Elementary Theory of Numbers, Harriet Griffin

ELEMENTARY THEORY OF NUMBERS
INTERNATIONAL SERIES IN
PURE AND APPLIED MATHEMATICS

William Ted Martin and E. H. Spanier
CONSULTING EDITORS
ahlfors Complex Analysis

bellman Stability Theory of Differential Equations
buck Advanced Calculus
busacker and saaty Finite Graphs and Networks
cheney Introduction to Approximation Theory
coddington and levinson Theory of Ordinary Differential Equations
dettman Mathematical Methods in Physics and Engineering
epstein
Partial Differential Equations
golomb and shanks Elements of Ordinary Differential Equations

graves The Theory of Functions of Real Variables
Greenspan Introduction to Partial Differential Equation?
griffin Elementary Theory of Numbers
hamming Numerical Methods for Scientists and Engineers
hildebrand Introduction to Numerical Analysis
householder Principles of Numerical Analysis
Elements of Pure and Applied Mathematics

Vector and Tensor Analysis
lepage Complex Variables and the Laplace Transform for Engineers
nehari Conformal Mapping
newell Vector Analysis
ralston A First Course in Numerical Analysis
rosser Logic for Mathematicians
rudin Principles of Mathematical Analysis
saaty and bram Nonlinear Mathematics
Simmons Introduction to Topology and Modern Analysis
Sneddon Elements of Partial Differential Equations
Sneddon Fourier Transforms
stoll Linear Algebra and Matrix Theory
struble Nonlinear Differential Equations
lass
lass
weinstock
weiss
Calculus of Variations
Algebraic
zemanian
Number Theory
Distribution Theory and Transform Analysis
ELEMENTARY
THEORY OF NUMBERS
HARRIET GRIFFIN
Associate Professor of Mathematics
Brooklyn College
McGRAW-HILL BOOK COMPANY,

NEW YORK
TORONTO
1954
INC.
LONDON

Copyright, 1954, by the McGraw-Hill Book Company, Inc. Printed in the
United States of America. All rights reserved. This book, or parts thereof,
may not be reproduced in any form without permission of the publishers.
Library of Congress Catalog Card
Number 53-12046
6789-MP-9876
24785
PREFACE
The arithmetic theory of numbers is ever a fascinating subject, the
fundamentals of which can be presented with ease and profit to the average undergraduate student of mathematics and to those who are preparing to teach mathematics provided that these students are stimulated
by a clear and logical treatment of carefully selected topics. It is the
aim of this textbook to offer such a development of the subject. The
facts and methods of proof are old, indeed, but the author believes that
her experience of teaching the theory of numbers for over fifteen years
has enabled her to choose the topics that not only develop the student's
an ample basis for

more advanced work. The simplicity and lucidity of the presentation
have been tested time and again in the classroom. Experiment has
shown that the student appreciates knowing on exactly what foundation
insight into the problems of this field but also furnish
Consequently, care has been taken

he can readily understand. Definitions are precise, but in elementary form.
Each new concept is illustrated.
Theorems are proved in detail. Brief historical comments
elucidate the material and provide a setting for it.
The exercises are
graded. They include a sufficiently large number of numerical problems
the reasoning
to orient
him
is
to be constructed.
at a level
and
in language
to develop the student's power of inspection,
some to test his understanding of simple theoretical questions, and others to challenge his originality.
It is to be emphasized that this book is just a text.
It is written for
the student rather than the teacher. It is neither erudite nor exhaustive.
The reader with a good grasp of algebra and the ability to concentrate
will be able to understand it.
It is hoped, moreover, that it will interest
him and induce him to pursue his inquiries further.

The author is indebted to her students in both the graduate and undergraduate divisions of Brooklyn College for their help in accomplishing
this work.
They have patiently perused mimeographed and lithoprinted
by their lively reactions the parts to be retained,

improved, or deleted. One of them, Miss Alice Osterberg, has given
expert assistance in reading the proof. Their contributions are gratefully
acknowledged.
material, indicating
Vi
PREFACE
It is apparent that there is more material in the text than can be

covered by the average class of undergraduates in only one semester.
Owing either to the importance of the information or to the method of
proof involved, however, with the exception of Theorem 3-20, on the
number
of primes not exceeding the positive integer n, it is suggested

that no theorem be omitted until Chap. 4 has been completed. Then,
for the ordinary course,
Theorems
5-7, 5-8, 5-9,
and
5-11,
Theorems
6-3
through 6-6 and 6-9 through 6-12, Theorems 7-11 through 7-16, as well
For the convenience of the reader the
as all of Chap. 8 may be omitted.
theorems just mentioned have been marked with an asterisk. Most
instructors will wish to include as much of Chap. 9, on quadratic residues,
as time permits.
Theorems from Chap. 10 may be selected at pleasure
provided that Theorem 10-4, on the Pythagorean triangle, is included.
The last two chapters will, doubtless, have to be omitted in a 45-hour
course.
Harriet Griffin
CONTENTS
PREFACE
chapter
THE FUNDAMENTAL LAWS
1.
1-2.
The development of the integers

The system of rational integers
1-3.
Definitions
1-4.
The
1-1.
2
6
principle of
chapter
Archimedes
2.
THE LINEAR DIOPHANTINE EQUATION
2-4.
The form ax + by
The Diophantine equation ax + by = n
A method for finding a solution of ax + by = n
The solution of the linear Diophantine equation in more than two
2-5.
Simultaneous linear Diophantine equations
2-1.
2-2.
2-3.
chapter
3-3.
3-4.
Unique factorization
3-5.
Fermat's method of infinite descent

The Euclidean algorithm
3-2.
3-6.
3-7.
3-8.
3-9.
The
The
least
common
3-12.
The highest power of a prime that

Some theorems on primes
Congruent integers
4-2.
Basic properties of congruences
4-3.
The
4-5.
is
53
54
56
58
62
residue classes
<j>
a factor of n!
36
37
40
46
PROPERTIES OF CONGRUENCES
4-1.
4-4. Euler's
23
33
34
an integer
Perfect numbers
4.
21
31
multiple
divisors of
chapter
19
variables
25
25
26
28
30
3-10. Scales of notation

3-11.
16
PROPERTIES OF INTEGERS
3.
The composite
The sieve of Eratosthenes
The number of primes
3-1.
14
function
Residue systems modulo
m
vii
CONTENTS
Vlll
chapter
5-1. Identical
5-2.
5-3.
5-4.
5-5.
5-6.
5-7.
5-8.
5-9.
Equivalent congruences
Linear congruences
Division of polynomials
The number and multiplicity of

The Chinese remainder theorem
6-2.
6-3.
of f(x)
The Mobius function
An
7.
7-1.
The X function
Belonging to an exponent modulo
Another test for a prime
7-7.
7-8.
7-9.
7-10.
7-11.
89
92
93
96
ON BELONGING TO AN EXPONENT
7-2.
7-6.
88
Mobius inversion formula
application of the
chapter
7-5.
Fermat's theorem
Applications of Fermat's theorem
Wilson's theorem
6-4.
7-4.
= 0(mod m)
(mod p 8 )
98
100
104
105
107
108
112
114
116
118
118
Primitive roots
Gauss' method for finding a primitive root modulo p
Primitive roots modulo p n and 2p n
Primitive X roots
Integers belonging to a divisor of X(2 n ) modulo 2 n
Integers belonging to a divisor of <j>(p n ) modulo p n
Integers belonging to a divisor of <f>(2p n ) modulo 2p n
Integers belonging to a divisor of X(m) modulo m
chapter
8.
INDICES
a prime modulus
8-2. Euler's criterion for the solvability of x n
125
129
8-1. Indices for
chapter
9.
The
9-2.
Quadratic residues
The Legendre symbol
The prime moduli of which an integer
The Jacobi symbol
The solution of x 2 = a (mod 2 n )
9-4.
9-5.
9-6.
chapter
10.
10-3.
Fermat's
10-4.
The area of an integral right triangle

The generalized Wilson theorem
The Pellian equation
10-5.
last
is
a quadratic residue
134
135
139
140
152
156
SOME FAMOUS PROBLEMS
10-2.
10-6.
c(mod m)
general quadratic congruence
The Waring problem

The equation x 2 + y 2 =
10-1.
QUADRATIC RESIDUES
9-1.
9-3.
66
68
70
72
75
79
80
84
85
THE THEOREMS OF FERMAT AND WILSON AND

THE MOBIUS FUNCTION
6.
6-5.
7-3.
...
the solutions of a congruence
Other simultaneous linear congruences

of solutions of the congruence fix)
The number
The solution
chapter
6-1.
THE SOLUTION OF CONGRUENCES
5.
and conditional congruences
z2
theorem
158
167
168
168
170
172
CONTENTS
chapter
11-1. Integral
domains and
11.
POLYNOMIALS
fields
Polynomials with respect to a prime modulus

11-3. A method for solving a congruence modulo p, a prime
11-2.
chapter
12-1.
The additive theory
of
12.
IX
177
186
188
PARTITIONS
numbers
12-2. Partitions with repetitions

12-3.
Diagrams
12-4.
Generating functions for the number of partitions
of partitions
190
191
192
193
BIBLIOGRAPHY
197
INDEX
199
CHAPTER

1-1.
The Development
of the Integers.
The
rational integers are the
many centuries of development of the concept of number.

Doubtless man learned first to distinguish oneness and otherness without
result of
abstracting the idea of
number
is
elements,
number
itself.
The
basic notion in the concept of
that of one-to-one correspondence.
and B, and
if
to each element of
If there are
there
is
two
sets of
assigned exactly
one element of B, while each element of B is thereby related to a single

element of A, the relationship is called a one-to-one reciprocal correspondence.
By means of such a relation a man could determine that he had
exactly as many rings as he had fingers even though he had not learned to
Any two sets that can be put into one-to-one correspondence are
count.
Thus the concept of number implies the
said to have the same number.
abstraction of that property which is common to sets that are so related.
However imperfect these concepts may have been, man did eventually
learn to count and to represent by marks the ideas now represented by
the symbols 1, 2, 3,
These numbers have the single property of
denoting quantity. They answer the question, How many units? They
are the natural numbers, and they were the only numbers known to the
Greeks until Diophantus (c. 275) extended the concept of number to
include fractions.
To be sure the Ahmes papyrus, which was written
before 1700 B.C., indicates that in their calculations the Egyptians
employed symbols that are equivalent to fractions with numerator one,
but such symbols, even to the Greeks of Euclid's time, referred to the
notion of magnitude rather than number. The art of calculating was
thus distinguished from the science of number. It is to be noted that
zero is not among the natural numbers.
The Greeks had no symbol for
It was probably not until the fifth century that the Hindus introzero.
duced a symbol for zero and the principle of position in writing numbers.
These were, indeed, great accomplishments in the field of arithmetic.
By the twelfth century the advancement of the Hindus in algebra almost
matched their achievements in arithmetic, for they were the first to
.
recognize the existence of negative quantities even though they did not
admit them as solutions of their problems. It was not until the sixteenth
century that European mathematicians reached this stage of development
1
of the idea of number, and it remained for Descartes (1596-1650) to grasp

completely the concept of signed numbers.*
The signed numbers have two
They compound
basic properties.
the
Thus, although each of the symbols + 1, 1, +5, and 5 represents but one number, that number has
two fundamental attributes. It is evident, moreover, that a one-to-one
correspondence can be set up between the positive integers and the natural
idea of quantity with that of quality.
numbers
1-2.
in the following
The System
bers, arithmetica, is
numbers
1, 2, 3,
*-* 1,
+2
The
+3
- 2,
<-> 3,
theory of numproperly concerned with the study of only the natural
Nevertheless,
1, 2,
rational integers 0,
itself, this
+1
manner:
of Rational Integers.
we
shall
make use
of the
system
of
because, besides being of interest in
set has properties not possessed
enable us to develop
classical
by the natural numbers that
many theorems expounding
the qualities of the
natural numbers through methods that have the facility characteristic

of
an algebraic presentation.
Throughout
this text, therefore, the
word
" integer," unless otherwise stated, will refer to a rational integer, and the
x, y, z, will
Moreover, we
represent these integers.
letters a,
b, c,
shall, as
a rule, observe the usual convention of omitting the positive
+1, +2, +3, ....

we study about these integers?
signs in writing
What
shall
Just consider the integers
through 10, and they will serve to illustrate some salient facts.
It is apparent that the product of three consecutive integers is divisible
by 6, that when two consecutive even integers are chosen, one is divisible
by 4. You will claim you have known all your life that we need only the
10 digits 0,1,
9 to write any integer, but can this be accomplished
in another way?
Have you ever noticed the remarkable fact that, of the
consecutive integers 8 and 9, one is a perfect cube and the other a perfect
square? Surely you have not overlooked the familiar right triangle
from
whose
sides
have the lengths
3, 4,
and
5.
Can you
Again, the
sum
of the positive
having this
Although for its size 6 has many divisors, you notice that
Observe how close the first few integers of the
5 has but 1 and 5.
Would you be interested in examining the law indicated
latter type are.
by the following equations?
divisors of 6
is
double
itself.
find another integer
property?
5=1+1+3
7=1+1+5=1+3+3
9
+ 1+7
1+3 +
Perhaps these few examples

other significant facts.
*
F. Cajori,
"A
10
1+3
6=1+5=3+3
8=1+7=3+5
3
will stimulate the reader to look for
Having made a discovery, he
History of Mathematics."
certainly will
some
want

to determine
and
why it
so, starting
is so.
That, indeed,
with elementary notions,
is
the attitude
we
we wish
up a
shall set
to foster,
basis for
show-
ing that certain laws do hold.
We
is familiar with the concepts of numsum, difference, product, equality, greater than,
absolute value, etc. We shall assume that he understands the fundamentals of algebra and the derivative with respect to x of a rational
We shall also suppose
integral algebraic function of the real variable x.
that he is familiar with the content of the following system of postulates
which the integers obey:
shall
assume that the reader
ber, uniqueness, relation,
1.
The
set of rational integers contains
+1
(positive real one)
and at
one other integer.

2. The law of closure for addition: For each pair of integers a and b,
in that order, the sum of a and b exists and is a unique integer c.
Thus
a
b = c.
b = b
a.
3. The commutative law for addition: a
4. The associative law for addition (a + b)
c = a + (6 + c)
5. The equation a + x = c has a solution for x that is an integer.
The second postulate tells us that the operation of addition exists for
the rational integers. Moreover, since +1 is in the set, we can generate
a subset of integers by merely adding
1 to itself and to each result so
least
Thus we have 1 + 1 = 2, 2 + 1 = 3,
We call all
these numbers positive integers, and 1, 2, 3,
are distinct.
Furthermore, if a + c = b and c is positive, we say that a a (b is greater than a)
The set of rational integers is
ordered by this relation "less than."
The fifth postulate says that subtraction exists, and we therefore write
the statement a + x = c in the form c a = x.
We shall show later
that we can derive the fact that there is a single integer
such that
= a for any integer a.
a +
6. The law of closure for multiplication: For each pair of integers a
and b, in that order, the product of a and b exists and is a uniquely determined integer c. Thus a b = c, or ab = c.
7. o( + l) = a for any integer a.
8. The commutative law for multiplication: ab = ba.
9. The associative law for multiplication: (ab)c = a(bc).
10. The distributive law: a(b + c) = ab + ac.
11. The cancellation law for multiplication: If ab = ac and a tA 0, then
obtained.
c.
is no integer x such that a < x < a

1.
statement means that there is no integer between a and a
1.
Hence, a and a + 1 are called consecutive integers. We say, then, that
the rational integers form a discrete set, for when they are arranged
12.
The
There
last
according to the relation "less than" and then separated into any two
parts without disturbing the array, there is a first integer in one part and
a last integer in the other.
>
sequence of integers rii, where i = 1, 2, 3,

such that
> has a least positive integer.
a = 6 and c = d, then a + c = 6 + d and ac = bd.
a = b, then either one may be substituted for the other in any
Any
13.
fti
n2
14. If
If
15.
>
'
relation.
has a property and if n + 1 has this property whenever the

n has it, then every positive integer has the property.
This last postulate is the one on which the method of mathematical
16.
If
+1
positive integer
induction, sometimes called "finite induction,"

this
method
is
sum of a finite number of

second postulate. Thus a
b
c = (a
b)
(a
b
c)
d, and, in the general case, a
the
+ +
(a
+ +
+b+
+
+6+
We
based.
shall use
For example, we can define

integers inductively on the basis of the
both definition and proof.
for
m)
c,
a+6+c+d=
+m+n=
n.
It is worth while recalling that the relation "equals" which we have

used so frequently in stating these postulates has the following properties
1. Equals is reflexive; that is, a = a.
2. Equals is symmetric; that is, if a = b, then b = a.
is, if a = b and b = c, then a = c.
which we have stated above is certainly not
the most concise one that might have been chosen. To be sure, some of
the statements can be derived from others. But the set is sufficiently
elementary for our purposes and permits us to develop the ideas we need
on a basis that will be readily understood by the student.
We shall prove first that there is a unique integer such that, for any
3.
Equals
integer
b,
transitive; that
is
The system
of postulates
The
b.
a particular integer a there
and
But c
this integer Oi
Oi
a.
fifth
is
postulate
shall
show that
exists
such that b
is
unique, for
if
Oi
us that corresponding to
+ a.
= (c + a) + Oi
= c + (a + Oi)
= c +a
=
This integer Oi
tells
an x such that a
b for
a.
any
We
shall call
b.
We
have
Hence,
there were a second integer O2 such that
any b, then 2 + Oi = 2 and Oi + 2 = Oi. But Oi +

There is, then, but a single integer,
2 =
2 + Oi, and therefore Oi =
2
= b for any b. We may also write
zero, designated by 0, such that b +
this statement in the form 6 6 = 0.
b for
Since
6, it
follows immediately that
if
6 is positive, 6
is
greater

than
Moreover,
and conversely.
zero,
and
shows that
+1
are consecutive integers.
Furthermore, for each integer a there exists an integer x such that

x = o. This integer x is called the negative of a. It is the inverse
Thus a +
of a with respect to addition, and it is designated by a.
_f_
a) =0.
This inverse is unique because the following argument shows

is always unique.
Suppose that subtraction is not unique and that when a and c are given,
both b\ and 6 2 are such that
(
that subtraction
6i
+ a)
62
Since
and
a) +
since
(-a)
it
(a
6i)
= (-a)
+
+
(a
6 2)
a]
62
follows that
[(-a)
a]
6i
[(-a)
6i
&i
and that
b2
Hence,
6
Consequently, subtraction is unique.

The inverse of a positive integer is said to be a negative integer. Accordingly, we write (+2) = 2 that is, the negative of positive two is nega;
Moreover, each of these negative integers is less than zero, for

For the same reason, ( a) = a.
( a) + a = 0, where a is positive.
This means that a is the inverse with respect to addition of a; that is, a
This statement implies that the positive
is the negative of negative a.
To summarize, we
integers are the negatives of the negative integers.
have proved that:
Theorem 1-1. There is a unique integer such that, for any integer 6,
tive two.
b.
Theorem
1-2.
Subtraction
is
unique, and each integer has a unique
inverse with respect to addition.
Has
it
ever occurred to you that
principles
a^0,
is
so.
Hence,
unique.
We
0*0 =
0.
for
any
that a
b because of the basic
b-a-\-b'0 = b-a.
Therefore, b
statement includes
Theorem
6-0 =
We know
stated?
consider the equation
uniquely
tion
we have
However,
c.
and so when
b a = c and
= c, and subtraca,
But
Notice that when 6
0.
0,
this
Thus we have:
1-3.
For any integer 6, 6
have seen that c + ( c) =0.
0.
Hence, ac
+ a(c)
0.
( oc) = 0, and this inverse is unique. Consequently, a(c)

When c = 1, this equation includes a( 1) = (a( + l)) = a.
Furthermore, subtraction can be accomplished by addition, for we can
But
ac
= ac.
prove
Theorem
1-4.
+ b).
(
According to the definition of subtraction, a b = x, where the

integer x is such that b + x = a.
We wish to prove that a + ( b) is the
b) for x in the last equation, we
proper value of x. Substituting a
(
have
b
[a +(-6)] - b + [(-b) +a]
= [b + (~b)]
+a
=
=
a
>
member
If a > b, then b > a, and conversely.

then a = b
c with c positive, and, by multiplying each
c).
of the equation by 1, we find that a = ( b)
(
Theorem
If
1-5.
b,
a) + c = [( b) + c)] + c, and finally a) +

that 6 > a. The converse is now evident.
Theorem 1-6. -(a - 1) = -a + 1.
We know that
-(a- 1) = (a- 1)(-1)
Hence,
= b,
so
= [a+(-l)](-l)
= -a + 1
no integer between a and (a 1) that is, if two
integers are consecutive, as are a 1 and a, so are their negatives.
We
As a
result there
must conclude,
is
therefore, that the integers are positive, zero, or negative.
EXERCISES
a
a
=
=
7.
On
4.
Show that
Show that
Show that
Show that
5.
If
6.
Prove that a(b
1.
2.
3.
8.
then the negative of a equals the negative of

cannot also be true that a > b.
a b cannot both be true.
if a < b and a = b + c, then c is a negative integer.
+ c and c is negative, prove that a < b.
if
if
b,
b.
b, it
c)
ab
ac.
the basis of postulate 11, prove that if ab =

and a ? 0, then 6=0.
Show that the theorem of Exercise 7 is a sufficient condition for the cancellation
law for multiplication.

9. Prove the law of signs for addition.
10. Prove the law of signs for multiplication.
1-3. Definitions.
with the principle of
The law
finite
of closure with respect to addition together
induction shows that
tion of addition to a finite
number
when we apply
the opera-
of integers of our set, the result is
We know
necessarily in the set.
into addition
and that the
to the use of integers,
product of a
also that subtraction can be converted
result
When we
unique.
is
restrict ourselves
also evident that our laws provide that the
it is
number
finite
them always
of
But
exists.
and
any
b are
mean by
An
an
when a
we
necessary to define what
it is
integer a
Under these conditions
be.
The operation
no integer
If
division.
integer b is said to divide
that a
b.
and therefore
integers,
not always
it is
possible to find an integer that will satisfy the equation ax
if
written in the form b
is
we
c exists,
write b
there exists an integer c such
a and
a and read
Jf
is divisible by
"b divides a."
also said that a
it is
it,
is
read,
"b does not divide a."
When 6^0 and c exists, c is unique, for if a = bci and a = bc then

When
bci = bc and c\ = c by the cancellation law for multiplication.
b = 0, c does not exist unless a = 0, in which case c is not unique.
When a = bc,b is said to be a factor of a, or a divisor of a a is a multiple
2,
and when
of b;
0, c is
the quotient of a by
referred to as the factor of a that
is
The
b.
complementary to
integer c
is
often
b.
and \a\ = a if a is negaand a = be, then \a\ > \b\,

for a = be implies that neither b nor c is zero and that \a\ = \b\ \c\.
But
Hence, \c\ = 1 + r, where r > 0, and
\c\ being positive, it is at least +1.
We recall that
\a\
\a\
=
=
|6|(1
Since b
Theorem
Theorem
Since 6
\b\
>
if
positive or zero
is
1-8.
If b
1-9.
If 6
when
When
\b\r.
0, \a\
>
integers d
b,
(ce)d
is
it
0,
thus
evident
that
\b\.
and if c divides b, then c divides a.

and e exist such that a = bd and b =
ce.
and therefore c a.
divides a, and b divides c, then b divides a + c.
divides a, and b does not divide c, then b does not
c(ed),
c.
a,
then a
we should have a
b
If b divides a,
1-7.
a and
Hence, a
divide a
r)
and when
\b\,
Theorem
\a\
It is clear, therefore, that
tive.
would divide
Theorem
-\-
= bd and a + c =
= be and 6e = 6c?
6c?
If 6
c.
did divide a
Hence, b(e
c.
d)
c,
c,
and
c.
1-10.
tive integers 0,
between two of the nonneganot divisible by b unless the two
If 6 ?* 0, the difference
1, 2,
|6|
1 is
integers are equal.
Choose ri and r 2 so that

< ri < |6| and
ments are often combined by writing
<
ri
>
r2
mb >
b
0.
If ri
But
r2
mb, then
ri
= mb
<
r2
< \b\.
< \b\.)
or
common
divisor, of
two
or
Suppose that
Therefore,
this result is contrary to the hypothesis,
\ (n - r 2 ).
A common factor,
(These state-
r2
ri, r 2
more
n >
|6|,
for
and consequently
integers
is
an integer
that divides each of the given integers.
greatest
common
divisor of a set of
two
or
more
integers, not all zero,
8
exists,
is, if it
mon
common
divisor of the set that
divisor of a set of
one of
refer to the
common
greatest
by every com-
The symbol d =
cti.
common
greatest
is
(a h
unchanged
a2
if
a r)
divisor of the set a\,
-15), and 12
(6, 12,
It is evident, too, that the greatest
divisor.
divisor of a set of integers
divisible
is
We notice then that if -\-d is a greatest common

integers, so is d.
It is conventional, however, to
the two integers -\-d and d that is positive as the
divisor of the set.*
any a
is
0,2,
common
of the set is replaced
by
used to denote that d is the

3 =
,
.
.
, ar ; for example,
(36,48).
A common multiple of two or more integers is an integer that is divisible

by each of the given integers.
A least common multiple of two or more integers is, if it exists, a common
multiple that
is
a divisor of every
It is evident that
then
is also.
integer that
common
fits
a ^
Again,
if
is
common multiple of the given

common multiple of a set of
a least
it is
usual in this case to refer to the positive
the definition as the least
multiple
is
unchanged
di. The least common

A unit is an integer that
by
integers.
integers,
if
any
common
multiple of
6,
The
multiple.
of the given integers
15, and 9
divides every integer.
az
is
least
replaced
is 90.
Since
+1 and 1
divide every integer, they are both units.
have but two units, +1 and 1.

be denoted by x. Then x divides +1, and
therefore \x\ < +1.
But because +1 is a unit and x is an integer, +1
divides x and thus +1 < \x\ since x 9^ 0.
Consequently, \x\ = +1, and
Theorem
1-11.
The
rational integers
If there is a third unit, let it
is
+1
The
or
1.
null element is
an integer that divides only
the definition of division that zero divides only

ever, a
no integer
0,
itself.
itself.
We know
For any
a,
from
hoAv-
and consequently any integer divides zero. Therefore,

from zero is a null element. Hence, we have:
different
Theorem 1-12. Zero is the null element of the rational integers.

The associates of an integer are the results of multiplying it by the units.
Thus +6 and 6 are a set of associates. When an integer is divisible by
some integer, it is obviously divisible by each of the associates of that
integer and consequently in any problem it is necessary to show only the
division by one of the associates.
A prime is an integer, not a unit, that is divisible by only its associates
units.
This definition implies that the greatest common divisor
a prime p and an integer a is 1, or the positive associate of p.
and the
of
* This definition and that of a least common multiple are so worded that they will
apply equally well in a domain of algebraic integers where we cannot say of two disFor instance, the set of algebraic
tinct integers that one must be less than the other.
integers of the form a + bi with a and b rational integers is not ordered by the relation
"less than."
An integer that is not the null element, a unit, or a prime is a composite.

Two or more integers are prime to each other, or relatively prime, if their
The integers 6, 9, and 14 are relagreatest common divisor is +1.
tively prime.
Two
or
more
integers are said to be prime each
common
to each,
or relatively
is +1.
The integers 3, 4, and 35 are relatively prime in pairs.
Theorem 1-13. If d is the greatest common divisor of a and b, the
quotients obtained by dividing them by d are relatively prime.
If d = (a, b), let a = a Q d and b = bod.
Let e be any common divisor
Then ao = ek, b = em, and a = dek, b = dem. Hence,
of a and b
de is a common divisor of a and b and must divide d; that is, d = des.
Therefore, es = 1.
But this is impossible unless e = +1. Conse-
prime in pairs,
if
the greatest
divisor of all possible pairs
quently, (o
1-4.
who
The
1.
Among
Principle of Archimedes.
the early mathematicians
contributed to the theory of numbers were Euclid
Archimedes
of numbers.
(c.
225
B.C.).
(c.
300
B.C.),
and
much interested in the theory

many propositions concerning
Euclid was very
He collected and organized

the integers in his " Elements. " On the other hand, although Archimedes
is considered one of the greatest mathematicians of all time, he did

with this branch of the subject and but one theorem in the theory of
numbers bears his name. This theorem is, however, a basic one.
We shall assume the principle of Archimedes extended to include the
rational integers.
This principle states that any integer a either is a
multiple of an integer b ^
or lies between two consecutive multiples of
b; that is, corresponding to each pair of integers a and 6^0, there exists
an integer m such that, for b > 0,
little
1)6
< (m
l)b
mb < a < (m
and, for b
<
0,
mb <
Theorem
integers a
1-14.
and
The theorem
6^0,
Euclid.
of
there exist two integers
= bm
<
m
<
Corresponding to
r such that
two
and
|6|
and m and r are unique.

Except for the unicity of m and r the theorem follows immediately
from the principle of Archimedes, for when b > 0, it is evident that the
inequalities yield
< a mb and a mb < b, and when b < 0, <
a
mb and a
mb < b. Thus r = a mb exists and fulfills the
required conditions. Suppose then that a bm\ + n and a = bm 2 + r 2
where
< r 1} r 2 < \b\. Hence, b{m m 2 ) = r 2 r h and b divides
,
10
r2
= r and
m =
Therefore, r\
r\.
2,
equation shows that mi
We shall
call r
Theorem
the remainder and
An
1-15.
integer a
b (mi
2)
and that mi
=
=
ra 2
= bm
6^0,
this
m the quotient in the division of a by b.

or
is
not prime to an integer
is
according as the remainder in the division of a by b

The theorem of Euclid shows that
Since
0.
<
<
or
is
is
6^0
not prime to
b.
\b\
and hence an integer d divides a and b if and only if d divides both r and
For example, to find out whether or not 152 is prime to 21, just divide
b.
152 by 21, getting the remainder 5. Since (5, 21) = 1, then (152, 21) = 1.
Theorem 1-16. All integers take the form 2n or 2n + 1.
According to the theorem of Euclid, any integer a can be expressed in
the form
< r < 2
a = 2n + r
so that r
An
even
either
is
even integer
is
The
odd.
or
is
1.
one that
a multiple of
is
An
2.
classification of the integers into
integer that
is
not
odd and even integers
was made by the Pythagoreans.
EXERCISES
Prove that the product of any two consecutive integers is divisible by 2.
integers can be written in the form n(n + 1).
Euclid, n has the form 2k or 2k
1, whence
the product has the form 2k(2k + 1) or (2k + l)(2k + 2) = 2 (2k + 1)(& + 1). In
/
either case the product has the factor 2.
2. Show that the sum of an integer and its square is even.
*
3. Show that all integers take the form Sn, Sn + 1, or Sn 1.
4. Prove that the product of any three consecutive integers is divisible by 3.
5. Prove that the square of an odd integer has the form 8n + 1.
6. Prove by induction: 1 + 3 + 6 +
+ n(n + l)/2 = n(n + l)(n + 2)/6
1.
The product of any two consecutive

But then, according to the theorem of
for
n >
0.
When n = 1, the left-hand member is 1 and the right-hand member is l(2)(3)/6 = 1.

The statement is thus true in the first case. Now assume that the statement is true
for n = k.
The first term of this series is 1, and the th term, given by the term
formula n(n
l)/2, is k(k
_o.
1J
1 + 3 +
Hence,
l)/2.
k(k
1)
k(k
l)(fc
where the right-hand member of the equation is the sum

series obtained from the formula for the sum of n terms.
We
must now show by a general method that the sum
the series can be correctly obtained from the
we
shall build
cated
sum
up the
of the first
series of k
A;
terms.
The
+2)
sum
of the first k
terms of the
terms of
To accomplish
this end,
of the first k
formula.
terms by adding the (k + l)st term to the indi+ l)st term is obtained from the term formula
(k
by substituting k
one
member
n and
for
l)(k
if we add this number to

member also to maintain an
But
2)/2.
to the other
it
Thus we have
equality.
,,,,
1+3+-..+ k(k 2+
l)
But the right-hand member

*(*
is (k
we must add
of Eq. (1),
11
+ !)(* +2) +
6
(*
+ l)(k+2)
(k
+ l)(k+2)
6
+
k(k
and
of Eq. (2) can be factored
l) (*+2)
+ !)(*
+2)
. n
(t
1WJ
ON
+ 1)(t+2)
.
(k
g+
.
l)(k+2)
2
(2)
simplified, giving
1\
l).
(Jfe
+ l)(*+2)(*+3)
is correct in the fcth case, (k + 1) (k + 2)

+ 3) /6 is the correct
+ 1 terms. Consequently, unless the sum formula gives this result
+ 1 terms, the sum formula is in error. Substituting + 1 for n in
+ l)(n + 2)/6, we find that it also gives + l)(fc + 2) + 3)/6.
Assuming that the law
sum
(k
(A;
of the first k
sum
for the
of k
fe
the formula n(n

We must conclude that this formula
(fc
(A;
gave the correct result for n = 1,

and, upon the assumption of its truth for n = k, we found it gave the correct result
Hence, being true for n = 1, it is true for n = 2, 3,
for n = ft + 1.
7. Prove by induction: 1 + 2 +
n = n(n + l)/2 for n > 0.
2
8. Prove by induction: l
+ 22 +
+ n 2 = n(n + l)(2n + l)/6 for n > 0.
9. Prove by induction: l 3 + 2 3 +
+ n 3 = n 2 (n + l) 2 /4 for n > 0.
10. Prove by induction that a n b n is divisible by a b if n is a positive integer.
If n = 1, a b is divisible by a b.
is
correct, for it
Assume that ak
bk
(a
ak+i
=
=
Now
b)F(a, b).
&fc +i
the quotient ak and the remainder ak b
b k+1
(a
divide ak+1
b k+1
by a
b,
getting
Thus
+
+
h)a k
6)a*
a 6
fc
6(a
fc
- 6 fc+1
- bk
)
But since a b is a factor of a k b k the distributive law

factor of a k+1 b k+1
Then according to postulate 16, a n
ioT/n > 0.
shows that a
b n is divisible
b is a
by a
n is a positive odd integer, a n + 6 n is divisible by a + 6.

Prove that if n is a positive even integer, a + 6 is a factor of a n 6 n
Prove by induction that 10 n + 3 4 n+2 + 5 is divisible by 9 if n is zero
v^ll. Prove that

12.
if
13.
^
or a
positive integer.
If
0,
then
+3(16)
+5 = 54
+ l)st
the difference between the (n
and this integer is divisible by 9. Now form

and the nth case and simplify by factoring,
getting the result

10+i
+3
+5 -
4-+ 3
The right-hand member
is
(10"
4"+ 2
clearly divisible
by
9.
5)
and hence the theorem
true for
7 2n
b.
7n
3 4 "+ 2
c.
15.
true for
n =
43
But, being true for n
1.
4- +2 (3)
=
5
and applying
is
a multiple of
1, it is
likewise
the positive integers.
Prove that:
+ 16n
14.
a.
two
all
is
+3
Substituting n
the distributive law, the above equation implies that 10

9,
10*(9)
1 is
divisible
by 64
for
n >
0.
+ 3 5 W - 5 is divisible by 24 for n >

+ 5 " +1 is divisible by 14 for n > 0.
0.
Prove that the cube of any integer
is
equal to the difference of the squares of
integers.
To
solve a problem of this type,
we
often set
up a few examples with the purpose
of
12
discovering the law
by
Thus
inspection.
=
=
=
=
23
33
43
p _
Q2
32
32
62
10*
l2
62
reminds us of the sum of the series 1+2

The sequence of integers 1, 3, 6, 10,
+
+ n = n(n + l)/2. Hence, we should like to show that n 3 is the difference
of [n(n + l)/2] 2 and [(n l)n/2] 2
This statement is correct, for
.
V n(n
Show
Show
Show
16.
17.
18.
or 5n
1) 1
that
if
that
if
7z
2x
that
if
an integer
(n
l)rc
n 2 (n 2
a multiple of
is
2î
n2
2w
i;
3,
5,
that the square of an integer that
is
not a multiple of 2 or 3
is
of the
form
1.
Prove that the sum of the odd integers from
20.
then 28z 2 13x 5 is a multiple of 9.

then 14z 2 + 19a: 19 is a multiple of 25.
both a square and a cube, it is of the form bn, bn + 1,
a multiple of
is
1 is
4.
Show
19.
24A;
square, n 2
through In
is
a perfect
21. Prove that every odd cube, n 3 is the sum of n consecutive odd integers.
Find
a corresponding law for an even cube.
22. We have assumed the principle of Archimedes, but just as some of our other
postulates are not independent, it is possible to prove the principle of Archimedes on
the basis of the assumptions already made.
= 0. If a = b,
and 6 > 0. If a < b, then
< a 
= 1, If a > b, there is a positive multiple of b that is less
then b = a < 26 and
,
than
But
a.
<
Thus
in the sequence of multiples of
(a
there
<
(a
>
ab
>
1)6.
1)6
Let a
2.
1)6
is
>
greater than
a.
Let this multiple of 6
Then
mb < a <
Case
6,
a least positive multiple of 6 that
is
be (m
<
and
>
(in
1)6
Then nonnegative
0.
cb
<
\a\
<
(c
c exists so
that
1)6
and
-cb > a
If
> (-c -
1)6
a < cb, let m = c 1, but if a = cb, then a < ( c + 1)6 and m = c.

Let the student complete the proof.
23. If a ^
and |6| > \a\, then 6 does not divide a.
24. What values can you assign to r in order that 4n + r include all odd primes?
26. Using the idea of Exercise 24, write another set of expressions whose values
include
all
26. If
n(n
odd primes.
is
l)/2.
a positive integer, the triangular numbers are given by the formula
Find by
trial
some
integers that are both square
and
triangular.
13
even integers closed under the operation of addition? Do the odd

Is the set of even integers closed under multiplication?
important property that pertains to multiplication does the set of even integers
27. Is the set of
integers have this property?
What
lack?
Peano stated the following postulates, together with the principle of finite
numbers
a. There is a number 1.
+
b. Every number n has a unique successor n
c. The number 1 is not the successor of any number.
d. If n + = m + then n = m.
Define addition and multiplication, and derive the commutative, associative, and
distributive laws for the natural numbers on the basis of these postulates.
28.
induction, to define the natural
CHAPTER
The Form ax
by.
A polynomial in the variables Xi, x 2
xT
a rational integral algebraic expression in these variables.
Thus
~
a n where the exponents of the variable x are
a Qx n
aix n l
2-1.
is
positive integers or zero,
bixy
bzxy
is
is
a polynomial in
a polynomial in x and
The
y.
x.
The
expression b\X 2
degree of a polynomial
is
the
degree in all its variables of its term or terms of highest degree. The first
polynomial is of degree n, and the second is of the fourth degree. The
second, however, is of only the second degree in x, but it is of the third
degree in y. If all the coefficients of a polynomial are integers, it is said
to be an integral polynomial.
x r is a homogeneous polynomial
A form* in the variables x h x 2
in these variables; that is, each term of the polynomial is of the same
The degree of a form is the degree in all its variables of any term
degree.
The polynomial Sx 2 y + 5xy 2 y z is a form of the third
of the form.
All the forms with which we shall be concerned will have
degree.
We shall make use of the form ax + by to show
integers as coefficients.
that the greatest common divisor of two rational integers (not both zero)
,
exists
and
is
a rational integer.
Theorem 2-1. The least positive integer in the set of integers defined
by ax + by, where a and b are not both zero, is the greatest common
divisor of the set.
by when
Consider the set of integers defined by the linear form ax
a and b are constants and x and y are variables whose values are all the
Since there is but a finite number of integers between zero and
integers.
by contains a positive integer,
any positive integer, and since the set ax
this set
Let
has a least positive integer.
L =
This integer
ax
it
be represented by
by
divides every integer of the set because, according to the
This technical use of the word "form" is not to be confused with the ordinary
we have made use of the term. When we say, for instance, that an
integer has the form Qk + 1, the word is synonymous with "mold" or "structure"
and in this case designates that the given integer is always a multiple of six, plus one.
Whenever "form " is used to mean a homogeneous polynomial, the implication will be
*
sense in which
clear
from the
text.
14

of Euclid, corresponding to the integers
theorem
there exist integers
and
n = ax
by\
and L
such that
n =
15
mh +
<
<L
Hence,
ax\
byi
a(zi
mxo)
= m(ax
by
m?/
and
Therefore, r
is
an integer
6(?/i
of the set, but, being less
than L,
must be
0.
The
least positive integer of the set thus divides every integer of the set
and
is
necessarily a
greatest
common
any common
therefore
common
divisor of the set.
But L
divisor of the set ax
is
in the set,
Hence,
divisor of the set divides L.
by, for it satisfies
is
and
the
the stated
by being a positive integer which is a common divisor of the

and which is divisible by every common divisor of the set.
Theorem 2-2. The greatest common divisor of a and b, where not
both are zero, exists and is the least positive integer in the set defined by
definition
set
ax
-f-
by.
and b are determined by the form ax + by, when x = 1,

when
and
x = 0, y = 1, respectively. Hence, Theorem 2-1 shows
0,
y
But L = ax + by and thus srny
that L is a common divisor of a and b.
common divisor of a and b divides L. Therefore, the greatest common
divisor of a and b exists and is L.
Theorem 2-3. If d is the greatest common divisor of a and b, then d is
The
integers a
the greatest
If
(a,
common
b),
divisor of the set, for ax
common
fore, the greatest
Theorem
by
by
divisor of a
divisor of the set ax
ax
d,
is
The
and
and
b d,
+
b,
we
b y).
On
for both a
d.
by.
see that d
is
common
the other hand, any
and
b is a divisor of d.
by necessarily divides
common
2-4.
d(a x
divisor of the set divides a
But any common
divisor of the set of integers ax
so that a
The
b are in the set.
Thus a common
integer d
is,
there-
divisor of the set.
greatest
common
divisor of the set of integers
unique.
Suppose that both d\ and d 2 are greatest common divisors of the set
+ by and that they are positive. Then, according to the definition,
Consequently, d\ < d 2 and d 2 < d\, so that di = d 2
d\ di and d 2 d\.
then,
that the greatest common divisor of a and b is the least
It follows,
positive integer in the set ax + by and that d = (a, b) can be expressed
as a linear function of a and b with integral coefficients.
Thus 4 =
=
=
12(-18) + 20(11).
(12, 20) can be written 4
12(2) + 20(-l)
The fact that the greatest common divisor of any two rational integers
a and b, where not both are zero, can be written in the form ax + by with
x and y rational integers is an important characteristic of the set of
ax
16
Not
all sets of integers have this property.

