=1
m Y
Y n
= (pi |qj )(qj |pi )
i=1 j=1
Ym Y n
= (−1)(pi −1)(qi −1)/4 (By the reciprocity of Legendre symbols)
i=1 j=1
Pm Pn
(pi −1)(qj −1)/4
= (−1) i=1 j=1
Pm Pn
= (−1) i=1 (pi −1)/2 j=1 (qj −1)/2
m m
!
(P −1)/2(Q−1)/2
X (pi − 1) P −1 X (qi − 1) Q−1
= (−1) ∵ ≡ (mod 2), ≡ (mod 2) .
i=1
2 2 i=1
2 2
114
Definition 3.5.2. Equations to be solved in integers are called Diophantine

equations.
Definition 3.5.3. The equation
y 2 = x3 + k, k ∈ Z \ {0}
is called Mordell’s equation.

A natural number-theoretic task is the description of all integral solutions to
Mordell’s equation. Mordell, in 1920, showed that for each k ∈ Z the equation
y 2 = x3 + k has only finitely many integral solutions. The following table
describes all the integral solutions for some k of Mordell’s equation:
k Integer solution k Integer solution
1 (−1, 0), (0, −1), (2, −3) -6 None
-1 (1, 0) 7 None
-2 (3, -5) 11 None
-4 (2, -2), (5, -11) 16 (0, 4), (0, -4)
-5 None -24 None
6 None -26 (3, -1), (35, -207)
Theorem 3.5.5. The Mordell’s equation:
y 2 = x3 + k (3.9)
has no integer solution if k is of the form
k = (4n − 1)2 − 4m2 (3.10)
where m and n integers such that no prime p ≡ −1(mod 4) divides m.

Proof. Assume there is an integral solution (x, y). Note that
k = (4n − 1)3 − 4m2

= (4n)2 − 3(4n)2 + 3(4n) − 1 + 4m2
≡ −1(mod 4)
Hence Mordell’s equation y 2 = x3 + k can be written as:
y 2 ≡ x3 − 1(mod 4) (3.11)
Here is a table of values of y 2 and x3 − 1 modulo 4:

y y 2 (mod 4) x (x3 − 1)(mod 4)
0 0 0 3
1 1 1 0
2 0 2 3
3 1 3 2
4 0 4 3
5 1 5 0
115
Note that y 2 ≡ 0(mod 4) or y 2 ≡ 1(mod 4).
Case 1: Suppose x is even.

Suppose x = 2m for some m ∈ Z. Then x3 − 1 = 8m3 − 1. Therefore
x3 − 1 ≡ −1(mod 4). But y 2 ≡ 0(mod 4) or y 2 ≡ 1(mod 4). Hence (3.11)
is not satisfied if x is even.
Case 2: Suppose x is odd.
Then we have
x ≡ ±1(mod 4)
Subclaim 1: Suppose x ≡ −1(mod 4).
Then x = 4m − 1 for some m ∈ Z. Then
x3 − 1 = (4m − 1)3 − 1 ≡ −2(mod 4).
Hence eqution (3.11) is not satisfied if x ≡ −1(mod 4).

Subclaim 2: Suppose x ≡ 1(mod 4).
Then x = 4m + 1 for some m ∈ Z. Then
x3 − 1 = (4m + 1)3 − 1 ≡ (mod 4).
Hence eqution (3.11) is satisfied if x ≡ 1(mod 4). Thus y is even and

x ≡ 1(mod 4).
Let a = (4n − 1). Then equation (3.10) can be writen as:
k = (4n − 1)3 − 4m2 = a3 − 4m2
Then equation (3.9) can be written as y 2 = x3 + a3 − 4m2 . or
y 2 + 4m2 = x3 + a3 = (x + a)(x2 − ax + a2 ) (3.12)
Now
x2 − ax + a2 ≡ 1 − a + a2 (mod 4)(∵ x ≡ 1(mod 4))

≡ 3(mod 4)(∵ a ≡ −1(mod 4))
≡ −1(mod 4)
Let x2 − ax + a2 = p1 p2 · · · pk , where pi ’ s are primes(not necessarily distinct).

Then
p1 p2 · · · pk ≡ −1(mod 4) (3.13)
Equation (3.13) tells us that all its prime factors cannot be ≡ 1(mod 4). There
fore some prime pi ≡ −1(mod 4). We denote the prime number pi by p. Then
p ≡ −1(mod 4). Now
p|p1 p2 · · · pk ⇒ p | x2 − ax + a2
116
⇒ p | y 2 + 4m2
⇒ y 2 + 4m2 ≡ 0(mod 4)
⇒ y 2 ≡ −4m2 (mod 4)
⇒ (y 2 |p) = (−4m2 |p)
⇒ 1 = (−4m2 |p)(∵ p - y)
⇒ 1 = (22 |p)(m2 |p)(−1|p)
⇒ 1 = 1(−1|p)(∵ p - m)
⇒ 1 = −1(by theorem 3.3.1),which is absurd.
• (−1)(n−1)/2 = 1 if and only if n ≡ 1(mod 4)

2
• (−1)(n −1)/8
= 1 if and only if n ≡ ±1(mod 8)
• (−1)(n−1)(m−1)/4 = 1 if and only if n ≡ 1(mod 4) or m ≡ 1(mod 4)
3.6 Problems
Problem 1. Determine whether 219 is a quadratic residue or nonresidue mod
383.
(219|383) = (3 × 73|383) = (3|383)(73|383)(theorem ) (3.14)
(3|383) = (383|3)(−1)(383−1)(3−1)/4 = −(383|3)

= −(−1|3)(∵ 383 ≡ −1(mod 3))
= (−1)(−1)(3−1)/2 = 1
(73|383) = (383|73)(−1)(383−1)(73−1)/4
= (383|73) = (18|73)(∵ 383 ≡ 18(mod 73)))
2
−1)/8
= (2|73)(9|73) = (2|73) = (−1)(73 =1
Hence from equation (3.14), we have
(219|383) = 1
Hence 219 is a quadratic residue mod 383.
117
Problem 2. Determine whether 888 is a quadratic residue or nonresidue mod

1999.
(888|1999) = (4|1999)(2|1999)(111|1999)
2
−1)/8
= (1)(−1)(1999 (111|1999)
= (111|1999)
= −(1999|111)(by Reciprocity law)
= −(1|111)(∵ 1999 ≡ −1(mod 111))
= −1
Therefore 888 is a quadratic nonresidue of 1999.
Problem 3. Determine whether −104 is a quadratic residue or nonresidue mod
997.
(−104|997) = (−1 × 2 × 13|997)

= (−1|997)(2|997)(13|997)
= −(13|997)
= (997|13) = −(9|13) = −1
Therefore −104 is a quadratic nonresidue of 997.
Problem 4. Determine those odd primes p for which 3 is a quadratic residue
and those for which it is a nonresidue.
Let q be an odd prime > 3. Let p = 3 so we have p = 3 ≡ 3(mod4). We
already know
(
(q | p) if p ≡ 1(mod 4) or q ≡ 1(mod 4),
(p | q) =
−(q | p) if p ≡ 3(mod 4) and q ≡ 3(mod 4).
So (
(q | 3) if q ≡ 1(mod 4),
(3 | q) =
−(q | 3) if q ≡ 3(mod 4).
But q ≡ ±1(mod 3). If q ≡ 1(mod 3), then (q | 3) = (1 | 3) = 1. If q ≡
−1(mod 3), then (q | 3) = (−1 | 3) = (−1)(3−1)/2 = −1. Hence
(
1 if q ≡ 1(mod 3)
(q | 3) =
−1 if q ≡ −1(mod 3).
So the value of (3 | q) = 1 if and only if [q ≡ 1(mod 4) and q ≡ 1(mod 3) ] or

[q ≡ 3(mod4) and q ≡ −1(mod3)]. This is equivalent to,(3 | q) = 1 if and only if
q ≡ ±1(mod12). Moreover the value of (3 | q) = −1 if and only if [ q ≡ 1(mod 4)
and q ≡ 1(mod 3)] or [ q ≡ 3(mod 4) and q ≡ 1(mod 3)]
118
3.7 Exercise
1. Determine whether 219 is a quadratic residue or nonresidue mod 383.
2. Determine those odd primes p for which (−3|p) = 1 and those for which
(−3|p) = −1.
3. Prove that 5 is a quadratic residue of an odd prime if p ≡ ±1(mod 10))
and that 5 is a non residue if p ≡ ±1(mod 10)).
4. Find all quadratic residues a mod p (in the range −p/2 < a < p/2) for
p = 17, 19 and 23.
5. Use Gauss’ Lemma to compute the following Legendre symbols
(7|11) (5|13) (−3|17) (5|19)
6. Evaluate each of the following Legendre symbols using quadratic reci-