Consider,
even rational integers. Although the integers 4 and
6 have the greatest common divisor 2 which is in this particular set, 2
cannot be written in the form 2 = 4x + Qy, where x and y are selected
from the set of even integers.
Theorem 2-5. The set of integers ax + by consists of all the multiples
of d = (a, 6).
It has been shown that every integer of the set ax + by is a multiple of
Moreover, every multiple of d = ax + by is in the set, for k(ax Q +
d.
rational integers.
for instance, the set of
by
a(kx
b(ky
).
bers of the set defined
Therefore,
by ax
all
and only multiples
of
d are
mem-
by.
It is interesting to notice that because d is not zero, the set ax + by

never consists of just one integer. Furthermore, the sum and difference
by are also in this set.
of any two integers defined by ax
Any set of at least two elements is called a modul if its elements obey
the associative and commutative laws for addition and if the equation
a + b = c is satisfied by an element of the set whenever two of the three
elements a, b, c are chosen from the set. It is now apparent that the
integers ax + by form a modul.
EXERCISES
Describe the set of integers 3z + 6y.
Use the form ax + by to define a set of integers all of which are even.
3. Use the form ax + by to define a set of integers that are multiples of 5.
Are
all multiples of 5 included?
4. Use the linear form in two variables to determine a set of even integers that are
multiples of 5.
by include all the integers?
5. When will the set ax
1.
2.
2-2.
The Diophantine Equation ax
was the one Greek mathematician
of note
by
n.
Diophantus
who devoted
(c. 275)
himself to algebra.
He solved quadratic equations in a single variable, but he found only one

answer and discarded all but positive rational numbers as solutions. He
even considered types of quadratics in two unknowns and two simultaneous equations of this kind. He is credited with enlarging the concept
Because he sometimes restricted his
of number to include the fractions.
solutions to integers, his name is now attached to the kind of equation
defined below.
Diophantus developed no general method for the solution
of these equations, however. It was not until the Hindus attacked the
problem that such general methods were devised.
A Diophantine equation is a rational integral algebraic equation in
which the coefficients of the variables and the absolute term are integers
and of which the solutions, or values of the variable or variables that

must be
satisfy the equation,
and two-dollar
bills
Theorem
paid in five-dollar
bills?
The
2-6.
Such an equation would be used
integers.
how many ways can $25 be
to solve the problem: In
17
linear
Diophantine equation ax
-\-
by
= n
has a
and only if the greatest common divisor of a and b divides n.

Since we have shown that d (a, b) divides ax + by for all integral
values of x and y, if ax + by = n has a solution, d divides n.
Because d has been proved to be an integer
If d divides n, let n = n d.
of the set ax + by, it follows that n = no(ax + by ) = a(n x ) + b(noy Q ).
This result shows that x = n Q xo, y = n yo is a solution of the equation.
Corollary.
A necessary and sufficient condition that a and b be relatively prime is that there be a solution of the linear Diophantine equation
solution
ax
if
by
1.
if (a, b) = d and ax
by = n
then this very pair satisfies a Q x
b y = n'
where a = a Q d, b = b d, n = n'd. Again, if (a, b) = 1, the equation
kby =
by = n has a solution and that same solution holds for kax
ax
The problem of solving the equation ax
by = n is, therefore,
kn.
reduced to finding a general method for solving the equation in the case
where the coefficients of the variables are relatively prime.
Theorem 2-7. If m divides ab, and m and a are relatively prime, then
The preceding proof
has a solution x
also implies that
m divides
b.
Since (m, a)
y
1,
the equation
mx
ay
Xi,
(m, a)
has a solution x
Then
z/i.
b(mxi
ayi)
mbxi
abyi
and
But
mb, and
m\
ab.
Observe that Theorem

1,
it is
necessary that a
sequently, 6
0,
Therefore,
= +1.
But
b.
when
2-7 holds even
0, for
then
ab implies that ab
if
0.
Con-
and m\b.
Notice also that if m ^ + 1 is prime to an integer n, then m does not

divide n.
On the other hand, if m does not divide n, the integers m and n
need not be relatively prime; for example, 6 \ 15, but (6, 15) = 3.
Corollary 1.
If m is prime to both a and b, then m is prime to ab.
Corollary 2.
If m is prime to a h a 2
, a^, then m is prime to their
,
product.
EXERCISES
*
1. Prove that the product of three consecutive integers is divisible by 6.
Let P = n(n + l)(n + 2). Then P is divisible by both 2 and 3. Hence, P
= 3m. But (2, 3) = 1, and 2 3m. Therefore, 2 m. Thus m = 2s, and
|
3 (2s)
6s.
2k
P =
18
4.
Show that n(n 2 1) is a multiple of 24 if n is odd.

Show that n 5 n is a multiple of 30.
If n 2 is divisible by 3 but not by 4, then n 2 1 is divisible by
6.
Prove that
2.
3.
A number
rational
is
4.
V2 is irrational.
if
and only
if it
can be expressed as a quotient a/b of two

Hence, assume that
a 2 we see that b a 2
Now apply Theorem
rational integers with the denominator 6 different from zero.
y/2 = a/b with (a, b) = 1. From 26 2 =

and show a contradiction.
6. Prove that V^3 and -^2 are irrational.
2-7,
7.
When m is
= +d
9.
=
=
If (a, b)
2.
Show
integer,
b)
1,
if (a, 6)
1,
6,
the greatest
6) is
common
common
2.
62
and a
divisor of a 3
+
+
6 is either 1 or 2.
63
and a 2
6 2 is
6.
The
common
greatest
divisor d of a finite set of integers
On can be expressed in the form
for
Xi,
1,2,
a 2 2
.
"
nX
n, are integers.
oi,
a 2 and a 3 and let d\

,
Then
(a h a 2 ).
exist so that
Any common
divisor of a h a 2
a 3 and hence a divisor of d 2

,
a divisor of
divisor of a 2
the greatest
aiyi
of di, a 2
c.
-\-
Let us start with the integers

2/ 2
(ac, be)
(a
the greatest
aiî
where the
then
0,
1,
2-8.
>
(a
Theorem
and
that
a divisor of a
and
irrational.
an
Thus d =
11. If (a, 6)
?/i
is
common divisor of a + b and a b is either 1 or 2.

b) = 26 and (a + 6) + (a 6) = 2a. Furthermore,
Notice that (a
(2a, 26)
0i, as,
it is
or d.
10. If (a, 6)
12.
a positive integer, prove that if \/m is not an integer,

of two fractions a/b + c/d in their lowest terms
sum
the
8. If
CL2V2
di
and a 3 must be a common divisor of d\ and

But d 2 is itself a common divisor
(di, a 3 ).
and a 3 and therefore d 2

,
three integers.
Since integers
Z\
is
the greatest
and
z 2 exist so
common
divisor of these
that
diZi
a 32 2
d2
a 2 y 2 )zi
a 3z 2
d2
a 3 3
d2
we have
(i2/i
and
finally
diXx
a 2 2
Using induction, we can similarly
raise the
number
of the integers in the
set to n.
Furthermore, as in Theorem 2-6, we can now show that when d =

a2
a n ) and d m, there exist integers x'i7 where i = 1,2,
n, that satisfy the equation a\X\ + a 2 x 2 +
+ a nx n = m and that
if such an equation has a solution in integers, d
m.
ak
Corollary.
If d\ is the greatest common divisor of a h a 2
then the greatest common divisor of d\, a 2 ,
.
a,k, a*+i is the greatest
(oi,
common
divisor of d\
and
a^+i.
Theorem
2-9.
If
19
a and b are relatively prime and x = x y = y is a

by = n, all solutions are given by the equa,
solution of the equation ax
= x + kb, y = y ka for all integral values of k.

f
Supposing that both pairs of integers Xo, yo and x y' satisfy ax
by = n, we have
tions x
axo
by
= n
ax'
by'
xo)
and
so that
a(x'
b(y
and
a{x'
b{y'
provided that 6 ^ 0, it follows that b (V

x = kb, or x' = x
/b5.
Substituting
therefore
a/
xo, we obtain
akb = b(y f y )
Since
(a,
b)
that
1,
After applying the cancellation law,
the original equation
That the
all
and
kb
for
integers x
we find
that
y'
ka.
If b
0,
clearly trivial.
is
kb,
+ by = n for
by substituting
ka satisfy ax
integral values of the parameter k can be determined
them
Xo)
x'
in the given equation.
A Method
2-3.
for Finding a Solution of ax
by
n.
If (a, b)
and d\n, we have shown that, by dividing each term of the equation
ax + by = n by d, we obtain an equivalent equation, that is, one which
is satisfied by all and only the solutions of the original equation.
Consequently, it will be sufficient to solve the equation ax + by = n when
(a, b)
If
1.
n, so that
n n
a, let
and x = n
and the equation
is
satisfied.
If
Jf
n,
then a
1 and we may
5*
suppose that
qia
+n
<
ri
<
|a|
w =
#20
7*2
<
r2
<
|a|
(oia
q 2a
r2
by
= n
exists, riy
<
\a\
<
\b\.
Then
and
Therefore,
ax
Since a solution oi ax
ri?/
az
r2
by substituting
T\
Jf
r 2,
If
-\-
r2
choose
ri)i/
0,
a multiple of a; that is,

thus determined. Then
r% is
and y
is
in the original equation, a value for x is found.
proceed as before, using y and
divisor since r\
<
z as
the variables and
Therefore, where
\a\.
qsri
r3
<
rz
<n
But
if
as the
ELEMENTARY THEORY OF NUMBE
20
and
#47-1
7*4
ny =
q*ri
r4
T2
<
<
r4
ri
we have
(gsî
+ r )
3
and, as above,
r 3z
riw
r4
we continue in this manner, we find that \a\ > r\ > r >

>
> 0. Each of the remainders r 2k+ and r 2 +2 determined when
respectively, are divided by r 2k -i is less than r 2k -\.
r 2 fc-3 and r
We see,
therefore, that each new set of two nonnegative remainders has an upper
bound r 2 k-i which is smaller than the upper bound of the preceding set of
two. Within a finite number of steps the process will necessarily end, for
Supposing that this happens when k = s, the
some r 2 k-i will divide r 2k
and u = m,
equation r 2s -\U + r 2s -%v = r 2s will be solved by letting v =
where r 2s = mr 2s _i. The original variables can then be determined by
If
r 2s _i
2fc
fc
substitution.
23a;
S7y
Solve the Diophantine equation 69x
Example.
Since
Solution.
111)
(69,
3,
we
14)y
(130) (23)
solve
Illy
9000.
the equivalent
equation
Thus
3000.
2Sx
(23
10
Hence,
Uy +
2Sz
10
9)2
10
9z
+ Uw
10
(9
9v
4)t>
5s
1)
IH-l
'
and
14t/
(14
so that
Again,
9z
S)w
and
bw
Therefore,
bw
(5
and
4^
Finally,
4z;
so that
(4
a^
1, w =
+
j
= 2,
l/
and
124.
Therefore,
quently,
the solutions of the original equation are of the form
all
2,
x
y
=
=
124
4
?/
37k
23/c
4,
a;
Conse-

If we wish only the
we take
21
positive integers that satisfy the given equation,
124
37/c
>
23/c
>
-v
and
so that
4
124
Hence, k can have only the values 3, 2, 1, and
0.
EXERCISES
number
Solve the equations and determine the
of solutions for
which both x and
y are positive.
16x + 7y = 601.
2. Ux - 45y = 11.
75z + 91y = 320.
4. 56x - 50y = 74.
40x - 63?/ = 135.
6. 123a; + 57y = 393.
77z + 165?/ = 3553.
Separate 1591 into two parts such that one part is a multiple of 23 and the other
1.
3.
5.
7.
8.
a multiple of 34.
More than
Diophantine equation in more than two
variables or a simultaneous set of such equations can be solved very
2-4.
Two
The
Solution of the Linear Diophantine Equation in
Variables.
single linear
when there is a solution, by the reduction of the augmented

matrix of the equation or system of equations to a canonical form. * This
method enables us to determine a linear transformation the application
of which substitutes for the original equation or equations a set of one or
more equations which is easily solved. We shall not develop any of this
theory, but we shall show in the case of a single equation in three variables
how a linear transformation can be used to facilitate the finding of the
expeditiously,
solutions.
Consider the equation

d\X
If (ai,
a 2 a 3)
,
d,
there
to find the solutions
by
d.
is
a 2y
a zz
= m
no solution unless d
we should
first
(1)
Supposing that d m,
of the equation
m.
divide each
member
Therefore, let us assume that (a h a 2 a 3 ) =

and sufficient that d\ divide a z z
,
(a h a 2 ), it is necessary
there be a solution of the given equation.
atf
has the solutions z

*
H.
New
diw,
dit
t
But
(a 3 , di)
a z w, where
J. S. Smith, "Collected Mathematics Papers," Vol.

York, 1894.
1,
Then
1.
if
d\
m in order
=
1,
that
and therefore
to is
one
Oxford University Press,
22
Hence,
solution.
gives
all
Now
diw, for
and only values
is
a multiple of
d\.
consider the linear transformation
=
=
x
y
We,
integral values of the parameter w,
all
which a 3 z
of z for
of course,
wish u and
+bv
+cv
biu
c\U
(2)
and we know that
to be integers,
c 2x
bic 2
biy
bic 2
b 2y
b 2 c\
Cix
b 2 Ci
Therefore, any choice of integral values for
bic 2
b 2 Ci
61, b 2 , Ci, c 2
that
makes
will force
u and
Applying
this transformation (2) to the original equation (1),
stituting z
v to
diw, the resulting equation
ai(biu
be integers since x and y have only integral values.
b 2 v)
a 2 (ciu
c 2 v)
and sub-
is
a 3 (ô
diw)
= m
or
(01&1
a 2 Ci)u
(ai&2
a 2 c 2 )v
a 3 diw
dit
(3)
since
a dz
= m
^1^0
Because we introduced two independent variables u and
v,
we
are per-
mitted to impose a second condition upon the coefficients of the transformation. We, therefore, set
aib 2
Since
(a,i,
a 2)
di, let a\
doidi,
dQidib 2
a>iC 2
a2
aoarfi,
= a 02 dic
and then
or
01?>2
Let
us, therefore,
6iC 2
6 2 Ci
choose
b2
CLo 2 C 2
= a 02 and
c2
02Ci
condition
determines 61 and C\ since (a i, a Q2 ) = 1. Choose a set

determined for b\ and Ci in which neither bi nor Ci is zero.
we multiply each member of the last equation by d\, we have
this equation
if
Then the
of values so
Now
floi-
becomes
aoi&i
and
a x bi
a%c x
= d

and therefore the transformed equation
d\U
takes the form
(3)
a^diw
23
dito
and
u =
which
Eqs.
a>zw
the same as the value found for
is
(2) for
of the
to
x and
y,
we
Then by
t.
eliminating u from
find that the solutions of the original equation are
form
x
y
z
=
=
=
bit
a 02 v
crfo
a 01 v
20
+ dzbiw
+ a ciw
3
diw
where v and w are the parameters. That all these values of x, y, and z
determined by integral values of the parameters satisfy the equation is
easily verified.
It
evident that the solution of a Diophantine equation in four
is
now be made
depend upon the solution of one with three

we have used equations in two variables
in the above development to solve an equation in three variables.
Example. Solve: Qx + 24?/ - 41z = 91.
Solution.
(6, 24) = 6, and Alz + Qt = 91 has the solutions z =
The equations aib 2 + a 2 c 2 = and 6ic 2 b 2 Ci =
1 Qw, t = 22 41w.
1 become 6b 2 + 24c 2 =
and bi + 4d = 1, so that b 2 = 4, c 2 = 1;
=
=
= t, it is easy to compute the answer in the
1.
Since
C\
u
bi
3,
form x = -3(22 - 41ti>) - 4v = -66 + 123w - 4v, y = 22 - 41w +
variables can
variables in the
v,
2-5.
to
same manner
as
Qw.
Simultaneous Linear Diophantine Equations. A set of two equamay or may not have a solution in integers. Con-
tions in three variables

sider the set
aix
bix
If (ai,
2,
a2
there
CJ3)
is,
+ay+az =
+by+bz =
2
mi
ra 2
mi
d 2 fails to divide
no solution for the set. But even when these conthere need not be a common solution.
Take, for
di does not divide
or
if (61, 62, &s)
of course,
ditions are fulfilled,

instance, the set
2x
2x
If
the second equation
is
+
-
When
together with an
equation, but because
subtracted from the
4y
Any y and z that,
+z=
+ 3z =
Sy
(4, 2)
2,
2s
first,
the result
is
= -
x, satisfy
the given set must satisfy this
the equation has no solution whatever.
each individual equation has a solution, we can always determine
24
common
exist,
by solving one equation and sub-
stituting these values in the second.
Thus, when the solutions of the
the
solutions,
they
if
equation
aix
a 2y
ri
=
=
= mi
a 3z
(5)
are
with parameters
&2?/
>32
= m we
2,
r2
r3
+
+
+
The given equations have

is
(5),
+
+
t\W
t2
(6)
s 3^
common
Wo
v
in bix
Bv = C
solution
Suppose the solutions
solvable.
w =
v =
When
s 2v
and w, upon substituting these expressions

obtain an equation of the form
Aw +
tion
SiV
-\-
exist
and only if this last equaand are of the form
if
Bit
Ait
these values are substituted in the solutions (6) of the

common solutions of (4) take the form
first
equation
the
x
y
z
where there
but one parameter
is
Example.
= X
= Y
= Z
+ Kit
+K
+ K^
2t
t.
Solve the set of equations: 6x
2x
24?/
Sy
412
7z
91,
2.
Substituting in the second equation the solutions of the

one arrived at in the preceding example, we have 2( 66 + 123w
Hence
4w) - 3(22 - 41tc + v)
7(1 - Gw) = 2, or 327w - Uv = 193.
=
-2
=
-77
w
+ life, and v
+ 327/c. Therefore x = -66 + 123(-2
Solution.
first
lift)
(-77
- 4(-77 +
=
327/c)
= -4 + 45/c; y =
124/c; 2 = 1- 6(-2 +
327/c)
27
- 41(-2 + 11/c) +
11/c) = 13 - 66/c.
22
EXERCISES
1.
Solve: 2x
3.
Solve: 10a;
5.
Solve the set: 2x
7.
A room has
to realize exactly
the children,
8.
$100,
If
5y
16?/
3s
17.
2.
4z
4.
2y
4y
7x 100 seats.
$10
if
the
Solve: 3z
6y
2z
11.
= 521.
= 22,
= 11,
6. Solve the set: x + 3y - hz = 29.
5a: - 5y - 3z = 33.
How many men, women, and children should be admitted
men will pay 50 cents each; the women, 20 cents each; and
48.
Solve: 127a:
319?/
43z
7z
cent each?
100 pieces of
how many
money
of each
in denominations of 50 cents,
denomination must there be?
$5,
and $10 are
to
amount
to
CHAPTER
3-1.
The Composite.
most familiar
proved it?
is
Perhaps one of the facts with which you are
that a composite has a prime factor, but have you ever
Theorem
3-1.
Every composite has a prime factor.
Because any negative integer can be expressed as the product of its
positive associate and the unit 1, we shall assume that the composite m
Since m is neither a unit nor a prime, it has a factor that is
is positive.
not a unit or an associate of m. Therefore, let m = /i/ 2 where both
Then /i < m, and /i is either a prime or another
factors are positive.
composite.
If /i is a prime, the theorem is complete, but if /i is a comThen
posite, it, in turn, has a factor other than an associate or a unit.
< /3 < /i. If /3 is not a prime, the line of reasoning
/i = fzfA, where
continues in the above manner, but only for a finite number of steps, for
,
since
m > /i > / >

3
we must
ciates
way
>
arrive at a positive factor f<m-\ that is divisible only by its assoThe integer 2n -i is, therefore, a prime, and by subunits.
and the
stitution
3-2.
it is
obviously shown to be a factor of m.
The Sieve
of Eratosthenes
(c.
230
B.C.).
to test whether or not a positive integer
It is evident that
is
one
a prime would be to
write all the integers from 1 through m; then to leave 2, and strike out
every second integer thereafter; next to leave 3, and strike out every third
integer thereafter; generally, to leave the next unstruck integer p, and
Each
strike out every pth integer thereafter.
integer except
that
is
not
For how long must this

process be continued before we know that m is a prime? Eratosthenes
answered this question by means of the following theorem and thus
crossed off
by
this process is obviously a prime.
presented a useful test for a prime:
Theorem
3-2.
A positive integer m is prime if it has no positive prime
factor less than or equal to /,
where I
is
the greatest integer such that I 2
than or equal to m.
Suppose that m is not a prime but is a composite. Then m has a prime
This prime factor p must be greater than / according to the
factor.
is less
25
26
hypothesis of the theorem. But when m = pf, the complementary factor

/ must also be greater than /, for if it were not, a prime factor of it, which
would necessarily be a prime factor of m, would be less than or equal to /.
Therefore, both p and / are at least equal to /
1, and m = pf shows
that
m >
(I
l) 2
This statement is contrary to the hypothesis which says that (I + l) 2 >

m, and hence m is a prime.
Example. To test 2503 to see whether or not it is a prime, we try to
divide 2503 by the primes 2, 3, 5, through 47, since in this case I is 50.
Because none of these primes is a divisor of 2503, this integer is a prime.
3-3. The Number of Primes.
Euclid included the following theorem
in his " Elements."
Theorem 3-3. The number of primes is infinite.
So that we may understand the method of this proof, let us recall that a
set of elements is said to be infinite if it has a proper subset, that is, a
subset different from the whole set, which can be put into one-to-one
correspondence with the whole set. Otherwise, a set is said to be finite.
Consequently, if it can be shown that, in addition to a finite subset of a
set, there is always another element that belongs to the set, the set is
infinite.
Suppose that the number of primes is finite. Then there is a largest

by taking the product of all the primes
prime p. Form the integer
from 2 through p and adding the integer 1 thus
(2
N is either a prime or a composite.
p)
But
if
which is contrary to the assumption.

This factor is not one
tive prime factor.
If
p,
5,
p, or
N
N
is
a prime,
is
a composite,
of the
according to the distributive law
it
it is
greater than
has a posiprimes in the set 2, 3,

it
would divide
1,
which
is
a prime larger than p. Because we have

shown that when we assume the number of primes is finite, we can always
find a positive prime that was not previously counted, the number of
impossible.
primes
Consequently,
it is
is infinite.
moreover, that Euclid's proof shows that if p h p 2

,
prime lies in the interval from p r
1
1.*
This interval provided one of the first
pr
answers to the question, Where is the next prime larger than a given
prime?
It is evident,
p r are the first

through pip 2
'
'
r primes, the next
* D. N. Lehmer has compiled a "List of Prime Numbers from

Carnegie Institution of Washington, Washington, D.C., 1914.
to 10,006,721,"
Theorem
27
The number of primes of the form An 1 is infinite.

number of positive primes of the form 4n 1 is finite, and
3-4.
Suppose the
p be the largest one of them. Determine all such primes from 1

through p, and form their product. Then construct the integer N by
multiplying this product by 4 and subtracting 1 from the result. Thus
let
(4
p)
one of the form An 1 that was not previously

counted, for it is larger than p. But if iV is a composite, it has an odd
prime factor. Consequently, this factor has the form 4n 1 or 4n + 1.
If it is of the form An 1, it is not one of the primes from 1 through p, or
by the distributive law it would divide 1, and that is impossible. ThereBut
fore, if the prime factor is of this form, it is a prime larger than p.
has the form An
if the prime factor of
1, there is a complementary
factor of the form An I, for the complementary factor must also be odd,
If
AT"
is
a prime,
it is
and the product

(4s
does not yield the form An

(4s
1)(4
1,
1)
4/c
whereas the product
l)(At
1)
Ak
We observe immediately that this complementary factor At I

prime
factor.
has a
The argument is now repeated with reference to this last prime factor.
Since the size of each of the positive factors having the form An 1
decreases as the argument proceeds, we must finally arrive at a factor
having the form An 1 that has only itself and 1 as positive factors.
This integer is, therefore, a prime. It is of the proper form, and it is not
one of the primes from 1 through p. It is, therefore, larger than p, and,
as a result, the number of primes of the form An 1 is infinite.
Dirichlet (1805 1859) showed by using advanced methods that the
arithmetic progression determined by an + b, where a and b are fixed
integers and (a, 6) = 1, while n has all integral values, contains an infinite
number of primes. Although for many years elementary proofs were
known for special cases of this theorem, such as An 1, no proof of
the general theorem that did not make use of the methods of analysis
employed by Dirichlet had been developed until 1949. About this time
certain mathematicians (Paul Erdos, Atle Selberg, H. N. Shapiro, and
H. Zassenhaus), working on the prime number theorem and in the theory
of algebraic numbers, developed methods that produce new proofs of
Dirichlet's theorem.
These are, however, beyond the scope of this text.
does.
28
Theorem
If 2 n
3-5.
a prime, n
1 is
is
a prime.
a composite and that n = nin 2 where 1 < n\ < n.

Then 2 nin 1 has the factor 2 n 1 which is neither 1 nor 2 n 1.
Consequently, n must be a prime if 2 n 1 is a prime.
Notice, however, that when p is a prime, 2 P 1 need not be a prime,
Suppose that n
is
when p =
for
11,
2 11
has the factor 23.
EXERCISES
1.
6n
Show
that
all
2.
Show
3.
Prove that n 4
4.
5.
primes except 2,
are represented
by the forms 6n
and
1.
that there
is
an
is
number
infinite
primes of the form 6n
of
composite when n
>
1.
1.
Show that if 2 n + 1 is an odd prime, n is a power of 2.

Determine whether or not 1781 and 4079 are primes.
We
Unique Factorization.
3-4.
the case where
have previously proved by means of

ab and (m, a) = 1, then m\b.
In
theorem can also be proved in the follow-
when
a rather long discussion that
m is a prime this
ing manner:
Theorem
3-6.
If
is
a prime and p does not divide a or
b,
then p does
not divide ab.
Take
a, 6, and p positive integers, and consider the case in which

and b < p. We shall prove that the prime p does not divide ab
by assuming the contrary and showing that we arrive at an impossibility.
According to the principle of Archimedes there exists a positive integer k
<
p,
such that for a
>
ka
where the equality sign
is
<
<
(k
l)a
omitted because p
<
<
(p
ka
<
is
a prime.
Therefore
and
But
if
p divides
ab,
ka)b
<
ab
then p divides pb kab and this positive integer is a

This argument leads to the conclusion
is ab.
smaller multiple of b than
is always a positive multiple of b that is divisible by p and is at

same time smaller than the one last found. Accordingly there would
be an infinite number of multiples of b between b and ab. The result is,
of course, impossible, and consequently p Jf ab.
Suppose now that not both a and b are less than p. Then
that there
the
a
b
= mip
= mp
2
+
+
ri
r2
<
ri,
r2
< p
29
Thus
= Kp
ab
rir 2
and r 2 are positive and

and therefore p \ ab.
The result implies that if p ab, p divides at least one of a and b.
Theorem 3-7. If p is a prime and p does not divide a where i = 1,2,
n, then p does not divide the product of the a
an
But p does
a n then p divides a or a 2 a 3
If p divides aia 2
In like manner p divides
an
not divide oi, and so it divides a 2 a%
Finally, p would have to divide a n but it does not, and
an
o 2 or a 3
and
if p
than
less
ab, it follows that
We know
p.
nr 2 where both
r\
this division is impossible,

\
t-,
{.
p does not divide the given product.

Theorem 3-8. Fundamental theorem of arithmetic. Except for
associated primes and the order of the factors, a composite can be factored
into primes in one and only one way.
Assuming that m is a positive composite, according to Theorem 3-1, m
has a positive prime factor pi so that m = piirii. If mi is not a prime, we
again apply this theorem and obtain mi = p 2 m 2 and m = pip 2 m 2
Since
m > mi > m 2 >
> 0, we need carry out this process only a finite
number of times until we have factored m into positive primes. Thus
therefore
m =
PiPi
'
Pn-
'
Suppose that there

that
m=
qiq 2
a second factorization of
is
r,
>
where n
P1P2
pn
Therefore, p\ divides the product qiq 2
= 1, 2,
r.
+ 1 and qi, and
primes qh where j
divisors of qi are
Pi
'
m into positive primes so
Then
r.
qiq2
'
qr
qr and must divide one of the

Suppose pi dividesî. The only
since pi and qi are positive primes,
Hence,
q\.
P2PZ
Pn
qiqz
'
'
qr
Continuing this process, we see that when n = r, each p where i = 1,

n, is identical with a corresponding qj} but when n > r, we
2,
{,
finally
have
Pr+l
and each
of the integers
'
Pn
p n must be a
Therefore, the
primes is unique.
It is obvious, then, that if we gather the equal primes together into the
power form p s any composite m can be written in the form pi n ^p 2 n2
n
p r *, where the factorization is unique except for the use of an associate
in the place of any prime and the order of the factors.
p r+i,
factorization of the positive composite
The reader may
unit.
m into positive
well
remember that an
expression shows factors of
all
algebraic factorization of an
integers represented
by the expression, and
30
the factorization may, therefore, be used to demonstrate a property of all

-\- I and a 1, we
never
1
represents a prime. In factoring
can show that, for a > 2, a
an integer that is written in polynomial form into the product of two
integers, however, one must recall that while the expression may not be
these integers.
For instance, by using the factors a

2
may be factorable in just

a particular integer represented by the
factorable algebraically into rational factors or
one
way
besides itself
expression
may have
and
1,
several factorizations.
It is
evident that although
+ 1 lacks factors in our domain, yet, for the particular

+ 1 = 26 has the factors 2 and 13. Again, an integer
the expression a 2
value a = 5, a 2
represented by a 2
may have
by the
factors other than those given
2
1 of a 1.
For instance, when
only rational factors a 1 and a
2
=
=
1
24 and this integer has the factors 3 and 8, 2 and 12, as
a
5, a
well as 4 and 6.
Hence, we must be careful to avoid drawing the conclu-
sion that an integer lacks factors just because the algebraic expression of
which
it is
a value
fails
to produce the factors.
EXERCISES
Show that if both x and y are odd, there is no z such that x 2 + y 2 = z 2
Show that if {a, b) = 1 and ab = c n then a = s n b = t n
3. Show that an integer can be represented as a difference of two squares if and only
Show also that the representation is unique when
it is of the form 2n + 1 or 4n.
1.
2.
if
is a prime.
Find the positive integers x that make x(x
Find the positive integers x that make x(x
the integer
4.
5.
Method
+ 42)
+ 84)
a perfect square.
a perfect square.
Although it was fairly

prime
of the form 4ft -j- 1
easy to show
the problem of
difference
squares,
represented
uniquely
as
of
two
can be
a
expressing a prime of the form 4ft + 1 as a sum of two squares attracted
the attention of mathematicians for many years. Fermat claimed in
1641 that he had proved the theorem by the inductive method, called the
method of infinite descent, which was exhibited in the proof of Theorem
3-6.
Thus he said, " If a prime 4n + 1 is not a sum of two squares, there
is a smaller prime of the same nature, then a third still smaller, etc., until
the number 5 is reached."* Since 5 is the smallest positive prime of the
form 4n + 1, and since it can be written 5 = l 2 + 2 2 the method leads
to a contradiction and consequently each prime of the form 4ft + 1 is a
sum of two squares. Whether or not Fermat actually proved the theorem
we do not know, but Euler is generally credited with developing the first
proof together with the proof of the unicity of the representation.
It
was not until 1843, however, that he established a point that he needed
3-5. Fermat's
of Infinite Descent.
in Exercise 3 of the last set that a
* L.
E. Dickson, " History of the Theory of
Numbers,"
Vol. 2, p. 228.
31
In the meantime by 1801 Gauss had applied
for a rigorous proof.
the theory of binary biquadratic forms to show the existence of the

representation. *
But Fermat's method
of infinite descent used, as
proof of the above theorem

it
is
more
could be applied with
of particular interest.
prove the impossibility of certain
facility to
which Theorem 3-6
he claimed, in the
He admitted that
an example, than to show affirmative

statements. He made use of the method in the one specimen of a
detailed proof that has been left by him when he showed that if integers
relations, of
is
are the sides of a right triangle, the area cannot be a perfect square.
a, b, c
The Euclidean Algorithm. An algorithm is a method of obtaining

a result by repeated applications of an operation and is such that an
element determined by one application of the operation is used at least
once in a succeeding application until the desired result has been found.
It is important to have a method for finding the greatest common
3-6.
divisor of
mon
two
divisor
is
integers a
and
b.
If
one integer
is zero,
either or both integers are negative, the greatest
same
as
it is
common
divisor
Taking both a and

up the Euclidean algorithm
a and b.
for the positive associates.
therefore, with a
the greatest
the greatest com-
evidently the positive associate of the other integer, and
>
we
b,
common
shall set
divisor of
is
if
the
b positive,
for finding
Applying the theorem of Euclid, we have
= rriib + r\
= ra ri + r 2
= m r2 + r3
a
b
ri
Because
>
r\
>
r2
cations of the theorem

of r k -i.
Then
r k is
rrtiTi-i
+ Ti
= m k r k -i +
= m k+x r k
r k -!
Ti-2
r k -2
<n 
we must
the greatest
>
<
<
rk
r*
>
Ti
rk
0, it is
<
<
Ti-i
r*_i
evident that after k appli-
arrive at a remainder r k that
common
divisor of a
and
b,
is
a divisor
for first of
all
In like manner because r k divides r k -i and

Using the steps of the algorithm in reverse
rfc-2, it must divide r _ 3
order, we observe, therefore, that r k divides each r* and finally that it is a
common divisor of b and a. But any common divisor of a and b divides
Proceeding in this manner,
ri, and by the second step it also divides r 2
we see that any common divisor of a and b divides r k Therefore, r k is the
greatest common divisor of a and b.
The first step of the algorithm shows that n can be expressed linearly
rk
r k -2 since it divides r&_i.

fc
* Ibid., p. 233.
32
terms of a and
in
b,
for
mj)
h so that r 2
ri
Moreover,
2 a.
r2
2r
ra 2 (a
raib)
+ mim
(1
2 )b
+Mb
+Nb
= Mia
n = N a
Ti-i
we
Thus from the equations
see that since

ri+ i
u-i
mi+iTi
r l+1
Qid
Q 2b
then
Hence, as was shown by a different method in Chap. 2, r k the greatest

common divisor of a and b, can be expressed as a linear function of a and
b with integral coefficients.
In the special case where b a so that ri = 0, it is evident that b is the
,
and b and that b = a(0)

b(l).
applying the unique factorization theorem, it is also apparent that
if the integers a i} where i = 1, 2,
n, are factored into powers of
distinct primes so that
greatest
common
divisor of a
By
en
vr^
=1
the exponents being positive integers or zero, the greatest
common
divisor
r
s
of the di is
3
Pj in
pj
] [
>,
where each
the factorizations of the
the smallest exponent that occurs for
a;.
Find the greatest
Example.
573
Sj is
=1
291
282,
Hence, the greatest
291
common
common
282
divisor
9,
divisor of 573
282
(31) (9)
and 291.
3,
(3) (3)
is 3.
EXERCISES
Find the greatest common divisor of 5040 and 4704.
Express the greatest common divisor of 168 and 525 as a linear function of these
numbers.
ba
3. If d = (a, b), then d is the number of integers in the sequence a, 2a, 3a,
Prove it.
that are divisible by b.
4. Show that the sum of a finite number of rational fractions in their lowest terms
cannot be an integer if the denominators are prime each to each.
2"
5. Prove that two integers having the form 2
+ 1 are relatively prime. From
1.
2.
this fact develop another proof that there are infinitely
the form 2 2 "
are called Fermat numbers.
mula generated primes
He
for all integral values of n.)
many
primes.
(Integers of
believed erroneously that the for-
33
Prove that the number of divisions required to find the greatest common divisor
two positive integers written in the scale of 10 by means of the Euclidean algorithm
6.
of
does not exceed
five
times the number of digits in the smaller integer.
3-7. The Least Common Multiple

Theorem 3-9. If d is the greatest common
integers a
and
and
that a
b so
a Q d, b
divisor of the positive
the least
b d,
common
multiple of a
b is a b d.
Any common
But
form ma.
because (a
multiple of a and b
it
a multiple of a and therefore has the
is
also a multiple of b
is
follows that b
1, it
b d ma d, and
common multiple
Thus
b d.
Hence, any
m.
and b has the form ka bod.

ka b d, and in order that an integer be a common
However, a
multiple of a and b, it is sufficient that it have the factors a and b d.
Therefore, the positive common multiple of a and b that divides every
of a
common
multiple
Corollary
multiple
Corollary
common
W-
if
integer
and only
is
common
the least
3-10.
If
common
and
the least
common
,
the least
is
multiple of their least
is
the least
that
is
it
is
a multiple of a 3
multiple of
divides
all
it is
is
an
cin-i,
is
their
common
the least
a multiple
and a 3 is a multiple of L.
oi, a 2 and a 3 for it is a com-
of a h a 2
common
multiples.
of integers
multiple of a n and the integer
ccmmon multiple of the set Oi, a

Any common multiple of the nonzero
the least
Corollary.
an
multiple
In like manner we can extend the theorem to show that:

Theorem 3-11. The least common multiple of the set
.
then
common
multiple of them, and
Thus any common multiple
L 12
is
common
besides being a multiple of Li 2 an integer
a2
Then L is a common
of L 12 and a 3
But any common multiple of a and a 2 is a
common multiple L i2 For the same reason, if
common multiple
Therefore,
multiple of a\ and a 2
multiple of a h a 2 and a 3
Let L be the least
ai,
is
a%.
multiple of a 1} a 2 and a 3
mon
multiple of the nonzero
the smallest positive integer that
if it is
multiple of them.
the least
of L.
common
a and b are relatively prime, their least
An
2.
and
Theorem
of L12
is
If
1.
the positive associate of their product ab.
is
integers a
2,
a n _i.
integers a h a 2
a multiple of the smallest positive integer of which each a z
is
a factor.
It is
obvious that when a {
]} p/
3
ni
>,
where
1,
2,
n, the
=1
r
least
common
multiple of the a {
is
J J p/**,
where Gj
is
the exponent of the
highest power of p3 occurring in the factorizations of the a z

-
-.
34
EXERCISES
1.
If
where a and
(a, b),
show that ab
b are positive,
d and the least common multiple of a and b.

k
2. Show that if 2 is the highest power of 2 that
2, 3,
then that integer of the set that
n,
only integer in the set that

3. Prove that \
\
by 2 k
divisible
is
equal to the product of
is
a factor of an integer of the set 1,

k
is 2 itself and is the
is
a multiple of 2
is
fc
1/n is not an integer.

necessary and sufficient condition that a positive common multiple
of the
n, be their least common multiple is that
nonzero integers a t where i = 1, 2,
Prove it.
the quotients M/ca be relatively prime.
4.
-,
6.
Show that
the product of n positive integers
common multiple and
common
the greatest
is
equal to the product of their least
divisor of all possible products that can be
formed by taking the integers n 1 at a time.

r, is a set of integers having (m,, m,) =
6. If rrii, where i = 1, 2,
r, and if A is the least common multiple of m h m 2
j = 1, 2,
and m r then the greatest common divisor of rm and Ai is the
m i+h
.
mon
di 2
The Divisors
d<,-i, d*,*+i,
we mean
and we designate
the
by r(ra).
it
di r
m_i,
least
com-
When we refer to the number of

number of positive divisors of that
Thus r (6) = r( 6) =4. We also
an Integer.
of
where
di }-,
divisors of an integer m,
integer
multiple of da,
3-8.
use the symbol <r(m) to represent the sum of the positive divisors of m.
It is evident that the number of divisors of a prime p is 2 and the sum
is just p + 1 if p > 0.
Moreover, the divisors of p a are 1, p, p 2
2
a + 1, and their sum is 1 + P + P +
of the divisors
is
Theorem
of
3-12.
m is (a\ +
<m
If
1)(2
Pl
1)
i+i
Pi
'
ai
pi p2
(r
p2
P2
pr
ar
r+i
Pr
their
number
a
-
the
Pr
and so
and the sum
1)
2 +i
az
number
of divisors
of the divisors of
m is
If
m =
J Pi
ai
,
it is
evident that each divisor of
m which is also a divisor
i = l
of p\ ai is a
term
in the expression
Pi
Pi
Moreover, only these terms are divisors

ner the terms of
1
P2
P2
'
of
'
both
Pi
ai
m and pi
(1)
ai .
In like man-
(2)
P2">
If we multiply these
and only the divisors common to m and p 2 a2
of
the
result
is
a divisor of pi ai p 2 a2
terms
the
each
of
together,
two sums
and furthermore these terms give all the common divisors of pi ai p 2 a and
This product of (1) and (2) is
m.
give
all
1
Pi
Pi
'
'
Pi"
Vi
P1P2
Pi
35
+ ' +
+
P2 +
2>2
Pi
ai
ai
'
'
Pi p2
a*
it has (i + l)(a 2 + 1) terms.

Continuing the reasoning in this manner, we see that the terms of the
expansion of the product
and
(1
give
Pi
all
'
and only
Pl 0(l
tt
P2
'
'
'
P2
'
(1
+ Vt +
1)
'
'
Pr')
(3)
There are
divisors of m.
r(ro)
terms in this
The sum
(3)
and
is,
(i
l)(a 2
1)
(,
and therefore that
result,
of the divisors of
m is the
the
is
sum
11 (*
t-i
!)
number
of the
of divisors of m.
terms in the same product
therefore,
r
<r(m)
= \\
i
(1
+P +
t
Pf) = J! g
-1
^~T
EXERCISES
r
Prove that the sum of the nth powers of the divisors of
1.
p.(i+D
n
i
P<
=l
Find the smallest positive integer with 6
2.
Since 6
=6-1 =2-3 =
the required
number
smaller than 2 5
(ai
l)(a 2
are either the pair
5,
l),
m =
JJ
i=l
p*"** is
divisors.
the exponents of the prime factors of
or the pair
1, 2.
Evidently 2 2
12
is
Find some integers having 10 divisors.

Find the smallest positive integer with 15 divisors.
-1 n -1
have
5. Prove that if n = ninz, with n\ > ni > 1, so that both 2 n_1 and 2 n
3
n divisors, then 2 n_1 > 2 n i- 1 3 n 2 _1
n -1 3"~ 1 5 n -1 is less
6. Prove that if n = nintfiz, with 1 < n\ < ni < n 3 then 2
than 2 2 n 3-i3H-i and both have n divisors. Consider the cases in which n 2 = n%, in
which ni = n 2 and in which n\ = ni n^.
7. Find by trial positive integers n such that the sum of the divisors of n is a
3.
4.
ri
perfect square.
Find all primes that are one less than a perfect square. Is there a prime that
than a perfect cube? Can you find a prime that is one less than n 4 ? Prove
a general statement to cover these results.
9. Find by trial positive integers n such that the sum of the divisors of n is a
8.
is
one
less
multiple of n.
Prove that a positive integer is the sum of consecutive positive integers if and
not a power of 2.
11. Prove that the number of divisors of a positive integer is odd or even according
as the integer is or is not a square.
10.
only
if it is
36
Prove that the product of the divisors of a positive integer n
12.
the
number
n 8/2 where
,
s is
Prove that if r is the number of distinct prime factors of n > 0, the number of
which n can be factored into two relatively prime factors is 2 r_1
13.
ways
is
of divisors of n.
in
Examine the
Numbers.
6, and you will notice

Observe also that the sum of the
divisors of 28 is 56.
A positive integer is said to be perfect if the sum of
its divisors is double itself.
Besides 6 and 28, the integers
496, 8128,
to
Hi* 3??ZNotice that all these
130,816, 2,096,128, and 33,550,336 are perfect.
perfect numbers are even.
Although no odd perfect number has ever
been found, mathematicians have not been able to prove that none exists.
It has been shown, however, that if one does exist, it is greater than 10
billion.*
On the other hand, we can prove:
Theorem 3-13. An even integer is perfect if and only if it has the
~
form 2 P 1 (2 P 1), where 2 P 1 is a prime.
If an integer is of the given form, which is due to Euclid, the sum of its
3-9. Perfect
the interesting fact that their
sum is
divisors of
12.
'
divisors
'
'
is
(1
2^- 1 )(l
2P
1)
2^(2^
1)
and hence the integer is perfect.