procity.
(511|881) (−257|541) (221|347) (105|1009)
119
120
Module 4
Cryptography
4.1 Introduction
The fundamental goal of cryptography is to allow two people, Alice and Bob, to
exchange messages over an insecure channel in such a way that their opponent,
Eve, cannot discover their content.
4.2 Terminologies
• Plaintext: The message you want to send, like “HELLO”. The plaintext
is written in some alphabet consisting of a certain number N of letters.
The term “letter” (or “character”) can refer not only to the familiar A−Z,
but also to numerals, blanks, punctuation marks, or any other symbols
that we allow ourselves to use when writing the messages. (If we don’t
include a blank, for example, then all of the words are run together, and
the messages are harder to read).
• Encryption: The process of disguising a message in such a way as to
hide its substance is encryption.
• Ciphertext: The disguised message, like XQABE.
• Message unit:The plaintext and ciphertext are broken up into message
units. A message unit might be a single letter, a pair of letters (digraph),
a triple of letters (trigraph), or a block of 50 letters
• Decryption: The process of turning ciphertext back into plaintext is
decryption.
• Encode: The processes of converting plaintext into a number, numbers is
called encloding.The first step is to “label” all possible plaintext message
units and all possible ciphertext message units by means of mathematical
objects from which functions can be easily constructed. These objects are
121
often simply the integers in some range. For example, if our plaintext and
ciphertext message units are single letters from the 26-letter alphabet A-Z,
then we can label the letters using the integers 0, 1, 2, . . . , 25, which we call
their “numerical equivalents.” Thus, in place of A we write 0, in place of
S we write 18, in place of X we write 23, and so on. As another example,
if our message units are digraphs in the 27-letter alphabet consisting of
A-Z and a blank, we might first let the blank have numerical equivalent 26
(one beyond Z), and then label the digraph whose two letters correspond
to x, y ∈ {0, 1, 2, . . . , 26} by the integer
27x + y ∈ {0, 1, ..., 728}.
• Decode: Turn number or numbers back into plaintext. There is nothing

secretive about encoding and decoding.
• Block Cipher operates on blocks of symbols. Examples: Digraph is a

pair of letters, Trigraph is a triple of letters.
• Cryptosystem: A pair of enciphering and deciphering algorithms. We

associate the process of encryption with a one-to-one function f from
plaintext message to ciphertext message, and the reverse decryption pro-
cess function f −1 is the map from ciphering text to plain text. We can
represent the situation schematically by the diagram
f f −1
P→C → P
Any such set-up is called a cryptosystem.
• Cryptanalysis is the process by which the enemy tries to turn cipher

text into plain text.
4.3 Types of Cryptography

4.3.1 Symmetric Key Cryptography
Symmetric Key Cryptography is an encryption system in which the sender and
receiver of a message share a single, common key that is used to encrypt and
decrypt the message. The Algorithm use is also known as a secret key algorithm
or sometimes called a symmetric algorithm
The key for encrypting and decrypting the message had to be known to all
the recipients. Else, the message could not be decrypted by conventional means.
Symmetric-key systems are simpler and faster; their main drawback is that
the two parties must somehow exchange the key in a secure way and keep it
secure after that. The following is one of the simplest forms of symmetric key
cryptography, called the Caesar cipher. Both the sender and recipient have the
“symmetric” key b.
122
Figure 4.1: Symmetric Cryptosystem (see [11]).
We encrypt the message P by
C = f (P ) ≡ P + b(mod N ),
we can compute
P = f −1 (C) = C − b(mod N ).
. For example, say we take P = 18 to C = 26, in modular N = 32, so we
substitute P and C in the forms of C ≡ P + b(mod N ) and get
26 ≡ 18 + b(mod 32).
By solving for key of b, and b = 8 and both sender and recipient know this
key. Now, when the recipient only gets the ciphertext C from the sender, he
can eventually decipher the message byP = C − b = 26 − 8 ≡ 18(mod 32). 2).
This is correct.
4.3.2 Affine enciphering transformation

An affine enciphering transformation is of the form
C ≡ aP + b(mod N )
where the pair (a, b) is the encrypting key, gcd(a, N ) = 1.

Example: C ≡ 13P + 3(mod 26). Encrypt B = 1 or D = 3 and get 13×1 + 3 ≡
16 and 13×3+3 ≡ 16(mod 26). C ≡ 3P +4(mod 26) is OK since gcd(3, 26) = 1.
In this case F = 5 goes to 3 × 5 + 4 ≡ 19(mod26) and 19 = T .
Decryption: Solve for P .
C − 4 ≡ 3P (mod 26)
and
3−1 (C − 4) ≡ P (mod26)
123
. Now
3−1 ≡ 9(mod 26)
(since 3.9 ≡ 1(mod 26)). So P ≡ 9(C − 4) ≡ 9C − 36 ≡ 9C + 16(mod 26). In
general encryption:
C ≡ aP + b(modN )
and decryption:
P ≡ a−1 (C − b)(modN )
. Here (a−1 , −a−1 b) is the decryption key.
4.3.3 Asymmetric Key Cryptography

Asymmetric cryptography , also known as Public-key cryptography, refers to
a cryptographic algorithm which requires two separate keys, one of which is
private and one of which is public. The public key is used to encrypt the
message and the private one is used to decrypt the message.
4.3.4 RSA cryptosystem.