The converse was first proved by Euler, but we shall present a method
due essentially to L. E. Dickson. Assume that m is of the form 2 k q,
where q is odd. If, further, m is perfect, by letting s represent the sum of
all the divisors of q except q itself, we have
2 k+i q
But 2 k+l
1 is
2 *+i
substituting this value in the

q
l)(g
+ s)
+
2 k+l divides q
odd, and hence

q
Upon
so that
2 k+l n
first
(2* +1
s,
equation,
we obtain
l)n
But by subtracting q from q + s we

s = n = q.
Then the equa l)q gives 1 = 2 k+1 1, and k = 0. In this case the
tion q = (2
But if we suppose that n is a
original integer 2 k q would not be even.
divisor of q that is not q and not 1, then n = s is at least the sum of the
However, it is impossible that n > n + 1. Accorddivisors n and 1.
Thus q is a prime.
ingly, n = l, and the only divisors of q are q and 1.
Consequently, n
find that s
n.
k+l
is
a divisor of
Now
let
q.
us suppose that
1
Since q = 2 k+l 1, the exponent k
even perfect number has the form 2 P_1 (2 P
is
a prime.
1), in
Therefore, every
which both 2 P
and
p are primes.
*
Am. Math. Soc, Vol. 49, No. 10, pp. 712-718,

Am. Math. Monthly, Vol. 56, No. 9, pp. 628-629, 1949.
A. Brauer, Bull.
hard,
1943.
H. A. Bern-
37
less than or equal to 257 are

Marin Mersenne, who
mathematician
called
Mersenne was interested in discovering which
lived from 1588 to 1648.
of these numbers 2 P 1 are primes and made certain decisions about
their primality even though he lacked the modern facilities for testing
numbers of this magnitude. It has taken many years to decide that
In 1944, six Mersenne
exactly 12f of the Mersenne numbers are primes.
=
the
primes
to
numbers corresponding
157, 167, 193, 199, 227, and 229
p
The
tremendous
task of investigating the charremained to be tested.
H. S. Uhler. He completed
numbers
was
carried
by
out
acter of these 6
finding
prime
no
among them. We now know that
the work in 1947,
=
only the primes p
2, 3, 5, 7, 13, 17, 19, 31, 61, 89, 107, and 127 deter.mine Mersenne primes of the form 2 P 1 and hence that only these 12
Mersenne numbers yield perfect numbers. The largest of these perfect
Integers of the form 2 P
Mer serine numbers*
with p a prime
after the
digits J when written in the scale of 10.

The author is indebted to D. H. Lehmer for the information that during
the year 1952 it was demonstrated by the electronic calculator SWAC
numbers has 77
that 2 P
is
more
quently, five
prime for p
521, 607, 1279, 2203,
integers are
now known
and 2281.
to be perfect.
Conse-
It is particu-
gaps between some of these primes.

Doubtless you have noticed that each of the first seven perfect numbers
given above ends in 6 or 28. It has been proved that all perfect numbers
of Euclid's type end in this way.
larly interesting to notice the large
an integer n the sum of whose positive

first integer n the sum of whose divisors
is Sn is 120.
Fermat found the second one, which is 672. The third is
The first integer the sum of whose divisors is four times itself
523,776.
Recently some new multiply perfect numbers have been
is 30,240.
multiply perfect number
divisors
is
is
The
a multiple of n.
discovered.
Two
integers are said to be amicable
if
their
sum
is
sum
the
of the
numbers is 220 and

284.
Another pair, 17,296 and 18,416, was found by Fermat.
3-10. Scales of Notation.
Have you thought of 5347 in the form of
the polynomial 5x z + Sx 2 + 4x + 7, where x = 10?
The
divisors of each one.
Theorem
form
m =
3-14.
are such that

*
+
<
Any
a\r n
~l
<
smallest pair of amicable
m can be written uniquely in the

a n where r > 1 and the coefficients
positive integer
+
r
and
+
<
a$
<
r for i
1, 2,
n.
R. C. Archibald, Scripta Mathematica, Vol. 3, No. 2, pp. 112-119, 1937.

H. S.
f D. H. Lehmer, Bull. Am. Math. Soc, Vol. 53, No. 2, pp. 164-169, 1947.
Uhler, Bull. Am. Math. Soc, Vol. 53, No. 2, pp. 163-164, 1947; ibid., Vol. 54, No. 4.
pp. 378-380, 1948; Scripta Mathematica, Vol. 18, No. 2, pp. 122-131, 1952; Proc.
Natl. Acad. Sci. U.S., Vol. 34, No. 3, pp. 102-103, 1948.
t H. Gupta, Am. Math. Monthly, Vol. 42, No. 3, pp. 163-164, 1935.
B. Franqui and M. Garcia, Am. Math. Monthly, Vol. 60, No. 7, pp. 459-462, 1953.
38
When m
rn
written in the form a
is
air"
-+-
conditions designated in the theorem,
-1
-\-
a n with the
said to be expressed in the
is
and r is called the base, or radix.

For a given m and r > 1, we know that
scale of r
m =
qir
-\-
an
qi
qtf
a n _i
ft
ft fir
=
=
fti-l
?n
<
where
>
<
a*
arrive at q n
r,
for
which
1, 2,
qn - 2
m =
This representation
rn
air 71
<
bj
<
"1
and
r(a r n-1
is
n- s
tion
is
>
0,
n.
less
Furthermore, we must finally

than r, because m > q > q 2 >
+a
+ ar +
a Qr
a xrn
a r2
an
a n -i
a,
and
a n -j
and then
aorn
Since r
=
=
0, 1,
s,
find that each 6 S _;
>
Ol
unique, for
rn
Therefore, r divides b s
we
a n -i
Then
0.
g n _i
with
ft^r
and
positive
is
_ s _!
b r
-1
b r8
+ a
vl'^
and
bs
b^ '
8
++?>
then
s,
s- 1
>
if
until,
a2
an
if
n =
(ln
fe 8
_i)
bs
an
Continuing in this manner,

= a but if
s, we have b
,
_s _ i
the remaining coefficients must
all
be
and the representa-
unique.
on the basis of this theorem that we know we can write an integer

one way in the Hindu- Arabic system, which uses the scale of 10
and the digits 0, 1, 2, 3, 4, 5, 6, 7, 8, and 9. The integer 363 in the
ordinary scale becomes 2423 in the scale of 5, while it is 101,101,011 in the
It is
in just
scale of 2.
Corollary.
one
way
as a
Any positive integer n can be

sum of distinct powers of 2.
expressed in one and only
We can apply this corollary to show that for weighing approximately

any load not exceeding 127 lb but seven weights of 1, 2, 4, 8, 16,32, and 64 lb
each are needed for the
scales, for
127
is
written 1,111,111 in the scale of
2.
39
It is interesting to notice that the Egyptians recognized that an integer

can be expressed as a sum of powers of 2 and that they made use of this
Suppose, for instance, that 237 was to be
fact in doing multiplication.
1
They
would
first determine that 45 = 32 + 8 + 4
45.
by
multiplied
multiples
of
237
the
process
of
the
corresponding
by
compute
then
and
Accordingly, the calculations would be:
doubling.
4
8
16
32
Then
7584
all
237
474
948
1896
3792
7584
that was necessary was to find the sum, 237
948
1896
10,665.
They carried out division in a similar manner. If 539 was to be divided

by 41, they used the process of doubling on the divisor until they could
find suitable multiples of it which, when added, would give a result smaller
than 539 but less than 41 units from it. Their work might be indicated
as follows:
41
But 328
or 539
164
13(41)
+ 41 =
+ 6.
533,
82
164
328
and so 539 = 8(41)
4(41)
1(41)
6,
The digital idea is the basis for many of our computing machines.
Indeed even the abacus, which was probably the earliest mathematical
machine, makes use of it. Each rod of the abacus corresponds to the
place a digit occupies
when the number
is
written in the scale of 10, and
Corresponding to the
machines use the contrivance of a
the beads on the rods correspond to the digits.

digital places,
some
of the calculating
gearwheels that rotate when certain levers are pressed. Each

has 10 teeth to correspond to the 10 digits, and upon a
cogwheel
such
complete rotation of a given wheel there is a mechanism that turns the
wheel corresponding to the next higher digital place through one-tenth
series of
of a rotation.
In setting up some of the modern electrical calculating machines it has

been found practical to use the binary rather than the denary system of
In spite of the increased number of digital positions necessary
notation.
for representing any number greater than one in the scale of 2 as compared with the scale of 10, fewer electronic tubes are required by machines
40
constructed on the basis of this system, for but two digits,
needed
The
in
and
1,
are
The mechanism is, therefore, decidedly simpler.

the Institute for Advanced Study in Princeton is of this
each position.
calculator at
type.*
EXERCISES
1. Prove that any positive integer can be expressed uniquely as a sum of distinct
powers of 3 with coefficient 1, 0, or +1. From this representation show that a set
of five weights is sufficient to weigh any load of at most 121 lb if a balance scale having
two pans is used.
Then write 42 and 352 in the
2. Set up multiplication tables for the scale of 5.
Check the answer by
scale of 5, and find their product when they are so written.
converting the result to the ordinary scale of 10.

3. If 42 and 352 are in the scale of 6, find their sum and product in that scale.
4. Explain the ordinary process of multiplication by writing 382 and 73 as polynomials in 10 and finding the product of the polynomials.
In what scale is 374 written if it is the square of an integer?

Prove that the fraction a/b with (a, b) = 1 and a < b can be expressed uniquely
Show that the decimal either terminates or repeats in cycles of not
in decimal form.
more than 6 1 digits. For what values of b will the decimal terminate?
7. If (a, b) = 1 with a < b and if in calculating the value of a/b in decimal form a
remainder b a occurs, show that half of the repeating cycle of digits has been found
and that the remainder of the cycle can be determined by finding in order the differences between 9 and the digits already established.
5.
6.
3-11.
The Highest Power
positive integer,
definition
<
<
is
a.
Theorem
is
of a
Prime That
3-15.
U? =
^~?
and
2,
For any a and
Let
a and
(3,
where n
a,
= -3.
b greater
[[]'
_
than
0,
-hi
lab]
"
&
so thaib
T)_
n = aa
a = 3b
/
+
+
<
<
"i
r 2
7*1
r2
<
<
a
b
Therefore,
n =
Mathematical Machines,
Sci.
ar 2
American, April, 1949, pp. 29-39.
If
<
n.
= aa
the largest integer a such that aa
equivalent to saying that
Thus
Is a Factor of n\.
is
This
with
41
and
[sl-'+N*]
However, r 2 is at most b
at most a(jb 1) + a
is
Corollary
If
1.
1,
and
r is at
a&
2,
If
2.
>
Corollary 3.
If ra, n,
Corollary
If
Letting
t,
4.
>
and
n = wi
and thus ar 2
1,
a2
s+,\
1,
then
aa
*)
where the n^
t ,
>[?] +
[>]
i;
(ai
1,
positive,
n2
= a we have n; =
n =
>
and a are
are positive, then
a positive prime,
is
p<
Corollary
most a
Therefore,
1.
ri
< n <
r2
+ [si
+
[
n with
for
Therefore,
a.
'
'
and
ai
a2
OL t
Hence,
[i]^ [?] + [?]+ +[?]

If p is a positive prime, let E p (m) be the exponent of the
highest power
prime p that is a divisor of ra. Using this symbol, we shall prove

the following theorem due to Legendre (1752-1833).
It will help you to
appreciate how E p (n\) increases as n increases.
Theorem 3-16. If both n and the prime p are positive, the exponent
of the
of the highest
EJnl)
power
of
K+
p that divides n\
[?]
--
is
\M-
42
Consider the set of integers

1, 2,
The
p,
shows that there are
=E
is
p,
p"
(4),
r-i
thereby obtaining the factor pL^-l.
But the last
integer of the
Lp
We
P-
o
2
new
set
set.
coeffi-
All other
Therefore,
1, 2,
[;])
can, as before,
that
Likewise,
<
we remove
p
s+1
Corollary.
Theorem
n = a Qp s
is
a multiple of p
is
,? J from
remove the factor pLp
2
new
showing that
set,
*M*[S] + &] + *('-

<
and the
the product of the integers of the
(4)
|j] p)
.o-[;]+*.0->
take out one factor p from each of these multiples of p that are in
the set
Hence,
(v-2p
multiples of p in this
integers of the set are prime to p.
E p (n\)
V\
by p
last integer of the set that is divisible
cient of p
Now
2p,
r-i
the factors pL? -
If
n =
3-17.
If
-{- +
r-iJ
pLp
4
so that
^
\
#P
>
0,
while
a&,
((a6)I)
is
[?])
.
-^
> a# p
0.
until
we
find that
Therefore,
(6!).
written in the scale of the prime p so that
a s then
,
>
di
= ap
1, 2,
Because n
a,i
<
for
n
Lpj
n
aip s
.
~l
43
<
a s with
<
<
p and
s,
a p s_1
aip s_2
a ps
~2
ai7?
s_3
as _ 2p
as_ 2
a s _i
fe]Therefore
a,_i
or
tf,(n!)
a ps
aip s
~1
'
'
(27)5
We may
scale
(138
of
a s _ip
(a
ai
a s _ip
as
3;
1.
[|?]
Hence
we have
6)/4
Oi
- a s -i
a 8 -i
- as
a.)
of the highest
27; 27
5 (138!)
53
--
(5)5
27
also use the second formula for
5,
ao
Find the exponent
Example.
+ 0;
Pp-
va ps
138
0(5 2 )
2(5)
power
^J =
2;
E p (n\).
of 5 in 138!.
(1)5
33.
Writing 138 in the
Therefore
3.
5; 5
J,(138!)
33.
t
Theorem
3-18.
The
expression n!/ai!a 2
t \,
where
a{
>
n,
i=i
is
an
integer.
We shall show that n!/ai!a
at
is
an integer by proving that the
highest power of any prime contained in the denominator
is
at least
equaled by the highest power of that prime contained in the numerator.

We know that
w* - [=] + [>] +
+ [] +
+
Lp
(5)
44
Since a\
therefore
+
-^
2
+ <H = n,
for
if
1, 2,
the integers that follow
sum
in the
it
*;'""; [?] +
m <
for
exceeds n,
may
1,
.fe]
.;;;
exceeds each
and
a*,
+ ...+[p]
fe]
are also
it
Hence,
---
*^-[|] + [p] +
where, of course, some
s+1
be
0, in
which case
all
Likewise,
0.
+
.7.'.
[r]
But by Corollary 4 above,
L?J - L^J
As a
result of
summing by columns
conclude from Eq.
(5)
by
divisible
Yv
'
the expressions
+ E p (a
ai \)
and thus the given expression
is
The product
1.
'
'
the
(6) for
EP (ai\), we
that
E p (n\) > E p
Corollary
l_^j
of
2 \)
+ E p (a
\)
an integer.
any n consecutive positive integers
is
n\.
(k + n - l)/n\ = (k + n - 1) !/
+ l)(fc + 2)
and therefore is an integer.
This corollary shows that the coefficients in the expansion of (a + b) n
with n a positive integer are themselves integers, a fact otherwise known
from the multiplication itself. As a matter of fact Theorem 3-18 proves
that the coefficients in the expansion of (6i + b 2 +
+ b r ) n for n > 0,
obtained by means of the multinomial theorem are integers, for any term
of the expansion takes the form
The
(k
expression k(k
1) \n\
ai!a 2
where a
a2
'
'
In particular
it is
Corollary
If
b r) p
2.
'
now
is
ar
ar
bi
ai
b 2a *
'
b ra -
n.
evident that:
a positive prime,
except the coefficients of the
multiples of p.
all
the coefficients of
&* p ,
where
1, 2,
(6i
.
b2
r,
are
45
with d\ + a 2
ar p, is
an integer, and furthermore since each a,- either is positive and less than
p or is 0, there is no factor p in the denominator. Hence, (p 1)!/
The
atlas!
expression p\/ai\a 2
ar
is
ar
\,
an integer, and the required
coefficient is
Mp.
EXERCISES
If aP
1.
by p 2
2.
3.
show
bp
where p
is
a positive prime,
is
divisible
by
p,
prove that
it is
divisible
Prove that
If
re
is
if
m,
n,
and a are positive
number and
a real
integers,
the largest integer that
[x] is
is less
than or equal to
x>
that:
a.
[(1
b.
[(1
+
+
+
\/S) n ], for n > 0, is odd or even according as n

2n
1 is divisible by 2 n+1 for n > 0.
\/3)
]
-\/o>]
+
+1
is
by
divisible
c.
[(3
4.
For which values of n
5.
Is there a positive integer
is [4 n
is
even or odd.
2".
(2
\/2)
n
]
n such that 3 200
divisible
by 112?
the highest power of 3 contained
is
inn!?
Applying Theorem 3-17, let (n l)/2 = 200, for the sum of the coefficients of the
powers of 3 necessary to express n in the scale of 3 is at least 1. Then n is at least 401.
However, 402 has the factor 3,
But the highest power of 3 contained in 401! is 3 196
so that 402!, 403!, and 404! have the factor 3 197 but 405 has the factor 3 4 so that 405!
Hence, there is no n such that n\ has the factor required.
has the factor 3 201
19
is the highest power of 5 contained inn!.
6. Find a positive integer n such that 5
With how many zeros does 100! end?
7. Show that 95! ends with 22 zeros.
8. Find the highest power of 12 contained in 500!.
9. Prove that the exponent of the highest power of 3 contained in (3 r 2)! is
.
(3*-
10.
11.
2r
l)/2.
Find the exponent of the highest power of 5 that is a factor of (5 r 1)!.

If m >
and r(ra) means the number of positive divisors of m, show that
*>+*>++*) -R] + [5] +

12. If
m >
and a(m)
is
the
sum
....+ [2]
of the positive divisors of m, prove that
a)+w + ---+--[f]+.[5]+8[;] +
...
+ .[2]
When a > and b > 0, prove that if m = ab, then m\ is divisible by

m = ab with a and b positive integers, prove that m! is divisible by the
common multiple of (a\) b and (6!) a
15. Prove that when (m, n) = 1 and m and n are positive integers,
13.
14. If
(m
+n -
1)!
m\n\
is
an
16.
integer.
Prove that when
and n are positive
integers,
(2m)!(2n)!
7n\n\(m
is
an
integer.
n)\
(a!) 6 .
least
46
n =
17. If
Oi
0,2
a>r
with
>
all en
and
(ai,
a2
ar)
<2,
then
I
d(n
ai!a 2
an
is
1)!
ar
integer.
m,
and a are
18.
If
19.
Prove that
n,
positive integers, under
what conditions
will
[t>[;]+*R]
.
(x+r-i
i)
(x
w hen
l)(x 2
is
1)
(z r
1)
is
a factor of (x n
l)(x n+1
l)
a positive integer.
3-12. Some Theorems on Primes.

We have already found some polynomials like 4as 1 that represent an infinite number of primes for
integral values of x, but we have noticed that not all the values of these
polynomials obtained when integers are substituted for the variable are
when x = 4. The polynomial

produces primes f or x = 0, 1, 2,
39, and 40. Mathematicians have not yet produced an
3,
integral polynomial of the second degree in one variable that can be
shown to represent an infinite number of primes, nor have they proved
On the other hand, certain
that such a polynomial does not exist.
quadratic forms represent an infinitude of primes.
If f(x) is a function of x whose value is a prime whenever x is a positive
integer, then f(x) is a prime-representing function.
Mathematicians have recently developed prime-representing functions*
that require the symbol [as], denoting the largest integer less than or equal
to x, but it is interesting to observe that no algebraic expression that
represents an infinite number of distinct primes, and only primes, has
been discovered. Euler (1707-1783) showed that Fermat (1601 - 1665)
2n
was in error in thinking that all integers of the form 2 + 1 are primes
1 = 4,294,967,297 has the factor 641.
It has
by showing that 2 32
been proved, however, that no rational function of x except a constant
can be a prime-representing function, f We shall prove the corresponding
well-known theorem about a polynomial.
Theorem 3-19. An integral polynomial of at least the first degree
cannot represent primes alone.
Suppose that, for x = x' where x' > 0, f(x) = a x n +
+ an
Substituting x = x'
mp in /(as), we obtain
represents the prime p.
For instance, 4x
primes.
x2
41
is
gives 15
extraordinary, for
it
',
f(x'
where the
mp) = a Q (x
mp) n
a x {x
mp) n ~
coefficients in the expansions are integers
and
in
+ a
each expansion
* W. H. Mills, Bull. Am. Math. Soc, Vol. 53, No. 6, p. 604, 1947.
Am. Math. Monthly, Vol. 58, No. 9, pp. 616-618, 1951.
t R. C. Buck, Am. Math. Monthly, Vol. 53, No. 5, p. 265, 1946.
E.
M. Wright,
every term except the
mp) = a
f(x'
first is
a multiple of
x' n
aix'
n-
Therefore,
p.
'
'
47
+ Kp
an
or
+ Kp
and f(x' + mp) is not a prime unless f(x' + ^p)
Hence, p f{x'
n 1, f(x''+ mp) =
Suppose that, for m = 0, 1, 2,
is p or p.
yield
the
equation f(x) = p of
of
m
can
or
other
value
no
Then
p,
p.
+ mp)
+ mp),
f{x'
= p
degree n would have more than n roots.

can yield p. Consequently, there
mp >
and /(V
mp)
In like manner, but n values of

is
a value of
for
which
x'
^ p.
Moreover, if the integral polynomial f(xi,x 2

, n ) represented
x n = a n the integral polynomial
primes alone, then for x 2 = a 2
,a n ) having but one variable would represent primes for all
f(x h a 2
The given polynomial is thus not a prime-reprepositive values of xi.
,
senting function.
In the proof given above
it
has been necessary to refer to the theorem
number of roots of a rational integral algebraic equation which

depends upon analysis for its validity. A property of the whole set of
complex numbers is accordingly made use of in order to show a characterSuch cases, in which we use a domain that
istic of the rational integers.
of
itself
the
one with which we are particularly conpart
includes as
rare
in
mathematics.
Can you name some of them?
cerned, are not
on the
EXERCISE
~
Prove that for integral values of x an integral polynomial a x n + a\X n
+
has an infinite number of distinct prime factors.
(Assuming the
l
+ a n of degree n >
number
is finite,
substitute multiples of a n \[ pi for x.)

i
*Theorem 3-20 (Legendre).

n is
=i
The number
of positive
primes not exceed-
ing the positive integer
7r(n)
= n
+r
Lf
lPi\
Z/ IViVA
M=
l
(-!)'
P1P2
'
Pr\
2
where p h p 2
p r are all the positive primes such that pi < n
where i = 1,2,
r.
Determine I so that / is the largest integer whose square is less than or
equal to n. Then find all positive primes p h p 2
p r that are less
,
48
than or equal to
n are
divisible
I.
Since exactly
by p h none
integers from
1,2,...,
of the integers
of these multiples of
In like manner,
prime.
Mi =
p h except p\
itself, is
through n are divisible by
LPzj
p2
Of these multiples
and
pi,
p 2 however,
of
= [il
IPii
number of integers from

two primes, pi and p 2
Assuming then that
the
the
first
Mi
pi
and hence
M
is
by both
are divisible
fe]
- \jl]
LP*}
LP1P2J
through n that are divisible by either of
1
.
LVkj
+ [-=-1+
LpiPz]
LP1P2]
iPk-iPk]
+(-rLP1P2
LP1P2P3]
pk \
the number of integers from 1 through n that are divisible by at least

one of the first k primes pi, p 2
Pk, we shall find the number of
integers from 1 through n that are divisible by the next prime p k +i and
that are prime to all the first k primes.
The number of integers from 1 through n that are divisible by p k +\ is
is
n
Pk+ij
Of these integers
Pk+i
Pi
examine the
the coefficients of m+i, which are 1,2,

or not the integers are divisible
operate on
pi, for
if
we
set of multiples of Pk+i,
p k+h 2p*+i,
how many
by
are also divisible

J
by
p\.
\-^-\
by p h p 2
we operated on n above.
or
p r we must
,
Hence,
[-1
lPk+i
Pk+i.
Pk
determine whether
IPk+i]
Likewise, if we wish to determine
of these integers are divisible
just as
Pk+i
P1P2
"
-1
Pk+
L
Pk-iPk
(-D
.Pk+i.
[pip 2
"
Pk:
49
number of integers from 1 through n that

by at least one of the primes p h p 2
the
is
also
r
*
LP1P2
Pt
and subtracting M'k from
jlI
_ r_5_i
_p k +ii
LPiPk+i]
Pk-
fjLl
PlP2
by p k+ i and
Using the fact that
are divisible
pp*+iJ
we have
+
f_=_i
IPkPk+ij
\^\ +
LPiPzPk+i}
(8)
p*p*+i.
number of integers from 1 through n that are divisible by p k+ but

Adding this number (8) to
not by any of the primes p p 2
Pan
that
are divisible by at least
1
through
from
the number (7) of integers
find
that
one of pi, P2,
Pk, we
as the
1}
=-+...+
fc+1
zlI
p*J
-fe]
+ r jl] _ [jl]
LP1P2J
Lpfc+iJ
JL -]+
+(-i)4
PkPk+ij
LP1P2
PkPk+ij
number
of integers from 1 through n that are divisible by at least one

This formula (7) with k = r, thereprimes p h p 2}
Pk, Pk+i.
But by Theorem 3-2 any positive integer
fore, holds for the first r primes.
less than or equal to n and greater than p r is a prime unless it is divisible by
one of these first r primes. Hence, n
r is the number of integers
and p r
Consequently,
from 1 through n that are prime to pi, p 2
this number counts the integer 1 and all primes greater than p r but it
does not include the primes pi, P2,
p r themselves. Therefore
is
the
of the
w(n)
= n
+r
number of positive primes that are less than or equal to n.

Sometimes the symbol <fr(n,r) is used to indicate the number of positive
integers not exceeding n and prime to the first r primes.
Then 4>{n,r) =
and
the
number
n
of
positive
primes
exceeding
not
n can be
r
is
the
written
ir(n)
<f>(n } r)
+r
r is the number of positive primes not exceeding y/n.

This formula obviously becomes impracticable when n is large. In
1870 Meissel developed another formula the use of which is less cumber-
where
50
some, but even though his method has been improved upon,* no expeditious method for finding the exact number of positive primes less than
a large n has been discovered.
On the other hand, due to the work of Legendre and Gauss (1777-1855)
in applying analysis to the theory of numbers, we have formulas which
approximate the number of primes not exceeding x. Legendre stated the
empirical formula
F(x)
logx
1.08366
which agrees very well with t(x) so long as x is not greater than
Gauss discovered "the integral logarithm of x,"
but
we have no
their
observation.
[*
dt
J 2 log
reason to believe that he developed a proof of
work on that
in developing
0) =
of
1,000,000.
it.
Basing
Riemann (1826-1866), mathematicians succeeded
an asymptotic formula that

1900 they had proved what
By
a refinement of Gauss'
is
is
called
"the prime number
theorem,"
lim
x
-*&- =
^ z/log
as 1948 Paul Erdos and Atle Selberg developed new and more
elementary methods for showing this limit.
As a matter of fact although many theorems about primes have been
demonstrated, we can still state a large number of theories that mathematicians believe to be true but which remain unproved. We have seen
that Euclid established an interval within which there must be a prime.
Again it has been proved that if p 1} p 2
Pn-i are the first n 1
primes, when n > 4, the next prime p n is such that
As recently
Vn
<
PlP2
'
'
Pn-l
Furthermore, Tchebysheff (1821-1894) developed a better result by

proving that for a real number n > \ there is always at least one prime
between n and 2n 2. But the problem of naming the next prime after
any given prime remains unsolved. Likewise, we have no formula for
finding even one prime greater than a given one.
It has been previously pointed out that Dirichlet (1805-1859) proved
that there is an infinite number of primes among the terms of any arithmetic progression in which the first term and the difference are relatively
prime.
Kronecker (1823-1891) later showed an interval within which
the next prime after a given one of such a progression must lie.
* A. Brauer,
Am. Math. Monthly,
Vol. 53, No. 5, pp. 521-523, 1946.
51
Kronecker remarked, but there is no proof, that every positive even

integer can be represented as the difference of two positive primes in
If this theorem is true, it means that there are
infinitely many ways.
infinitely many pairs of primes that differ by 2, and hence no matter how
far out we go in the sequence of consecutive positive integers there will
always be primes that are as close to each other as it is possible for them
to be.
Of course the frequency of the appearance of such primes decreases
We can show, moreover, that if we take
as the primes increase in size.
integers,
we can find as many consecutive
large
positive
sufficiently
which
are
composites, for none of the integers
all
of
integers as we please
n\
>
2,
n\
3,
n\
Hence, as we move out in the sequence of positive integers, there must be consecutive primes whose difference is larger
than any assigned integer.
Euler mentioned that Goldbach (1742) had stated the empirical
theorem that every even integer greater than 2 can be represented as the
sum of two positive primes. Although the truth of this conjecture has
been verified in many cases, it has never been proved. Goldbach also
said that every odd integer greater than or equal to 9 is the sum of three
odd primes. In 1937 Vinogradov proved by analytical means that this
theorem is true for sufficiently large odd integers.
If we examine a table of primes, we notice that there is at least one
2
prime between any two consecutive squares n 2 and (n
l)
but whether
or not this statement is always true, we do not know.
Again, we do not
know whether or not there is an infinite number of primes of the form
is
a prime
when n
1.
(2n) 2
1.
52
Table of Primes
2003
2011
2017
2027
2029
2039
2053
2371
2377
2381
2383
2389
2393
2399
2411
2417
2423
2749
2753
2767
2777
2789
2791
2797
2801
2803
2819
1663
1667
1669
1693
1697
1699
1709
1721
1723
1733
2063
2069
2081
2083
2087
2089
2099
2111
2113
2129
2437
2441
2447
2459
2467
2473
2477
2503
2521
2531
2833
2837
2843
2851
2857
2861
2879
2887
2897
2903
1381
1399
1409
1423
1427
1429
1433
1439
1447
1451
1741
1747
1753
1759
1777
1783
1787
1789
1801
1811
2131
2137
2141
2143
2153
2161
2179
2203
2207
2213
2539
2543
2549
2551
2557
2579
2591
2593
2609
2617
2909
2917
2927
2939
2953
2957
2963
2969
2971
2999
1087
1091
1093
1097
1103
1109
1117
1123
1129
1151
1453
1459
1471
1481
1483
1487
1489
1493
1499
1511
1823
1831
1847
1861
1867
1871
1873
1877
1879
1889
2221
2237
2239
2243
2251
2267
2269
2273
2281
2287
2621
2633
2647
2657
2659
2663
2671
2677
2683
2687
3001
3011
3019
3023
3037
3041
3049
3061
3067
3079
1153
1163
1171
1181
1187
1193
1201
1213
1217
1223
1523
1531
1543
1549
1553
1559
1567
1571
1579
1583
1901
1907
1913
1931
1933
1949
1951
1973
1979
1987
2293
2297
2309
2311
2333
2339
2341
2347
2351
2357
2689
2693
2699
2707
2711
2713
2719
2729
2731
2741
3083
3089
3109
3119
3121
3137
3163
3167
3169
3181
233
239
241
251
257
263
269
271
277
281
547
557
563
569
571
577
587
593
599
601
877
881
883
887
907
911
919
929
937
941
1229
1231
1237
1249
1259
1277
1279
1283
1289
1291
1597
1601
1607
1609
1613
1619
1621
1627
1637
1657
71
283
293
307
311
313
317
331
337
347
349
607
613
617
619
631
641
643
647
653
659
947
953
967
971
977
983
991
997
1009
1013
1297
1301
1303
1307
1319
1321
1327
1361
1367
1373
73
79
83
89
97
101
103
107
109
113
353
359
367
373
379
383
389
397
401
409
661
673
677
683
691
701
709
719
727
733
1019
1021
1031
1033
1039
1049
1051
1061
1063
1069
127
131
137
139
149
151
157
163
167
173
419
421
431
433
439
443
449
457
739
743
751
757
761
769
773
787
797
809
179
181
191
193
197
199
211
223
227
229
467
479
487
491
499
503
509
811
821
823
827
829
839
853
857
859
863
2
3
5
7
11
13
17
19
23
29
31
37
41
43
47
53
59
61
67
461
463
521
523
541
1993
1997
1999
CHAPTER
4-1.
We
Congruent Integers.
have shown that when n and
9^
are given, the integer n can be expressed uniquely in the form
= qm
<
Now we separate all integers n into
<
\m\ classes
\m\
according to the remainders
upon being divided by m. We say that two integers

r
congruent
modulo
m if and only if the integers produce the same least
are
nonnegative remainder upon being divided by m ^ 0.
Gauss (1777-1855) introduced this idea of congruence, and it was he
who suggested the notation a = 6(mod m), which is read, "a is congruent
The value
to b modulo m," or "a is congruent to b for the modulus ra."
is
that
emphasis
is
placed
upon the imporof this concept and its symbol
that they yield
tant integers in the equations
+r
q m + r
giw
and
<
when
Since the definition requires that

in the
above equations be
difference a
a
r2
S3 b (mod
<
\m\,
m), for
r2
r\
Theorem
4-1.
their difference
by m.
b is divisible
then a
Therefore,
identical,
is
b
.
if
= qim
m(qi
q 2)
r\
it
<
\m\
b (mod
Conversely,
and
+n
m) the remainders
follows immediately that the
b
r2
q2
and
if
m+
ri
r2
km, then
where
< n,
b
r 2 is divisible
by m.
Thus we have:
Two
integers are congruent
divisible
by
m^
modulo
if
and only
if
0.
If a and b have distinct remainders r x and r 2 where

< r h r 2 < \m\,
upon being divided by m, then a and b are said to be incongruent modulo m.
The difference between two incongruent integers modulo m is, therefore,
not divisible by m. In this case we write: a ^ b(mod m).
When two integers are congruent to each other modulo m, each is said
,
to be a residue of the other for that modulus.
Thus, because 12
2(mod 10), 2 is a residue of 12, and 12 is a residue of 2 modulo 10.

The totality of integers congruent to a given integer for the modulus
constitutes a residue class modulo
m.
53
=
m
54
These definitions imply that each integer belongs to exactly one residue
modulus and that each residue class modulo m contains
one and only one of the integers 0, 1, 2,
\m\ 1.
Hence, there
class for a given
modulo m.
are exactly \m\ residue classes
Any set of \m\ integers selected so that no two of them belong to the
same residue class modulo m forms a complete residue system modulo m.
Thus for the modulus m the set of integers 0, 1, 2,
\m\ 1 or the
.
chosen to represent the classes of which the

individual integers are members.
The class represented by contains all
and only multiples of m, and the class represented by r consists of all the
set
1, 2, 3,
integers of the
\m\ is often
km
form
Any
r.
km
Since, as
\m\
km
a complete residue system
1 is, therefore,
one and
set of \m\ integers containing
only one integer represented by each of the forms km,
we have already shown, an
integer
according as the remainder obtained upon
or
is
is
1,
km
not prime to
by
2,
modulo m.
not
prime to m, the set of remainders from 1 through \m\ 1 which are prime
These
to m represents all and only the integers that are prime to m.
integers, prime to m, are thus separated into residue classes modulo m
that are in one-to-one correspondence with the positive integers from 1
through
Any
of
\m\
the modulus
division
is
or
is
that are prime to m.
set of integers
them belongs
its
prime to
and selected so that one and only one
to each of the residue classes of integers prime to
for
For the
constitutes a reduced residue system modulo m.
modulus 5 the set 1, 2, 3, 4 is a reduced residue system, but for the

modulus 6 the integers 1 and 5, as well as the set 1 and 1, form such a
system.
It is evident also that the residue classes for the
modulus
are iden-
with the residue classes modulo m, for when the sign of

changed, we need change only the sign of q in the equation
tical
= qm
Thus any congruence that holds

the other one.
<
for either
<
is
\m\
m or m as modulus holds for
It is convenient, therefore, to use only positive integers
and we
shall hereafter adhere to this convention without

statement of the fact in the discussion.
4-2. Basic Properties of Congruences.
The relation of congruence
has some properties similar to those of equality:
1. For any modulus m, a = a(mod m).
2. If a = b (mod m), then b = a (mod m).
3. If a = 6(mod m) and b = c(mod m), then a = c(mod m).
4. If a = 6(mod m) and c = d(mod m), then a c = b d(mod m).
as moduli,
making a
From
d
specific
the equations a
km,
tm,
we have
+ Lm.
5.
If
b (mod
m) and
= d(mod
m), then ac
bd(mod m).
6.
If
7.
If
55
= b (mod m), then a n = 6 n (mod m), where n is a positive integer.

a = b (mod m) and /(rr) = a x n + aix n_1 +
+ an is an
then
integral
coefficients,
with
function
of
x
rational
f(a) =
integral
/(6)(mod m).
show that the congruence relation is reflexive,

symmetric, and transitive. It is also clear that when a + b = c(mod m)
and b = d(mod m), then a + d = c(mod m). In a congruence the laws
thus permit the substitution of a number or expression in the place of a
The
first
three statements
The application of the fifth

statement allows us to conclude that when ac + b = 0(mod m) and
Accordingly, we infer that for
c S3 e (mod m), then ae + b = 0(mod m).
a given modulus a congruent number or expression may be substituted
term congruent to
for a factor of a
"
When
it
for the given modulus.
term
of a congruence.
the relation, substitutions of the above types are

covered by the familiar postulate, " Equals may be subIt is, however, not true that
stituted for equals in any operation."
integers which are congruent for a modulus m may always be substituted
two
just
equality"
is
of those
one for the other in a congruence modulo m. If a s = b (mod m) and s =

= b (mod m), for a s need not be
2 (mod m), it need not happen that a
congruent to a* modulo m. Notice that 5 2 = 4 (mod 7) and that 2 =
9(mod 7) but 5 9 = 6(mod 7), so that 5 2 ^ 5 9 (mod 7).
The congruence relation also has the following properties that pertain
1
to division:
m and a = b (mod m), then a = 6 (mod d).

and a = b (mod m 2 ), then a = b (mod L), where L
is the least common multiple of m x and m 2
is such that (c, m) = 1, then a =
3. If ac = be (mod m) and c ^
b (mod m).
4. If ac = be (mod m) and (c, m) = d, then a = 6 (mod m ), where
1.
If
is
a divisor of
2.
If
b (mod mi)
m =m
d.
We
can find the remainder when 2 30 is divided by 17

by simple operations on congruences. Since 2 4 = 16 (mod 17) and 16 =
l(mod 17), we have 2 4 = l(mod 17). Raising each member of the
congruence to the seventh power, we obtain 2 28 = l(mod 17). But
Examples.
1.
and therefore 2 30 = -4(mod 17), or 2 30 = 13(mod 17).

2. We know that 10 = l(mod3).
Accordingly, a (10) n + a^lO)"- +
+ a n = a Q + ai +
+ a n (mod 3). Thus a number written in
the scale of 10 is divisible by 3 if and only if the sum of its digits is divisible
by 3.
22
4(mod
17),
EXERCISES
1.
21 10
is
2.
Find the remainder when 7 10

divided
by
is
divided by 51;
is
divided
when
3 10
is
divided
by
51;
51.
Find the remainder when 5 21
by
127.
Do
the
same
for 5 66
when
56
3.
Prove that 2 U
4.
If a
has the factor 23 and that 2 23 1 has the factor 47.

prove that the greatest common divisor of a and m is the
State this result in terms of the integers of a
greatest common divisor of b and m.
residue class modulo m.
5. If a + b = c(mod m) and b = d(mod m), show that a + d = c(mod m).
6. If ab = c(mod m) and b = d(mod m), show that ad = c(mod ra).
4
= l(mod 5) and 4 = 9 (mod 5), is 2 9 = l(mod 5)? Explain.
7. Since 2
8. Prove that an integer is divisible by 9 if and only if the sum of its digits is divisible
by
9.
= b (mod
ra),
9.
Prove that an integer
last three digits
is
divisible
is
by
divisible
by 8
if
and only
if
the
number formed by
its
8.
10. Prove that an integer is divisible by 1 1 if and only if the sum of the digits in the
odd-numbered places diminished by the sum of the digits in the even-numbered places
is divisible by 11.
11. If an integer N is written in the scale of r and then its digits are rearranged in
is divisible by r 1.
any way to form the integer M, the difference N
4-3. The Residue Classes.

The properties of congruences stated
above show that as far as the operations of addition, subtraction, and
multiplication are concerned the elements in any two residue classes for
the modulus ra combine to give results that are always in the residue class
designated by carrying out the very operations on any convenient repreThus if an element a of class A is added to an
sentatives of the classes.
element b of class B and the result a + b is in the class C, then if any
element of A is added to any element of B, the result will be in class C.
Because subtraction is always possible, it follows, moreover, that when
a + b is an element of class C, any element a + b + km of C can be
expressed as the sum of an element of A and one of B.
It is important to notice, however, that although the result of multiplying any element of class A by an element of class B is always in the
same class D, say, yet each element of D need not be a product of an
element from A and an element from B. Take, for instance, the residue
When any element of the class represented by 2 is
classes modulo 10.
multiplied by an element of the class of 3, the result is an element in the
Yet the particular element 16, which is conclass represented by 6.
gruent to 6 modulo 10, cannot be written as the product of two factors,
one from the class of 2 and the other from the class of 3, for if
16
an odd number would divide
(2
10fc)(3
6); that
is,
100
16.
4 (mod 6) as well as 2 2 =
the same least positive residue 2 ^ 0(mod 6) can be
It is interesting to notice also that 2
4(mod
multiplied by either one of the distinct least positive residues 2 and 5 to
produce a number of the

positive residues of
class of 4
any prime, we
modulo
6.
If
we examine
the least
find that no such thing happens
when
we do not choose the first factor from

To prove this statement, let p be
6(mod
(a, p) = 1, C\ = c 2 (mod p).
residue class modulo p.
6(mod
p)
and ac 2
When we
Hence,
select the integers a
integer x such that ax
the class of zero for the given prime.

a prime, and suppose that aci
b (mod
and
m),
C\
and
c2
and ask whether or not there
it is
ac 2 (mod p),
and since
must come from the same
Therefore, aci
p).
57
evident that
we
We have,
a problem in division, the inverse of multiplication.

b by a for the modulus m,
shown that when we divide
is
an
are dealing with

therefore,
have
do not belong to the same residue class for that modulus.
It may happen, of course, that all answers are in but one residue class as
But it is also posis true in the case of the congruence 5x = 1 (mod 6)
sible that there be no answer whatever, for a solution of 2x = l(mod 6)
would demand that 2x = 1 + 6/c and that 2 divide 1. Consequently, we
must proceed with care, for division modulo m is not always possible and,
when it is, need not yield a unique result.
One of the important ideas of arithmetic is that if ab = and a ^ 0,
then b must be 0. Suppose that
it is
possible to
results that
ab
0(mod m)
Is it necessary that either a or b be in the class of
When the modulus is
we
for the
modulus m?
0(mod 6) and
modulo 6. But when the modulus
that neither 2 nor 3 is congruent to
is a prime p and ab = 0(mod p), both a and b cannot come from the set
of integers 1, 2, 3,
p 1, for the product of any two of these
integers is prime to p.
Consequently, we see that in this case a product
.
is
not congruent to
6,
notice immediately that 2
unless at least one factor
is
in the class of
for the
When the modulus m is composite, however, by factorthat m = n n^ where 1 < n\ < m, it follows that ttin =
given modulus.
ing
mÔso
0(mod m).
We call any integers ni and n 2 neither one of which is in the class of

modulo m, but whose product is congruent to
for the modulus m,
divisors of zero modulo m.
The existence of divisors of zero for a composite modulus again reminds us of the need of caution in applying the
,
idea of division to the notion of congruence.
EXERCISES
that although 2(6) = 26 (mod 14), 26 cannot be factored into integers such
in the class of 2 and the other in the class of 6 modulo 14.
2. Find numbers in the class of 10 modulo 11 that can, and some that cannot, be
expressed as a product of two integers, one from the class of 2 and the other from the
1.
Show
that one
class of 5
is
modulo
11.
58
How many
3.
solutions do each of the following congruences have?
m
=
a
=
2x
2x
2x
Sx
4.
Find some divisors of zero modulo
6.
Compare the
the powers of
6 (mod 10)
3 (mod 4)
3 (mod 5)
6(mod
15)
12.
residue classes of the powers of
3, 5,
and 6 modulo
Can you
15.
2, 4,
find a
and 7 modulo 15 with those of

law that governs them?
Function.
Leonhard Euler (1707-1783) worked in
pure and applied mathematics. His voluminous publications were concerned with algebra, the calculus of finite differences, the
differential and integral calculus, the calculus of variations, astronomy,
and analytical mechanics besides the theory of numbers. In the latter
field he discovered the theorems which taken together make up the
4-4. Euler's
many
fields of
supplied the proof and generalization of
quadratic reciprocity law,
Fermat's theorem, showed that every prime of the form 4n + 1 is

expressible as a sum of two squares in exactly one way, as well as making
many less startling but nevertheless important discoveries, one of which
is
the
function.
an integer m ^
is the number of positive
and prime to m. Thus 0(m) is the
number of integers in a reduced residue system modulo m, and 4>(m) =
Because of this last fact, it will be sufficient to use only positive
0( ra).
The
indicator <f>(m) of
integers less than or equal to \m\
m in
integers
considering the
Examples.
0(1)
It is evident that
Theorem
4>(p
n
)
To
4-2.
= p n~Kv -
when p
p
If
function.
0(5)
1,
is
is
4,
0(6)
2.
a positive prime,
4>(p) is
a positive prime and n
1.
a positive integer,
is
i).
find 4>{p n ), consider the set of integers

1, 2,
Each of these integers is

of them are divisible by
p,
2p,
either divisible
by p
p
or
r
,
is
prime to
But p n ~
p.
n~ 1
n~ l
= p (p 1) of the
Therefore, p p
p.
through p n are prime to p.
r, is arithmetic
Any function of the variables Xi, where i = 1, 2,
if it assumes only integral values for the sets of integral values of the
The function <f>(x) is
variables Xi for which the function is defined.
integers
from
arithmetic, as are the integral polynomials.
single-valued arithmetic function f(x)
plicative
We
shall
if,
for
any a and
show that the
is
said to be regular or multi-
which are relatively prime, f(ab)
function
is
multiplicative.
f(a)f(b).
Theorem
4-3.
If
59
a and 6 are relatively prime, 0(a6)
0(a) 0(6).
We suppose that the integers a and b are positive and arrange the
integers from 1 through m = ab in the following array suggested by the
residue classes
modulo
a:
2a
+
+
L)o+
(6
2a
+
+
a
2a
3a
fc
ba
2a
+
+
+k
a
2a
3
3
(6-
l)a
Because an integer is prime to m if and only if it is prime to both a and b,

we shall determine first the number of integers in the above array that
are prime to a and then find how many of these are also prime to b.
We know that there are 0(a) integers prime to a in the first row.
Moreover, each integer in the column headed by an integer k, from 1
through a, is of the form sa + k. Furthermore,
sa
k(mod
a)
if the integer k at the top of a column has a divisor in common

every integer in that column has that divisor in common with a,
Therefore,
with
and
a,
if
prime to
is
a,
every integer in that column
prime to
is
are then 0(a) columns of integers that are prime to
a.
There
How many
a.
of
these integers are prime to 6?
Consider the set of b integers in any column,

k
No two
2a
of these integers are
...
(b
congruent modulo
sa
(t
s)a
=
=
ta
\)a
b,
for
/c(mod
0(mod
b)
= 0(mod
b)
if
b)
and
t
since
2,
= 1. But
1 modulo
b)
(a,
.
integers of
and
any column
to the integers
1, 2,
6,
range through the residue system
so that unless
s,
s(mod
b).
0,
The
1,
some order congruent modulo b

But then exactly 0(6) of them are prime
are, therefore, in
b.
too.
Since 0(6) integers in each of the 0(a) columns of integers prime to a
6, the number of integers from 1 through ab that are
are also prime to
prime to both a and
Theorem
.
(P.
If
6 is 0(a) 0(6).
m=
ni
p1 p2
are positive primes,
r,
1)
4-4.
'
'
(Pr
1).
U2
0(m)
Pr
Tlr
,
pi'
where the p { for i

~
~
n2 "" 1
Vr
(Vi
p2
,
il
lr
1, 2,
""!)
60
Since
m has
been factored into powers of distinct primes,
0(m)
By
we
repeating this process,

4>(m)
from which we
Corollary.
0(pi ni )0(P2 n2
'
Vr
nr
)
find
<t>(p^)<t>(p2
n2
HPr nr )
infer the desired result.
m>
If
2, <t>(m) is
even.
Example. The number of positive integers

to 360 is 0(360) = 2 2 3 (2 - 1)(3 - 1)(5
less
1)
than 360 and prime
96.
EXERCISES
1. Show that the formulas for the number of divisors of an integer

and the sum of
the divisors are multiplicative functions.
2. Show that, for n > 1, the sum of the positive integers less than n and prime to n
is (n/2)<f>(n).
Show that the sum
3.
of the squares of the positive integers less than
n and prime
to
is
-(i-)+Sa->a-*> *-
0-3
f(i-a
where pi, Pi,

and p T are the distinct positive prime factors of n.
4. Prove that if n = p\Pip%, where the pi, with i = 1, 2, 3, are distinct primes, then
the product of all the positive integers less than n and prime to n is
.
(n
1)!
^~
[]
1)1
cp*p<
- 1)!
"
pi<r -l)(Pfc-l)
[1
i
=i
i<j
Use the method so developed to find the product

ar
prime ton = pi ai p2 a2
Pr
Set
6.
method
If
6.
up a method
for finding
to find the solutions of
G(n)
>
4>(d),
of all positive integers less
than n and
by
<j>(x)
trial all integers
x such that
<f>(x)
Use your
n.
=16.
show that G(n)
is
multiplicative.
d\n
Note that the symbol
>
^>(d) is
sum over
read "the
the divisors of n of
</>(d),"
and
d|n
we understand that we use only
Theorem 4-5.
2,
through
'
(1
r,
Urn =
the positive divisors of n.

ni
m that are prime to p h p

-
U2
pi p 2
p^ *
are distinct positive primes, the
(1/Pi)).
2,
pr
nr
where the p
number
p*ism(l
of integers
(l/pi))(l
if
from
1,
(I/P2))
61
sometimes important to find the number of positive integers not

m and prime to some but not all of the
prime factors of m. For instance, we may wish to know the number of
positive integers less than 360 and prime to 2 and 5.
It is
greater than the positive integer
When
from
m=
ni
pr
nr
m that
are prime to p h p 2
m
Some
of integers
that are divisible by pi is m/pi, leaving then

(1/pi)) integers that are prime to pi.
(m/pi) = ra(l
Proceeding by induction, suppose that the
through
number
evident that the
it is
through
n*
pi p 2
number
and p k
from
of integers
is
0-5)0-)-"(-s)
We
however, are divisible by pk+i.
of these integers,
wish, there-
number the number of integers from 1

by pk+i and at the same time are prime to
fore, to subtract from the above
through
Pi, P2,
from
set
that are divisible

.
To
Pk-
find this
m that
through
number,
how many
is
or
is
or
is
are
-TYl p k +i
Vk+l
not prime to
not prime to
integers of the set
as its coefficient c
They
are multiples of p k+i.
p k+h 2p k+h 3p k+ i,
Since any integer cp k+1
consider the integers of the
first
is
pi,
p2
p,
p2
.
and p k according
and p k we must ask
.
m
1, 2, 3,
'
are prime to pi,
p2
and p k
we consequently know
(1),
Pk+i
'
Referring to our inductive assumption
that
m
Vk+
pj\
vj
(2)
Pk)
of integers from 1 through m/p k+ i that are prime to p h p 2

and p k and is, therefore, the number of integers from 1 through m
divisible by p k+ i and also prime to p h p 2
Subtracting this
pk
last number (2) from the number (1) of integers prime to p h p 2
pk
we have
is
number
the
m
\
Pi)\~pJ
"
p k)
m
Pk+ii
Pi) \
\
\
Pi) \
pi)
pi)
Ph)
V ~
Pk) \
"
Pk~^)
62
as the
number
p and
k
2,
prime factors of m.
or all of the distinct
Example.
m that are prime to p h p
through
Thus by induction we may write this formula so that

the number of integers from 1 through m that are prime to
includes
it
some
from
of integers
also p*+i.
m =
If
23
32
that are prime to 2 and 5
Theorem
n from
m is
n and
number
the
23
is
32
of integers
5 3 (1
m =
If the positive integer
4-6.
through
divisor of
53
)(1
-J)
of integers
the greatest
is
through
3600.
number
kd, the
having the property that d
from
common
<f>(k).
Consider the integers 1, 2,

There are
.
.
.
.
, d,
, 2d,
, kd.
k multiples of d in this set, but the integers id and m = kd have the greatest common divisor d if and only if t and k are relatively prime.
Since t
.
has the values
through
Theorem
m^
0,
2,
1,
4-7.
then
d h d 2j
If
0(d)
0(di)
there are exactly
k,
m the
that have with
common
greatest
integers
0(/c)
divisor
from
d.
d r are the distinct positive divisors of
0(d 2 )
4>(d r )
\m\.
d\m
Each
one and only one

divisor.
Consequently, if we pick from this set the integers n such that (n, m) = d
where \m\ = didj, there will be exactly <f>(dj) of them. As di ranges
through all the positive divisors of m, so does dj, and each integer from 1
through \m\ will thereby have been put into one and only one class
defined by the greatest common divisor it has with m.
Recalling that the
of the
integer
di,
where
in the set 1, 2,
1, 2,
has with
\m\
as greatest
r,
common
{,
symbol
of the
>
<j>(d) is
and
2.
sum
numbers indicating the
Examples. 1.
such that (n, 90)
78,
read, "the
If
m =
6
is
sizes of these classes is
90, the
0(15)
m of <(d)," the sum
over the divisors of
number
The
of integers
>
tf>(d)
n from
through 90
integers are 6, 12, 24, 42, 48, 66,
8.
84.
The
positive divisors of 70 are
1, 2, 5, 7, 10, 14,
m>
1,
and
35,
and
70.
More-
1; 0(5)
4; 0(7)
6; 0(10) = 4; 0(14) = 6;
1; 0(2)
=
=
integers is 70.
and
the
sum
of
these
24;
0(35)
24; 0(70)
4-5. Residue Systems Modulo m.
Take any integer a prime to
over, 0(1)
let
r h r 2)
rm
Form
be a complete residue system modulo m.

ar h ar 2
No two
of these
integers are congruent

ari
arj (mod
the products
ar m
modulo m,
m)
for
if i
^j
and
PKOPERTIES OF CONGRUENCES
63
then
Ti
which
Moreover,
modulo
ra.
if
n, r 2
r^m)
1,
system modulo m, for this

which are incongruent modulo
ra,
a reduced residue system modulo ra and

ar h ar 2
is
Consequently, these integers
contrary to our assumption.
is
represent the ra residue classes
is
r,-(mod ra)
(a, ra)
the set
ar^m)
also a reduced residue
set contains exactly
and each integer

prime to ra.
Example. The set 0, 5, 5 2, 5 3,
,5-11, or 0, 5, 10, 15, ...
55, is a complete residue system modulo 12, while 5, 25, 35, 55 is a reduced
residue system modulo 12.
Again if r 1} r 2
r w is a complete residue system modulo ra,
</>(ra)
integers, all of
is itself
a
for
any a
in the set,
is
and
rh
r2,
rm
another complete residue system, for there are m integers

two of them were congruent modulo m, when i ^ j,
if
+n=
ry(mod m)
then
Ti
Sometimes
ry(mod m)
convenient to use the integers of smallest numerical

value to represent the residue classes modulo m. If we write any integer
it is
a in the form
= km
<
< m
system we keep the values of r which do not exceed ra/2, that is,
through ra/2 or (ra l)/2 depending upon whether
ra is even or odd.
But those values of r which exceed ra/2 are replaced
by the negative integers of least numerical value to which they are congruent for the modulus ra. Since
for this
the integers from
ra
this
= r(mod
complete residue system modulo
19
and when
odd,
ra is
ra
when
ra is even, is
ra
it is
ra
U, 1, A,
ra,
ra)
-1
>
m -1
ra
,
3
j
2,
64
For the modulus 14 the set 6, 5, 4,

1,0,
a complete residue system, but for the modulus 15 the
system using least numerical values is 7,-6,
1,0, 1,2,
Examples.
1, 2,
is
7.
In Theorem 4-3 we showed that the integers

k,
d,
2d,
(&
l)d
of the arithmetic progression in which (b, d) = 1 form a complete residue

system modulo b.
Furthermore, we can generate a complete residue system modulo ab
by using the form ax + y, where x takes all the values in a complete
residue system
r 2}
ri,
for the
modulus
b,
and y takes
modulo
(3)
ka
(4)
form thereby gives ab

modulo ab, for if
It is evident that the
a.
values in a complete residue system
all
kh k2
These
integers.
integers are, moreover, incongruent
ar s
rs )
fa
kj
ari
kj
A'*(mod ab)
then
Ufa
and a
But
(fa
since
kj(mod ab)
so that
kj),
no two
(mod
a)
of the integers of (4) are congruent
modulo
a,
kj.
Hence,
a(ri
rs )
0(mod
ab)
rs
0(mod
b)
and
Ti
so
that,
as
above,
n =
rs
Consequently,
two integers formed
as
described cannot be congruent modulo ab unless they are identical, and

set, therefore, forms a complete residue system modulo ab.
Again when (a, b) = 1, if we use the form ax + by, letting x have the
values of (3) and y have the values of (4), the resulting ab integers form a
complete residue system modulo ab, for if
the
ari
bkj
ar s
a(ri
rs )
b(k
bk
(raod.
ab)
then
But then a\
fj
rs
(fa
fa),
and therefore
Av)(mod ab)
fa
fa.
Also
Hence, the given ab integers are distinct
(rt
modulo
ab.
r s ), so that
65
It is easy to show also that if x has the values in (3) that are prime to b
while y has the values in (4) that are prime to a, then when (a, b) 1, the
by form a reduced residue system modulo ab.
integers ax
EXERCISES
Use the form ax + by with (a, b) = 1 to show that <(a&) = <t>(a)<t>{b).
Prove Theorem 4-5 by setting up the integers 1 through m = p^pf*
Pknk s
n
n
in an array of s complete residue systems modulo pi ip2
Phnk
n~ x
3. Show that a
+ y generates a complete residue system modulo a n if x has the
values in a complete residue system modulo a while y has the values in a complete
~
residue system modulo a n l
4. If f(x) is an integral polynomial and if there are \p(m) integers prime to m in the
,/(m), prove that when (a, b) = 1, xp(ab) = Ha)Hb).
set/(l),/(2),
5. Find the number of integers prime to m in the set:
1.
2.
a.
,
6.
2,
3,
1-22-3
~T' -2~'
For m >
'
'
'
set
up
m(m +
m(m +
1).
1)
2
all
the permutations
positive integers not greater than m.
at a time with repetitions allowed of the
fc
Then the number
of these sets of k integers
whose greatest common divisor is prime to m is 4>k{m). Find a formula for <}>k(p n ), and
show that this function is multiplicative.
7. Without using an enumeration according to size, show that if a, b, and c are
positive integers and a = be, there are in a complete residue system modulo a exactly
(Let ci, c 2
c c be a complete residue system
c integers that are divisible by b.
modulo c. Then consider the set be i, 6C2,
bc e .)
8. Can you find an integer the powers of which set up a complete residue system
modulo 13? Can all integers prime to 13 be used to form such a set?
9. By expanding (1 + 1 +
+ l) p prove that if p is a prime, a p = a (mod p)
and hence that when (a, p) = 1, a^ {p) = l(mod p).
.
CHAPTER

and Conditional Congruences. In Sec. 4-3 the considmanner in which residue classes combine led to setting up
a congruence in which we used the symbol x to represent an integer.
For instance, when we asked whether or not there is an integer which
when multiplied by 2 would give an integer in the class of 4 modulo 6, we
were dealing with a congruence of the form 2x = 4 (mod 6). In such
congruences we shall hereafter think of x not as just a symbol for an
unknown integer but as a variable whose values are the rational integers.
By substituting such values in the congruence 2x = 4 (mod 6) it has
already been verified that both 2 and 5 are values of x that satisfy this
5-1. Identical
eration of the
a,2,
.
xn
we shall say that if the constants

a n are substituted, respectively, for the variables Xi, x 2
of whose values are integers) in the congruence
In the general case
congruence.
ai,
(all
fi(x 1} x 2 ,.
where
/i
and f2 are
If
two
ah x2
(x h x 2 ,.
,^)(modm)
integral polynomials in these variables,
fi(a h a 2 ,.
then x\
=f
,x n )
a2
,a n )
xn
/2 (ai,a 2 ,.
integral polynomials /i
an
is
and f 2
,fl)(mod
and
if
m)
a solution of the given congruence.

in the variables x h x 2
xn
are such that the coefficients of like terms are congruent to each other for
the given modulus m, then these expressions are said to be identically

congruent for the modulus m, and
fi(x h x 2 ,.
calling this
,x n )
we sometimes
f2 (x h x 2 ,.
write
,x n
)(modm)
congruence an identical congruence although very often we
use only the ordinary sign of congruence to express this relation.
Corre-
,x n ) with
spondingly, an integral rational algebraic function f(xi,x 2 ,.
if and
integral coefficients is identically congruent to zero for the modulus
.
by m. All the congruences that

involve only constants, such as a = 6 (mod m), are necessarily identical
congruences. 9x 2 2x + 5 = Sx 2 + Ax l(mod 6) is an identical
congruence in the single variable x. It is evident that an identical cononly
if all its
coefficients are divisible
66
67
gruence will be satisfied regardless of the integral values that are assigned
to the variables.
On
the other hand, we shall call a congruence of the above form, but in
left- and right-hand members /i and f 2 are not identically con-
which the
gruent for the modulus m, a conditional congruence.

We shall be concerned chiefly with congruences of the form fi(x) =
We
m), where fi(x) and f 2 (x) are integral polynomials in x.
(x)(mod
2
f
the properties of congruences that when r is a solution of such
also
a congruence, every integer in the class with r for the modulus
km) (mod ra). This whole
satisfies the congruence, for fi(r) = fi(r
know from
class of integers is considered just
consequently a solution
r(mod
is itself
one solution of the congruence, and
written in the form of a congruence,
ra).
Example. The congruence 2x = 6 (mod 10) is satisfied by 3 and 8.

Hence, the complete solutions are written in the form x = 3 (mod 10) and
x
8(mod
10).
Let us recall that in algebra* two polynomials fi(x) and f2 (x) are said
to be identically equal if and only if they are equal for all values of x, and
hence if and only if corresponding terms have the same coefficients. In
particular, a polynomial vanishes identically if and only if it vanishes for all
In
values of x, which means if and only if all its coefficients are zero.
contrast, consider the congruence x 3 x = 0(mod 3), and observe that
although for the modulus 3 the polynomial x 3 x is congruent to zero for
x, yet not all its coefficients are congruent to zero modulo 3.
members of the congruence x 3 2 = x + l(mod 3) have the
same values modulo 3 for all values of x, but the members are not identically congruent modulo 3.
In other words a congruence may be
satisfied by all integers and still not be an identical congruence according
Examples of such conditional congruences are
to the above definition.
all
values of
Again, both
xb
x
Sx 2
x
2x
= 0(mod
= 0(mod
2x
4(mod
6)
x2
2(mod
5)
5)
6)
Likewise,
and
first one having two incongruent soluwhile the second has no solution whatever.
are conditional congruences, the

tions
On
modulo
6,
the other hand,
x2
*
(x
M. Bocher, "Introduction
pany,
New
York, 1931.
2)(x
2)
0(mod
4)
to Higher Algebra," Chap.
1,
The Macmillan Com-
68
and
Qx 2
15
3(mod
6)
are identical congruences.
EXERCISES
Determine whether the following congruences are identical or conditional, and
find the solutions
2x 2
xz
x2
x4
6.
4
1
= 0(mod 5).
= (x - l)(x -
5-2. Equivalent
if
if
operation listed
4-2),
trial
+ Sx = 5(mod 7).
+ x = x - z (mod
1.
3.
5.
by
there are any.
2).
2){x
3)(x
x4
4.
2x 3
4)(mod
6).
5).
Having performed on a congruence an
Congruences.
among
+ x m 0(mod 10).
+ 3x + x = 0(mod
2.
the properties of the congruence relation (Sec.
a second one of these operations can be performed on the result so
as to restore the original congruence, the

reversible.
The
first
operation
is
said to be
following operations on a congruence can be reversed:
1. Adding to or subtracting from each member of a congruence congruent integers or other expressions that are identically congruent for the
given modulus.
2.
is
Substituting F(x) for f(x) in a term f(x)g(x) of a congruence
if
F(x)
identically congruent to f(x) for the given modulus.
or, when possible, dividing the coefficients of each

congruence by an integer that is prime to the modulus.
4. Multiplying or, when possible, dividing the coefficients of each
member of a congruence as well as the modulus by the same integer.
If any of these operations is performed a finite number of times on a
congruence fi(xi,x 2 ,.
,x n ) (mod mi) and the result,x n ) = f2(x h x 2 ,.
ing congruence is giix^x*,.
,x n ) = ^2(^1,^2,.
,n)(mod m^), then
For instance, when
the two congruences are said to be equivalent.
f(x) = F(x)(y&o& m) identically, then f(x) g (x) = k(x)(mod m) is equivalent to F{x)g(x) + m[h(x)] = k(x)(mod m).
Accordingly, any congruence
3.
Multiplying
member
of a
fi(x h x 2 ,.
,x n )
f2(xi,x 2 ,.
,x n
)(modm)
,x n )
can be reduced to an equivalent congruence of the form f(xi,x 2 ,.
= 0(mod m). For example, the congruence x 2 + 10 = 7a:(mod 6) is
4 = 0(mod 6), 5x 2
equivalent to each of the congruences x 2 x
.
5x
= 0(mod
6),
and 3x 2
Sx
12
0(mod
18).
,x n ) =
a congruence has been written in the form f(xi,x 2 ,.
0(mod m), the degree of the congruence is defined as the degree of the term
When
or terms of highest degree in f(xi,x 2 ,.
,x n )
whose
coefficient or coeffi-

modulus m.
cients are not congruent to zero for the
6x 3
is
of degree three,
of
7x
0(mod
12)
Sx 2
5x
0(mod
12)
Thus
although
12z 3
is
10a;
69
only the second degree and can be written
3x 2
5x
0(mod
12)
when two conditional congruences are equivalent, any

them necessarily satisfies the other. Take,
2
for instance, the congruence x = 4 (mod 5) that has the solutions x =
The congruence 5x z + 3x 2 = 2 (mod 5)
2 (mod 5) and x = 3 (mod 5).
It is evident that
integer that satisfies one of
has the same solutions, for the two are equivalent. Again, consider the
congruence 2x = 4 (mod 6). Every integer that satisfies this congruence
also satisfies the equivalent
lies in
the fact that
all
congruence x
2 (mod 3).
The
distinction
the integers that satisfy the latter congruence are
in the class of 2 for the
modulus
3,
but these integers are separated into
The first conthe classes of the residues 2 and 5 for the modulus 6.
gruence, therefore, has two incongruent solutions modulo 6, while the
second one has but one solution modulo 3.
It is not true, however, that two congruences that have the same soluBoth 2x =
tions for a given modulus are necessarily equivalent.
4 = 0(mod 6) have the solutions x = 2 (mod 6)
4 (mod 6) and x 2 x
5 (mod 6), yet they are not equivalent.

single conditional congruence may be equivalent to a set of simul-
and x
When f(x) = 0(mod m) and m = raim 2

mr
r, and i ^ j, then the conwith i, j = 1, 2,
gruence can be broken up into the set of congruences f(x) = 0(mod mi).
Conversely, if (ra*, my) = 1, the set of congruences f(x) = 0(mod m t ) can
be combined to form f(x) = 0(mod m). We shall say in this case that
For
f(x) = 0(mod m) is equivalent to the set f(x) = 0(mod m^).
example, 5x = l(mod 12) is equivalent to the set of two congruences
5x = l(mod 3) and bx = l(mod 4). It follows that if x = (mod 12) is
Cona solution of the first congruence, then x G also satisfies the last two.
versely, any simultaneous solution of the set of two congruences implies
that the original congruence has the same solution, for the moduli 3 and 4
are relatively prime, and thus when 5x 1 is divisible by both 3 and 4,
taneous congruences.
where
(m, m/)
1,
divisible by their product.

But when two congruences are themselves not equivalent, the existence
of a solution of either one of them may often be determined by showing
the existence of a solution of the other one.
In that case we speak of the
it is
problems of the existence of the solutions as being equivalent.
Notice
70
that if the congruence 2x = 3 (mod 9) has a solution, then 2x = l(mod 3)

has a solution, and conversely. This statement is valid, for any solution
x = (mod 9) of the first congruence would have to be a multiple of 3,
and consequently each x /3
the second congruence.
satisfies
other hand, the second congruence
is
satisfied
by
all
On
the
integers in the class
modulo 3, and none of these integers is a multiple of 3. However,

= x'(mod 3) is a solution of 2x = l(mod 3), x = 3'( m od 9) must
be a solution of 2x = 3 (mod 9). Thus the questions of the existence of
of 2
where x
solutions of these congruences are mutually dependent.
Example.
7
0(mod
To determine whether
15)
and 9x
Sx
or not the congruences 3x 2
= 0(mod
11
15) are equivalent,
we
-f-
try
modulus that will change the leading

congruence into the one we want. What value of
to find a multiplier prime to the

coefficient of the first
y will satisfy Sy
We reject
3 since
9(mod
it is
15)?
The
not prime to
solutions are 3,
15.
8,
and 13 modulo 15.
Multiplying each
member
of the
congruence by 8, we have 24# 2 8x + 56 = 0(mod 15), which

11 = 0(mod 15).
Thus the congruences are
reduces to 9x 2 Sx
first
equivalent.
EXERCISES
1.
2.
3.
4.
5.
6.
2
7.
8.
x 3 - 2 = 0(mod 3) equivalent to x z + 1 = 0(mod 3)?

Show that 2x 3 - x 2 + 2 = 0(mod 5) is equivalent to 2x 3 + 4z 2 = 3 (mod 5).
Show that x 2 - 2x + 1 = 0(mod 3) is equivalent to 2x 2 - x + 2 = 0(mod 3).
Show that 3a: 2 - 6 = 0(mod 15) is equivalent to x 2 - 2 = 0(mod 5).
Are 2x 2 + x - 4 = 0(mod 5) and x 2 + 3x - 2 = 0(mod 5) equivalent?
Is the congruence x z 2 = 0(mod 3) equivalent to the congruence 2x 3 x
= 0(mod 3)? Are the solutions the same?
Develop two congruences equivalent to x 2 9 = 0(mod 12).
Are the congruences x z x = 0(mod 3) and x 4 + 2x 2 = 0(mod 3) equivalent?
Is
Are the solutions the same?

2
9. Note that 2x = 8 (mod 10) and x
Are they equivalent?
tions.
2>x
= 0(mod
10)
have the same solu-
Can the double of an integer give the

5-3. Linear Congruences.
remainder 7 when it is divided by 52? Has 2x = 7 (mod 52) a solution?
Theorem 5-1. When a is prime to m, the congruence ax = l(mod m)
has one and only one solution modulo m and this solution is prime to m.
The congruence ax = l(mod m) is equivalent to the equation ax +
my = 1, which we have shown always has a solution in integers when
There can be but one solution modulo m, moreover, for if
(a, m) = 1.
both xi and x 2 satisfy ax = l(mod m),
axi
ss
ax 2 (mod m)
Xi
.x 2
and
(mod m)

Furthermore, when ax\
71
l(mod m), it is clear that X\ is prime to m.

prime to m, the congruence ax = 6 (mod m)
has one and only one solution modulo m.
Now consider the congruence ax = b (mod m) with (a, m) = 1. We
know that ax = l(mod m) has a solution x = i(mod m). Then
Theorem
5-2.
When
is
axib
6 (mod
m)
shows that # = Xi6(mod m) satisfies the original congruence.

Again, as proved above, there can be but one solution modulo m, but
notice that it is not necessarily prime to m.
Theorem 5-3. If d is the greatest common divisor of a and m, the
congruence ax = b (mod m) has a solution if and only if d divides b.
When d does divide b, there are exactly d incongruent solutions modulo m.
Let (a, m) = d, a = a d, and m = m d. If ax = b (mod m) has a solud divides
tion, it is evident that
cd,
On
b.
the other hand,
if
d divides
b,
let
and reduce the congruence to the equivalent congruence
a x
c(mod
m ) = 1, and there is but one solution x = #i(mod m ).

km and
Consider this class of integers all of which are of the form Xi
obviously satisfy the given congruence. We wish to know whether these
In this case (a
integers constitute one or

like to
know
for
modulo m.
class
We
and only
solutions
modulo m; that
of k these integers are in the
is,
we should
same residue
see that
Xi
if
more
which values
kmo
Xi
0(mod m)
sm (mod m)
if
(k
s)
and only if k
Consequently, when k ranges
s = 0(mod d).
through d 1, the integers xi + kmo represent exactly d solutions
that are incongruent for the modulus m and all solutions of the given
congruence lie in one of these d classes modulo m.
Example. Solve: 15x = 12(mod 36).
Since (15, 36) = 3 and 3 12, we reduce the congruence to hx =
4(mod 12) of which there is one solution x = 8(mod 12). Hence, the
solutions of the original congruence are x = 8, 20, 32 (mod 36).
that
is, if
from
EXERCISES
State with reasons the
number
of distinct solutions of the following congruences,
find the solutions.

1.
3.
5.
3x m 5(mod 9).
6z = 3(mod 18).
12x s 36 (mod 56).
2.
4.
5x = 3(mod 27).
49z = 23(mod 125).
and
72
5-4. Division of Polynomials.
and f2 (x) are
If fi(x)
+m
integral polyno-
where g(x) is
an integral polynomial, by f 2 (x), obtaining integral polynomials q(x) and
mials and
possible to divide /i(x) or/i(x)
if it is
g(x),
r(x) so that
fi(x)
} 2 (x)q(x)
an identical congruence, and
(mod m)
r(a;)
is continued until r{x) is

congruent
to zero modulo m,
f
then q(x) is said to be the quotient and r(x) the remainder in the division
modulo m of fi(x) by f 2 (x). When r(x) = 0(mod m) identically, the
division is said to be exact and both f 2 (x) and q(x) are factors modulo m
Moreover, fi(x) is a multiple modulo
offi(x) or divisors modulo m oifi(x).
is
either lower in degree than
of
we
ordinary long division,

a divisor
As
modulo
2a;
l(mod
x2
there
On
6)
no value of a that
is
divide
modulus
6,
but
it is
2x
is
we
by
a;
3 using
is,
and therefore we say that x
of
b that
will satisfy this congruence.

2
x by 2x
2x
4a;
modulo 6
is
not exact, for
5\
\x 2
\x 2
lOx
4x
4x
l(mod
10
11
If
divide x 2
lOz
Hence, 4z 2
a remainder 12 which
3.
the other hand, the division of
possible,
2a;
1 by Sx 2 modulo 6, for if
would make (Sx 2) (ax -f- b) +
an identical congruence, then 3a = l(mod 6). But
there were a quotient ax
is
find that there
we cannot
stands,
it
6 of
we
if
for the
course, in the class of
or
(x)
f 2 (x).
Notice, for example, that
is
the division
if
is
(2x
try to divide
4a;
5) (2x
+x+
2)
6)
5(mod
6)
5 (mod 6).
by 2x
modulo
6,
we
find
we cannot
carry out the division far enough to obtain a remainder of the required
form, for
2x
5[
4x 2
4x 2
10a;
11a;
But 2y
5 (mod 6) has
identical congruence
However,
instead of
no
2
solution,
2x
5a;
l(mod
6)
and the best we can do
we change the form

5, we have
if
2a;
4a;
(2x
of the
5)(2x)
5x
is
to write the
l(mod
6).
above divisor using 4a;
4x
5| 4a;
+
+
2
2
4a;
\-x +
4x
4a;
0(mod
6 ==
Hence,
4a;
+ +
a;
We could
5a;
73
6)
= (-4x - 5)(-x +
obtain a like result by
first
l)(mod 6).
adding Qx 2 to 4a; 2
+x+
for
we
find that
2a
10z 2
10x 2
x .+
5s
25a;
26a;
2a;
l(mod
2a;
We,
therefore,
have
4a;
Notice, furthermore, that
by
3a;
1,
(2x
5) (5a;
when the modulus
6)
= 0(mod
is 6,
6)
l)(mod
we can
6).
divide
3a;
getting three distinct quotients and three distinct remainders,
for
3a;
3a;
+
+
=
=
(3a;
(3a;
2
2
l)(x)
+ x + l(mod 6)
+ 3a; + l(mod 6)
l)(3a;)
and
3a;
When
the modulus
is
(3a;
l)(5x)
5x
l(mod
6)
a prime and the polynomials are not constants,
it
easy to show that the division modulo p of fi(x) by fz(x) can always be
accomplished, for any term present in either one of these polynomials has
is
is prime to the modulus and the congruence ay =

has
exactly
one solution when (a,p) = 1. The remainder r(x)
p)
will, therefore, be congruent either to zero modulo p or to an expression
that is at least one degree lower than the degree of the divisor.
In this
case, moreover, both q(x) and r(x) are unique, for if
a coefficient that
6(mod
fi(x) =J2{x)qi{x)
ri(aO(mod p)
r 2 (x)
and
fi(x)
= f2(x)q
(x)
(mod
p)
then
f2(x)[qi(x)
If q\{x)
q 2 (x) f
0(mod
q 2 (x)]
r 2 (x)
ri(x)(mod p)
p), let its leading coefficient
be
0(mod
p),
the leading coefficient of f 2 (x) be a ^ 0(mod p).

Then the leading coefficient of f2 (x)[qi(x) q 2 (x)] is a 6o ^ 0(mod p), and the degree
and
let
of this expression is at least that of

r 2 (x)
ri(x).
identically
This
is
impossible,
and likewise that
7*1(0;)
f2 (x), thereby exceeding the degree
and we
infer that qi(x)
r 2 (o;)(mod p).
q-ii-v)
(mod
of
p)
74
EXERCISES
1.
why
Explain
division
modulo
x2
+3
cannot be divided by
4a;
modulo
Carry out the
8.
5.
Divide x 2 - 2x + 5 by 2x - 3 modulo 7.
Divide 3x 2 - 2x + 4 by 2x - 1 modulo 15.
2
4. Divide x 3 - 2x + 5x - 1 by 2x - 3 modulo 11.
5. Divide 2x 2 2x + 5 by 2x 3 modulo 6, and then change the divisor to the
form 4x 3 and carry out the division modulo 6.
2x
1 by 2>x - 1 modulo 11.
6. Divide Ax z - 3z 2
n
and g(x) = b Q x + friz'" 1 +
7. If fix) = a x
ai^- 1
+b
are integral polynomials with t < n, and if (b m) = 1, do polynomials g(x) and r(x)
exist so that /(a:) h= g(x)q(x) +r(a;)(mod m) with r{x) lower in degree than g{x)l
If so, are these polynomials unique modulo m?
2.
3.
++
Theorem 5-4. If x = r(mod m) is a solution of the congruence f(x) =

0(mod m), where the polynomial f(x) = a Q x n + aix n ~ +
+ an
with a ^ 0(mod m), then x r is a factor of /(#) for the modulus m, and
l
conversely.
According to the remainder theorem of algebra,

is divided by x r.
Consequently,
remainder
f(r) is the
when f(x)
(x
r)q(x)
- f(r) =
(x
r)q(x)
f(x)
+ f(r)
or
f(x)
identically,
But
f(r) is
and q(x) is a polynomial

an integer, and furthermore
n-1
/(r)
b 1x n
~2
= 0(mod
Therefore, the coefficients of f(x)
the given congruence.
&_i.
m), for r satisfies
f(r)
are
Moreover, it is evident from a consideration of the process of

long division or from the following argument that the coefficients of q(x)
Because the leading coefficient of f(x) is an integer, the
are integers.
product (x r)q(x) shows that 6 = a is an integer. If bi were not an
where rb Q is an integer, would not
integer, the coefficient a\ = b\ rb
be an integer. In like manner the fact that 6 -_i and bi r6 -_i, where
i = 1, 2,
n, are integers implies that bi is also an integer.
Thus
Hence,
q(x) is an integral polynomial.
integers.
f( x )
f( r)
and so
f(x)
identically,
=
=
(x
r)q(x)
(x
r)q(x)(mod m)
(mod m)
r is a factor of f(x) for the
f(x)
(x
f(r)
= 0(mod
showing that x
Conversely,
if
r)g(a:)(mod m)
then
and x
r(mod m)
is
m)
a solution oi fix)
0(mod m).
modulus m.

It is easy to
quotient
5-5.
is
show that when x
r is
75
a factor modulo
of f(x), the
unique.
The Number and
Multiplicity of the Solutions of a Congruence.
We have shown that the congruence 2x 4 = 0(mod 6) has the solutions

x = 2 (mod 6) and x = 5 (mod 6). Therefore, Theorem 5-4 shows that
x 2 and x 5 are factors modulo 6 of 2x 4. Thus 2x 4 =
2(x 2) (mod 6), and 2x 4 = 2(x 5) (mod 6). Notice, however,
that the product (x 2)(x 5) is not a factor modulo 6 of 2x 4. On
if the modulus is a prime, the proof of the following
theorem brings out the fact that when ri, r 2
r s are incongruent
integers modulo p that satisfy /(#) = 0(mod p), the product (x ri)(x
the other hand,
(x r s ) is a factor modulo p of f(x)

The eminent mathematician Lagrange (1736-1813) improved the work
of his teacher Euler in the calculus of variations, did extensive work in the
r 2)
solution of algebraic equations,
and expanded the theory
of differential
In the theory of numbers he was the first to prove Wilson's

theorem and the first to prove that every integer is a sum of at most four
equations.
He gave a
complete proof of the method of solving the equation

Again, he made important contributions to
the theory of quadratic residues and binary quadratic forms. The next
theorem is credited to him and reflects his interest in the solution of
squares.
x2
by 2
in integers.
equations as well as congruences.

Theorem 5-5 (Lagrange's Theorem).
If
is
a prime and f(x)
+.+ o in which a ^
integral polynomial a x n
ai^ n_1
then the congruence j(x)
0(mod
p) has at
is
an
0(mod
p),
most n incongruent solutions
modulo p.
We have observed that the congruence f(x)
0(mod p) need have no

ri(mod p), it follows from Theorem
If it should happen
ri is a factor of f(x) for the modulus p.
that (x ri) B1 where 1 < rii < n, is a factor modulo p of f(x) and if this
power is the highest power of x ri contained as a factor in f(x), then
x = ri (mod p) is said to be a solution of multiplicity n h and
but
5-4 that x
solution,
if
there
is
a solution x
f(x)
(x
ri)
ni
gi(x)(mod p)
with qi(x) of degree n n\ and leading coefficient b = a (mod p).