The RSA cryptosystem, which was invented by Rivest, Sharmir and Adleman
in 1978. RSA works based on tremendous prime numbers. The steps are shown
in the following:
1. Choose large prime numbers p and q and calculate n = pq.
2. Recall ϕ(n) = (p − 1)(q − 1) when n = pq because p, q are primes.
3. Randomly pick e satisfying the property of 1 < e < ϕ(n) with gcd(e, ϕ(n)) =
1.
4. Find d such that de ≡ 1(mod ϕ(n)) through the Euclidean algorithm.
5. At the end of the process, the public enciphering key is kE = (n; e) and
private deciphering key is kD = (n; d).
Example 1:
1. Say we pick two prime numbers p = 3863, q = 83, so n = pq = 320629.
2. Calculate ϕ(n) = (p − 1)(q − 1) = 316684.
3. Pick e = 997, which satisfies gcd (e; ϕ(n)) = 1.
4. We need to find d such that de ≡ 1(modϕ(n)). By applying the extended

Euclidean algorithm, we obtain d = 263321.
5. So the public enciphering key is kE = (320629; 997) and the private deci-
phering key is kD = (320629; 263321).
124
• Suppose a person A wants to send the plaintext message P = 5 to the

person B.
• First, A encrypts the message with B’s public key kE = (320629; 997) and
gets the cipher text C = 5997 = 145463(mod 320629), and only sends the
ciphertext C = 145463 to B.
• Once B receives the ciphertext C from A.
• He applies his own private key kD = (320629; 263321) to the ciphertext
C = 145463 and gets 145463263321 (mod 320629).
• After calculation, at the end, he gets the plaintext message P = 5, which
is the correct plaintext.
Example 2:
• Bob choose two secrete primes p = 7 and q = 17
• Bob computes n = p.q = 7 × 17 = 119.
• Bob computes ϕ(n) = (p − 1)(q − 1) = 96
• Bob chooses a public encrypition exponent e with the property that
gcd(e, (p − 1)(q − 1)) = gcd(e, 96) = 1
Take e = 5.
• Find d ≡ e−1 (mod 96). Then d = 77
• Public Key: (119, 5), Private Key: (119, 77)
RSA Encryption
• Alice converts her plaintext into an integer
m = 19 satisfying 1 ≤ m ≤ 119.
• Alice uses Bob’s public key (n, e) = (119, 5) to compute
c ≡ me (mod N ), c ≡ 195 ≡ 66(mod 119)
• Alice sends the ciphertext c = 66 to Bob.
RSA Decryption
• Bob takes the ciphertext c = 66 and computes
cd ≡ (mod n), 6677 ≡ 19(mod 119)
• The value that he compute is Alice message m = 19.
125
4.3.5 Hash Functions

The ideal cryptographic hash function has four main properties
• it is easy to compute the hash value for any given message
• it is infeasible to generate a message that has a given hash
• it is infeasible to modify a message without changing the hash
• it is infeasible to find two different messages with the same hash.
4.3.6 Signatures
Authentication when people send the message, we often need to verify the iden-
tity of the sender. We often call this the signature. We can construct a digital
signature similar to a physical signature while communicating the message. Sup-
pose Alice wants to send message P to Bob. She sends her signature SA along
with her message. Let fA be the enciphering transformation for Alice and fB be
the same for Bob and make them public. First, Alice composes Bob’s fB with
her own private key fA−1 to obtain fB fA−1 (SA). When Bob gets the message, he
first applies his own private key fB−1 to obtain fA−1 (SA). Then he applies Alice’s
public enciphering transformation fA , and obtains SA. Since he can decipher
the signature, he concludes this message was indeed sent by Alice.
4.3.7 One way functions

The main idea behind asymmetric-key cryptography is the concept of the trap-
door one-way function.
• f is easy to compute(y = f (x)).
• f −1 is very difficult to compute(y = f −1 (x))
When n is large, n → p × q is a one-way function. Given p and q it is always

easy to calculate n; given n, it is very difficult to compute p and q. This is the
factorization problem.
4.3.8 Trapdoor One-Way Function

Suppose y = f (x) be a one way function. Given y and a trapdoor, x can be
computed easily.
When n is large, the function y = xk (mod n) is a trapdoor one-way function.
Given x, k, and n, it is easy to calculate y. Given y, k, and n, it is very difficult
to calculate x. This is the discrete logarithm problem. However, if we know the
0
trapdoor, k 0 such that kk 0 ≡ 1(mod ϕ(n)), we can use x ≡ y k (mod n) to find
x.
126
4.4 Problems
Problem 5. In the 27 letter alphabet(with blank =26), use affine enciphering
transformation with key a = 13, b = 9 to encipher the message “HELP ME”.
The affine enciphering transformation is
c(P ) = aP + b(mod 27)
P c(P ) = 13P + 9 C
H 13 × 7 + 9 = 100 ≡ 19(mod 27) T
E 13 × 4 + 9 = 61 ≡ 7(mod 27) H
L 13 × 11 + 9 = 152 ≡ 17(mod 27) R
P 13 × 15 + 9 = 204 ≡ 15(mod 27) P
blank 13 × 26 + 9 = 347 ≡ 23(mod 27) X
M 13 × 12 + 9 = 165 ≡ 3(mod 27) R
The required message is “ THRPXDH”.
Problem 6. In a long string of ciphertext which was encrypted by means of an
affine map on single-letter message units in the 26-letter alphabet, you observe
that the most frequently occurring letters are “Y” and “V”, in that order. As-
suming that those ciphertext message units are the encryption of “E” and “T”,
respectively, read the message “QAOOYQQEVHEQV”.
The affine deciphering transformation is
P ≡ a0 C + b0 (mod 26)
Given that “Y” and “V” are encryptions of “E” and “T” respectively. So we
have the following equations:
4 ≡ 24a0 + b0 (mod 26) (4.1)
19 ≡ 21a0 + b0 (mod 26) (4.2)

Subtracting equation (2) from equation (1), we obtain
−15 ≡ 3a0 (mod 26)
This implies that
11 ≡ 3a0 (mod 26) ⇒ 99 ≡ 27a0 (mod 26)
⇒ 99 ≡ a0 (mod 26)
⇒ 21 ≡ a0 (mod 26)
Putting a0 ≡ 21(mod 26) in Eqn(2), we obtain
4 ≡ 504 + b0 (mod 26) ⇒ b0 ≡ 20(mod 26)
Hence the affine deciphering transformation is
P ≡ 21C + 20(mod 26)
127
C P ≡ 21x + 20(mod 26) P

Q 21 × 16 + 20 = 356 ≡ 18(mod 26) S
A 21 × 0 + 20 = 20 ≡ 20(mod 26) U
O 21 × 14 + 20 = 314 ≡ 2(mod 26) C
Y 21 × 24 + 20 = 524 ≡ 4(mod 26) E
E 21 × 4 + 20 = 104 ≡ 0(mod 26) A
V 21 × 21 + 20 = 461 ≡ 19(mod 26) T
H 21 × 7 + 20 = 167 ≡ 11(mod 26) L
So the required message is
“QAOOYQQEVHEQV” → SUCEESSATLAST
Problem 7. You are trying to cryptanalyze an affine enciphering transforma-

tion of single-letter message units in a 37-letter alphabet. This alphabet includes
the numerals 0–9, which are labeled by themselves (i.e. , by the integers 0–9) .
The letters A-Z have numerical equivalents 10–35, respectively, and blank=36.
You intercept the ciphertext “OH7F86BB46R36270266BB9” (here the O’s are
the letter “oh”, not the numeral zero). You know that the plaintext ends with
the signature “007” (zero zero seven) . What is the message?
The deciphering transformation is
P ≡ a0 C + b0 (mod 37)
Given that “0(zero)” and “7” are encryptions of “B” and “9” respectively. So
we have the following equations:
0 ≡ 11a0 + b0 (mod 37) (4.3)
7 ≡ 9a0 + b0 (mod 37) (4.4)
Eqn(3)-Eqn(4) gives:
−7 ≡ 2a0 (mod 37) ⇒ 30 ≡ 2a0 (mod 37)

⇒ 570 ≡ a0 (mod 37)
⇒ 15 ≡ a0 (mod 37)
Putting a0 ≡ 15(mod 37) in Eqn(3), we obtain:
0 ≡ 11 × 15 + b0 (mod 37) ⇒ 0 ≡ 165 + b0 (mod 37)

⇒ 0 ≡ 17 + b0 (mod 37)
⇒ −17 ≡ b0 (mod 37)
⇒ 20 ≡ b0 (mod 37)
Hence the deciphering transformation is
P ≡ 15C + 20(mod 37)
128
C P ≡ 21x + 20(mod 26) P