If rii < n and if the original congruence has another solution x
r 2 (mod p) with r 2 incongruent to n modulo p, then
/(r 2 )
But since /(r 2 )

0(mod p) and,
= 0(mod p)
(r 2
and
rx
ri)
ni
gi(r 2 )(mod p)
r 2 (mod p), it is
as above,
qi(x)
(x
r 2 ) m q2(x)(mod p)
evident that gi(r 2 )
76
so that
f(x)
(x
ri)
ni
(x
r 2 ) n2 g 2 (V)(mod p)
identically.
= O(mod
p) has one or more other distinct solutions for the

we continue in this manner until either all the solutions which
are fewer than n in number have been found or we have at most n linear
In the latter case we find the identical
factors modulo p of f(x).
If f(x)
modulus
p,
congruence
f(x)
+n +
a (x
rtl
ri)
(x
r 2 ) n2
(x
r k ) nk (mod p)
(1)
n k = n, and hence there are n solutions, for a

where i = 1, 2,
k, is counted ni times.
If we now substitute for xm (1) any integer s not congruent to any r^
where i = 1, 2,
k, for the modulus p, we find
where n\
solution of multiplicity
/()
But s
0(mod
Ti
p),
rii,
- rîs -
a Q (s
r 2 ) n2
(s
r k ) n *(mod p)
(2)
^ 0(mod p), and a ^ 0(mod p). Consequently,

= s(mod p) is not a solution of f(x) = 0(mod p).
and x
f(s)
There-
can be no more than n solutions.

show, furthermore, that the solutions are the same regardless of the order in which they are found.
Moreover, the multiplicity of each solution is unique regardless of the
order in which the factors are obtained, for if
fore, there
It is easy to
f(x)
(x
rxYq^x)
(x
r 1 ) v q 2 (x)(mod p)
where neither qi(x) nor q 2 (x) is divisible modulo p by x

u > v, we have the identical congruence
(x
But
is
if
ri) [(x
ri)
u- v
qi(x)
q 2 (x)]
= 0(mod
p)
u~v
the leading coefficient of the expansion of (x
not congruent to zero modulo
p,
ri)
rh
qi(x)
and
if
qi(x)
the leading coefficient of the last con-
gruence written in the expanded form cannot be congruent to zero modulo

The congruence states, however, that when it is written in the form
p.
~
a n ss 0(mod p), all its coefficients are multiples
a xn
aix n l
Consequently, the leading coefficient and, in like manner, each

~
of the other coefficients of the expansion of (x ri) u v qi(x) q%{x) are
congruent to zero modulo p. Therefore, we have the identical congruence
of p.
(x
Substituting
r\
for x,
we
(ri
ri)
w- y
giO)
q 2 (x)(mod p)
gi(ri)
# 2 (ri)(mod p)
find
ri)
w- y

As a
= 0(mod
result q^ri)
which
a factor
r\ is
when n >
obvious, of course, that
It is
modulo p
of #2(2),
is
0(mod
and x
p),
77
have no more than p
p) can
Again,
us observe that
let
we
if
the congruence f(x)
p,
distinct solutions
modulo
p.
place no condition on the a
0(mod p) and assume that the congruence has an (n

= s(mod p) that is distinct from each r where i = 1,
of f(x)
l)st solution
2,
4-,
k,
for
the modulus p, the congruence (2) requires that a Q be congruent to zero

modulo p. Consequently, the identical congruence (1) implies that
every coefficient of f(x)

conclude:
congruent to zero modulo
is
Hence, we
p.
Theorem 5-6. If the congruence a x n + a\x n ~ x +

0(mod p), with p a prime, has more than n solutions (a solution
an
of multi-
n, is
m being counted m times), each aiy where i = 0, 1,
congruent to zero modulo p and the congruence is an identical congruence.
Example. Find by trial the solutions of x*
x2 x
2 = 0(mod 5).
4
2
If f(x) = x
Hence, x
1 is a factor of
x x
2, /( 1) =5.
3
By using synthetic division we find f(x) = (x
f(x) modulo 5.
1) (x
plicity
+
+
Then/i(-3) =
-45, and therefore f(x) = (x -f l)(x + S)(x 2 - 4x + 4) (mod 5). It is
now evident that f(x) = (x + l)(x + 3) (a; 2) 2 (mod 5) and that
besides the solutions x = l(mod 5), x = 3 (mod 5), there is a double
solution re = 2 (mod 5).
Theorem 5-7. If p is a prime, the congruence f(x) = 0(mod p) of
degree n < p has a solution # = a (mod p) of multiplicity r < n if and
only if /(a) = 0(mod p), /'(a) = 0(mod p),
/^(a) = 0(mod p),
and/^(a) ^ 0(mod p).
We recall that x = a (mod p) is a solution of multiplicity r of /(z) =
0(mod p) if and only if (x a) r is a factor modulo p of the polynomial
r+1 is not
Furthermore, we
a factor modulo p of /(x).
f(x), but (# a)
x2
2x
3)
(mod
5).
Let/iO) =
x*
x2
2x
3.
= a xn
~
n~ 2
a n as f(x) = atfix n l
l)x
a n _i,
i(^
the application of Taylor's theorem to the polynomial f(x) sets up the
notice that
ai# n-1
we
if
/(a)
If /(a)
but / (r) (a)
identity
f(x)
define the derivative with respect to x of f(x)
+ (x-
a)f(a)
0(mod p),/'(a)
0(mod p) for
+
=
r
(x
a)
O(modp),
<
n, then, the
+..-+.(*.
and/â) =
{i)
(a)/i\
a)
(a;
- aY J^- + +(*-
/(x)
Or
and
a) r Q(x)(mod p)
a)
-^
O(modp),
being integers,
have
/(*)
-^p
(mod
p)
we
78
If (x
a) r+1
were a factor modulo p
(x
and the
a) r+1 q(x)
then
of f(x),
(x
a) r Q(x)(mod p)
identical congruence
(x
a) r [(x
a)q(x)
Q(x)]
0(mod
p)
implies that
(x
a)q(x)
Q(x)(mod
p)
n -r
Hence, [f (r) (a)/r\]

(x - a)
[f^(a)/n\] has the factor x a modulo p, and so/ (r) (a) = 0(mod p), which is contrary to the hypothesis.
To prove the converse of the theorem, first call to mind the rule for
differentiating a product of two functions, and remember that when

polynomial has been factored into two polynomials, this rule enables us
to set up an identity between two forms of f'(x), for if
>
f(x)
a Qx n
ciiX"-
+- +
a*
(b Q x
xs
b r )(c
'
+C
'
8)
then
f'(x)
a n^ n_1
kXcowr-
ai(n
1
+"+ &n-i
+ c _i) + (b rx ~ +
l)x n
"
~2
(b
xr
& r_i)(co^
+
On
this basis the
method
of induction
shows that
it is
are polynomials in
d k (uv)
dk u
~ ~
~dx~
dx~k+
Consequently,
if
we may,
x,
f(x)
dv d k
~l
c.)
If
u and
therefore, write
u
+
dx dx ^
1
valid to apply
Leibnitz's rule for the kih derivative of a product to this case.

v
k(k
1)
d 2 v dk
~2
2!
+
dx~ d^F
2
'
'
dkv
"
dx~k
a) r q(x) (mod p), then the formula verifies

0(mod p) for k < r, because clearly each
a when s < r. But (x a) r+l is not a
and so we infer from q(a) ^ 0(mod p) and
(x
the statement that f (k) (a) =

d s (x a) r /dx s has a factor x
factor
modulo p
^ [(X
X)]
~dx^
that/ (r) (a)
To make
of f(x),
^ ^+
ir)
'
'
'
0(mod
it
plain that the restriction which
r ( r!)(
^ ~
WW +
( r!)
^W
p).
Theorem
5-7 places on the
degree of f(x) = 0(mod p) is a necessary one, consider the congruence

10
a;
x b = 0(mod 5).
Its only solutions are x = 0(mod 5) of multi-
and the simple solution x = l(mod 5). Nevertheless, f (k) (0)

and f (l) are congruent to modulo 5 for all positive values of k.
When we are solving a congruence by trial, It is often helpful to make
use of the following two theorems to eliminate some of the possible
plicity 5
(k)
solutions
'
79
*Theorem 5-8. If r satisfies the congruence a Q x n

a n = 0(mod m), then r is a factor modulo m of a n
a\X n
~x
Since r
is
a solution of the congruence, a x n
r)Q(^)(mod
= a (mod m).
(ic
modulo m of a n
*Theorem 5-9.
m), where Q(x)
Hence, a n
n~ 1
= r6 n _i(mod
aix n
++ =
m) and
6 n _i(mod
m), and thus r
is
a factor
the congruence f(x)
If r satisfies
0(mod m), then
a divides f(a) modulo m.

The hypothesis implies that f(x) = (x r)Q(rc)(mod m).
tuting a for x, we derive f(a) = (a r)Q(a)(mod m).
r
Substi-
Example. Solve /(x) = ^ 3 - 4x 2 - 11a;

6 = 0(mod 12) by trial.
Applying Theorem 5-8, the possible solutions are obtained by consider-
ing the congruences rx

r
6(mod
12) for r
1, 2,
The values
,11.
4 and 8 are immediately ruled out, for the corresponding linear con-
gruences have no solutiom-
Then we
values r
(r
\)y
rowed to
= 8, and we
apply Theorem 5-9. The

7 and 10 are thereby eliminated, for there is no y such that
find that /fl)
= 8 (mod
The
12).
2, 3, 5, 6, 9,
and
possible solutions are, therefore,
modulo
11
values in f(x) shows that each one
12,
and the substitution
nar-
of these
a solution of the given congruence.
is
The Chinese Remainder Theorem. The problem of finding an

integer that yields certain remainders upon being divided by given
integers was familiar to the Chinese as early as the first century.
The
solution is now easily effected by using congruences.
When the integers
mh m2
m n are relatively prime in pairs, we wish to find an integer
5-6.
x such that
x
=
=
ai(mod mi)
a 2 (mod mi)
a n (mod
mn
Let
\\
t=i
rrti
and Mi =
M/rrii,
where
1, 2,
n.
Then
set
up
the n congruences
Mix
In each case (M{, mi) =
of each congruence (4).
1,
l(mod
and there
Now
mi)
(4)
exactly one solution x
is
.x\(mod
X = M iXidi + M2X2CL2 +
+ M nx n a (mod M)
and substitute X for x in x = ai(mod mi). Since M and each M
Mi
every integer of
and MiXi is congruent to

the form X + kM satisfies x =
are congruent to
0,
z)
consider
n
for the
except
modulus mi,
a! (mod mi).
In like
80
manner the
integers in the residue class of
modulo
M satisfy
all
the
given congruences.
But
this class of integers
congruences, for
and Xi
if
is
the only simultaneous solution of the set of
in the given congruences (3)
X=
and since the
shows that
Xi(mod
prime in
ra z are relatively
-
X=
pairs, the
m,-)
pairs,
Xi(mod M)
Hence, we have shown that

Theorem 5-10. If the im, where
prime in
there were a second solution Xi, the substitution of
congruences x
2,
1,
a; (mod
n, are relatively
have one and only one
m)
M=
simultaneous solution modulo
55a:
S5x
=
=
=
6,
when
Hence,
X=
385).
M* = 55
x = 77
3 = 35
2x = l(mod 5) has the solution x = 3 (mod 5)
7) or Qx = l(mod 7) has the solution x = 6 (mod 7)
11) or 2x = l(mod 11) has the solution x = 6 (mod 11)
l(mod
l(mod
l(mod
27(mod
5-7.
2,
5, 7,
M
77x
m^
Find all integers that give the remainders

and 11, respectively.
Example.
divided by
\\
385
5) or
(77) (3) (2)
(55) (6) (6)
(35) (6) (5)
Other Simultaneous Linear Congruences.
a method for finding a solution,
when
(mod
385),
We shall
or
Xm
demonstrate
of certain linear simul-
it exists,
taneous congruences whose moduli are not relatively prime in pairs by

proving the following theorem by induction:
Theorem 5-11. The set of n linear congruences x = a;(mod mi) has
if and only if the greatest common
and j having the values 1, 2,
divisor of
a solution
m^
mj,
corresponding a
of the set,t all
least
common
a,j.
common
When
the integer
n,
satisfies
solutions take the form
multiple of the
m,-
and
is
any
any pair
with
^ j,
of moduli,
divides the
each congruence
where L is the
X + Lt,
integer.
Taking the two congruences

x
x
=
=
ai(mod mi)
a 2 (mod m 2 )
f Oystein Ore gave the general form of the solution in

6, pp. 365-370, 1952.
No.
Am. Math. Monthly,
Vol. 59,
with d\2
(mi,
2 ),
let
81
us suppose that x satisfies both of them.
Then
since
=
=
x
Xq
we
ai(mod
a 2 (mod
m
m
x)
2)
infer that
=
=
ai(mod
di 2 )
a 2 (mod
c? i2 )
a-i
a 2 (mod d i2 )
Xq
and that
if d i2 divides a\ a 2 we can show that there is a solution

congruence
that satisfies the second one.
Every solution of
of the
and
if
any
of these integers
the first congruence is of the form a\
nay,
satisfy the second congruence, the values of y are determined by the
Conversely,
first
congruence
cti
miy
a 2 (mod
miy
a2
2)
or
But
since
dn
(a 2
i),
there
is
ai(mod
2)
at least one value y of y that produces
miy of the two congruences.

Moreover, if there are two integers x and x\ that satisfy the given
congruences (5), substituting them in these congruences shows that
a simultaneous solution a\
x1
=
=
x (mod mi)
x (mod
2)
Xi
^ (mod L)
X\
and hence that
where
is
the least
integer of the
common
multiple of mi and
+ Lt is a
form x
common
Furthermore, any
2.
if
Xo
both the moduli mi and m 2

Suppose now that the theorem is true for a set of k 1 congruences of
for Lt is congruent to
is,
the given form.
Then
for
if
there
is
common
the least
If
tainly
dij
i t* j.
is
dji
common multiple of mi,
2,
X
X +L
solution
simultaneous solutions of the set are of the form

is
solution of the congruences
k -it,
m^-i,
and
and only
where L k -i
all
the parameter.
is
a solution of a set of k congruences of the required form, cersatisfies
the
first
(mi, mj) divides a*
congruences, and thus

a,-
for
t,
1,
2,
it is
.
But
X=
ai(mod m*)
X=
and
fc
(mod m&)
1, 2,
true that
1,
and
82
Hence,
X=
X=
a k (mod d ik )
a (mod d ik )
(mod dik )
a*
and therefore
di
Consequently,
Conversely,
dij
divides a*
if
dy divides
solutions of the
where
fc
first
for
aj
di
i,
for
aj
1, 2,
i,
where
k,
2,
1,
k,
^ j.
common
congruences exist and have the form
X +
moreover, possible to
determine a value for the parameter t so that this expression will produce
To prove this statement, consider the
a solution of x = a (mod m k ).
congruence
Lk-it,
is
a simultaneous solution.
It
is,
fc
X + Lk-it =
in the
Lk-it
We
a k (mod
k)
form
X (mod m k
ak
can show that this congruence has a solution
ing that the greatest

If
is
common
divisor of
a prime factor of any of the
L k -i and
=
rrik
(mod
k)
divides ak
by prov-
w!i be the exponent of the
ra 4-, let
The highest power of this

power of p that is in any one of m h m 2
m k -i. Suppose it is p m/ the power of p that occurs in m r where 1 < r <
Then the exponent of the highest power of the prime p that is
h 1.
Let this integer
contained in D = (m k L k -i) is the smaller of m'k and m!r
But
be m's
highest power of that prime contained in m*.
in Lk-i is the highest
ar
= 0(mod m
X -
ar
= 0(mod
r)
and hence
m/
)
is divisible by the greatest common divisor of m k and m r it

by that power of p determined by the smaller of the two
But from the above congruences
exponents m'h and raj., that is, by m's
X a r is divisible by the power of p designated by the smaller of the
exponents m!k and raj.. Hence,
Since a k
is
ar
divisible
X -
ar
= 0(mod
ar
X -
ak
- 0(mod
m ')
and
ak
0(mod p
vls ')
Therefore,
Furthermore, since this congruence
(m k L k -i), the integer

,
ak
is
is
m>)
any prime factor of D =

by the product of the very
true for
divisible
83
a k is divisible
powers of the primes that are contained in D that is,
by D. Thus there is a value of t that forces a solution of the first k 1
congruences to satisfy the kih one, x = a (mod m&).
Again, any two solutions of the set of k congruences x = a* (mod mi)
;
fc
must be congruent modulo the least common multiple

m k for two solutions X and Xi are such that
of mi,
2,
X =
X =
Xi(mod L -i)
Xi(mod m k )
X =
Xi(mod Lk)
fc
and hence
Lk
where
2,
is
,
the least
common
Moreover,
k.
congruence of the given set

Example. Solve
Let d!
evident
(18,
that
dx
15)
\
(7
3,
L k -i and m k and
form X + L k
common solution.
multiple of
integers of the
all
if
x
x
x
=
=
=
7(mod
10(mod
l(mod
d2
(18,
d2
10),
is
(7
14)
hence of m h
satisfy each
18)
15)
14)
1),
dz
2,
d3
(10
14)
(15,
1.
It is
Therefore,
substitute 7
-f-
1).
solution exists.
Substituting 7
second congruence,
18 in the
18*
3(mod
15)
l(mod
15)
we
find
so that
Since 90
18
is
the least
common
multiple of 18 and 15,
we
90s in the third congruence and have
= -24(mod
90s
14)
and therefore
Hence, the
common
solution
is
3 (mod 14)
X=
295 (mod 630).
EXERCISES
1.
2.
3.
a.
6.
Find by trial all the solutions of 3z 4 - 3z 3 - 2z 2 + 3z + 6

0(mod 7).
Find the solutions of x + 10x 5 + 4.x 4 - x 3 + x + 6 = 0(mod 11).
Decide whether or not the following congruences have multiple solutions.
x3
x2
2z 3
+3 =
+1 =
0(mod
0(mod
5)
3)
84
Find the
4.
common
solutions of the following sets of congruences.
m 2(mod
= 4(mod
= 9 (mod
x
x
x
= 11 (mod 21)
= 2 (mod 12)
s 4(mod 10)
a
=
=
x
x
a.
b.
c.
Find a multiple
6.
respectively,
5-8.
If
by
of 7 that
2, 3, 4, 5,
The Number
and
14)
12 (mod 46)
l(mod 31)
16 (mod 28)
has the remainders
1, 2, 3, 4,
and 5 when
it is
divided,
6.
of Solutions of the
Congruence
f(x)
r, are distinct primes and

the p^ where i = 1, 2,
nr
p r there is a solution of the congruence
.
11)
15)
0(mod m).
m =
ni
pi p 2
n*
J{x)
where /(#)
is
0(mod w)
an integral polynomial,
if
and only
(6)
if
there
is
a simultaneous
solution of the equivalent set of congruences
Six)
=
=
0(mod pi ni )
0(mod p 2 )
/(x)
0(mod p r nr )
S{x)
if there is a solution of each of the individual congruences of (7), we

can build up a simultaneous solution of the set by means of the Chinese
remainder theorem. Suppose that integers Xi satisfy /(x) = 0(mod Pi ni ),
Then find the common solution of the
where i 1, 2,
r.
congruences
x = i(mod pi ni )
(8)
But
Let this solution be x

set (7)
and the
= Xi(mod
m).
The
integer
X\
satisfies
original congruence (6) because for each
Xi
x z (mod pi ni )
and therefore
SiX,)
=f(xi)(mo&pf)
But
= 0(mod
pi**)
x)
0(mod
p^-i)
SiX,)
0(mod m)
Sixi)
so that
SiX
and
both the
85
the set of congruences (8) we replace the solution of just one of the
congruences of (7), say the first one, by a solution x[ distinct from X\
If in
modulo pi ni the solution x

,
x
x
=X
(mod m)
= xi(mod pi ni
= ^(mod pfi)
of the resulting set,
be distinct modulo
were the same,
will
2, 3,
m from the solution
(mod m)
=X
(mod
Xi =
Xi
of (8), for
if
the solutions
would imply that
Xi
pi
ni
1, 2,
Then
xi
xJ(mod pi ni )
whereas we assumed that these integers are incongruent solutions of

= 0(mod pi ni ). Hence, if each congruence of (7) has kt solutions, by
inserting them in (8), we see that there are kik 2
k r solutions modulo
fix)
m of f(x) =
0(mod m).
It is evident too that
there
no solution
is
Theorem
5-12.
if
If
one congruence of
(7) fails to
have a
solution,
0(mod m).
Thus we have shown that:

the congruences f(x) = 0(mod Pi ni ), where i =
of f(x)
1,
2,
r,
have
fc
>
solutions, respectively, there are exactly
]
i
k{
=l
incongruent solutions modulo
of f(x)
= 0(mod
m), where
m=
J[ pi
=i
ni
and the
pi are distinct primes.
We have, therefore, reduced the

f(x) = 0(mod m) to that of solving
power
5-9.
f(x)
problem
of solving a
congruence
a congruence whose modulus
is
of a prime.
The Solution of f{x) = 0(mod p ). Any integer that satisfies

0(mod p s ), where p is a prime and f(x) is an integral polynomial,
s
evidently satisfies the congruence

f(x)
= 0(mod
p*- 1 )
The
converse, however, is not true, but it is obvious that if the second

congruence fails to have a solution, f(x) = 0(mod p s ) can have no solution.
Suppose that f(x) = 0(mod p 8-1 ) has a solution x = a;' (mod p*~ ).
Under what conditions will one of these integers, x' + kp 5-1 be a solution
of f(x) = 0(mod p s )?
We must determine which values of k, if any, will
l
86
satisfy the congruence
kp 8 '
fix'
n~ l
= 0(mod
When/(x) = a x + a\X
+
+ a n is a rational integral function
we have defined the derivative of fix) with respect to x as a Q nx n ~ 1 +
n~ 2
+
+ a n -\. The expansion of f(x + h) according to
diin l)x
Taylor's theorem is finite and is of necessity valid for integral values of x.
n
of x,
Consequently,
fix'
kp 8 -^)
f(x')
k V 8 -y{x')
fc
2 *- 2
0^
+
_+_
frn
f
n S -n J
(n)
(x')
\_>_
nl
mQ( J p s\
an identical congruence, the expressions (r) (re') /r! having been shown
~
Moreover, if s > 2, then p rs r > p 8 f or r > 2 and hence
to be integers.
all except the first two terms of the expansion are congruent to
for the
modulus p s so that
is
But
+ kp
+
8-1
that fix')
if
x'
kp 8- 1 )
fix'
= fix') + kp - f'ix')imod p
satisfy fix) = 0(mod p
this congruence
)
),
by p s and
kp 8 xf'ix') must be divisible

that fix')
chosen to satisfy the congruence fix')
know
to the congruence
When
tp
t
-1
,
to
is
and therefore the
kf'ix')
common
0(mod
kp s
-l
last
therefore k
f'ix
shows
must be
0(mod p 8 ).
We
congruence can be reduced
p).
and p is 1, the congruence

has exactly one solution k = /ci(mod p). This
s_1
(mod p s ) of fix) =
value of k yields exactly one solution x = x'
A)ip
kf'ix')
the greatest
= (mod
divisor of fix')
p)
0(mod p 8 ).
When
divide
t,
the greatest
common
the solution x
divisor of fix')
= a/ (mod p ~
= 0(mod p
8
and p
of fix)
But
is p,
and p does not
0(mod p ~
s
does not
p does divide t, fix') is a

multiple of p s and when k has the values 0,1,
p 1, each of the
8
8- 1
s
is a distinct solution modulo p of fix) = 0(mod p ).
p integers x' + kp
We have therefore proved:
'
+a and a/ satisfies
Theorem 5-13. If fix) = a x n + aix 1 +
s_1
and (1) if ifix'),p) = 1, there is
the congruence /(a?) = 0(mod p
)
exactly one value of k such that x = x' -\- &p s_1 (mod p s ) is a solution of
8
8
/(re) = 0(mod p ), but (2) if ifix'),p) = p, no solution of fix) = 0(mod p )
s
can be derived from x' unless x' itself satisfies fix) = 0(mod p ), in which
s
case x' + /cp s_1 yields exactly p incongruent solutions modulo p by letting
k have all the values in a complete residue system modulo p.
According to this theorem, to solve fix) = 0(mod p 8 ), we first solve
fx) = 0(mod p) and then determine by means of linear congruences
yield a solution of fix)
).
if
71

modulo p the solutions,
= 0(mod p 3 ), ....
Example.
2x 2 - 3x -
To
1
Since /(4)
=
19,
if
solve 2x 2
0(mod
t
19)
Sx
and
13ft
0(mod
= O(mod
19 2 ),
first
find the solutions x
Also f'(x)
1.
Hence, we must solve

= 4 (3) (19)
that x
there are any, of f(x)
87
0(mod
308 (mod
4x
19 2 )
3,
19).
4,
),
f(x)
take f(x)
7(mod
19).
and therefore /'(4) = 13.

But k = 3 (mod 19), so
a solution
is
of
the
original
congruence.
Let the student show that x = 235 (mod 19 2 ) is the solution derived
from x = 7 (mod 19).
After showing that x = 3 (mod 4) is the solution of f(x) = 0(mod 4),
apply the Chinese remainder theorem to find the solutions of f(x) =
0(mod 2 2
19 2 ).
EXERCISES
Solve the following congruences.
1. 3a: 2
3.
5.
7.
9.
2x
17a:
+3 =
-
0(mod
72).
= 0(mod 500).
- x + 20a: + 4 = 0(mod 1089).
2a:
- 2a; - 2 = 0(mod 2000).
x x -2x -8x - Q = 0(mod 357).
6a:
a:
20
+ 7a: + 16a: - 2 = 0(mod 45).

- 19a; + 32a: + 34 = 0(mod 75).
- 24a: - 9 s 0(mod 1350).
5a:
2a:
+ - 6a: - 13 = 0(mod 340).
- 5a: 2 - 12s + 6 = 0(mod 1700).
2.
4a: 3
4.
a:
6.
8.
10.
a;
a:
CHAPTER
THE THEOREMS OF FERMAT AND WILSON AND THE

MOBIUS FUNCTION
6-1. Fermat's Theorem.
Pierre de Fermat (1601-1665) was a profound scholar who was not especially attracted to mathematics until a
translation of the work of Diophantus excited his interest.
Once having
embarked on the study of mathematics, he influenced its development in
all the branches known at the time.
With Pascal he originated basic
ideas in the theory of probability, and independent of Descartes he
developed the analytical geometry. He also helped to lay the foundaBut he is renowned for his work in the theory of
tions for the calculus.
numbers. Unfortunately he failed to publish the many theorems he
discovered about the integers, and when he announced his results in
letters to other mathematicians, he usually concealed his methods of
proof.
as early as 500 b.c. that 2 P 2 is divisible by p if p

remained for Fermat in 1640 to state that if p is a prime
~
and (a, p) = 1, then a p l = l(mod p). We do not have his demonstration of this fact, but by 1761 both this theorem and the more general one
stated below had been proved by Euler.
The Chinese also believed that if n is not a prime, 2 n 2 is not divisible
by n. Even Leibnitz (1646-1716) thought this statement to be true.
Not until the latter part of the eighteenth century did mathematicians
The Chinese knew
is
a prime, but
it
The conjecture was finally dissigns of resolving the problem.

proved in 1819 when Sarrus found that 341, which is not a prime, is a
Since then it has been shown that there is an infinite
factor of 2 341 2.*
number of composites n that will divide 2 n 2.
Theorem 6-1. If the integer a is prime to m, a 0(m) = l(mod m).
r 0(M) is
We showed in Chap. 4 that when (a, m) = 1 and r h r 2
a reduced residue system modulo ra, the set ar h ar 2
arô) is also a
reduced residue system modulo ra. The integers in the second set are,
Hence,
therefore, in some order congruent to those in the first set.
show
*L. E. Dickson, " History of the Theory of Numbers," Vol.

Lehmer, Am. Math. Monthly, Vol. 43, No. 6, p. 347, 1936.
88
1,
p.
92.
D. H.
THE THEOREMS OF FERMAT AND WILSON
ar 2
rn(mod m)
r 2 (mod m)
ar^m)
n>(m)(mod m)
ari 53
89
Therefore,
(t>(-
m)
r 1r 2
= nr
r^ (m)
r^ (ro)
(mod m)
and
,<Kro)
Corollary
1.
If
Corollary 2.
If
56
Examples.
p
p
l(mod m)
is
a prime and a
is
a prime and a
48
7); 2
l(mod
prime to
is
any
l(mod
p,
av
~l
integer, a p
is
= l(mod p).
= a (mod p).
105).
EXERCISES
1.
2.
7?
3.
4.
5.
6.
for
Find the remainder when 7 48 is divided by 180; when ll 50 is divided by 180.

What do you notice about the least positive residues of the powers of 3 modulo
Is the same fact true of the powers of 2 modulo 7?
Find the remainder when 2 100 is divided by 11 and when 2 105 is divided by 11.
Show that n n 1 is divisible by 7 when n is prime to 7.
Find the solutions of the congruence x A 1 = 0(mod 12).
Prove that n 37 n is divisible by the product of the primes 2, 3, 5, 7, 13, and 37
any
positive n.
7.
Prove that a n
8.
Can you
b ls is divisible
find integers all of
by 133
whose
if
a and
b are
prime to 133.
digits are 9 that are divisible
by 7?
By
other
primes?
Fermafs Theorem
The solution of ax = b (mod m), where
6a^ (w)-1 (mod m).
6-2. Applications of
Theorem
is
6-2.
(a,
m)
1,
According to Fermat's theorem

a <t>(m)
i(
a^ m) b
6(mod m)
mo d m)
Hence,
and therefore x
Theorem
6-3.
greater than
2,
6a^ (w)_1 (mod
m)
is
If f(x)g(x)
~
xp l
x p-i
the solution of ax
l(mod
b (mod m).
where p
is a prime
each of the congruences f(x) = 0(mod p) and g(x) =
0(mod p) has the maximum number of incongruent solutions modulo p
permitted by its degree.
Fermat's theorem shows that the congruence
has exactly p
distinct solutions
0(mod
modulo
p),
(1)
p)
p.
When
>
2, if
we
factor
90
x p-i
modulo p into
tively, the
and
f(x)
g(x) of degrees r
and p
r,
respec-
congruence
f(x)
= 0(mod
p)
(2)
p)
(3)
has no more than r distinct solutions, and

g(x)
has at most p
f(x)g(x)
0(mod
^ 0(mod
r incongruent
p) has exactly p
1 ==
l)(a;P-2
p.
But
since
distinct solutions, neither of the
congruences (2) and (3) can have fewer than

incongruent solutions modulo p.
For example, if p > 2,
X V-1
modulo
solutions
1
maximum number
its
3P-I
l ==
0(mod
of
1)
Hence, the congruence

xP -2
has exactly p
.
_j_
^p-3
-{-
2 incongruent solutions
p)
modulo p and they are
2, 3,
P -i.
If p is a prime greater than 2, each of the congruences
0(mod p) andx^- /2 + 1 = 0(modp) has exactly (p - l)/2
that are incongruent modulo p.
Corollary
x (p-d/2
>
2,
solutions
For p
1.
x p-i
and therefore each

x (p-i)/2 4-
modulo
p.
0(mod
(a^-D/ 2
of the congruences x
p) has exactly (p
l)(x^-' 2
(p ~ 1)/2
1)
1
0(mod
p),
and
l)/2 solutions that are distinct
Furthermore, since
(p
a)^- 1)/2
= (-^-^(mod
p)
p is of the form 4fc + 1 and (p, a) = 1, both a and p a satisfy the

same congruence, but if p is of the form 4k 1, a satisfies one of the
congruences while p a satisfies the other.
Corollary 2. If p is a prime greater than 2 and d divides p 1, the
congruence x d 1 = 0(mod p) has exactly d solutions that are incongruent modulo p.
if
EXERCISES
1.
Write the solution of Sx
modulo 35.
8
2. Find the solutions of x +
x* + 1 = 0(mod 19) and x - 1
20 (mod 35), and reduce
= 0(mod 17) and

= 0(mod 19).
x8
it
to a least positive residue
= 0(mod
17),
and
also of
How many solutions has the congruence x 3 1 = 0(mod 13)? Find them.
Prove that the congruence x 2 + 1 = 0(mod p) in which p is a prime of the form
4n + 1 has two distinct solutions modulo p.
3.
4.
91
Corresponding to any conditional congruence f(x) =

an 0(mod p), where p is a prime, a ^
a x + ai.x
p
>
0(mod p),n
p, saidf(x) is not a multiple of x x modulo p, there is a
and
in
degree
has the same distinct
than
lower
that
is
congruence
p
congruence.
original
the
as
solutions modulo p
*Theorem
6-4.
n- 1
Using Fermat's theorem, we showed that
xp
has
all
0(mod
distinct solutions permitted
by x p
p)
by the modulus.
Moreover,
we have a remainder whose degree
if
we
than
the
remainder
Q(x)
and
R(x)
are
both
the
quotient
that
we
observe
p,
integral polynomials and that
divide f(x)
x until
f(x)
is
(x p
x)Q(x)
R(x)(mod
an identical congruence. Consequently, every

= 0(mod p) must satisfy
is less
p)
distinct solution of
}{x)
R(x)
and conversely.
Example. 2x 4
0(mod
p)
x 2 l(mod 3), and therex 2 1 = 2x(x z x)

2
1 = 0(mod 3) has the same distinct solutions
fore the congruence x
x = l(mod 3) and x = 2 (mod 3) as the congruence 2x A x 2 1 =
0(mod 3). Although these integers are simple solutions of the second
congruence, they are double solutions of 2# 4 x 2 1 = 0(mod 3).
Notice that if f(x) = 0(mod p) has p distinct solutions modulo p,
R(x) == 0(mod p) is an identical congruence, for it is lower than p in
degree and yet has p incongruent solutions modulo p. Hence, f(x) is a
multiple of x p x modulo p.
Conversely, if f(x) is a multiple of x p
x modulo p, f(x) = 0(mod p) has p distinct solutions modulo p. Thus
we have:
Corollary.
The congruence f(x) = 0(mod p) with p a prime has p
distinct solutions modulo p if and only if fix) is a multiple of x p x
modulo p.
*Theorem
a n = 0(mod
gruence b Q x n
The
6-5.
p) with a
n~ l
+ bix
congruence
0(mod
a xn
and p a prime
p)
f(x)
bn
0(mod
is
p) in
a\x n
~x
equivalent to a con-
which
l(mod
p).
Let x satisfy the congruence

a x
and multiply
x Q a Qx n
+ Xodtf"- +
1
Obviously, x f(x)
(x Q , p)
by
f(x)
1.
s=
l(mod
p)
Xo so that
0(mod
x an
p)
is
xn
bix n
equivalent to fix)
6 n (mod p)
0(mod
p),
for
92
Theorem 6-6. The congruence /(V) = a Q x + a-^x " +

0(mod p) with p a prime, a f 0(mod p), and n < p has
rl
71
'
"
exactly n
modulo p if and only if the remainder R{x) obtained

upon dividing x p x by f(x) modulo p is identically congruent to zero
for the modulus p.
Dividing x p x by f(x) modulo p, suppose that we find the identical
distinct solutions
congruence
xp
f(x)Q(x)
R(x)(mod
p)
which R(x) is an integral polynomial of degree less than n. But then

every solution of f(x) = 0(mod p) must satisfy R(x) = 0(mod p). Consequently, if f(x) = 0(mod p) has n distinct solutions modulo p, R(x) is
identically congruent to zero modulo p.
If, conversely, R(x) is identically congruent to zero for the modulus p,
the congruence
in
xp
is
identical
f(x)Q(x) (mod p)
and from Theorem 6-3 we

modulo p.
= 0(mod
infer that f(x)
p) has
exactly n distinct solutions

6-3.
Theorem
Wilson's
Theorem
6-7.
If
Since the integers
is
a positive prime, (p
2,
1,
+1=
1)!
modulo p constitute the solutions of the congruence x v

-1
there are exactly p 1 linear factors modulo p of o^
x p-i
is
i)( x
an identical congruence.
It
2)
(z
p).
l(mod
by x
p),
Hence,
1.
l)(mod
therefore, satisfied
is,
0(mod
of a reduced residue system
p)
= 0(mod
p)
with the result that
-1 =
If
it is
follows that (p 1)!

evident that the theorem holds.
is
odd,
it
Corollary.
sum
1, 2,
Because
If
is
l)!(mod p)
1
= 0(mod
a positive prime and r
1,
p),
2,
of all possible products of distinct integers selected
the set
(_i)P-i(p
1)
~
xp l
modulo
1 is
1 is
p, the coefficients of like
powers
exception of the coefficient of x p

2,
except for
1+2+
of integers selected
l)(x
powers of x
nomials are in the same residue class modulo
when p >
sum
rata
if
p =
2,
2, the
time from
a multiple of p.
identically congruent to (#
cients of all the positive
and
p.
2)
in these
(#
two poly-
Therefore, the coeffi-
of x occurring in the
are congruent to
product with the
modulo
p.
But
~
the coefficient of x p 2 in this product is the
~
the coefficient of x p z is the sum of products
its sign,
two
at a time in all possible
ways without
repetition

from
1,
X P-r-i
2,
for r
2,
1,
Likewise, except for sign, the coefficient of
1.
.
is
2,
a time from the same
selected r at
93
sum
the
of products of integers
All these
set.
sums
are, therefore,
Equating the absolute terms would, of course, give
multiples of p.
Wilson's theorem.
Theorem
6-8.
If (n
Suppose that n
>
1 is
are factors of (n
Hence, n
<
0(mod
n),
then n
is
a prime.
a composite and that
(n
Then n = n\n 2 where
1)!
1)!
<
n2
rii,
n,
0(mod
and consequently both
But neither
1)!.
n)
can divide
of these integers
a prime.
Theoretically Wilson's theorem and
and n 2
ri\
1.
is
when n
prime, but
that the test
is
large, the
amount
converse afford a test for a
its
of calculating involved is so great
entirely impracticable.
is
EXERCISES
Use Wilson's theorem to show that 23 is a prime.
Show that, for p > 5, (p 1) + 1 has a prime factor different from the prime p.
3. If p is a prime of the form 4.n + 1, prove that (2n)! is a solution of the congruence x 2 = l(mod p).
4. If p is a prime of the form 4n 1, show that (2n 1)! is a solution of the congruence x 2 = l(mod p).
p_1 is
5. If p is a prime, prove that each of the coefficients of the expansion of (1 z)
one greater than some multiple of p.
k p are both complete residue systems for
6. If r\, r 2
r p and h\, k 2
r p k p form a complete residue system
the modulus p, a prime, can rjci, r 2 k 2
modulo p?
r p _i is any reduced residue system modulo p, a prime,
7. Show that if ri, r 2
1.
2.
p-1
1 rt
= l(mod
p).
i=l
r p _i is a reduced residue system modulo p, an odd prime, then p
any integral rational symmetric function of the n, where i = 1, 2,
p 1, whose degree is less than p 1.
9. Develop another proof of Wilson's theorem by making use of the solutions of
x 2 = l(mod p) and ax = l(mod p), where (a, p) = 1 and p is an odd prime.
Notice
that of the integers 1, 2, 3,
p 1 only 1 and p 1 satisfy the first congruence
and that when a is selected from the set 2, 3,
p 2, there is a solution of the
second congruence that is in this very set and is distinct from a.
8.
If
n, r 2
divides
If the p
where i = 1, 2,
r, are
nr
and if m = pi ni p 2 n2
we saj that
pr
if any exponent n > 1
ju(m) =
that m(^) = ( l) r if each w< = 1 and
that p(m) = 1 if each n = 0, which means that /*( + 1) = 1.
When it is
so defined, /x(m) is called the Mobius function.
6-4.
The Mobius
Function.
i}
distinct positive primes
94
*Theorem
Take (a,
The function n(m) is multiplicative.

n
ni
nr
1, and let a = pi p 2
Pr
6-9.
b = qi si q 2 *
powers
of
distinct
into
positive primes.
factorizations
are
the
where
q
=
=
where
where
or
S/,
.
.
r,
.
If any n iy
i
1, 2,
1, 2,
j
,
, t, is
But then /*(a&) = 0.
greater than 1, jn(a)/*(b) = 0.
r
If all rii and s,- are 1, then /*(a) = ( l) and /*(&) = (l 1)*, so that
r+f
r+t
=
(~l)
=
(i-l)
But
M (a)/*(6)
If either a = 1, or a = 1 and & = 1, then ju(a&) = p(a)n(b).
b)
St
mW
*Theorem
6-10.
or
> ii{d) is
according as
greater than or
\m\ is
d\m
equal to
If
1.
m-
1,
J mM
m(1)
1.
rf|i
m=
= for any divisor d of m that

has a factor p* where i = 1, 2,
r, we need consider only the
divisors of pip 2
But these divisors are found by combining
prfactors chosen in all possible ways from these r primes.
Thus we find
If
+pi ni p 2 W2
n
Pr % since p(d)
'
'
,Ci(
Co -M(l)
-1)
-2>
,c 2 (--I) 2
But r Co-KCi(-l)
hence p(di)
n{d 2 )
'
viViVi)
rC r (-l) r
C 2 (-l)
"(Pi)
H{pip 2
++ n(d
'
'
Pr)
+rCr(-l) r = (1-
'
s)
where
0,
di,
d2
1)'
.
0,
and
d s are the
positive divisors of m.
*Theorem
6-11.
If
is
a positive integer,
fi(n)
1.
n=l
The preceding theorem shows that
mWi) + /ift) +
di|l
But
will
ju(l) will
d 2 |2
+ J
dm|m
"W) =
be a divisor of each integer of the set
occur
1,
2,
m, so that
m times in the above sum; 2 will be a divisor of

[1YI
of the
times.
Generally, d will be a divisor

of
of the integers
-r
-r
times in the sum.
from
through m, and therefore
95
fi(d) will
occur
Hence,
n=
d\n
and
m
=l
^Theorem 6-12 (The Mobius Inversion Formula). If, for m

f(m) is an arithmetic function and if an arithmetic function g(m)
>
0,
is
so
defined that
g(m)
f(d)
d\m
then
f{m)
^ "to (f)
(4)
We shall build up the right-hand member of the equation (4) to be

proved by first observing that as d ranges through all the positive divisors*
According to the hypothesis, it
of m, the integer m/d does likewise.
follows that for each positive divisor m/d of m
a\m/d
and
Kd)g
(j)
mW
/(a)
a|m/d
Hence,
2^
d\m
)?
fe)
M(rf)
d\m
XX
d\m a\m/d
/(a)
a\m/d
M(rf)/(a)
96
Reversing the order of summation, we have
d\m
a\m/d d\m
But to say that a ranges through the positive divisors of m/d while d takes
on the values of the positive divisors of m is the same as saying that a
ranges through the positive divisors of m while d is a positive divisor of
m/a. Consequently,
d\m
a\m
=
But
y(d)
>
unless a
d\m/a
/(a)
m(0
d\m/a
m, so that
d\m
An
6-5.
shown that
if
m>
m =
0,
>
We
Mobius Inversion Formula.
Application of the
4>{d).
have
Consequently,
d\m
</>(ra)
5>3
d\m
d\m
Thus
if
0(m)
ra
7?i
/?l
=
I
(
1,
then
<j>(m)
1,
but
Pi
if
m=
h
H
Pr
ni
pi p 2
"
"
n2
pr
nr
,
then
+
Pr-l??r
PlZ>2
P1P2P3
+(-1)'
PlP2
'
'
)
Pr)
and
^..(i-lî-i)..
.(1-1)
EXERCISES
1.
Is it necessary for the truth of
functions?
Theorem 6-12 that
/(ra)
and g(m) be arithmetic

2.
If F(ri)
f] f{d) and no f(d)
is 0,
then/(n)
Prove
this state-
d\n
din
ment by using a method analogous

logarithm of each
d
[] F(n/d)^ K
97
member
to that of Sec. 6-5,
and then prove
it
by taking the
of the given equation.
m
3.
Prove that
for
any positive
integer m,
m
of
V
n=
(- - T-l)
l
together with
v(n)/n
>
Theorem 6-11.1
<
1.
Consider the value
CHAPTER
The
7-1.
X Function.
We
have proved that
for
any integer a prime
to
a 4>(m)
In solving congruences, moreover,

a positive power, smaller than
l(
mo d m)
we have
exhibited
some
cases in which
a particular integer a
<f>(m), of
is
sufficient
modulo m even when a ^ 1. Take, for example, the congruence x 2 ss l(mod 5), which has the solution 4, as well as 1, modulo 5,
and the congruence x z = l(mod 14), with solutions 9, 11, and 1 modulo
to produce
It
14.
is,
therefore, important to ask
whether there
is
a positive integer
and yet such that when any integer prime to m is raised

We shall
to that integral power, the result is congruent to 1 modulo ra.
show that when m takes any form except 2 n for n = 0, 1, or 2, andp n or
2p n for p an odd prime and n > 0, there is such a positive integer. For
example, consider the case where m = 15. Here <f>(m) = 8, and x 8 =
l(mod 15) is, of course, satisfied by all integers prime to 15. But each
one of these integers also satisfies the congruence x A = l(mod 15).
Now take any integer a prime to 2. Then a is odd and has the form
smaller than
2k
-r-
1.
<f>(m)
Hence,
(2k
l)
4fc(fc
1)
or
a2
= l(mod2
3
)
But
(
a 2)2
(l
a 22
a 22
l(mod
2 4)
2t
l(mod
2k)
2 3 s) 2
or
2 4s
+ 2V
so that
In like manner,
if
a
it
follows that
or
a2
"' 1
m l(mod2 A:+1
Hence, we conclude that
>
if
2n ' 2
99
2,
l(mod 2 n )
which means that

*<2)/2
>
Consequently, for n
power
this integral
modulus 2 n
of
an integer smaller than <j>(2 n ) such that

any integer prime to 2 yields the residue 1 for the
2 there
is
we proceed
Accordingly,
= l(mod2 w
name
to give a
to this
number
n
0(2 )/2 as well as to other numbers closely related to <f>(m).
It was R. Carmichael who used the symbol X(ra) to designate the arith-
metic function which E. Lucas had defined as follows:

1.
2.
3.
4.
= 2 n and n = 0, 1, or 2, X(ra) = <f>(m).