O 15 × 24 + 20 = 380 ≡ 10(mod 26) H
H 15 × 17 + 20 = 275 ≡ 16(mod 26) G
7 15 × 17 + 20 = 314 ≡ 2(mod 26) C
F 15 × 15 + 20 = 245 ≡ 23(mod 26) N
8 15 × 8 + 20 = 140 ≡ 29(mod 26) T
6 15 × 6 + 20 = 110 ≡ 36(mod 26) BLANK
B 15 × 11 + 20 = 185 ≡ 0(mod 26) L
4 15 × 4 + 20 = 80 ≡ 6(mod 26) A
R 15 × 27 + 20 = 425 ≡ 18(mod 26) 1
3 15 × 3 + 20 = 65 ≡ 28(mod 26) S
2 15 × 2 + 20 = 50 ≡ 13(mod 26) D
9 15 × 9 + 20 = 155 ≡ 7(mod 26) 7
The required message is “ AGENT 006 IS DEAD 007”
Problem 8. Suppose there is a cryptosystem with 27-letter alphabets in which
A–Z have numeral equivalents 0 − −25,a blank=26. Suppose a study of a large
sample of ciphertexts revals that the most common diagraphs are (in order)
“ZA”, “IA” and “IW”. Suppose that the most common digraphs in the English
language (for text written in our 27-letter alphabet) are “E ” (i.e. , “E blank”
) , “S ”; “ T”. You know that the cryptosystem uses an affine enciphering
transformation modulo 729. Find the deciphering key, and read the message
“NDXBHO”. Also find the enciphering key.
We know that plaintexts are enciphered by means of the rule
C ≡ aP + b(mod 729)
, and that ciphertexts can be deciphered by means of the rule
P ≡ a0 C + b0 (mod 729)
, here a, b form the enciphering key, and a0 , b0 form the deciphering key.
digraph Numerical equivalent(27x + y(mod 729))
“ZA ” 27 × 25 + 0 = 675(mod 729)
“E ” 27 × 4 + 26 = 134(mod 729)
“IA” 27 × 8 + 0 = 216(mod 729)
“S ” 27 × 18 + 26 = 512(mod 729)
“IW” 27 × 8 + 22 = 238(mod 729)
“ T” 27 × 26 + 19 = 721(mod 729)
By hypothesis we have the following equations:
675a0 + b0 ≡ 134(mod 729) (4.5)
216a0 + b0 ≡ 512(mod 729) (4.6)
129
238a0 + b0 ≡ 721(mod 729) (4.7)
Eqn(5)-Eqn(6) gives:
459a0 ≡ −378(mod 729) (4.8)
Note that (459, 729) 6= 1. So (4.8) has no solutions. Eqn(5)-Eqn(7) gives:
437a0 ≡ −587(mod 729) ≡ 142(mod 729) (4.9)
To find the inverse of 437 in Z729 .
729 = 437 × 1 + 292

437 = 292 × 2 + 145
292 = 145 × 2 + 2
145 = 2 × 72 + 1
1 = 145 − 2 × 72
= 145 − 72(292 − 145 × 2)
= −72 × 292 + 145 × 145
= −72 × 292 + 145(437 − 1 × 292)
= 145 × 437 − 217(729 − 1 × 437)
= 145 × 437 − 217 × 729 + 217 × 437
≡ 362 × 437(mod 729)
Multiplying both sides of equation (9) by 362, we obtain:
a0 ≡ 362 × 142(mod 729)

≡ 51404(mod 729)
≡ 374(mod 729)
Putting a0 ≡ 374(mod 729) in equation(5), we get:
b0 ≡ 134 − 675 × 374

− 252316 = −383 ≡ 647(mod 729)
digraph Numerical equivalents(27x + y(mod 729))

“ND ” 27 × 13 + 3 = 354(mod 729)
“XB” 27 × 23 + 1 = 622(mod 729)
“HO” 27 × 7 + 14 = 203(mod 729)
We have
P ≡ a0 C + b0 (mod 729)
Therefore,
P ≡ 374C + 647(mod 729)
130
≡ 374 × N D + 647(mod 729)

≡ 374 × 354 + 647(mod 729)
≡ 133043(mod 729) ≡ 365(mod 729)
≡ 27 × 13 + 14(mod 729) = “N O00
P ≡ 374C + 647(mod 729)

≡ 374 × XB + 647(mod 729)
≡ 374 × 662 + 647(mod 729)
≡ 233275(mod 729) ≡ 724(mod 729)
≡ 27 × 26 + 22(mod 729) = “ W 00
P ≡ 374C + 647(mod 729)

≡ 374 × HO + 647(mod 729)
≡ 374 × 203 + 647(mod 729)
≡ 76569(mod 729) ≡ 724(mod 729)
≡ 27 × 0 + 24(mod 729) = “AY 00
The required message is “ NO WAY”. Note that a = a0−1 and b = −ab0 . We

have
374 ≡ a0 (mod 729)
729 = 374 × 1 + 355

374 = 355 × 1 + 19
355 = 19 × 18 + 13
13 = 6 × 2 + 1
Note that a = a0−1 and b = −ab0 . We have 374 ≡ a0 (mod 729)
1 = 13 − 2 × 6
= 13 − 2(19 − 1 × 13)
= −2 × 19 + 3 × 13
= −2 × 19 − 3(355 − 18 × 19)
= 3 × 355 − 56 × 19
= 3 × 355 − 56(374 − 1 × 355)
= −56 × 374 + 59 × 355
= −56 × 374 − 115 × 374
= −155 × 374(mod 729)
So a ≡ −115(mod 729) ≡ 614(mod 729). Also b = −614 × 647 ≡ 4(mod 729).
131
1. Alphabet ⇒ 27-letter Alphabet(For example A–Z, blank)

2. Plaintext ⇒ digraphs. For example
digraph Numerical equivalent(27x + y(mod 729))

“ZA ” 27 × 25 + 0 = 675(mod 729)
“E ” 27 × 4 + 26 = 134(mod 729)
“IA” 27 × 8 + 0 = 216(mod 729)
“S ” 27 × 18 + 26 = 512(mod 729)
“IW” 27 × 8 + 22 = 238(mod 729)
“ T” 27 × 26 + 19 = 721(mod 729)
Consider an N -letter alphabet.

1. Note each digraph has numerical equivalents in ZN 2
2. Instead of associating eqach digraph to element in ZN 2 we associate each
digraph with a veector [x, y]T where x is the numerical equivalent of the
first letter and y is the numerical equivalent of the second alphabet. For
example, consider the alphabet A = {A − −Z, blank}.
“ZA00 ⇒ (25, 0)T

00
“E ⇒ (4, 26)T
“IA00 ⇒ (8, 0)T

2 3
Problem 9. Find the inverse of the matrix A = in Z26 .
7 8
det(A) = D = 2 × 8 − 3 × 7 = −5 = 21(mod 26)
Note that (21, 26) = 1 so D has inverse in Z26 and D−1 = 21−1 = 5. Therefore
−1
D d −D−1 b

A−1 =
−D−1 c D−1 a

5 × 8 −5 × 3
=
−5 × 7 5 × 2

40 −15 14 11
= =
−35 10 17 10
Problem 10. Solve the following system of simultaneous congruences:
2x + 3y ≡ 1mod 26
7x + 8y ≡ 2mod 26
The given system is equivalent to matrix:

2 3 x 1
≡ (mod 26)
7 8 y 2
132
Therefore
−1
x 2 3 1
≡
y 7 8 2

14 11 1 10
≡ ≡
17 10 2 11
Problem
11.
Suppose we are working in the 26-letter alphabet, use the matrix
2 3
A= to encipher the message “NO”.
7 8
Note that the matrix corresponding to the digraph unit “NO” is

13
P =
14
Therefore
2 3 13 16
C = AP = = → “QV ”
7 8 14 21
Problem
12.
2 3
A= to encipher the message “NOANSWER”.
7 8
The matrix corresponding to “NOANSWER” is

13 0 18 4
P =
14 13 22 17
Therefore

68 39 102 59 16 13 24 7
C = AP = =
203 104 302 164 21 0 16 6
The coded message is “QVNAYQHI”
Problem
13.
2 3
A= to decipher the message “FWMDIQ”.
7 8
The matrix corresponding to “FWMDIQ” is

5 12 8
C=
22 3 16
We have
C = AP
Therefore

14 11 5 12 8 0 19 2
P = A−1 C = =
17 10 22 3 16 19 0 10
The decoded message is “ATTACK”
133
Problem 14. Suppose that we know that our adversary is using a 2 × 2 enci-
phering matrix with a 29-letter alphabet, where A–Z have the usual numerical
equivalents, blank=26, ?=27, ! =28. We receive the message
“GFPYJP X?UYXSTLADPLW, ”
and we suppose that we know that the last five letters of plaintext are our ad-
versary’s signature “KARLA.” Decipher the message
“GFPYJP X?UYXSTLADPLW ”
Problem 15. Note that

0 3
“AR” = → “DP ” =
17 15

11 11
“LA” = → “LW ” =
0 22
We have
C = AP
Therefore

3 11 0 11
=A
15 22 17 0
So −1
−1 0 11 3 11 21 19
A = =
17 0 15 22 22 18
Now
C = AP
Therefore
P = A−1 P

21 19 6 15 9 26 27 24 18 11 3 11
=
22 18 5 24 15 23 20 23 19 0 15 22

18 17 10 26 19 13 14 28 0 11
=
19 8 4 0 26 14 13 10 17 0
= “ST RIKE AT N OON !KARLA.”
134
Bibliography
[1] Apostol T.M., Introduction to Analytic Number Theory, Narosa Publishing

House, New Delhi, 1990.
[2] G. H. Hardy and E.M. Wright: Introduction to the theory of numbers;

Oxford International Edn; 1985
[3] A. Hurwitz & N. Kritiko: Lectures on Number Theory; Springer Verlag
,Universitext; 1986
[4] T. Koshy: Elementary Number Theory with Applications; Harcourt / Aca-

demic Press; 2002
[5] D. Redmond: Number Theory; Monographs & Texts in Mathematics No:
220; Marcel Dekker Inc.; 1994
[6] P. Ribenboim: The little book of Big Primes; Springer-Verlag, New York;
1991
[7] K.H. Rosen: Elementary Number Theory and its applications(3rd Edn.);
Addison Wesley Pub Co.; 1993
[8] W. Stallings: Cryptography and Network Security-Principles and Prac-
tices; PHI; 2004
[9] D.R. Stinson: Cryptography- Theory and Practice(2nd Edn.); Chapman &
Hall / CRC (214. Simon Sing : The Code Book The Fourth Estate London);
1999
[10] S.Y. Yan: Number Theroy for Computing(2nd Edn.); Springer-Verlag;
2002
[11] Jonathan Katz and Yehuda, Introduction to Modern Cryptography, CRC
Press.
135
4.5 Syllabus
Semester 1
MTH1CO5: NUMBER THEORY

No. of Credits: 4
No. of hours of Lectures/week: 5
TEXT 1 : APOSTOL T.M., INTRODUCTION TO ANALYTIC NUMBER