= 2 n and n > 2, X(ra) = 0(ra)/2.
= p n p being an odd prime, X(p n ) = <(p n ).
nr the
= 2 npiWl2>2n2
Pr
p where i = 1,
odd primes, then X(m) is the least common
m
If m
li m
If m
If
distinct
X(pi0,
X(Pr
'
4-,
2,
r,
being
multiple of X(2 n ),
n ').
m has the form 2 n for n =
n
or 2p n for
0, 1, or 2, the form p
and p an odd prime, the X function has the same value as the
function, but when m has the factor 2 n with n > 2, or 2 2 and an odd
prime factor, or two factors that are powers of distinct odd primes, the
X function is at most half of the
function.
n
n2
=
When the p^ where i
r, are odd primes, if m = 2 pi ni p 2
1, 2,
nr
=
know,
therefore,
that
1, we
p r and (a, m)
Thus, when
>
<f>
a X(2)
ni
au P
i
But X(m)
X(ra)/X(2
=
=
i( mo d2)
l(mod pi ni )
a multiple of each of the functions X(2 n ), \(pi ni ), and thus

and \(m) /\(pi ni ) are integers. Consequently,
is
n
)
a xC2))x<)/x(2.)
i(
mo d2
a X(p^))X(m)/X( Pi ")
l(
mo d p
TC
and
(
Finally, since 2
n
,
pi ni
pr
nr
a x(m)
..)
are relatively prime in pairs,
mo d m)
for n = 0,
i(
Therefore, if m is not of the form 2

1, or 2, and not of the form
n
n
and p an odd prime, then the X function gives a better
p or 2p f or n >
result than does the
function.
It is on this account often advantageous
to have:
n
cf>
For (a, m) = 1, a Mm) = l(mod m).

Example. Although 0(2800) = 0(2 4 )0(5 2 )0(7) = 960, X(2800) is the
least common multiple of 4, 20, and 6 and is only 60.
Hence, for (a, 2800)
Theorem
1,
a 60
7-1.
(mod 2800).
100
The new formula
is useful in finding the remainder in the division by

power of an integer prime to m whenever the exponent of the
power lies between X(ra) and <f>(m). Thus, to find the remainder when
7 14 is divided by 180, we start with 7 12 = l(mod 180), for X(180) = 12.
Then 7 2 = 49(mod 180), and 7 14 = 49 (mod 180).
7-2.
Belonging to an Exponent Modulo m. Since there is always at
least one positive power of a particular integer a prime to m that is congruent to 1 modulo m, there must be a least positive integer d such that
ad = l(mod ra). We call this integer d the exponent to which a belongs
modulo m, or we say that a belongs to d modulo m. For example, 4 3 =
l(mod 7), while 4 2 = 2 (mod 7), so that 4 belongs to 3 modulo 7. Again,
3 4 = l(mod 5), and there is no smaller positive power of 3 which is congruent to 1 modulo 5.
Hence, 3 belongs to </>(5) modulo 5.
Theorem 7-2. If an integer a, prime to m, belongs to d modulo m, and
k
= l(mod m), then d divides k.
if a
< r < d, then
Suppose that k = qd + r, where
of the
that
ak
(a d ) 3
ar
l(mod m)
a r (mod m)
is
But d
is
the least positive exponent such that a d
l(mod
ra),
and there-
Consequently, d k.
Corollary.
The exponent d to which an integer a, prime to m, belongs
is a divisor of 4>(m) and of X(m).
modulo
This corollary shows that we need try only divisors of X(m) to find the
exponent to which an integer belongs modulo m. For instance, to find
the exponent to which 7 belongs modulo 55, we try only the exponents
fore r
0.
2,
4,
5,
36(mod
10,
and
55), 7
20,
for X(55)
32(mod
55), 7
Thus 7 2 = -6(mod 55), 7 4 =

34(mod 55), and 7 20 = l(mod 55).
20.
10
Hence, 7 belongs to X(55) modulo 55.

In 1844 A. L. Crelle gave a device for finding the exponent to which an
To employ this method, first set up the
integer a belongs modulo m.
m 1 in a row, and under 1 put r h the least positive
integers 1,2,
residue modulo m of the integer a under 2 put the least positive residue
Ti of n + a = 2ri(mod m); under 3 put the least positive residue of
Then the resulting table
r 2 + a = r<t + ri(mod m)
.
r2
r3
r4
gives in order in the second
...
...
m-
row the residues
r h 2ri, 3fi,
(m
r m -i
of
l)ri
101
modulo m. According to this scheme the integer r\ congruent to a

modulo m is under 1 the integer congruent to a 2 = r x ri(mod m) is
under t\\ and likewise the integer congruent to a 8 = t ri(mod m) is
under t.
Thus,
m=
if
modulo
7,
7 and
we form the
we wish
12
14
Then
3 (mod 7),
to find the exponent to which 3 belongs
table
and so we move to 3
in the first row.
We
find 2
under 3, and hence 3 2 = 2 (mod 7). Moving to 2 in the first row, we find
Continuing in this manner, we find
6 under it, and have 3 3 = 6 (mod 7).
35
modulo
7.
34
6
4,
powers
5,
and
36
l(mod
The
7).
integer
Moreover, the residues of the
of 3 are in order 3, 2, 6, 4, 5,
repeated as the exponent of 3
1,
and
3,
first
it is
therefore, belongs to
six positive integral
evident that this cycle
is
increased beyond 6 through the positive
is
integers.
an integer a prime to m belongs to d modulo m, the least positive

a d of a taken in order constitute the
residues of the powers a, a 2 a 3
These residues are distinct modulo m, for if
period of a modulo m.
~ =
l(mod m). But since s and t
a s = a* (mod m) with s > t, then a s
d, the difference s t is less than d and
have only the values 1,2,
cannot be divisible by d. Hence, we have proved
Theorem 7-3. If an integer a belongs to d modulo m, the integers a,
a d are distinct modulo m.
a2 a3
When (a, m) = 1, s > 0, and a s = b (mod m), then b is said to be a
power residue of a modulo m. Thus Theorem 7-3 shows that if a belongs
to d modulo m, the first d power residues of a modulo m are incongruent
modulo m. Clearly, they constitute the power residues of a modulo m.
We say, then, that there are exactly d incongruent integers that are power
a
residues of a modulo m, omitting modulo m" after "incongruent integers/'
for there can be no confusion about the required modulus.
If
EXERCISES
Find the exponents to which 5 and 7 belong modulo 11.
Find the remainder when 7 182 is divided by 675.
3. Prove that if a and b are prime to 1729, then a 36 6 36 is divisible by 1729.
4. If p is a prime, (a, p) = 1, and a d = 6 d (mod p), where d is the least positive
exponent for which the congruence is true, then d is a divisor of p 1.
5. If a belongs to d modulo p an odd prime, and if d is even, then a d/2 = 1 (mod p).
6. If p is an odd prime, and if a k = l(mod p), where k is the least positive
integer for which the congruence is true, then 2k is the exponent to which a belongs
1.
2.
modulo
p.
102
a, where
< a < p 2, belongs to the exponent 3
modulo p.
belongs
to
6
p,
8. If the integer a belongs to d modulo p, a prime, show that the product of all the
distinct residues of the powers of a is congruent to 1 or 1 according as d is odd or even.
K(m) ~ (mod m) is a solution of ax = b (mod m) if
9. Show that x = ba
(a, m) = 1.
7.
Prove that
modulo
the integer
if
a prime, then a
Compare the fact that the powers of 2 will generate all the solutions of x 4
= l(mod 5) with the corresponding property of the root i of the equation x 4 1 = 0.
What do you notice about the other solutions?
10.
Theorem
7-4.
If
an integer a prime to p belongs to d modulo
prime,
there are exactly
modulo
p.
When p =
p,
incongruent integers that belong to d
<f>(d)
2, d has only the value 1 and then the

modulo 2.
When p is an odd prime, consider the congruence
xd
l(mod
integer
belongs to
p)
Then each integer
and suppose that the integer a belongs to d modulo p.

of the set
a,
is
prime to p and
satisfies
a2 az
,
ad
the congruence, for
(a 8 )*
(a d ) s
l(mod
p)
According to Theorem 7-3, furthermore, no two of these integers are conConsequently, these powers of a give all the solutions
p.
of x d = 1 (mod p) for the modulus is a prime and there are, therefore, no
more than d solutions. Hence, any integer that belongs to d is congruent
modulo p to an integer of the set a, a 2 ... a d and none of these integers
belongs to an exponent greater than d.
However, if (s, d) = 1, a s belongs to d modulo p, for if we assume that
a 8 belongs to k < d,
gruent modulo
(a 8 ) k ss
l(mod
p)
== l( mo(l
p)
and
a sk
But
since a belongs to d
than
On
modulo
p,
the other hand,
if (s,
= n ^
d)
and a 8 does not belong to d modulo

Since there are
of the
d modulo
p.
so that s
1,
(a 8 ) do ss a Sod 3=
<j>(d)
Hence, d
sk.
k,
and k
is
not
less
d.
l(mod
d
p, for
<f>{d)
,
1, 2,
is less
ns
= nd
then
p)
positive integers less than d
powers a 8 where
d,
than
d.
and prime to
d,
exactly
belong to the exponent
103
d is a divisor of p 1 and an integer a belongs to d

modulo p, a prime, the solutions of x d = l(mod p) are congruent to
2
a, a
... a d modulo p.
Corollary
If
1.
d is a divisor of p 1 and an integer a belongs to d

modulo p, a prime, then a s belongs to d modulo p if and only if the exponent s is prime to d.
If d is a divisor of p 1 and an integer a belongs to d
Corollary 3.
modulo p, a prime, then a s where (s, d) = n ^ 1, belongs to the exponent
Corollary
If
2.
d/n modulo p.
To prove Corollary 3, notice that if p = 2, d can be only
In the case of an odd prime, if d = d n, we saw that
Now
(a s ) d ==
l(mod
p)
l(mod
p)
<
with
<
Then d
do.
follows that do
Theorem
7-5.
This
t.
and a s belongs to
do
if s
impossible
modulo
When
But
st.
is
(s,
then
s n,
when
<
(s
is
odd.
(ab)
if
1, if
(a') '(&')'
modulo m,
a6 belongs to k
st
and k
<
ks, so
kt,
so that s
it
do,
a belongs to
and
s,
belongs to
so that
l(mod m)
l(mod m)
k.
kt
Since
b ks
l(mod m)
In like manner,
that t\k.
(ab)
Moreover,
st.
(ab) ks
and therefore
and
1,
so that
(ab) k
then k
do)
p.
t)
st
Therefore,
do.
modulo m, then ab belongs to st modulo m.

We know that a s = l(mod m) and = l(mod m),
and
for a
suppose that
(a s Y
But
1,
(s,
kt
l(mod m)
1, it
follows that
st
and
<
fc.
Consequently, k = st.
Theorem 7-6 (H. G. Erlerus, 1841). When pi and p 2 are odd primes,
if m = ai(mod pi) and m = a 2 (mod p 2 ), and if in addition ai belongs to
di
modulo
common
Since
pi
and a 2 belongs
to d 2
modulo p 2 then
,
multiple of d\ and d 2 for the modulus pip 2
m =
dl
l(mod
pi)
m<* 2
l(mod p 2),
(m dl) L/di
l(mod
(m d *) L/d2
l(mod p 2)
and
multiple of di and d 2 then

,
pi)
and
if
belongs to the least
is
the least
common
104
Therefore,
mL =
But
if
m belongs to
modulo
l(mod
P1P2)
P1P2, the last
congruence implies that k
L.
Again, from
m =
l(mod
P1P2)
m =
l(mod
pi)
we
infer that
k
1,
k, d 2
k, and L
k.
Hence, k = L.
Examples. Because 3 belongs to 5 modulo 11 and 10 belongs to 2 for
the same modulus, and since the exponents 5 and 2 are relatively prime,
it is evident from Theorem 7-5 that 30, and hence 8, belongs to the exponent 10 modulo 11.
Again, 7 belongs to 10 for the modulus 11, and 5 belongs to 4 modulo 13.
Thus, according to Theorem 7-6, the integer 18 belongs to 20 modulo 143.
Therefore, di
EXERCISES
1. Find an integer that belongs to 2 modulo 19 and one that belongs to 3 modulo 19.
Using these results, find an integer that belongs to 6 modulo 19.
2. Set up the least positive residues of the powers of 2, 3, and 6 modulo 17.
3. Find all the integers that belong to 16 modulo 17.
10
= l(mod 31).
4. Solve the congruences x 3 = l(mod 7) and x
5. Show that 2 belongs to 12 modulo 13, and thus find the exponent to which 8
belongs modulo 13. Do any other integers belong to this exponent modulo 13?
Then determine
6. Find the integer to which 7 belongs modulo 5 and modulo 11.
the integer to which 7 belongs modulo 55.
n
7. When p is a prime, if a and b are prime to p, and if a = b (mod p ), show that
pr
pr
n+r
n+1
p
=
=
induction
and
hence
by
that
a
6
(mod
(mod
6
a?
).
),
p
p
8. When the modulus m is composite, prove theorems analogous to Corollaries 2 and
3 of
Theorem
If pi
9.
7-4.
and p 2
is
an odd prime, does Theorem 7-6 hold?
Generalize the
theorem.
7-3.
Another Test for a Prime. If we can find one integer a prime to

m and satisfying the condition
the integer
am
is it
true that
6
6, and so a
6 l(mod
2
21
is
m is a prime?
~l
l(mod m)
This question
is
easily answered, for X (21)
l(mod 21) for all integers a that are prime to 21. Thus
2
Hence, 8 20 = l(mod21), and yet
21), and 8 = l(mod 21).
not a prime.
Is it
even true that
m is a
prime
a m-i
if,
i(
for all a that are
prime to m,
mo d m)
Again we can answer negatively by showing a case
in
which the hypothesis
but where m is a composite.
3(11) (17) satisfies the congruence
Any
is fulfilled
a 80
because X(561)
80.
105
integer a prime to 561
(mod 561)
Consequently,
a 56o
i(
mo d
561)
Thus the exact converse
of Fermat's theorem is not true.

important test for a prime can, nevertheless, be derived from these
The following theorem was first proved by Lucas* in 1876, and
ideas.
from it powerful methods for deciding the primality of integers having
certain forms have been developed
Theorem 7-7. If an integer a is prime to an integer m greater than 1,
~
and if a m l = l(mod m) although there is no positive integer k less than
m 1 and such that a k = l(mod m), then m is a prime.
An
Assuming that
But 0(m)
m is
< m
m>
1
1 is
composite, since
because
(a,
m)
m has a factor between
1,
and m.
Therefore,
a prime.
Actually, by virtue of Theorem 7-2, in applying this test for a prime we

need evaluate only those powers a k for which k is a divisor of m 1.
Example. To test 47 for a prime, we can make use of the fact that 2
belongs to 23 and 46 belongs to 2 modulo 47 and determine by Theorem
7-5 that 2 46 = 45 (mod 47) belongs to 23 2 = 46 for the modulus 47.
Or we may reason that since the divisors of 46 are only 1, 2, 23, and 46,
and because 45 23 = (47 - 2) 23 = (-2) 23 = -l(mod 47), so that 45 does
not belong to 1, 2, or 23, then the integer 45 belongs to 46 modulo 47.
Hence, by virtue of Theorem 7-7, 47 is a prime.
7-4. Primitive Roots.
Is there a positive integer k smaller than X(m)
and satisfying the condition that, for any integer a prime to m, a k =
l(mod m)? The answer has been completely determined because Gauss
showed that exactly <(</>(m)) integers belong to </>(m) if m is 2 n with
n = 0, 1, or 2, or if m is p n or 2p n where n > and p is an odd prime, and
R. Carmichael showed that for all other moduli there is at least one integer
that belongs to X(m).
Consequently, there is no positive integer k
smaller than X(m) and such that, for all integers prime to m, a k =
l(mod m). We shall proceed to develop these ideas.
We call an integer that belongs to 4>{m) modulo m a primitive root of m
or a primitive root modulo m.
It is evident that 1 is a primitive root of 1
and 2 and that 3 is a primitive root of 2 2
There are no other primitive
roots of these moduli.

*
E. Lucas,
Am.
J. Math., Vol.
1,
p. 301, 1878.
106
There are exactly 0(p 1) incongruent integers that

modulo p, an odd prime.
Of course when d is a divisor of p 1, as far as we have proved, no
integer need belong to d modulo p, but we have shown that if one integer
does belong to d for the modulus p, exactly <f>(d) incongruent integers
Theorem
7-8.
are primitive roots
modulo p belong
integers from
to
Therefore,
d.
through p
if
we
represent the
let \p{d)
number
of
that belong to d modulo p,
m)
<
0(d)
Besides, each integer from 1 through p that is prime to p belongs to one

and only one exponent for the modulus p, and this exponent must be a
divisor of
It is
1.
0(<
where the
with
di,
But we have
1.
4>(di)
thus evident that
0(d 2 )
1, 2,
also
+
.
r,
Hdr)
= V -
are the distinct positive divisors of
shown that
<f>(d 2 )
= p -
<f>(d r )
Because no yp{d ) can exceed the corresponding 0(d/), if any \p(di) were less
than the corresponding 0((&), these statements could not both be true.
z
Therefore, for
all
i,
f(di)
In particular
cf)(p
1)
0(cfc)
the divisor d is p 1 itself, there are exactly

incongruent integers that belong top 1 modulo p.
when
4>(cf>(p))
p is an odd prime and d is a divisor of p 1, 0(d) of the

through p 1 belong to d modulo p.
Example. We shall find the primitive roots of 17 and the integers
Using Crelle's method, we see
that belong to the divisors of 0(17) = 16.
that for the modulus 17 the powers of 3 give the residues
Corollary.
If
integers from
3, 9, 10, 13, 5, 15, 11, 16, 14, 8, 7, 4, 12, 2, 6, 1
It is clear, then, that the primitive roots of 17
of 3
having exponents prime to 16 are
reduce to
3,
9(mod
33
determined by the powers

35
37
39
3 11
3 13
3 15
modulus
belongs to the exponent 8 modulo 17,
10, 5, 11, 14, 7, 12,
Notice that 3 2
and
3,
17)
respectively, for the
6,
and
17.
for
Moreover, the other integers that belong to 8

(2, 16) = 2 and ^- = 8.
modulo 17 are of the form 3 s where (s, 16) = 2. Hence, s = 6, 10, and
14.
The integers are, therefore, 3 6 3 10 and 3 14 and they reduce to 15,
In like manner we can find the integers that belong
8, and 2 modulo 17.
Thus 3 4 and 3 12 belong to
to 4 modulo 17, for in this case (s, 16) = 4.
There is just one integer that belongs to 2, and it is
4 modulo 17.
3 8 = 16(mod 17).
,
107
EXERCISES
Find all primitive roots of p = 5, 7, 11, and 13.
Prove that the product of all the integers that belong to a particular exponent
d > 2 modulo p, an odd prime, is congruent to 1 modulo p.
341
2 is divisible by 341 even though 341 is not a prime.
3. Show that 2
4. Find by trial composites m other than 561 such that a m_1 = l(mod m) for all
integers prime to m.
1.
2.
7-5. Gauss'
Method
Root Modulo
for Finding a Primitive
p.
To
find
a primitive root of a prime p, first choose any integer a\ prime to p, and

If d\ is not p 1,
find the exponent di to which a\ belongs modulo p.
find the least positive residues of
oi, i
None
modulo
p.
satisfies
the congruence
select
z
,
a\ dx
of these integers is a primitive root of
x dl
Now
ai
l(mod
any positive integer a 2
less
p since each one
p)
than p and not one of the residues
powers of a. Then if a 2 is not a primitive root of p, a 2 belongs to

some d 2 modulo p.
The exponent d 2 cannot be a divisor of d\, for if d\ = kd 2
of the
hd %
l(
mo d
p)
a 2 (mod p) would be a solution of x dl = l(mod p), which is

impossible because the powers of a\ determine all the solutions of this
and
a:
congruence.
If d 2 is a multiple of d h but not p 1, we have found an integer that
belongs to an exponent modulo p that is greater than d\.
d 2 ) = I, then a x a 2 belongs to d\d 2 modulo p.

lib = (di, d 2 ) and b is neither 1 nor di, factor b into powers of distinct
nr
Then separate b into two relaprimes so that b = pi ni p 2 n
pr
tively prime factors bi and b 2 by taking pi ni as a factor of b i or b 2 according
If the power of Pi is the same
as di or d 2 contains the lower power of p^
in both di and d 2 take pi ni as a factor of either bi or b 2
If (di,
Then according
Theorem
belongs to di/bi, and

However, ((di/bi), (d 2 /b 2 )) = 1, for if
to Corollary 3 of
a 2h2 belongs to d 2 /b 2 modulo p.

di = bdoi and d 2 = bd 02 then (d 0h d 02 ) =
and d 2 /b 2 = bid 02 and we know that (b 1} b 2 )
,
for
if
p that
(b 2
1.
Moreover, di/bi
1.
But
(6i,
i)
= bd
= 1 also,
2
i,
the highest power of any prime in b h it is the highest power of

In like manner,
a factor of d\ and thus p does not divide d i.
= 1. Consequently, the product ai bl a 2bi belongs to did 2 /bib 2 =
is
is
d 02 )
7-4, ai bl
modulo
d
and
d\
2
d\d 2 /b,
p,
and
this
exponent
is
the least
common
multiple of
108
Thus under any circumstances we have found an integer that belongs

After setting up the power
to an exponent modulo p larger than d\.
residues of this integer, we continue as before by choosing a positive
integer less than p but not among any of the power residues thus far
determined. After proceeding for a finite number of steps, we must find
an integer that belongs to p 1 modulo p.
Example. To find a primitive root of 41, first find the power residues
These residues of 2, 2 2
2 10 are in order 2, 4, 8,
of 2 modulo 41.
10
=
Since
2
l(mod
the remaining residues
40.
10,
20,
41),
16, 32, 23, 5,
are the negatives of these integers, and 2, therefore, belongs to 20 modulo
.
41.
Because the integer 3 does not occur among the power residues of 2, we
determine the power residues of 3 modulo 41. They are 3, 9, 27, 40, 3,
9, 27, 40. Consequently, 3 belongs to 8 modulo 41.
We know
Since (20, 8) = 4, we use the factors 4 and 1 as &i and b 2
Consequently,
that 2 4 belongs to 5 and 3 belongs to 8 modulo 41.
.
7 (mod 41) belongs to the least
16
is,
therefore, a primitive root of 41.
The power
residues of 7
7,
common
multiple of 5 and 8 and
modulo 41 are
8, 15, 23, 38, 20, 17, 37, 13,
35,
40
22, 31, 12,
2, 14, 16, 30,
5,
of the above integers taken in order.

Roots Modulo p n and 2p n
Theorem 7-9. If p is an odd prime, there are just
and the negatives

7-6. Primitive
gruent primitive roots
modulo p n
n
<f>(<j>(p
))
incon-
Suppose that the integer a is a primitive root modulo p. This statement means, of course, that every integer in the class with a belongs to
We shall show that we can choose
the exponent p 1 for the modulus p.
Specifically,
primitive
root of p n
is
a
class
that
residue
this
in
an integer
.
if
a p-i
we
shall
prove that a
is
shall
mo d
2
)
a primitive root modulo p n but

,
aP- 1
we
= i(
l(mod
p'2
if
prove that k can be so chosen that the integer a
kp
satisfies
the condition
(a
kpY~
l(mod p 2 )
this integer is a primitive root of
and that
First, let
l(modp
).
us suppose that
Then
when a belongs
n
.
to
modulo
p,
av
(a
kp) p
~l
=
=
and consequently
is divisible by p.
a?- 1
ap
(a
+~(p 2
kp)
there
l(mod p
r*
= p n~
l(mod
divides
(fcp^-^mod p 2 )
modulo p 2
if
and only
if
/cp will
ftp)*'-
l(mod p 2 )
kp of p such that
p_1
).
an exponent
4>(p
congruent to
is
It is true that since
to
always a primitive root
is
modulo p but
(a
is,
kp
we, therefore, choose k prime to p, then a
If
not only belong to p
that
~2
(mod p
/cp
p~
l)a p
109
But
congruence
if s
r*
n)
But does
1,
More than
p
1).
that, since
it is
so,
is
belong
a divisor of
necessarily true that
that p 1
has the form p s (p 1) with s = 0, 1, 2,
or
did not exceed n 2, we could raise each member of the
l(mod p n ) to the p n ~ 2 ~ s power and obtain
and
Hence,
if.
1.
r<^ = l(mod p n ).
smaller than 4>(p n )?
If
p)
modulo p
(p
p)
(r,
belongs to
modulo
p, it is clear
pn-2(p-l)
l( m()( J pn)
We shall show, however, that the last congruence cannot be true and
hence that r cannot belong to an exponent less than </>(p n ) modulo p n
To do so, use the fact that
.
r v-i == i(
in the
(r p
-i
form
rp
(1
n -*
~l
cp,
where
mo d
p)
(c,
<p)
+ cp) pn_2 (niod p")

n-2 _ ^
ii
+ p^Cp + ^ ^_
1.
Then
cp )2
+
+
The (m
l)st
term
p
n-2(
of this
n-2
expansion
is
j)
n-2
- m+
1/
m!
obviously a multiple of p n
is
it
n-2
1)
(m
1,
that
is,
(yn-2
^T-=
_m+
j)
if
m>
is
an
integer.
;,n_2
(mod p n )
1)
and
(cp)
and
p
if
(m, p)
But n
m =
fcp'
1,
for
with
1):
p ^' >
>
>
kp
n.
But
fcp*
t + 2.
Each
term
p
after the second of the expansion is, therefore, a multiple of p n
Hence, f or n > 3
(k,
n
p)
the expression will not be a multiple of p" unless p-'2 -'

unless n
-{-
r p-2(p-l)
_j_
pn-l c
mo(J p n)
110
and because
(c,
p)
1,
rp-HP -i) = i(
n =
If
2,
we know
mo d
n
)
that
rp
l(mod p 2 )
i(
whereas
r p( P
Consequently,
Now
-i)
r is a primitive root
2
)
>
modulo p n where n
,
1.
consider the congruence

3>-i(p-i)
Each
mo d
l(mod p n )
==
of the integers
y
yP n_1 (p 1)
y2
congruence, and each of them

incongruent
modulo p n for if
moreover,
satisfies this
prime to
is
They
p.
are,
assuming u
s
>
r v(
mo d
pn)
w, u
1, 2,
that r w_ "
we conclude
n~ 1
(p
1)
l(mod p n ).
This result is
than p
Hence, these </>(p n )
(p 1).
integers are distinct solutions modulo p n of the conditional congruence.
But only integers prime to p can be solutions of this congruence, and since
there are exactly <t>{p n ) integers from 1 through p n that are prime to p, we
have found all the solutions.
Furthermore, the integers that belong to 4>{p n ) modulo p n must be
v,
impossible because u
n~ 1
v is less
among the solutions of the given congruence. By arguing just as we did

Theorem 7-4, we can show that r belongs to {p n ) modulo p n if and
s
in
only
if
s is
prime to p
n~ 1
(p
1).
There
are,
then, exactly
n
<f>(4>(p ))
incongruent integers that are primitive roots modulo p

Example. We have already shown that 3 is a primitive root of 17.
But 3 16 = 171 (mod 289). Accordingly, 3 is a primitive root of 17 2 since
.
3 16
l(mod
Furthermore, the powers of 3 having exponents

complete set of distinct primiThere are exactly 0(272) = 128 of them.
tive roots of 17 2
The theorem shows, moreover, that because 3 16 ^ l(mod 289), 3 is a
289).
prime to 0(17 2 )
(17) (16) constitute the
primitive root of 17 3
17 4
....
EXERCISES
Find all the primitive roots of 5 2 and 5 3

Find the primitive roots of 49. Find also all the integers that belong to the
exponent 6 modulo 49. Find a primitive root of 343.
3. If p is an odd prime, prove that any primitive root of p n is a primitive root of p.
1.
2.
4. Prove that
d modulo p, and
n ~h
if
if
is
is
111
a prime and an integer a prime to p belongs to the exponent

the highest power of p that divides a d \ then a belongs to the
,
modulo p n for n > h.

exponent dp
n
5. If p is an odd prime, prove that the product of two primitive roots modulo p is
not a primitive root modulo p n
n
n
n
n
8
6. If r belongs to <j>(p ) modulo p and (s, <t>(p )) = d ?* 1, does r belong to <f>(p )/d
.
modulo p n ?
Theorem
roots
7-10.
modulo 2p n
Any
There are exactly
cf>(<j>(2p ))
integer that satisfies the congruence
l(mod 2p n )
.<f>(2pn)
necessarily odd, and so a primitive root

~
odd integer a belongs to 4>{2p n ) = p n l (p
of
is
tive root of
n
,
if
2p n must be odd.
n
it
1) modulo 2p
,
pn-i( P -i)
i(
mo d
a p--i(p-D
i(
mo d p n
implies that
But
is
if
an
a primi-
for
and
incongruent primitive
a belonged to d
l(mod 2p n ).
The
n~ 1
n
then since a is odd,
(p 1) modulo p
congruence
is,
of
course,
impossible, for a is
last
for
the primitive roots
We must, therefore, look
<
a
a primitive root of 2p n
of 2p n among the primitive roots of p n
.
2p n )
an odd primitive root of p n it is a primitive root of 2p n for

(pn)
n
and if r belonged to an exponent

1 is divisible by 2 as well as by p
r+
n
n
it would belong to that exponent
smaller than 4>(2p ) modulo 2p
modulo p n
Moreover, if r is an even primitive root of p n the integer r + p n which
But
if r is
is
also a primitive root of
because
it is
in the
same residue
class as r
modulo p n is odd and is of necessity a primitive root of 2p n

We have shown, therefore, that a primitive root of p n is a primitive
root of 2p n if and only if it is an odd integer.
It is evident, furthermore, that two odd primitive roots of p n that are
in the same residue class modulo p n are congruent modulo 2p n
On the
other hand, if two of the <f>(<f>(p n )) odd primitive roots modulo p n selected
one from each of the residue classes modulo p n were congruent modulo
There are, therefore, exactly
2p n they would be congruent modulo p n
.
incongruent primitive roots of 2p n

Example. We have shown that 3 is a primitive root of 17 2
n
(f>(<l>(2p ))
It is also
a primitive root of 2(17 2 ) = 578, for it is odd.

In like manner 3 3 and 3 5 are primitive roots of 578. But 3 7 =
166(mod 289) and is, therefore, an even primitive root of 289. It is not a
primitive root of 578, but the odd number 166 + 289 = 455, which is in
the same class as 166 modulo 289, is a primitive root of 578.
Again, 3 9 =
112
a primitive root of 578, and although 3 11 = 152(mod 289)

is not a primitive root of 578, it yields 152
289 = 441, which is. There
are in all 0(272) = 128 incongruent primitive roots of 578.
49(mod 289)
is
EXERCISES
Find all the primitive roots of 50 and 250.
Find the primitive roots of 98. Determine the integers that belong to the
exponent 3 modulo 98. Find a primitive root of 686.
3. Prove that the congruence x^ 2 ?") = l(mod 2p n ) has exactly 4>{2p n ) solutions
modulo 2p n if p is an odd prime.
4. If r is a primitive root of 2p n p being an odd prime, show that r 8 belongs to
1.
2.
Prove, furthermore, that if (s,(j>(2p n )) = d, then

belongs to <p(2p n )/d modulo 2p n
5. Show that if p is an odd prime, the product of two primitive roots modulo 2p n is
not a primitive root of 2p n
<f>(2p
if
and only
if (s,<j>(2p ))
1.
rs
R. Carmichael called integers that belong to
7-7. Primitive X Roots.
X(m) for the modulus
When m =
2 n-2
modulo
2n
"primitive X roots of
>
where n
for
to be congruent to
if
m.*
the integer 3 always belongs to X(2 n )
2,
3 belonged to a smaller exponent, 3
modulo 2 n
That
2 "" 3
would have
not true is evident

from the following argument: According to the binomial expansion,
(22
I)*-
so that, for
=
n
>
3,
2n
~3
and hence
it is
in this case 3
2*- 4 (2"- 3
2n- 3
2n " 3
Because a primitive root
is
pk
*,
2n - 2
(mod
2")
3,
X(2 3 )
2 and
necessarily a primitive X root,
we have now
a primitive X root for every modulus of the form p n

On this basis R. Carmichael proved by induction that
is
mi =
where the p
2 -i(mod 2 n )
with p a prime.
there is always a primitive X root of m.
tive X root of
1)2 4
is
l(mod 2 n ). When n
2 modulo 2 3
obvious that 3 belongs to
shown that there
this relation
22
i}
ni
pi p2
with
n2
Pk-i
2,
1,
He showed
nk - 1
.
and
that
if
is
a primi-
r 2 is a primitive X root of
are distinct primes, then an
k,
integer r that satisfies both the congruences
si r 2 (mod
ri(mod mi)
and
is
a primitive X root of
suppose that
m =
belongs to
mip k nk
s for
R. Carmichael, Bull
")
To show
that this statement
the modulus m.
r s ss
*
pk
is
true,
Then
l(mod mi)
Am. Math. Soc,
Vol. 16, No. 2, pp. 232-238, 1909-1910.
113
and
rs
But
can be
is
the least
exactly A(m).
is
l(mod p k Uk )
modulo mi and
and \(pk nk ) divide
r belongs to A (mi)
Therefore, both A (mi)

s
m) =
to \{pk nk ) for the

s.
modulus p& n *.
This means that the smallest
common
multiple of A (mi) and \(pk nk ).

This integer
true, moreover, that r X(m) = l(mod m) since
It is
is a primitive A root modulo m.

Carmichael also showed that when r is a primitive A root of m, the
powers of r whose exponents are prime to A(m) give </>(A(m)) incongruent
primitive A roots of m, and the product of these A roots is congruent to
These powers of r do not necessarily yield all the primitive
I modulo m.
A roots of m, but the same powers of another primitive A root will either
repeat in some order the results obtained from r or give </>(A(m)) different
primitive A roots of m distinct from those generated by r.
Although the theory of numbers is a branch of mathematics that we
evaluate on the basis of the profundity of its truths and the variety and
simplicity of its methods rather than on its applicability to practical
problems, yet it is interesting to observe that in 1935 H. P. Lawther
showed how the theory of primitive roots and primitive A roots can be
applied to the problem of splicing telephone cables.*
Example. To find a primitive A root of 21, we first find a primitive
root of 7 and also of 3.
It can be easily verified that 3 belongs to 6
modulo 7 and that 2 belongs to 2 modulo 3. We then find the common
solution of the congruences
x = 3 (mod 7)
x = 2 (mod 3)
(r,
The
If
Therefore, r
1.
solution
is
17 (mod 21), and this integer
we now find the powers
we have a set of
17 5 = 5(mod 21).
0(6)
=2
of 17 that
is
a primitive A root of 21.
have exponents prime to
primitive A roots of 21.
They
A (21)
6,
are 17 and
There are but two incongruent primitive roots of 7, and they are 3 and
When we use the integer 5 with the only primitive root of 3 to form
the set of congruences
x = 5 (mod 7)
x = 2(mod 3)
5.
is x = 5 (mod 21).
But the set of powers, 5 and
two primitive A roots of 21 already found. This situation,
however, does not mean that there are no other primitive A roots of 21,
for the number 2 belongs to 6 modulo 21.
Moreover, 2 and 2 5 =
II (mod 21) form a new set of two primitive A roots of 21.
we
find that the solution
5 5 repeat the
,
H. P. Lawther,
Jr.,
Am. Math. Monthly,
Vol. 42, No. 2, pp. 81-91, 1935.
114
EXERCISES
Prove that
a primitive X root of m, the powers of r having exponents prime

m.
2. Prove that if X(m) > 2, the product of all the incongruent primitive X roots of
congruent to 1 modulo m.
1.
if r is
to X(ra) are primitive X roots of
is
7-8. Integers Belonging to a Divisor of X(2 n )
*Theorem
7-11.
If
>
and
not
d,
divides 2
2,
2n
Modulo
n~ 2
then there are
modulo
exactly d incongruent integers that belong to d
2 n but
,
if
2,
modulo 2\
We have shown that the integer 3 is a primitive X root of 2 n if n > 3.
On this basis we shall show that for the modulus 2 n any integer having
the form 8/c + 3 cannot belong to an exponent smaller than 2 n_2
If
there are just three incongruent integers that belong to d
>
3,
(8/c
3)
2 " -3
(3
8/c)
2 "- 3
2n ~ 3
2 "- - 1
3
2*- 3
(2 0
(3)
~
W 3
2- -1
3
3
3(2
/c)
(2
/c)
- (mod
3
2")
Hence,
(8k
But
since 3
2n " 3
l(mod 2 n ),
2 "- 3
3)
8/c
32
"~ 3
(mod
2")
3 belongs to 2 M
"2
modulo 2 n
if
>
3.
Consequently, in this case there are at least 2 n_2 integers belonging to

X(2 n ) modulo 2 n for, in the expression 8/c
3, k can have the values
n~ 3 2"- 3 thereby
8/c
1
and,
in
k
can
be
2
3,
1, 2, ...
0, 1, 2,
yielding
all
form that are
positive integers of the required
less
than 2\
If n = 3, it is obvious that 3 and 5 belong to X(2 3 ) = 2 modulo 2 3

But besides these two integers the integer 7 = 2 3 1 is a primitive
.
X root of 2 3
modulo 2 3
Consequently, three incongruent integers belong to X(2 3 )
Again, if n >
having the form
3,
8/c
we can prove that
1,
in
which k
2"- 3 (2 3 /c)
is
for the
prime to
modulus 2 n
2,
integers
all
belong to the exponent
2"~ 3 for
,
(8/c
On
(8/c
l)
2 "" 3
the other hand,
I)
2"
-4
=
=
if
l(mod 2
n >
(2
2 "- 3
/c)
2")
4,
2"- 4 (2 3 /c)
2"- 5 (2 4
3
2
1)(2 A0
+
=
(mod
(2
2n_4
A-)
(mod
2 n)
2"" /c(mod 2 n
)
1
Hence, if (/c, 2) = 1 and n > 4, the integers 8k 1 belong to the expoMoreover, in this case values of k that are prime
nent 2 n_3 modulo 2 n
~
2 n_3 in 4>(2 n s ) ways,
can
from
the integers 1,2,3
2
be
chosen
to
.
and thus the form 8k
and 7 belong to
~4
(2V
also 2 4
l)
2n " 4
if
2 n_3 integers from
2r with
l(mod
1 through
not only the integers 9
15 belongs to 2 modulo 2 4
4,
2)
(r,
1,
2n)
must, therefore, belong to an exponent which
is
for this modulus.
we
Similarly,
But
if
(8Jfe
divisor of 2 n
observe, moreover, that
and that 2 4 r
yields 20(2 n_3 )
modulo 2
2 modulo 2 4 but
2" that belong to
We
2 n_3
115
find that, for

2n_s
l)
>
n~a
= 1 + 2 (2
= l(mod2 n
>
s
s
r)
3,
8
(2 r)
2n
~'(mod 2 n )
If,
in addition,
s
(2 r
l)
2 "" s_1
2)
(r,
>
and n
2-- s - 1
s
(2 r)
1,
n ~ s- 2
{2
n- s~ l
l)(2 s r) 2
(2Y)
2n ~ s-1
(niod 2 n )
2 w -V(mod 2 n )
so that in this case 2 s r
belongs to 2 n
~s
modulo 2 n
There are then
n~ s
ways of choosing r prime to 2 so that the integers 2 s r 1 are

Thus at least 2</>(2 n ~ s ) = 2 n ~ s incongruent integers
between 1 and 2 n
belong to 2 n_s modulo 2 n for n > s + 1 > 4.
~
But if n = s + 1 > 4, the integers 2 n
1 belong to 2 modulo 2 n and
n
n
1 belongs to 2 modulo 2
Consequently, for
it is also evident that 2
n > 4, 20(2) + 1=3 incongruent integers belong to 2 modulo 2 n
It is, furthermore, apparent that 1 belongs to 1 modulo 2 n for n > 3.
~
Consequently, if n > 3, we have shown that, of the 2 n 1 integers 1
)
<f>(2
through 2 n that are prime to

2
n~ 2
of
=
=
~
to 2 n 2 modulo 2 n
~
2 n 3 belong to 2"~ 3
~
~
2 n 4 belong to 2 W 4
2n
them belong
n- 3
2</>(2
)
n~ 4
)
2<K2
20(2 n
2,
~s
)
~s
+ 1=3
20(2)
1 belongs to
But
22
23
accounted for
all
Therefore, the
number
the 2
n~ 1
ft
When
to
>
ft
~s
belong to 2
modulo 2 n
2n
~2
2 n ~\
and we have thereby

and prime to 2.
positive integers less than 2 n
each case is exactly the number

belong to the specified exponent modulo 2"
of integers given in
of incongruent integers that
when
belong to 2 n
3.
3,
modulo 2 3
three incongruent integers belong to 2 and one belongs
116
The congruence x 2k
Corollary.
2 k+1 solutions
'2 n
l(mod
<
where
),
<n
2,
n > 2.
Theorem 7-11 shows that the congruence x 2 = l(mod 2 n ), for n > 2,
has 2 2 solutions modulo 2 n since the solutions must be prime to 2 and
2k
belong to the divisors of 2. For a like reason x = l(mod 2 n ), with
2
3
has 1 + 3 + 2 + 2 +
1 < k < n
2,
+ 2 k = 2 k+l solutions
n
modulo 2 if n > 2.
*Theorem 7-12. If d is a divisor of X(2 n ) and n = 0, 1, or 2, there is
exactly one positive integer less than 2 n that belongs to d modulo 2 n
has
if
7-9. Integers Belonging to a Divisor of
(f>(p
Modulo p n
p is an odd prime and d is a divisor of 4>{p n ),

where n > 1, there is an integer that belongs to d modulo p n
n
u where k is
(p ), d = kp
k
Take an integer a that belongs to k modulo p. Then a = l(mod p).
If a k = l(mod p 2 ), there is an integer r that is congruent to a modulo p
and is such that r k ^ l(mod p 2 ), for
*Theorem
7-13.
If
(a
rap^
a*
Hence,
if
(m, p)
(a
(mp) (mod p 2 )
fc
1,
mp)*
l(mod p 2 )
1.
mp. Then r belongs to k modulo p, and r k =

Keeping in mind that the exponent to which r
+
p)
belongs modulo p n has the factor k, we see that for n > 2 and (w, p) = 1,
Therefore, let r
cp with
3= 1
+ ka - mp +
+ ka ~ mp(mo&
(c,
(r
&),p-
(1
cp)
wpn
and
(r*)^""
~\mod p n )
-l
wcp n (mo& p n )
l(mod p n )
=z
But
(^p--!
Therefore, r belongs to kp
Furthermore,
if
== i(
n~ l
mo d
n
)
modulo p when n
ft modulo p and
>
1.
a belongs to
fc
ss
l(mod p s_1 )
l(mod p s )
but
a*
there
is
an
a(mod
p*),
where
r k ss
for since a*
(a
+ mp
2p
s
s_1
with
k
ss 1
(,
1, 2,
1,
such that
l(mod p s ),
p)
-\-
=
=
1,
ft_1
fca
s_1
mp*~ 1 (mod p s )
+ ka k ~ mp ~ (mod p
l
But an integer
117
prime to p can be so chosen that

ka k
~l
mp
~l
tp s
~x
{mod p s )
for this congruence is equivalent to
ka k
Hence,
+ mp ~
s
{a
Let r
i
1,2,
rap s_1
Then
~l
m=
k
(mod p)
l(mod p s )
ss
belongs to the exponent
/c
for each
p*,
where
s.
In like manner if r k = l(mod p s+1 ), we can find an integer

with (g, p) = 1 such that
(r
gp
gp
l(mod p s+1 )
as
each of the moduli p% where i = 1, 2,

Furthermore,
s, and hence v belongs to k for each of these moduli.
cp s with (c, p) = 1, if (w, p) = 1 and w > s,
since v k = 1
This
v is in
the class with
r for
(v
=
=
^
=
a) WP ---i
but
(^)p
n_s
~s
+ cp wpn ~\mod
+ wcp^^mod p w
s
n
)
l(mod p n )
l(mod p n )
~s
for n > s > 0.