THEORY, Narosa Publishing House, New Delhi, 1990.
TEXT 2: KOBLITZ NEAL A., COURSE IN NUMBER THEROY AND CRYP-
TOGRAPHY, SpringerVerlag, NewYork, 1987.
Module 1
Arithmetical functions and Dirichlet multiplication; Averages of arithmeti-
cal functions [Chapter 2: sections 2.1 to 2.14, 2.18, 2.19; Chapter 3: sections
3.1 to 3.4, 3.9 to 3.12 of Text 1]
Module 2
Some elementary theorems on the distribution of prime numbers [Chapter
4: Sections 4.1 to 4.10 of Text 1]
Module 3
Quadratic residues and quadratic reciprocity law [Chapter 9: sections 9.1 to
9.8 of Text 1] Cryptography, Public key [Chapters 3 ; Chapter 4 sections 1 and
2 of Text 2.]
136
4.6 Notations
N = the set of natural numbers

C = the set of complex numbers
[x] = the greatest integer less than or equal to x
{x} = x − [x]
d|n = d divides n
X
= summation over all divisors of n
d|n
X
= summation over all prime divisors of n
p|n
#(A) = the cardinality of the set A

(c, d) = the greatest common divisor of c and d
137

Ilovepdf Merged (17) Removed 0

Uploaded by

Copyright:

Available Formats

You might also like

Ilovepdf Merged (17) Removed 0

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ilovepdf Merged (17) Removed 0

Uploaded by

Copyright:

Available Formats

NUMBER THEORY

SCHOOL OF DISTANCE EDUCATION

STUDY MATERIAL FIRST

M.Sc. Mathematics (2019 ADMISSION ONWARDS) CORE

Dr. Anil Kumar V

Dr. Preethi Kuttipulackal

1 Arithmetical Functions and its Averages 1

2 Some elementary theorems on distribution of prime numbers 57

2.7 An asymptotic formula for the partial sums . . . . . . . . . . . . 86

Arithmetical Functions and

1.1 Arithmetical functions

1.1.1 The Möbius function µ(n)

factor > 1.Obviously,

Proof. We consider two cases.

Case 2: Suppose n > 1.

[µ(p1 p2 p3 ) + · · · + µ(pk−2 pk−1 pk )] + [µ(p1 p2 · · · pk )]

1.1.2 Euler totient function ϕ(n)

ϕ(n) := #{a ∈ N : a ≤ n, (a, n) = 1}.

Theorem 1.1.2. If n ≥ 1 we have

A(d) := {k : (k, n) = d, 1 ≤ k ≤ n}.

Claim 1: If d1 and d2 are distinct divisors of n, then A(d1 ) ∩ A(d2 ) = ∅.

x ∈ A(d1 ) ∩ A(d2 ) ⇒ x ∈ A(d1 ) and x ∈ A(d2 )

Claim 2: {1, 2, . . . , n} = ∪d|n A(d).

x ∈ S ⇒ (x, n) = d for some d, 1 ≤ d ≤ n

This implies that

Claim : #(A(d)) = ϕ(d).

Observe that if d is a divisor of n, then n/d is also a divisor of n. The divisors

Hence equation (1.4) can be written as

This completes the proof of the theorem.

1.1.3 A relation connecting ϕ and µ.

Proof. By the definition of ϕ, we have

{d, 2d, . . . , (n/d)d}

Hence equation (1.5) can be written as:

This completes the proof of the theorem.

1.1.4 A product formula for ϕ(n)

where p is a prime divisor of n.

Proof. Here we consider two cases.

Since there are no prime divide 1, we define

Hence the theorem is true for n = 1.

where pi ’s are distinct primes and α1 , α2 , . . . , αr are integers ≥ 1. Consider

= ϕ(n) (by theorem 1.1.3 )

This completes the proof of the theorem.

Theorem 1.1.5. Euler’s totient has the following properties:

(a) ϕ(pα ) = pα − pα−1 , for prime p and α ≥ 1.

(c) ϕ(mn) = ϕ(m)ϕ(n) if (m, n) = 1.

(d) a|b implies ϕ(a)|ϕ(b).

Putting n = pα in equation (1.6), we obtain:

Note that each prime divisor of mn is either a prime divisor of m or of n, and

(c)By part (b) we have

(d)Given that a|b.

Here we consider 3 cases. Case 1: Suppose b = 1.

1.1.5 Dirichlet product of arithmetical functions

Definition 1.1.5. If α is a complex number, the power function N α is an

Definition 1.1.6. The unit function u is defined as

Definition 1.1.7. The identity function I is defined as

Remark 1. Theorem 1.1.1 can be stated in the form ϕ = µ ∗ u.

Remark 2. Theorem 1.1.3 can be sated in the form µ ∗ N = ϕ.

Theorem 1.1.6. Dirichlet multiplication is commutative and associative. That