This result
always an integer that belongs to a divisor of 4>(p n )
Therefore, v belongs to kp n
implies that there
(1
is
modulo p n
modulo p n
*Theorem 7-14.
.
If d is a divisor of <f>(p n ), there are exactly <f>(d)

incongruent integers that belong to d modulo p n
We have proved that there is an integer a that belongs to d modulo p n
As a result it is clear that the integers a, a 2 ... a d satisfy
.
Moreover, a
if
(s,
d)
modulo p n
where
1,2,
Therefore,
1.
<j>(d)
Consequently,
l(mod p
n
)
belongs to d modulo p n if and only

of the above powers of a belong to d
.
d,
if \j/(d) is
the
number
of incongruent integers
that belong to d modulo p n \J/(d) > <f>(d).

However, every integer from 1
through p n that is prime to p belongs to just one divisor of 4>{p n ) modulo
,
n
,
and so
f(di)
where the
di,
with
+
i
ffa)
1, 2,
Hdm) = v n~\v -
m, are the divisors of
l)
4>{p
over,
<f>(d 1 )
and hence each
\f/(di)
4>(d 2 )
<f>(di).
<t>(d
m)
= p-\p -
1)
).
More-
118
7-10. Integers Belonging to a Divisor of 4>(2p n ) modulo 2p n

7-15.
There are exactly <f>{d) incongruent integers
*Theorem
2p
belongs to
If r
<t>(2p
modulo
modulo 2p n the incongruent
integers
modulo
that satisfy the congruence

*(2p)
are
modulo 2p n
of (f>(2p n )
2p n that belong to any divisor d
r,
r2,
r*
(2pn)
.
These are
l(mod 2p n )
the integers from
all
through 2p n
that are prime to 2p n

In the manner previously employed we can show
that r s is a primitive root of 2p n if and only if (s, <f>(2p n )) = 1, and that
.
if
4>{2p
(s,
))
1,
rs
belongs to
n
<f>(2p
)/d modulo 2p n
then, always at least one integer r s that belongs to
modulo 2p n
d
n
<j>(2p ))
(s,
But
for as s takes the values
if
ranges over
belongs to
d,
all
<f>(2p
and so does
n
<f>(2p
the integer
<f>(2p ),
is,
n
<j>(2p )/d.
a divisor of 4>(2p n ), modulo 2p n then

,
(5<f)M2 P )/

modulo 2p n
Consequently, all integers that belong
to d modulo 2p n are congruent modulo 2p n to just one integer of the set
2
r, r
r3
r^ 2pn)
By choosing s so that (s, <t>{2p n )) = <f>(2p n )/d, r s
will belong to d modulo 2p n
But there are exactly cf>(d) integers s in the
n
n
set 1, 2, 3, ...
<j)(2p ) that have with 4>{2p ) the greatest common
the congruence
b satisfies
gruent to a power of
Hence, exactly 4>(d) incongruent integers belong to

)/d.
d modulo 2p n
7-11. Integers Belonging to a Divisor of X(m) Modulo m.
One of the
divisor
<j>(2p
author's students, Bernard Sussman, and the author have developed the
following proof of a
of
method
for finding not only all the primitive X roots
but also
It
has been shown thus far that
all
the integers that belong to any divisor of X(m) modulo
if
m =
n
]
J pj
>,
a primitive X root
a,-
can
common
be determined for each of the moduli pf* and the

congruences
solution of the
set of
= ^(mod p/0
,r
j = 0, 1,
root of m.
We have seen, moreover, that
x
is
a primitive X
solution
generates
by means
of its
powers
v
,
(1)
with
this
common
prime to X(w),
exactly <f>(\(m)) incongruent primitive X roots of m.

If there is a primiX
tive
root of
distinct from those already found, that too will generate
<f>(\(m))
incongruent primitive X roots of
*R. Carmichael, Quart.
none
of
J. Math., Vol. 44, pp. 94-104, 1913.
which
is in
a pre-
ceding
Thus the number
set.
multiple of 0(X(ra)), but
119
of incongruent primitive X roots of
m is
our problem to find this multiple and to

the primitive X roots modulo m.
it is
show how to determine all

Let \l/(d) (mod 2 n ) be the number of incongruent integers that belong to
n
n
Then if we understand that
d, a divisor of X(2 ), for the modulus 2
p = 2, the number of a that can be used in the congruences (1) is
r, is
^(X(2 n ))(mod 2 no ) and the number of aj} where j = 1, 2,
.
<t>(k(pj
nj
))
Consequently, there are
^(X(2"))(mod2 w
<>)
Y[
y-i
cf>(\(pj
0)
each set having a single solution modulo m, and

each of these solutions is a distinct primitive X root modulo m.
But the primitive X roots so determined are not necessarily all that
ni
exist.
Let dij, where i = 1, 2,
,q3; represent the <& divisors of \(pj ).
ni
=
1 divisors dy of the \(pj ), where j
Choose any set of r
0, 1, 2,
nj
r, that contains exactly one divisor of each \(pj ) and such that the least
common multiple of the dy is the number X(m). Then find integers Uj
that belong to the dij modulo pj n >, and find the common solution Y
modulo m of the set of congruences
sets of
congruences
(1),
Y is
s ^(mod
This integer
pfi)
0, 1, 2,
... ,r
a primitive X root modulo m, for
YHm)
i(
mo d
YHm)
l(
mo d m
and that
py
(2)
evident that
it is
w ')
But if F belonged to an exponent u < X(m) modulo m, then F w =

l(mod m) and F w = l(mod Pj ni ), where j = 0, 1, 2,
r.
Thus m
would be divisible by each d ih and hence by their least common multiple,
which is X(m). Therefore, u is not less than X(m), and F belongs to X(m)
modulo m. It is evident also that if there are h3 integers tq that belong
.
to da
modulo
pj
nj
,
there will be J] kq congruences (2) for one choice of

3
=1
r
Consequently, there are exactly
the d^.
\l/(di
) (mod
2 n )
<f>(da)
=1
determined by one choice of the
incon-
gruent primitive X roots of

j
is
0, 1,
r,
having least
obviously a multiple of
common
dy,
multiple X(m), and this
where
number
<f>(\(m)).
r
Each
set of
\l/(d i0 )
(mod 2 n )
\\ 4>(dn) incongruent primitive X roots of

3
=1
120
separates without repetition
an element into subsets
of
of
</>(X(m))
Each such subset is generated by the powers, with exponents

prime to X(ra), of any member of the subset, for if gi and g 2 are primitive
X roots of m determined by one selection of the d and if g 2 ^ <7i (mod m),
where t is prime to X(ra) and
< t < X(m), and if
integers.
i3
gi
where
Si
and
s2
Sl
gi
(mod m)
s<l
are prime to X(ra), then

g-f*
S2t
g2
(mod m)
But because the powers of g 2 repeat the

by g 2 for a certain t, say ti,
S2
original 4>(\(m)) roots generated
1*1
g2
**i
g 2 (mod
This conclusion implies that our assumption

be exactly
m)
is false.
There
will,
there-
fore,
*(<fco)(mod2-) II *(*/)
0(X(m))
such sets developed from each of g distinct choices of r + 1 divisors di3

where i = 1, 2,
r, having the least common
.
.
q3 j = 1, 2,
multiple X(m). No primitive X root of m generated by the powers of a
primitive X root (3 that has been determined by one choice di l3 of the di3
can be found among the primitive X roots of m determined by a distinct
,
and dij differ by at least one divisor

corresponding to a X function of a power of a particular prime. Further-
set of divisors di 2 j, for the sets dij
more, assuming that d^h is distinct from di2h and that /3 belongs to d^h
modulo ph nh and 7 belongs to d^n modulo ph nh let us suppose that (3 V =
7 (mod m) with v and w prime to X(m). Then /3 U = 7 w (mod p^**). But
,
i>
and
w are prime to X(ph nh and hence to the divisors of \(ph nh

)
(pv)*i x *
d
((3
hh) v
l(mod
p^
However,
nA
)
Hence,
7 >)<V
= l(mod prt
wd^.
Therefore, d^h divides d^.
and thus
d,-,*
d it h divides
If
divides
d^h,
and
finally d^h
there are q sets of divisors
dij,
d;^,
which
In like manner,
is
the primitive X roots of
m determined by
them are, therefore, distinct modulo m.

But each integer which belongs to X(m) modulo
m and which is, thereprime to each of the p3 ni belongs to exactly one exponent that is a
Hence, from all possible choices of sets of
divisor of \(p3 nj ) modulo p3 nj
r,
r + 1 positive integers d i3 where i = 1, 2,
q3 j = 0, 1, 2,
fore,
121
which the least common multiple of the da in a set is X(m), we find all
the incongruent primitive X roots modulo m, for if a belongs to X(ra)
for
modulo m, then
a\(pfi)
i(
mo d
py
and hence a belongs to an exponent
dij
0, 1,
modulo pf that divides

1'
\(pj
ni
).
If, however, the least common multiple of these dq were u < X(m), a
would belong to u modulo m. Thus when \f/(X(m)) (mod m) is the number
of incongruent integers that belong to X(m) for the modulus m, we have:
*Theorem
the pj} where j

primes, there are exactly
7-16.
If
0,
1,
2,
r,
are distinct
^(X(m))(modm) = V
i(dU)(mod
2") [] <K<fe)
y=i
9
r
incongruent primitive X roots modulo

is
i
m=
]| p/', where the

i=o
taken over the number q of distinct sets of
1, 2,
qj,
of the X(py nj )
whose
least
common
summation
divisors d#, with
multiple
is
X(m)
and where each of the q sets contains just one divisor of each of the
ni
These primitive X roots of m can be separated into subsets, of
\(pj )>
0(X(m)) integers generated by any member of the subset.
It is, furthermore, obvious that the method with an analogous proof
holds for finding the integers that belong to any divisor of X(m) for the
modulus m.
Examples. At the end of Sec. 7-7 we showed that the primitive roots
of 7 and 3 yield exactly two distinct primitive X roots of 21.
But it is
evident that X(21)
6 can be built up, not only as the least
common
=6
multiple of 0(7)
and <(3) = 2, but also as the least common multiple
of the factors 6 of 0(7) and 1 of 0(3), and again as the least common multiple of the factors 3 of 0(7) and 2 of 0(3), and in no other way from
factors of 0(7)
and
0(3).
Using the fact that 3 and 5 belong to 6 modulo 7 and that

modulo 3, the sets of congruences
x
x
= 3(mod
= l(mod
7)
3)
x
x
=
=
show that 10 and 19 are primitive X roots

of 10 yield 0(X(21))
10 5
The
19(mod
5(mod
l(mod
of 21.
belongs to
7)
3)
Notice that the powers
2 primitive X roots of 21, and they are 10 and
21).
integers 2
and 4 belong to 3 modulo
7,
and 2 belongs
Therefore, the sets of congruences to be solved are
x
x
=
s
2(mod
2(mod
7)
3)
=
=
4(mod
2(mod
7)
3)
to 2
modulo
3.
122
= 2, x = 11 (mod 21), respectively, and again we

(mod 21). There are then but three sets of two
21, and therefore exactly six incongruent integers that
Their solutions are x
observe that 2 5
11
primitive X roots of
belong to 6 modulo 21.
As a second example consider the problem

X roots of 168
least
common
factor of X(7).
of finding the primitive
= 2, and X(7) = 6, we
2,
which X(168) = 6 can be set up as the
multiple of one factor of X(2 3 ), one factor of X(3), and one
These sets of factors are listed in the following table.
must determine
23
all
Since X(2 3 )
7.
possible
ways
X(3)
in
Number
Case
Factors of A(3)
Factors of X(2 3 )
Factors of A (7)
of incongruent
primitive X roots of
168 determined
(1)
(2)
(3)
(4)
(5)
(6)
2
2
2
2
6
2
6
(7)
Case
1.
The
sets of congruences to
x
x
x
x
x
x
x
x
x
The
=
=
=
=
=
=
=
=
=
3 (mod 8)
be satisfied are
=
=
=
=
=
=
=
=
=
x
x
x
2 (mod 3)
3 (mod 7)
5 (mod 8)
x
x
x
2 (mod 3)
3 (mod 7)
7 (mod 8)
2 (mod 3)
x
x
x
3 (mod 7)
3 (mod 8)
2 (mod 3)
5 (mod 7)
5 (mod 8)
2 (mod 3)
5 (mod 7)
7 (mod
2(mod
5 (mod
8)
3)
7)
primitive X roots of 168 so determined are 59, 131, 101,
5,
143,
and 47, respectively, and they occur in sets of 0(6) = 2, which are
5 and 5 5 = 101 (mod 168); 47 and 47 5 = 143 (mod 168); and 59 and
59 5 = 131(mod 168).
Case
The
2.
The congruences
solutions are x
are
x
x
x
= l(mod
= l(mod
= 3(mod
73 and x
x
x
x
8)
3)
7)
73 5
=
=
=
l(mod
l(mod
5(mod
145(mod
8)
3)
7)
168).
Case
3.
The congruences
x
a;
The
solutions are x
Case
4.
ss
l(mod
2 (mod 3)
7)
17 and x
a;
ss
=
=
x
x
x
a:
3=
a;
s=
3=
x
x
x
8)
3(mod
x
a;
are
s
=
The congruences
17 5
3(mod
l(mod
3(mod
5(mod
l(mod
3(mod
7(mod
l(mod
3(mod
The congruences
x
x
x
The
solutions are x
Case
6.
x
x
=
=
7)
a;
s=
8)
a:
3)
x
x
7)
3)
x
x
7)
a;
8)
x
x
x
a;
3e
3=
a;
3s
=
=
3(mod
l(mod
5(mod
3=
5(mod
l(mod
5(mod
=
=
=
7(mod
l(mod
5(mod
157, 61, 31,
8)
3)
7)
8)
3)
7)
8)
3)
7)
and
103, respectively,
115 (mod 168); 31 and 31 5
168).
=
=
=
x
x
x
8)
65 and x
168).
65
l(mod
8)
2 (mod 3)
4(mod
137(mod
7)
168).
are
3(mod
l(mod
4(mod
x
x
x
=
=
=
=
=
=
8)
a;
3e
3)
a;
7)
a;
7(mod
l(mod
4(mod
3(mod
l(mod
2(mod
8)
5(mod
l(mod
2(mod
8)
7(mod
l(mod
2(mod
3)
7)
3)
7)
x
x
x
=
=
5(mod
l(mod
4(mod
8)
3)
7)
8)
3)
7)
8)
3)
7)
modulo 168 are 163, 67, 37, 109, 79, and 151.
= 109(mod 168); 67 and 67 5 = 163(mod
151(mod 168).
solutions
the sets 37 and 37 5
79 6
l(mod
7)
3=
7)
are
2(mod
=
=
=
=
8)
5(mod
89(mod
3)
2 (mod 3)
The congruences
x
x
x
The
=
=
=
l(mod
2 (mod 3)
8)
solutions modulo 168 are 115, 19,

and they form the sets 19 and 19 5
103(mod 168); 61 and 61 5 = 157(mod
5.
3=
=
=
are
The
Case
123
They form
168); 79
and
124
Case
The congruences
7.
x
x
x
x
x
x
x
x
x
The
are
= 3(mod
= 2 (mod
= 2(mod
= 5(mod
= 2(mod
= 2(mod
= 7 (mod
= 2 (mod
= 2 (mod
8)
hh
3 (mod 8)
3)
7)
=
=
2 (mod 3)
4 (mod 7)
8)
3)
s=
7)
5 (mod
2(mod
4(mod
=
= 7 (mod
= 2(mod
= 4 (mod
8)
3)
7)
8)
3)
7)
8)
3)
7)
modulo 168 are 107, 11, 149, 53, 23, and 95, and they form
= 107 (mod 168); 23 and 23 5 = 95(mod 168); 53 and
149(mod 168).
solutions
the sets 11 and ll 5
53 5
EXERCISES
1.
2.
3.
4.
5.
Find
Find
Find
Find
Find
the primitive X roots of 32.

the integers that belong to 4 modulo 32.
the integers that belong to 6 and to 9 modulo 27.
the integers that belong to 6 and to 9 modulo 54.

the primitive X roots of 72.
CHAPTER
INDICES
8-1. Indices for a
When p
Prime Modulus.
we have shown that
primitive root of p,
r,
is
2
.
a prime and r is a
~l
are incongruent
rp
Hence, the power residues of r modulo p are in some order the

p 1 and form a reduced residue system modulo p.
It is evident, then, that any integer n that is prime to p is congruent to
one of the above powers of r for the modulus p. Thus
modulo
p.
integers
1, 2,
n
Gauss
To
called the
exponent
express this idea,
we
s of r
(mod
p)
the index of n modulo p relative to the base r.

= ind r n, but we usually omit the base r
write s
when
there is no danger of confusion as is true in the case of the congruence r indn = ft (mod p).
As thus defined, the index of n modulo p is unique for the modulus
and
1,
if
for
>
if
t,
-*
l(mod
2(mod p
(mod
p)
then
p)
and
It
is,
1)
therefore, convenient to use the least positive exponent s such that
s
Obviously the index of n
r (mod p) as the index of n modulo p.
n
modulo 2 is useless.
Of course, the index of n determined by the primitive root r may be
For
different from that determined by another primitive root of p.
instance, for the modulus 7, ind 3 2 = 2(mod 6) but inds 2 = 4(mod 6).
Theorem
8-1.
If
m=
n(mod
p), their indices relative to
a particular
primitive root of a prime p are the same modulo p 1, and conversely.

The proof follows immediately from the fact that if r is a primitive root
of p, the congruences r indm = r indn (mod p), and ind r m = ind r /t(mod p 1)
imply each other.
Theorem 8-2. If p is a prime, and m and n are prime to p, then
md mn =
r
ind r
m+
ind r
n(mod p
1).
125
ELEMENTARY THEORY OF
126
Since
r indn (mod p)
and
m
it
53 r indm
(mod
p)
follows that
mn =
indm+indn (mod
mn =
indmn (mod
mn =
ind r
p)
But
p)
Therefore,
ind r
Theorem 8-3.
n(mod p
k ind r
If
m+
ind r
a prime and n
is
w(mod p
\)
prime to
is
p,
then ind r n k
1).
If
r indn (mod
nk
r fcindn (mod p)
nk
r indn^ m()CJ p)
ind r n k
k ind r
p)
then
Also
Hence,
n(mod p
1)
It is evident from these laws that the index of an integer plays a role
which is analogous to that played by the logarithm of a number. This
analogy is further emphasized by the following formula for changing the
base of a system of indices from one primitive root of p to another
Theorem
ind r2 n ind ri
Let
ri
8-4.
If
r 2 (mod
and
r2
a prime and n
is
is
prime to
p,
then ind ri n
1).
be distinct primitive roots modulo
Hence, taking indices to the base
Then
p.
rh
ri
ind r2 n ind n
1)
ind ri n
ind r2 n ind ri r 2 (mod p
1)
ind ri n ind ri
r 2 (mod
or
Theorem
ind n
r2
ind r2
8-5.
ri
If ri and r 2 are primitive

l(mod p 1).
ri
ind ri
r2
ind ri r2
ind r2
r x ss
roots of p, a prime,
then
r 2 (mod p)
ind r2
r2
(mod p
1)
and
ind r r 2 ind r2
,
If
we
r\
l(mod p
1)
construct a table of power residues of a primitive root
r of p,
we
127
INDICES
can use indices in the solution of any congruences of the form ax n

b (mod p), with (a, p) = 1, for the theorems given above show that
ind a
+ n ind x =
ind 6(mod p
1)
Therefore,
n ind x
and unless d
satisfies the last

b (mod p).
ax n
(n,
ind 6
divides ind 6
1)
a(mod p
ind
ind
1)
no value of ind x
is no solution of
a,
congruence, and consequently there
But
if
(ind b
ind
a),
there are exactly d incon-
gruent values modulo p 1 of ind x that satisfy n ind x = ind 6

ind a (mod p 1). By letting (p l)/d = mo we can express these
incongruent solutions modulo p

,
ra
2ra
as
(d
l)ra
Therefore,
j.
r i+A;mo(
mo ^
p)
0, 1,
modulo p of the given congruence.

power residues of r, the d values of x are easily
gives exactly d incongruent solutions
By
referring to a table of
determined.
congruence requires that we have

is used as a modulus.
In 1839
Jacobi* constructed such tables of power residues for each prime and
power of a prime less than 1000. Since then men like Goldberg, Wertheim, and Cunningham have contributed to the task of finding the power
residues of the least positive primitive root of every prime less than
M. Kraitchikf has listed a primitive root of each prime less than
10,160.
25,000, but many errors have been found in his tables.
It is quite evident that, instead of defining the index of an integer only
in the case of a prime modulus, we might have used any other modulus m
for which primitive roots exist, for in these cases the residues of the powers
of a primitive root likewise form a reduced residue system modulo m.
However, tables for such indices are less frequently available than are
They are, moreover, not essential
those with modulus p, a prime.
because, as we have shown in Chap. 5, we can reduce the solution of any
congruence to the problem of solving congruences with prime moduli.
Examples. 1. Solve: 5x = 7 (mod 11).
Using the primitive root 2 of 11, we set up the following table of
Of course,
this
method
of solving a
a table of indices for each prime that
indices:
ind
10
10
*L. E. Dickson, ''History of the Theory of Numbers," Vol.

t Ibid., Vol. 1, p.
202.
1,
p. 185.
ELEMENTARY THEORY OF NUMBE
128
Then
ind 5
ind x
ind 7 (mod 10)
ind x
7(mod
ind
a;
3 (mod 10)
= 8(mod
or
10)
and
Therefore,
2.
Solve: 7x*
= 3(mod
ind 7
11)
11).
3 ind x
=
=
ind 3 (mod 10)

8 - 7(mod 10)
ind x
7 (mod 10)
7(mod
3 ind x
Hence,
and
3.
Solve:
3a:
2(mod
ind 3
4 ind x
4 ind x
But
(4,
10)
and 2
2,
11)
11).
3,
Jf
=
=
ind 2 (mod 10)

3 (mod 10)
so that there
is
no solution
of the given
congruence.
4.
Solve: bx 2
3(mod
11).
ind 5
=
=
2 ind x
2 ind x
There are two solutions since
(2,
10)
ind 3 (mod 10)

4 (mod 10)
2 and 2
4.
ind x
2(mod
10)
4(mod
11)
ind x
7(mod
10)
7(mod
11)
and
5.
Solve: 7x
Any
4(mod
121).
solution of this congruence
ind 7
ind x
ind x
Hence, x
the form x
10(mod
10
11),
and
11/c.
7(10
all
must
=
=
satisfy 7x
4 (mod 11).
But
ind 4 (mod 10)

5 (mod 10)
solutions of the original congruence have
Therefore,
11/b)
4(mod
77k
55(mod
7k
5(mod
121)
or
121)
Hence,
11)
129
INDICES
and
ind 7
ind k
ind k
= 7(mod
10)
7(mod
11)
87(mod
ind 5 (mod 10)
so that
and
Therefore,
8-2. Euler's
moduli
Solvability of x n
for the
Criterion
121)
c(mod m).
For
we can determine a convenient
that have primitive roots
for the solvability of a congruence of the
test
= c(mod m) with
= b (mod m) with a
form x n
m) =
1.
Because any binomial congruence ax
prime to m can be reduced to this form by multiplying each member
of the congruence by the solution of ax = l(mod m), the problem is also
(c,
and
solved in the latter case.
Theorem 8-6. If m is any modulus for which there is a primitive root

and (c, m) 1, the congruence x n = c(mod m) has a solution if and only
m)/d =
l(mod ra), where d = (n,4>{m)). Furthermore, when there
if c^
is one solution, there are exactly d solutions modulo m of the given
congruence.
Suppose that r is a primitive root of m and that c = r s (mod m). If

x = c(mod m) has a solution x = r (mod m), it follows that r kn =
s
Since k satisfies this linear conr (mod m) and kn = s(mod 0(m)).
Hence,
gruence, d = (n,<f>(m)) must divide s.
n
fc
Conversely,
<j>(m)/d
s4>(m)/d
<}>(m)\s/d
(mod
Til)
if
c <Km)/d
i(mod m)
r s<t>(m)/d
i(
then
Therefore, scf)(m)/d
result,
is
values modulo
cf>(m)
s(mod
of k.
exactly d incongruent values

solutions of x n
and s/d
is
4>{m)) is satisfied
by
a multiple of
the congruence nk
mo d m)
<t>(m),
an
integer.
As a
just d incongruent
Corresponding to these integers, there are

m of x that form the complete set of
modulo
c(mod m).
If p is a prime and d = (n,p 1), there are (p l)/d
incongruent values modulo p of c, prime to p, such that x n = c(mod p)
Theorem
8-7.
has a solution.
According to Theorem 8-6, the congruence x n = c(mod p) has a solution
~ 1)/d =
if and only if c (p
l(mod p) where d = (n, p 1). But the con~ 1)/d
ip
= l(mod p) has a solution. There are, thus, exactly
gruence x
130
(p \)/d incongruent solutions

are the values of c for which the
It is evident also that
integer for a
same
c is
modulus
if
modulo p
one
first
an integer
c is
and these
of 'this congruence,
solvable.
is
a residue of the nth power of an
m having a primitive root,
and
if
(n,<(m)), the
a residue of the dth power of an integer for that modulus, for
^m)/d == i( mo d
if
the congruence x d
c(mod m) has a solution.

Hence, if d ^ n, a residue of an nth power modulo m is always a residue
It
of a power that is smaller than n and is a divisor of both n and 4>(m).
is thus clear that when p is a prime of the form 5k + 2, 5k + 3, or 5k + 4,
the test for a solution of x 5 = c(mod p) is the same as the test for a soluSince the last congruence is always solvable, an
tion of x = c(mod p).
integer c, prime to p, is always a residue of a fifth power for prime moduli
But if p = 5k + 1, c is a
of the form 5k + 2, 5k + 3, and 5k + 4.
residue of a fifth power modulo p if and only if c k = 1 (mod p)
If p is a
prime of the form 4A; 1, the very test for a residue of a fourth power
modulo p, c 2fc_1 = l(mod p), is the same as that for a second power and
hence in this case the set of residues of fourth powers modulo p is identical
with the set of residues of second powers. If p = 4k + 1, then 4 =
Again, if p is of the form
(4, 4Jc) and no such statement can be made.
3k + 2, every integer not a multiple of 3 is a cubic residue modulo p, but
= l(mod p).
if p is of the form dk + 1, c is a cubic residue if and only ifc
Finally, if the modulus is an odd prime p = 2k + 1, c is a residue of a
second power if and only if c = l(mod p). It is to the study of these
c
fc
fc
quadratic residues that
we turn
in the next chapter.
EXERCISES
2
Prove that the least positive residues of l 2 2 2
(p l) modulo p, where
p is an odd prime, repeat themselves exactly twice.
2. If p is a prime and n is prime to p 1, prove that the integers l n 2 n 3 n
n
Thus show that if p is a prime of
(p l) form a reduced residue system modulo p.
3
the form 3n + 2, the integers l 3 2 3 3 3
(p l) form a reduced residue system
1.
modulo
p.
Prove that the odd prime divisors of x A + 1 are of the form Sn + 1.

4. If p is a prime, determine when the existence of a solution of x 6 = c(mod p)
dependent upon the existence of a solution of x a = c(mod p) with n < 6.
6. Determine whether or not there is a solution and, if so, solve the congruences:
3.
is
a.
x3
b.
x*
6.
if
(a,
7.
a.
b.
= 5(mod
= 7(mod
13)
13)
if r is a primitive root of p, then r^' 1 ^ 2 = l(mod p), and thus that

ind
a) = ind a [(p l)/2] (mod p 1).
1,
p)
(p
Set up a table of indices for the prime 13, and solve the following congruences:
Show
that
5x
5x 2
m 4 (mod
= 6 (mod
13)
13)
INDICES
8.
9.
10.
131
Find the index of 5 for each of the primitive roots of 13.

Find the index of 5 modulo 13 relative to the base 6 by using a table to the base 2.
Use the following table of power residues of the primitive root 5 of 193 to solve
the congruences:
a.
b.
c.
d.
Sx
3x
7x
x3
=
es
2 (mod 193)
191 (mod 193)
157 (mod 193)
=
= 64(mod
193)
132
Table of Indices for the Prime 193

n
ind
192
34
84
4
68
118
104
8
102
168
10
35
11
12
17
152
14
138
16
183
13
141
15
ind
85
136
31
18
10
145
20
69
21
188
22
25
23
162
24
25
186
26
175
27
60
28
172
29
123
30
119
31
82
32
170
33
75
34
65
35
105
36
44
37
ind
38
179
39
33
40
103
41
151
42
30
43
24
44
59
45
169
46
4
47
29
48
28
49
16
50
36
51
115
52
53
77
54
94
55
184
56
17
14
57
37
58
157
59
148
60
153
61
63
80
12
65
142
66
109
67
18
68
99
69
47
62
116
64
ind
54
70
139
71
177
72
78
73
91
74
39
75
86
76
21
77
95
78
67
79
167
80
137
81
144
82
185
83
122
84
64
85
32
86
58
87
88
93
89
147
90
15
91
53
92
38
93
166
94
63
95
146
96
62
97
158
98
50
99
159
100
70
ind
n
ind
n
ind
n
ind
n
ind
n
ind
19
11
101
103
107
104
51
105
189
111
107
154
108
128
109
160
110
134
102
149
106
ind
111
89
112
48
113
41
114
71
115
163
116
191
117
117
118
182
119
135
120
187
121
174
122
81
123
43
124
150
125
ind
126
114
127
13
128
46
129
108
130
176
131
132
143
133
57
134
52
135
61
136
133
137
110
138
88
139
190
140
173
143
132
144
112
145
124
146
125
147
100
148
73
149
155
150
120
ind
ind
20
26
141
113
142
ind
151
155
83
156
101
157
140
159
161
160
154
129
158
126
152
55
153
ind
161
19
171
ind
74
162
178
163
23
164
27
165
76
166
156
167
79
168
98
169
90
170
66
n
ind
171
121
172
92
173
165
174
49
175
106
176
127
177
40
178
181
179
42
180
45
181
182
87
183
131
184
72
185
6
186
8
187
22
188
97
189
164
190
180
ind
56
191
ind
130
192
96
INDICES
133
Table of Indices for the Prime 193

ind
2
25
125
4
46
14
56
{Continued)
37
185
153
8
186
158
15
16
87
49
17
52
18
67
142
20
131
10
18
ind
11
12
90
64
13
127
ind
21
76
22
187
23
163
24
43
25
22
26
110
27
164
28
48
29
47
30
42
ind
31
17
32
85
33
39
34
35
10
36
50
37
57
38
92
39
74
40
177
41
113
42
179
43
123
44
36
45
180
46
128
47
61
48
112
49
174
50
98
51
104
52
134
53
91
54
69
55
152
56
181
57
133
58
86
59
44
60
27
61
135
62
96
63
94
64
84
34
66
170
67
78
68
4
69
20
70
100
71
72
184
73
148
161
75
33
76
165
77
53
78
72
79
167
80
63
83
155
84
85
15
86
75
87
182
88
138
89
111
90
169
n
ind
n
ind
n
ind
n
ind
74
65
19
114
ind
81
122
82
91
73
92
172
93
88
94
54
95
77
96
192
97
188
98
168
99
68
100
147
101
156
102
8
103
104
40
105
35
106
175
107
103
108
129
109
66
110
137
111
106
112
144
113
141
114
126
115
51
116
62
117
117
118
6
119
30
120
150
121
171
122
83
123
29
124
145
125
146
126
151
127
176
128
108
129
154
130
191
131
183
132
143
133
136
134
101
135
119
136
16
137
80
138
14
139
70
140
157
141
13
142
65
143
132
144
81
145
19
146
95
147
89
148
59
149
102
150
124
151
41
152
12
153
60
154
107
155
149
156
166
157
58
158
97
159
99
160
109
161
159
162
23
163
115
164
189
165
173
166
93
167
79
168
9
169
45
170
32
ind
171
160
172
28
173
140
174
121
175
26
176
130
177
71
178
162
179
38
180
190
ind
181
178
182
118
183
184
55
185
82
186
187
120
188
21
189
105
190
139
191
116
192
n
ind
n
ind
n
ind
n
ind
n
ind
n
ind
n
ind
n
ind
n
ind
31
11
24
CHAPTER
QUADRATIC RESIDUES
9-1. The General Quadratic Congruence
Theorem 9-1. A solvable quadratic congruence a y 2 + a y + a 2
0(mod ri), where a ^ 0(mod n), can be reduced to the form x 2
a(mod m) in which (a, m) 1.
By multiplying the modulus and each member of the congruence
x
by 4a
we
?/
dry
a2
0(mod
=
=
n)
obtain the equivalent congruence
4a
y + 4a
aii/
4a a 2
0(mod 4a
n)
which, upon completing the square, becomes
2a y
ai
4a n
z(mod m)
4a a 2
6(mod m)
(2a y
Now let
ai)
4a a 2 (mod 4a n)
and
i
The
original congruence is thereby reduced to the

z
form
b(mod m)
Suppose that (b, m) = d = e k, where e 2 is the largest square contained

Then ek divides z. Therefore, let
in d, and that m = m d, b = bod.
2
ekw(mod m)
and the congruence takes the form

e
k 2w 2
b (mod
kw 2
m)
or
If (k,
m =
)
s,
member
is
of the last
s
6 there is no solution.
no solution unless s = 1.
congruence by k, and let
unless
sequently, there
(mod
= kw (mod m
134
But
(6
If s
ra
1,
1.
Con-
multiply each
QUADRATIC RESIDUES
135
Then
x2
and
if
we
s= b
k(mod
set
b k ss
a(mod
we have attained the required form x 2 = a (mod m ) with (a, ra ) = 1.

The problem of solving any quadratic congruence is, therefore, reduced
to that of solving binomial quadratic congruences
and
linear congruences.
Examples. 1. Solve: x 2 = 6 (mod 15).

Let x = 3z(mod 15). Then 3z 2 = 2(mod 5). Multiplying by 3, which
2
is prime to 5, and letting z = 3w(mod 5), we have w = 6 (mod 5), or
2
=
=
=
4(mod
l(mod
Hence,
w
(mod
2
w
5).
1,
3, 2
5),
5), and finally,
=
6(mod
x
15).
9,
2
2. Solve: x = 24(mod 60).
Since (24, 60) = 12, let x = 6z(mod 60). Then Sz 2 = 2(mod 5). Let
z
z
= 3^ (mod 5), so that w = l(mod 5).

= 3, 2(mod 5), and finally, x = 18, 48,
2
Therefore,
12,
42(mod
w =
1,
4 (mod
5),
60).
EXERCISES
Solve the congruences:
a.
x2
b.
x2
9-2. Quadratic Residues.
=
=
28 (mod 84)
64(mod420)
When
(a,
a (mod m) has a solution, the integer a
m)
is
and the congruence x 2
a quadratic residue modulo m, or
a quadratic residue of m, but if the congruence has no solution, a is a

quadratic nonresidue modulo m, or a quadratic nonresidue of m.
For
example, 2
is
a quadratic residue of 7 because 3
satisfies
x2
2 (mod
7),
is no solution of
but 2 is a quadratic nonresidue modulo
x 2 = 2 (mod 5). Can you find a number whose square gives the remainder 4 when it is divided by 15?
If an integer a is prime to m > 0, its quality of being a quadratic
residue or nonresidue modulo m is called its quadratic character with
respect to m.
Obviously all integers in the class with a modulo m have the
same quadratic character with respect to m.
The problem of determining the quadratic character of a is, therefore,
equivalent to that of testing the solvability of the congruence x 2 =
a (mod m). We have already shown that by factoring m into powers of
primes we can reduce the discussion to the question of solving the congruence x 2 ss a (mod p n ) with p a prime, and finally to the case x 2 =
5 because there
a(mod p). Theorem 5-13 shows that when p is an odd prime, these last
two congruences either are both insolvable or have the same number of
solutions,
and therefore
in this case
only the quadratic residues of p.
it will
be necessary to study in detail
136
In the case of the prime 2, any odd integer is a quadratic residue of 2,

and the even integers are excluded from the discussion. But for f(x) =
x 2 a, f'(x) = 2x, and thus d = (2x',2) is always 2. Hence, Theorem
5-13 shows that either a solution of x 2 = a(mod 2 n_1 ) satisfies x 2 =
a (mod 2 n ) and yields two solutions of the second congruence or no solution
of the second congruence can be developed from the particular solution x'
The problem of determining quadratic residues of powers of
of the first.
the even prime must, therefore, be considered separately from that of
powers of odd primes.
Euler (1707-1783), Lagrange (1736-1813), and Legendre (1752-1833)
were all very much interested in the theory of quadratic residues, but
Gauss (1777-1855) was the one who contributed most to this subject.
The fact is that Gauss was one of the greatest mathematicians of all
It is really no wonder that his name is connected with so much
time.
that has been produced in the theory of numbers, for it was his favorite
He considered it the " queen of mathematics." His "Disstudy.
quisitiones arithmeticae," published in 1801, is the classic of the theory of
numbers and exhibits very well the elegance of form and rigor of presentaSome of the topics to which he made great
tion for which he is noted.
contributions are quadratic forms, biquadratic residues, and the theory of
congruences.
Theorem 9-2
If p is an odd prime, the integer a

(Euler's Criterion).
~
a quadratic residue of p if and only if a (p 1)/2 = l(mod p).
This theorem is included in Theorem 8-6. Nevertheless, we shall give
an independent proof for the case when n = 2.
is
a p -i be a reduced residue system modulo p, and supLet ai, a 2

pose that a is a quadratic nonresidue modulo p. Then each congruence
,
diX
a(mod
p)
1, 2,
has just one solution x = a, (mod p) distinct from a* and prime to

Moreover, no two of these congruences have the same solution, for if
aidj
a (mod p)
a k a,j
a (mod p)
p.
and
i f^
then
di
= a^mod
p)
whereas these integers are distinct modulo p. The integers a are thereby
separated into (p l)/2 pairs, and the product of these pairs implies that
z
i2
'
cip-i
a (p_1)/2 (mod p)
(p
But from Wilson's theorem

aid?
dp-i
1)!
= l(mod
p)
QUADRATIC RESIDUES
137
Therefore,
= -l(modp)
If a is a quadratic residue modulo p and x = r(mod p) is one solution of
ss a(mod p), then x = p r(mod p) is the other solution, for (p r)
(p-i)/2
x2
s= r
(mod
p).
Hence, the congruence

diX 3= a (mod p)
a (mod p) as a solution if and only if a* = r(mod p) or a* = p

r(mod p). The remaining p 3 values of a* set up (p 3)/2 pairs of
distinct integers such that the product of each pair is congruent to a
has x
S3
modulo
p.
Consequently,
0,10,2
,'
r(p
^zia^^modp)
'
r)
But
r(p
r) s=
r = a(mod
2
p)
Therefore,
1)!
33=
a (p-D/2
(p
-tf(p-/J(modp)
and
i(
m odp)
Furthermore, if a (p_1)/2 = l(mod p), a must be a quadratic residue of

(p_1)/2 = l(mod
This condition is, therefore,
p).
p, for if it were not, a
an
odd prime and a is prime to p.
a test for a quadratic residue of p if p is
3
=
is
Examples. Since 5
a quadratic nonresidue modulo
6 (mod 7), 5
3
=
l(mod 7), 2 is a quadratic residue modulo 7.
7, and because 2
It is interesting to observe that if the modulus is a composite m, the
following theorem gives a necessary condition for a quadratic residue of m
Theorem 9-3. If a is a quadratic residue modulo m > 2, then a^ m)/2 =
l(mod m) and a X(w)/2 = l(mod m).
Suppose that r, prime to m, satisfies the congruence x 2 = a(mod m).
Then because <f>(m) is even if m > 2,
:
(r
2)*(m)/2
= a ^^ )/2 (mod
m)
But
r *()
i(
mo d m)
a <^(m)/2
i(
mo d m)
Hence,
It is
obvious that
if
m>
2, <(ra)
may
be replaced by X(m) in the above
proof.
This
result,
however, does not provide a sufficient condition for a
quadratic residue of m, for although
l(mod
48), still
7(mod
(7,
48)
1,
48) has no solution.
\(48)/2
2,
and 7 2
138
Theorem
The quadratic
9-4.
residues of an
odd prime p coincide with
the even powers of any primitive root of p.

Consider the congruence x 2 = a(mod p) with
a primitive root of p, because the powers

residue system modulo p, either
r,
r2,
r 2k (mod p)
r 2k+1 (mod p)
(a,
p)
rp
~l
1.
Then
if r is
form a reduced
or
In the
evident that a
a quadratic residue of p, for

Applying Euler's criterion to the second case, if
first case, it is
a (mod p).
(r
is
2fc+i)(p-i)/2
i(
(r
2
)
mc.dp),
l)/2
But then (2k
the exponent of r must be a multiple of p 1.
would have to be an integer, and that is impossible. Hence, in the second
case a
is
a quadratic nonresidue of
Thus the
p.
set of quadratic residues
even powers of a primitive root of p.

Corollary 1. The odd powers of any primitive root of an odd prime p
coincide with the quadratic nonresidues of an odd prime p.
Corollary 2. There are exactly (p l)/2 incongruent quadratic residues and the same number of incongruent quadratic nonresidues of an
odd prime p.
Corollary 3. The product of two quadratic residues or two quadratic
nonresidues of an odd prime p is a quadratic residue of p, but the product
of a quadratic residue and a quadratic nonresidue of p is a quadratic
of
consists of the
nonresidue of
When
p.
is at hand, it is convenient to use the even

powers of a primitive root of p to set up the quadratic residues of p, but
if a primitive root of p must be computed, the method implied by the
following theorem is usually the more expeditious one for finding quadratic
a table of indices
residues
Theorem
9-5.
The
integers
2
,
22
M
I=
are the incon
gruent quadratic residues of the odd prime p.

Because a 2 = (p a) 2 (mod p), we need use only the integers
.
these integers
no two
of
is
them
to determine the quadratic residues
modulo
p.
2
,
Each
22
of
evidently a quadratic residue of p, but, more than that,

are congruent
modulo
di
p, for
if
a 2 2 (mod p)
then
(ai
a 2 )(i
a 2)
= 0(mod
p)
QUADRATIC RESIDUES
139
and p divides at least one of a\ a 2 and a\

a2
But since both a\ and
a 2 are positive and less than p/2, neither a x a 2 nor a\
a 2 is divisible
by p. These (p l)/2 integers, therefore, yield all the quadratic
residues of p.
Examples.
Because 2
of 13 are 2 2
4,
24
3,
is
26
a primitive root of 13, the quadratic residues
8
12, 2
Using the method of Theorem

42
ee 3, 5 2
12, 6
10(mod
9-5,
9,
2 10
they are
10,
l
and 2 12
1,
22
l(mod
4,
13).
32
9,
13).
EXERCISES
1.
Is 15 a quadratic residue of 17?
Find all quadratic residues of 29 and 31.

3. Prove that the product of the distinct quadratic residues of a prime p = 4n 1
congruent to 1 modulo p, whereas if p = 4n + 1, this product is congruent to 1
2.
is
modulo
4.
p.
If
is
is
a prime, prove that the congruence x G
+ 7x
i
36
= 0(mod
p) has a
solution.
5.
x2
If
a prime of the form 4n
1 is divisible
by
1,
prove that there
always an x such that
is
p.
a quadratic residue of 7, find all solutions of x 2 = 6 (mod 7) for each 6.

a quadratic residue of 7, find all solutions of x 2 = b (mod 7 2 ) for each 6.
8. Determine whether or not there are solutions, and if so, solve the congruences
= 9 (mod 13 2 ) and x 2 = 5 (mod 13 2 ).
9. Prove that if the prime p = 2 n
1, every quadratic nonresidue of p is a primi6. If b is
7.
x2
If b is
tive root
modulo
p.
p,
Make use of Fermat's theorem to show that if a is a quadratic residue of a prime

then a(p-D/2 = l(mod p).
11. If p is a prime greater than 3, prove that the sum of the quadratic residues of p
is
divisible
10.
9-3.
tic
by
p.
The Legendre Symbol.
Legendre was chiefly interested in
ellip-
functions and the theory of numbers, but he also wrote a book on
geometry which was so well received that at the time it rivaled Euclid's
" Elements" in popularity.
In 1830 he published two volumes on the
theory of numbers that organized his own researches and those of his
predecessors in this subject. In this work he partly proved the remarkable law of quadratic reciprocity.
If
is
an odd prime and
according as a
is
(a,
p)
1,
by
letting
(-j = -f-lor
a quadratic residue or a quadratic nonresidue of p,
Legendre introduced a symbol well fitted for expressing the quadratic

character of a with respect to p.
This symbol enables us to express
succinctly some important facts with which we are already familiar.
1.
1.
It is evident that
if
is
an odd prime and
(a,
p)
1,
then
=
J
140
2.
If d\
a 2 (mod p), a\ and a 2 being prime to the odd prime
p,
then
1,
then
(?) - (?)
3.
Euler's criterion shows that
J
4.
a (p 1)/2 (mod
1,
where
is
an odd prime and
(a,
p)
p).
Corollary 3 of
(ai} p)
if
Theorem
1, 2,
9-4 implies that

.
n,
then
if
is
an odd prime and
^\
\^\
(?)
5.
If ai
and a 2 are prime to
p,
) (
expresses the fact that &i and a 2 have the

respect to the odd prime p.
same quadratic character with
Furthermore,
-(?) indicate
as well as
1,
= 1 and
=
J
that ai and a 2 have opposite quadratic characters with
respect to p.
9-4. The Prime Moduli of Which an Integer Is a Quadratic Residue.
Having solved the problem of determining the quadratic residues of a
prime,
we now ask
if
we can
find the
a quadratic residue.
prime moduli of which a given
the integer
is odd, it is, of course, a

quadratic residue of the prime 2 and so the question must be settled only
integer
is
odd primes.
Suppose that a
If
for
Then because
= k
)
b,
^=
with k 2 the largest square in a and b

) (
)>
>
0.
the quadratic character of a with
respect to an odd prime p is determined by the quadratic character of 1

and the positive prime factors of a that occur in it to an odd power. Let
these primes be q h q 2
qn
~ \ v
Then
)\v)\v)
\v)
and therefore our question about the prime moduli of which a

ratic residue
where q
can be answered by studying the symbols

is
is
a quad-
J,
and
an odd prime.
V,
Taking these symbols

for every
prime
p.
in order, it is first of all
evident that
= +1
QUADRATIC RESIDUES
Theorem
the form 4n
An
The
9-6.
1
integer
141
a quadratic residue of
is
and a quadratic nonresidue
of all
all primes of
primes of the form
3.
The congruence x 2 = 1 (mod p) has a solution if and only if ( l) (p-d/2

= l(mod p). If p = 4w + 1, then (p - l)/2 = 2n, but if p = 4n + 3,
Hence, 1 is a quadratic residue

of the first set of primes, and it is a quadratic nonresidue of the second set.
Example. There is no solution of x 2 = l(mod 31), but there are
two solutions of x 2 = l(mod 29), and they are x = 12 and x =
clear that (p
it is
= 2n
l)/2
-12(mod29).
Theorem 9-7 (The Lemma
prime to
q,
2q,
Find the
p.
[(p
The
Take p an odd prime and q

modulo p of the integers
the number of these residues that are
of Gauss).
least positive residues
l)/2]q.
greater than p/2, then
1.
If
is
(-) = (-!)"
integers
q, 2J, 3q,
^q
(1)
p.
Their least positive residues
and incongruent modulo
are prime to p
modulo p
are, therefore, distinct integers of the set

1, 2,
(2)
Let
ai,
(Li,
au
represent the least positive residues greater than p/2 of the integers in
while
(1),
bi,
b2
bv
denote those least positive residues which are
The
(p
less
Then
than p/2.
l)/2.
integers of the set
ai,
a2
au
are prime to p, less than p/2, and are incongruent

these integers are distinct from the bi, for if
bi
a/(mod p)
a,j
1, 2,
v;
modulo
1, 2,
Moreover,
p.
then
bi
However, both
therefore sq
set
1, 2,
+
.
h and
p)
a3 are congruent modulo p to integers of

-
= 0(mod
(p l)/2.
tq
= 0(mod
(1), and
where s and 2 are distinct integers of the
Hence, s + t = 0(mod p). But since both
p),
142
and
by p.
6t
a,-
61, b 2 ,
form a
less than p/2, the sum s + t cannot be divisible

cannot be divisible by p. Consequently, the integers
and
are positive
Thus
set of (p
positive
and
integers
1, 2,
less
.
bv ,
(p
b v (p
616.
ah p
a2
au
l)/2 incongruent integers modulo p all of which are

than p/2. They are, therefore, in some order the
As a
l)/2.
di)(p
a 2)
result,
(p
au )
au
^^
(mod
p)
(mod
p)
and
(-l) u bA
But the
b v aia 2
and the a; are the residues
bi
6162
b v a\a 2
au
'
of the
2q
products in
(1).
Therefore,
g(mod
p)
P- h ^-^(mod
niv
p)
!
and
(-1) U ^-7T^^ (P
By
multiplying by
1) M
~ 1)/2
= ^-o-^!(modp)
[(p
= l) M (mod
p)
and dividing by
f(p-l)/2
l)/2]!,
we have
But
g( p
- 1)/2
(mod
p)
Hence,
according to Euler's criterion.
(j)-(-l)(modp)
but since both
and
l) u have only the values + 1
and
1,
it
follows
that(j) = (-l).
Theorem
9-8.
-[?H
Let (p
q, 2a,
...
If
l)/2
s(?
s.
is
an odd prime and q

V
+
Also
q
)
r2
odd and prime to
then
p,
and
if
(-1)*.
p\
the least positive residues modulo p of
let
be in order n,
is
rs
Then
QUADRATIC RESIDUES
= v
2q
= p
M+"
_P
sq
= p
sq
_P_
rs
Adding these equations and using the

2
[(p - l)/2] = (p - l)/8, we have
p
But these
i
1, 2,
Let
a*
u,
bj
in
1, 2,
l)/2, are the a z
v,
of
where
Theorem
9-7.
pM + A + B
(3)
Theorem 9-7 we also showed that the p a together with the

some order the integers 1, 2,
Hence,
(p l)/2.
?
fey
= pu A
subtracting Eq. (4) from
v2
1
(q
(3),
l)
we
2).
+B
(4)
find
p (M
u)
2A
(5)
u is even; that is,

if both p and q are odd,
Hence, when q is odd and prime to the odd prime
Consequently,
u(mod
(p
Then
B.
p-
By
+r
r2
that
fact
y-i
l__l 9 =
But
fry,
= A and
=i
are in
ri
where k = 1, 2,
where j
and the
integers r k
.
pM +
143
applying Theorem 9-7
M=
p,
by
we have
= (-!)- (-1)"
On
the
hand,
other
if
2,
then
2q
0,
0,
V
(P
l)g/2 l
P
J
since p = 2& +
p-
LP
[
0.
Hence,
0,
and Eq.
1,
p<
= -(2k + l)M(mod
u = w(mod 2)
2)
(5)
shows that
144
Therefore,
by Theorem
9-7,
(_1)-
The
Corollary.
form Sn
when p = Sn
3, (p
then (p 2
1,
l)/8
l)/8
of all
= Sn
2
= Sn Qn
2
(_ 1 )(22)(24)/8 = +lj
()-
all primes of the

primes of the form Sn 3.
is
and a quadratic nonresidue
When p = Sn
Examples
integer 2
(-1)(p D/8
and
2n,
and
1,
but
= +1,
-
(I) -
but
<- 1) (28)(30)/8
-1.
Theorem 9-9 (The Quadratic Reciprocity Law).

-
tinct
odd primes, (
^J
C?) = (-l)^
If
p and q are
dis-
'^.
We shall present a proof of this theorem that is based upon a geometric

demonstration given by F. G. Eisenstein (1823-1852).
y
v^f
^<;
Taking rectangular coordinate axes and a convenient unit, mark

units from
along the x axis to represent the integers 1, 2, ...
(p l)/2, and in like manner along the y axis to represent 1, 2, ...
Then OB = (p - l)/2, and OD = (q - l)/2. Let OA =
(q - l)/2.
p/2 and OC = q/2. We then call all points both of whose coordinates
are integers lattice points.
Within the rectangle OAEC, but not on its
off
boundary, there
The equation
are, therefore,
of the line
Iff
2
OE is py =
parallel to the y axis in the point
integer,
is
the
number
and on or below OE.
lattice points.
qx,
k,
and
)
it
intersects
Therefore,
of lattice points
on x
if
any
line
a positive
is
k above the x axis
However, when k takes the values
1,
2,
...
145
QUADRATIC RESIDUES
(p
i)/2, since (q,p)
lattice points
number
on
1,
kq/p
~ l)/2, we
OEC is
(<7
triangle
not an integer and so there will be no

Consequently, the
of lattice points within the triangle
In like manner by taking py
is
OE and within the rectangle OAEC.
find that the
[?H?] +
OAE
is
qx and the lines y
number
"
t}
for
have shown that \S\
1, 2,
[ViH
Hence, the number of lattice points within the rectangle
We
of lattice points within the
= (-1) M and
that
(^J
OAEC
(-1)*.
is
There-
fore,
d) (,)
and consequently when p and
_<_,,
g are distinct
odd primes,
This famous theorem was discovered at different times by Euler,

He accomfirst one to prove it.
Legendre, and Gauss, but Gauss was the
He
plished this feat in 1796, when he was but eighteen years of age.
appraised the theorem so highly as to call it the "gem of higher arithmetic " and developed six different proofs of it. Among the leading
mathematicians who have also proved the theorem are Cauchy, EisenIndeed, the
stein, Jacobi, Kronecker, Kummer, Liouville, and Zeller.
interest that it has continued to arouse is evidenced by the fact that it was
proved in about fifty ways* during the nineteenth century. More than
that, the number of proofs keeps growing, but, of course, not all of them
are essentially different.
Because
of the
importance of the quadratic reciprocity law, it is worth

it that does not depend upon the geometric
The following proof is a modification of one of Gauss'
while giving a second proof of
meaning
of
proofs:
* P.
Bachmann, " Grundlehren der neueren Zahlentheorie."
146
Taking p and q as odd primes with
M -[
N
we wish
&
(p
than the prime
p.
course,
if
+r+
most s + 1.
=
sp
M + 2V =
no
+p
2p
because p q
p
<
2p.
Assuming that
for
the expression for
sp
at
is
ft
be
<
M whose value
of the series in terms of p,
q,
We
last
<
with
and
s.
at
is
term
of
<
M which
[**-'
most
if
less
Of
1.
]
p,
then
^+
Therefore,
1.
l)/2 the integer

is s,
observe that
both k and g are
integer, for
most
written
(p
q.
can
Iff
an
- =0, and
LPJ
[P
so that kq
ff
/bg/p is
q]
P.
2p
Furthermore, the value of the
pq
Moreover,
s,
p and
L q J
l)/2,
But
g.
-fe]
to prove that
1, 2,
2q
<
l)g
1)g
(P
is
is
at
Dffl
2p
2p
(k
(2
D/2,
is
the last term of
we shall find the number k of this term
Since
1,
Hence,
k< (lVP <k +
and therefore
k
is
number
the
where
term
all
<
of the last
<
(q
-[*>]
term
l)/2.
of the expression for
of this series that has the value s
nonnegative
<
(q
M having the value
It follows then that the
l)/2, the
1 is
number
of
number
s,
of the last
Consequently, for
terms of
M that have the
QUADRATIC RESIDUES
>
pî - [f
that have the value
M-
2g
B]
(q
Moreover, the number of terms of
is
(p
l)/2
Therefore,
-I^f]
Hi
P.
Oyg
2
([?]
LQ
L 2 J
1
J
["
[^!]+^^
[|]-[f]
N+p 1 q
+ JV =
2p
+
+
ikf
3p
[?])
Hence,
l)/2
147
^^
^^
>
and the quadratic reciprocity law
follows.
Corollary
If at least
1.
one of the primes p and q
of the
is
-(J)--*Ms)-primes
both
and
Corollary
the
If
2.
q are of the
form 4n
3,
-ry
) (
form 4^
1,
then
6)- (!)-
=
Examples.
Since 15
Also
=-)
2.
1.
= 2(mod
^) =
(-1) 8
= +1, and
17),
we factor 2 and have =(
=
= +1, and
(~J
there
is
(-l)tt>U>/8
+L
77
)*
Hence,
a solution of the congruence.
Is 67 a quadratic residue of 89?
Since
67
S -22(mod
Furthermore,
for
15 (mod 17) for a solution.
Test z 2
89
is
of the
-oq-)
89),
= +1,
form 8n
we
for 89
find
is
(^) = (^) (|) (j{).
of the
and (oq)(tt)
form 4n
+1
+ l;(^j=+l
according to Corol-
148
lary
L^J =
Thus
above.
I J = +
1.
Factoring immediately
= +1, and
x2
= 67 (mod 89).
is a solution of x
33 a quadratic residue of 89?
Is
we have
(g)
(|) Qjj
[~
y) =
(3)
= +1.
Consequently, (oq)
above,
Therefore, there
3.
However,
= !
Q0 (|)
(g^j = -1, and, from
Therefore,
33 (mod 89) has no solution.

The quadratic reciprocity law
But
= 1, and
the congruence
s=
problem
also enables us to solve the
previously stated of determining the odd primes of which a given integer

It is apparent that if we wish to
which a composite is a quadratic residue or nonresidue,
the composite must be factored into primes and the conclusions drawn by
is
a quadratic residue or nonresidue.
find the primes of
means
Theorem
of Corollary 3 of
If q is of
+ 1,
the form 4n
5,
then
1,
Hence,
if
of q as well as being odd,

of
q.
and
to be
is
if I
For example,
is
if
l(mod
1,
2)
of
(-) = +1.
quadratic residue of
) is
and one
Thus p
q.
must be an odd quadratic nonresidue
in order that
shall, therefore, consider first
an odd prime
the prime p must satisfy the congruence
and one
if (
is
p must be a quadratic residue
to be 1,7?
q
We
9-4.
the case where the given number
to be
1,
all
4(mod
Hence, p
primes of the form
5)
1,
9(mod
10A;
10); that
1.
On
is,
is
the other hand,
then p must satisfy

x
l(mod
2,
2)
of the congruences
the form
3,
7 (mod 10),
10/c
If the given
and
and 10k +
prime
is
is
3 (mod 5)
all
primes of
7.
of the
form 4n
3,
then l^J
yH
(-i)l*-/i
149
QUADRATIC RESIDUES
and so (^) = (-l)^-
1 )' 2
( Y
In this case
if
(\
is
to be
l(mod4)
and
(^) = +1
= 3(mod4)
and
1^1 = -1
1,
then either
+1,
either
or
If
is
to be
= 3(mod4)
and
(^
and
{-) =
= +1
or
Accordingly,
if
l(mod4)
7 and
+1, then p must
to be
and one
is
l(mod
of the congruences
and one
satisfy
4)
xsl,2, 4(mod
giving
-1
s L
9,
25 (mod 28), or p
the congruence
satisfies
3 (mod 4)
3, 5,
7)
of
6(mod
7)
with the result that p = 3, 19, 27(mod 28). Hence, 7 is a quadratic

residue of all primes of the form 28/c
1, 28/c
3, 28k
9, 28/c
19,
27.
In like manner we find that 7 is a quadratic
28k
25, and 28k
nonresidue of all primes of the form 28k
11, 28k
13, 28k
5, 28k
15, 28fc
17,
and 28k
suppose that the given integer
+L
Then
(|)
a quadratic residue of p,
23.
Now
4n
=
it
(?) (j)
of the
is
form 2q with the prime
(_l)(f-/
(j).
follows that either
= l(mod
8)
and
= 3(mod8)
and
(-]= ~
?J
= +1
Ox'
If
2g
is
to be
150
If either
m l(mod8)
and
m 3(mod
and
(?
or
then 2g
is
When
is
even
or
the given integer
evident that
f^J =
so that
(^J
is
of the
1,
'
fl
p.
But \S\
( j.
P\
The exponent
3 (mod 8).
Hence,
==
(p
will
be
and
(-) =
and
(-) =
-1
= 4n
(_i)(p-d/2
+
+
4p
1
3, it
5)/8
when
is
either
when
p
The
conditions for
5,
7(mod
XP/
to be
8)
are
now
obvious.
Finally, let us suppose that the given integer
primes
(!)
form 2q with the prime
= (-1)^-^
(-1)(p +4 P -5)/8
and only if p
if
8)
and
qi
q2
with q 1
By combining
= 4n
the
first
and
q2
two cases
qi and
by the following statements:
They satisfy x = l(mod 2).
determined by those of which both
is
4s
it is
the product of two odd
3.
Then
i^\ =
clear that the primes
q 2 are quadratic residues are
characterized
1.
2.
They
They
are quadratic residues of
q\.
l(mod 4) and are quadratic residues

satisfy x = 3 (mod 4) and are quadratic nonresidues of q 2
There are also primes p described by the conditions:
1. They satisfy x = l(mod 2).
3.
satisfy x
of q 2
or they
2.
3.
They
They
are quadratic nonresidues of

satisfy x
l(mod
3 (mod 4)
q\.
and are quadratic residues
of q 2 or they
,
and are quadratic nonresidues of q 2

Examples. 1. Find the odd primes of which 14 is a quadratic residue.
If 14 is to be a quadratic residue of the odd prime p, then p must satisfy
two congruences selected in the manner already exhibited from each one
satisfy x
4)
QUADRATIC RESIDUES
151
of the following sets:
a
=
a;
ss 5, 7 (mod 8)
sa 3, 5,
a?
1,
1,
3(mod 8)
2, 4(mod
7)
and
From
the
first set
we
find that
6(mod
7)
1,9, 11, 25, 43,
51(mod
and from
56),
the second, p = 5, 13, 31, 45, 47, 55(mod 56).

2. Find the odd primes of which 35 is a quadratic residue.
(
to be
= (-)(-)> an d both symbols must
+1.
and one
giving p
Hence, p
satisfies
be
+1
or both
for
one of the congruences
= +l(mod
10)
of
1, 3, 9, 19,
25,
27(mod
28)
1,9, 19, 29, 31, 59, 81, 109, 111, 121, 131,
satisfies
one of the congruences
and one
of
3,
5, 11, 13, 15, 17,
with the result that p

127 (mod 140).
7(mod
13,
139(mod
140), or p
10)
23(mod
28)
17, 23, 33, 43, 67, 73, 97,
107, 117, 123,
EXERCISES
1.
Evaluate:
2.
Is there
3.
Find the values of
4.
Find the primes of which 11 is a quadratic residue.

Find the primes of which 6 is a quadratic nonresidue.
Prove that 10 is a quadratic residue of all primes p =
5.
6.
(^), (^), (|), (|)-
a solution of x 2
q for
21 (mod 41)?
which
(A) = +
1.
1, 3, 9,
13, 27, 31, 37,
39 (mod 40).
7.
Prove that
is
all
primes of the form 6n
and a
quadratic nonresidue of primes of the form 6n 1.

8. Find the primes of which 15 is a quadratic residue.
Show that a quadratic residue of an odd prime p is also a quadratic residue of p n

Prove that there are infinitely many primes of the form An + 1.
(Assume
the number of these primes is finite and use them to construct an integer 4 2 + 1.
Consider the form of the prime factors of this integer.)
11. Show that there is an infinite number of primes of the form Sn + 1.
12. Prove that 3 is a primitive root of every prime of the form 2 2 " + 1 by considering
the quadratic character of 3 with respect to such a prime.
9.
10.
152
The Jacobi Symbol.
9-5.
if
1,2,
m is
,r,
P =
Let
pip 2
p r where the
with
pi,
Then
are positive odd primes, not necessarily distinct.
any integer prime to P, the Jacobi symbol
following manner:
\p) ~ xpJ
771
-p
is
defined in the
'
where the symbols to the right of the equality sign are Legendre symbols.
When P = pip 2
p r and Q = qiq 2
q s with the pi and qj,
s, positive odd primes, the properties of the Jacobi
where j = 1, 2,
symbol are expressed by the following theorems:
Theorem 9-10. If m is prime to both the positive odd integers P and
'
'
Applying the definition given above,
("pj^l )( )
-(?) - fe)
0) h(?)(s)'=
(!) (f) " d "*"*" (?) (l) - fe)
Theorem
9-11.
/mn\ _ (m\
-
If
'
'
'
i)
fe)
and n are prime to the positive odd integer P,
n\
yv)
\p) \pf
(mn\
(mn\ (mrb\
- (?)
if
m=
.
r,
9-12.
n(mod
Because
.
H- (")
fe>
Theorem
(mn\
If
,.
- fe) fe)
(s)
(mn\
fe) fe)
m and n are prime to the positive odd integer P and
P), then
m=
_,
(p-J
n(mod P)
p)*
m=
implies that
n(mod p ), where
%
1,
Jacobi symbol.
Theorem
9-13.
If
is
a positive odd integer,
?
-=-
According
to
the
definition
=r-
1) (P-D/2
(
V[( Pi -l)/2]
(
1)
2,
the theorem follows immediately from the definition of the
*
,
where
1,
2,
r.
But
P =
p xp 2
"
pr
QUADRATIC RESIDUES
(pi
(P2-
'
1)}
{1
(Pr-
J= (-l)+ J=
i,k
P =
and thus
1)},
1)}{1
153
(pi-1)(p-1)+
+fl (P.-D- Butpi-l
i<k
- O(mod
so that
2),
P=
>
i
(p.-
l)(mod
P-
Hence,
4).
=l
V ^^ (mod
i
Therefore,
2).
(^j\ = (-l)^' 2
Theorem
9-14.
If
a positive odd integer,
is
pi |j = (- l)<
1 >' 8
.
1)J{1
(?>2
1)}
{1
(p,*
1)}
+ 2
"
1)
^-
(Pi
=l
r
2
(Pi
i)fe
1)
+"
+ n
0(mod
follows that
8), it
P =
2
!)
l)(mod
But
since
>
Pi2
(mod
Theorem
8).
9-15.
If
Hence,
P and Q
Y"\
fe
z=i
i,&=i
fe
and
64),
=1
(^\ =
(-1)Cp*-u/8.
are positive, relatively prime odd integers,
(0--^-(s)te)---(^"factoring
p)
in like
manner and forming the
pairs
where
r we find
) (
1, 2,
and
1,3
1, 2,
For a
fixed
*.
j,
(p*
l)(
a -
1)
i=l
fe
1)
(p<
1).
But we
i=l
saw that \
i
(p
l
1)
= P
l(mod
4),
and because
qj
1 is
even,
it is
154
evident that
fe
(lk
1)
(P
i)(*
i)
J {(
1)
1)(
l)(mod
8)
=i
Therefore,
(^
X(
1)
(p<
p " !)fe-
i)}
l)(mod8)
- (PBecause
P -
1 is
even, (P
1)
fe
l)(mod8)
Jfo-
1)
1)
(P
1)(Q
l)(mod
8).
Hence,
Therefore,
-<-
According to the definition of the Jacobi symbol,
w
)
= +1
or
when an even number
1.
are
(^
(?)
)
is
In the
+1 when
first
all
case the
congruences
x2
have a solution
ra(mod
for each
p,-)
1, 2,
= m(mod
But
in the second case the congruences (6) fail to

pi,
(7)
has no solution.
However,
if
-=5
is
(6)
a solution of
(7)
have a solution
Hence,
we have a necessary but not a
a quadratic residue of P.
P)
tain
p)
p i} and consequently there

x2
and therefore
if
for cer-
the Jacobi symbol
sufficient condition that
= 1,
it is
m be
evident that con-
gruence (7) has no solution.

Let us take some examples to illustrate the differences in the use of the
Consider the congruence x 2 =
Jacobi and the Legendre symbols.
Using
135 (mod 173) in which 135 and 173 are odd and relatively prime.
155
QUADRATIC RESIDUES
Jacobi symbols,
'
zr=~ )
find
(773
-'
- (I) -
(w)
Therefore,
we
= +*
is
a prime,
Since 173
Legendre symbols.
we
+1
and
"' (iS) = +1
and we can conclude that there

Since 173
(735
is
is
a prime, this
(g|)
\j35j
- +1
* (I)
'
a Legendre symbol
is
a solution of the congruence.
can, moreover, solve the
Thus
(jjjlj )
-QL
problem by using only
But
j (jfg) (jfg
(AX?) -
and (-3-
r)
= 1.
= 1.
Hence, (770)
+1.
(A)(t)But
(?)
= (I) " - 1 and
so
'
- -1
(A)
and therefore the given congruence has a
Now
consider the congruence x 2
In this case
prime.
/2l\ =
+1?
)
Hence, \kfo
Qj {^fj
but this
is
Furthermore,
Hence
-
solution.
21 (mod 253), where 253
a Jacobi symbol, and
tion of
() (i)
x = 21 (mod
2
and
11).
since
(it) =
-1
(tt) -
Hence, there
is
'
is
not a
Qj =
= +1, and
=
(^fj
+1.
we can reach no
However,
conclusion as to the existence of a solution of the congruence.
(S)
+1,
(m)
'
there
no solution
is
of
no solu -
the given
congruence.
EXERCISES
1.
Apply both Jacobi and Legendre symbols

S3 35 (mod 71) has a solution.
to determine
whether or not the con-
gruence x 2
2.
3.
Evaluate the following symbols and interpret the results
Use Legendre symbols
q=
>
r^ry
to determine all the quadratic residues of 41.
>
^rz
156
The
9-6.
the congruence x
(a, 2)
ss
o(
= a(mod
2)
It is evident that if (a, 2) =

has the solution x = l(mod 2), but
= a (mod 4) has a solution if and only
the congruence x 2
1,
l(mod
Theorem
a ss l(mod
a
Solution of x 2
mod
2 n ).
1,
if
if
which case there are two solutions 1 and 3 modulo 4.

2
9-16.
If n > 3 and x = a(mod 2 n ) has a solution, then
4), in
8).
Suppose x
satisfies
Then x
the congruence.
a(mod
2 n ),
and there-
a(mod 8)
But x is odd, and its square is, therefore, congruent
Hence, a = l(mod 8).
to 1 modulo 8.
Theorem 9-17. If a = l(mod 8), there are exactly four distinct solutions modulo 2 n where n > 3, of the congruence x 2 = a(mod 2 n ).
We shall first show by induction that the congruence x 2 = a(mod 2 n )
with a s= 1 (mod 8) has a solution. We know that under the given condiAssuming that, for a given k, x 2 =
tion x 2 s= a(mod 8) has a solution.
We, therefore,
a (mod 2*) is satisfied by x we infer that x 2 a = 2 k h.
wish to determine t so that #o + 2 k ~H satisfies x 2 = a(mod 2 k+1 ), that is,
fore x
ss
so that
(x
2*x
+
+
2
2
-1
2A;
2 (h
and
A;
~2
s=
a (mod2 k+l )
=
=
0(mod
0(mod
a;
2 k+l )
2 k+1 )
finally
/i
Because
which x
this
= 0(mod
Xo t
2)
congruence always has a solution, there
2 fc-1 satisfies x 2
= a(mod
2 k+l ).
is
a value of
Thus the congruence x 2
for
a(mod 2 n ) has a solution if a s= l(mod 8).

But whenever x 2 = a(mod 2 n ), where n
> 3, has one solution x =

has exactly four solutions, for suppose that x\ and x 2 satisfy
Then
the congruence.
a:i(mod 2 n ),
it
Xi
a; 2
0(mod
(mod
2 n)
and
(xi
and because both
x\
[(#i
x 2 )(xi
x 2)
2n)
and x 2 are odd,

xi
However,
x 2 )/2]
x2
Xl
rr 2
[(î
a; 2
se
)/2]
0(mod 2 W ~ 2 )
iCi,
and
therefore
one
of
(iCi + x 2 )/2
~
Hence, one of the congruences (xi x 2 )/2 = 0(mod 2 n 2 ) holds.
~
But when x\
Therefore, x 2 = î(mod 2 n l ), or x 2 = Xi(mod 2 n_1 ).
2
n
=
Consequently,
all
four integers
satisfies x
a(mod 2 ),
X\ does also.
2
n
n~ l
=
a(mod
incongruent
and
they
are
2
satisfy x
X\, xi + 2
),
(iCi
2 n_2
x 2 )/2 and
modulo
2n
is
odd.
As a
result, the other is divisible
by
QUADRATIC RESIDUES
157
= a (mod 2 n ), for a must be

modulo 2 3 and thus a has just 2 n ~ 3 distinct values modulo
~
odd positive integers less than 2 n separate into
Therefore, the 2 n
2n
2 n_3 sets of four such that all four integers in a set satisfy one and only one
~
of the 2 n 3 congruences x 2 = a (mod 2 n ) determined by the permissible
But there
congruent to
are exactly four solutions of x 2

1
values of
a.
Example. In the case of x 2 = a(mod 16), a can have only the values
1 and 9 modulo 16.
The solutions of x 2 = l(mod 16) are 1, 7, 9, and 15
modulo 16, and those of x 2 = 9 (mod 16) are 3, 5, 11, and 13 modulo 16.
EXERCISE
First find the values that a can
64),
and then
have
in order that there
find all the solutions of these congruences.
be a solution of x 2
= a (mod
CHAPTER
10

10-1.
The Waring Problem.
that every positive integer
is
In 1770 Waring published the statement

of not more than 4 squares, not more
sum
than 9 cubes, not more than 19 fourth powers. He gave no proof of his
assertion and may have had only calculations to support it, but the
problem implied by his statement has ever since challenged the best
mathematicians and has been only recently solved. It is generally
agreed that Waring meant to imply that for every positive integer k there
exists a smallest positive integer g(k) such that any positive integer n can
be expressed as a sum of at most g(k) positive kth. powers.*
Certain results connected with this problem of representing a positive
integer as a sum of like powers of integers had long been conjectured even
though not much progress in proving them was made until the eighteenth
Fermat, for instance, was much interested in the theorem that
century.
every prime of the form 4n + 1 can be expressed as a sum of two squares
and, barring the use of negative integers, in but one way.
Thus 5 = l 2 +
2
2
2
It remained for Euler, however, to demonstrate
2 and 13 = 2 + 3
Moreover, that any integer n is a sum of two
the theorem satisfactorily.
squares if and only if it has the form n = 2 a 2 P, with t >
and P a
product of different primes of the form 4s + 1, had been determined in
the seventeenth century. Again, mathematicians had asserted that any
integer not of the form 4 r (8s + 7) is expressible as a sum of three squares
and that integers of this form fail to be expressible as such a sum. Furthermore, by 1770 Lagrange had proved that every integer is a sum of at
.
most four squares.
From the time Waring enunciated his theorem, it took 139 years to
prove that every integer is a sum of at most nine cubes. Although
Liouville proved in 1859 that there exists a smallest integer #(4) such that
every n is a sum of at most #(4) fourth powers, still it has not yet been
proved that in this case #(4) actually is 19. In 1909 Hilbert proved the
general theorem that for each k there exists a positive integer g(k), independent of n, such that every integer n is a sum of at most g(k) kth. powers,
but his proof merely shows the existence of algebraic identities for deter*
G. H. Hardy,
"Some Famous Problems

158
of the
Theory
of
Numbers."
159
mining g(k) and sheds no light on the actual value of g(k). Since then
Hardy and Littlewood have developed by analytical means a formula
that determines an upper bound for g(k) for every k.
From these few remarks we can obtain some idea of the magnitude of
this problem, and certainly a perusal of a few of the original proofs will
give an appreciation of the ingenious adaptation of the tools of the theory
complex variable to the problems of the integers. We

are not concerned here with the presentation of any of these powerful
methods, but we shall give two of Euler's* proofs that every prime of the
form 4n + 1 can be represented uniquely as a sum of two squares and then
reproduce a proof, due to Euler and Lagrange, f that uses only the ideas
of the classical theory of numbers to show that every integer can be
of functions of a
sum
expressed as a
Let us recognize
that the identity
first
(a 2
most four squares.
of at
b 2 )(c 2
For instance,
(2
ever, that (2 2
)(3
1)(2
1)
52
(ac
two sums
expresses the product of

2
d2)
bd) 2
of squares as a
32
(ad
be) 2
sum
7
2
.
two squares.
We notice, howof
4 2 and so in this case the

,
formula gives but one representation of the product as a sum of two positive squares.
Does the formula ever fail to give at least one solution
when (a, b) = 1 and (c, d) 1? In this case ac = bd, and ad = be.
But if ac = bd, then a = d and b = c, and so the two given sums are
2
2
Hence, in the single
identical.
If ad = be as well, then a = b = 1.
2
case (1 + 1)(1 + 1) = 2 -f-0, the formula fails to give a sum of two
positive squares.
It is also
apparent that when
(a, b)
and
d)
(c,
1,
the squares in
the expression for the result of the product need not be relatively prime,
for example,
(ac
or
bd,
if (c,
d)
Euler's
ad
(8
be)
1 )(9
1,
2 2)
then
k, it is clear
first
(a, b)
70 2
+
1
25 2
and
On
(c,
d)
that k divides both ac
the other hand,
if
k,
1,
for
if (a, b)
bd and ad
proof of the fact that a prime of the form 4n
represented as a
sum
instructive to study
it
be.
1
can be
two squares is a little cumbersome, but it is

and to compare it with the second, more elegant
of
proof which Euler published about 25 years later.
The second
proof
enormous improvement in the directness of the presentation that a mathematician often attains when the initial proof is reviewed.
Lemma 10-la. If a prime p = c 2 + d 2 and if there is a q > 1 such
that pq = a 2 + b 2 with (a, b) = 1, then q is a sum of two relatively prime
exemplifies the
squares.
*
L. E. Dickson, "History of the
t Ibid.,
Vol. 2, p. 281.
Theory
of
Numbers,"
Vol. 2, pp. 230-231.
160
If
the prime p
c (a
d 2 then
)
d2)
b 2c 2
2
fr
d)
(c,
(c
and
1,
if
pg
pq
a 2d 2
(be
a2
2
a p
52
we have
= mp
But
2
c {a
2
)
a 2 (c 2
ad) (be
ad)
Consequently, the prime p divides at least one of be ad and be + ad.

Moreover, be ad 7^ 0, for if be = ad, then a = c and b = d. Under
these conditions p and pq would not be distinct.
If be ad = tp, let
+r
= td-\-s
tc
and
Then
cr
be
ds
ad
- ds
be
tc
and
id 2
so that
cr
ad
d2)
en
k(c 2
d 2)
t(c
Hence,
= ds
= dn
cr
r
But
if
be
ad
/cp,
and
the equations
6
kc
kd
cr
be
kc 2
ds
ad
kd 2
ds
be
ad
cr
= ds
and
give
and
Hence,
cr
and
In this case
r
In the
first
pq
and
= dn
and
= en
case
a2
=
=
=
td-\-
en) + (tc +
+ n )(c + d
p(t + n
(
(r2
in the second
pq
=
=
=
en) +
+ n )(c +
p(fc + n
2
(fed
(k 2
(fcc
d 2)
dn) 2
dn) 2

Therefore, q
tively
prime
is
sum
two positive squares.
of
if (a, 6)
1,
for
These squares are
in the first case
if
161
n)
(t,
u,
rela-
the equations
+ dn, a = td-\-cn show that u divides both a and b. Likewise,

= 1 in the second case.
Lemma 10-2a. If PQ is a sum of two relatively prime squares and Q
is
tc
n)
(A;,
not a sum of two relatively prime squares, then

is not a sum of two squares.
has a prime factor
that
Consider Lemma 10-la, and let P = pip 2

Pk, where each prime
=
Then
sum
of
two
squares.
with
h,
is
i
a
1,2,
pi(p 2
PkQ)
Pi,
= PQ is a sum of two relatively prime squares, and it follows immediately
that P2
PkQ is a sum of two relatively prime squares. By repeating
this process, we find that Q is a sum of two relatively prime squares.
We must conclude, then, that if PQ is a sum of two relatively prime
squares and Q is not, P must have a prime factor that is not a sum of two
'
'
'
'
'
squares.
Lemma
is
sum
10-3a.
of
If
a prime p divides a 2
2
,
where
(a, b)
then p
1,
two squares.
b 2 with
Let the prime p divide a 2
not a sum of two squares. Set
(a, b)
and suppose that p
1,
= mp ri
< n <
= np
r2
<
is
and
r2
<
Then
2
7"i
?*2
a2
= Qp <
/cp
~2
Any common divisor of ri and r must divide Q, and thus the last equation
can be reduced to ai 2 + bi = Pp, with (a h hi) = 1. According to
Lemma 10-2a, it is now evident that P has a prime factor pi that is not a
sum of two squares, and furthermore, pi < p/2. Using the fact that pi
2
we can repeat this process. The method, therefore,

always produces a sum of two relatively prime squares, a; 2 + b { 2 <
2
(pi-i )/2, which has a prime factor p < p;_i that is not a sum of two
squares.
But this statement is contrary to fact, for the prime factors of
all sums of two sufficiently small relatively prime squares are themselves
divides a\ 2
bi
a
l
sum
of
two squares.
2.)
sum of two
Theorem
(3
22
13,
32
Consequently, the prime factor p of a
22
10,
2
5,
must be a
squares.
10-1.
Every prime
squares in just one way.
of the
form An
is
sum
of
two
162
The integer 1 is a quadratic residue of all primes of the form 4n

1.
Hence, there is an integer a such that a 2 = l(mod p), where p = A.n
+ 1. It is immediately evident that p divides a 2 + 1, and accordingly
is
expressible as a
Now
sum
each of the pairs a and
two squares.
of
suppose that p
a2
and
b, c
b2
d,
d 2 where it is evident that, of

one integer is even while the other is
c
Then
odd.
d2
h2
c)(a
+c) =
\d
b)(d
so
that a
n(d
a2
and
(a
Let
(a
(m, n)
c,
= 1.
b)
r,
if
b)
= rm and
with
rn,
Therefore,
m(a
Hence,
we
let (a
c,
b)
c)
s,
b)
we have
-\-
= ms
ns
and
If
a and
But
if
case both
(r
both even or both odd, it is clear that r is even, and so is s.

In the latter
c is even, then r is odd, and so is s.
and n are also odd. Moreover,
c are
only one of a and
m
s
)(m 2
n 2) =
mr +ms +nr +ns

(a - c) + (d + 6) + (d 2
6)
(a
c)
Hence, the integer

(r
)(m 2
a2
+
2
62
cP
= p
and s are even, p has been factored into the integers (r 2 + s 2 )/4
and m + n 2 both of which are greater than 1. But if r and s are odd,
r and s cannot both be 1, nor can both m and n be 1, for in either case
a = d and b = c. Consequently, when r and s are odd, p is equal to the
product of the integers (r 2 + s 2 )/2 and (m 2 + ^ 2 )/2, neither of which is 1.
Since such a factorization is impossible, we must conclude that the prime
p = \n + 1 has a unique representation as a sum of two squares.
Let us start the second proof of Theorem 10-1 with the statement that
since 1 is a quadratic residue of every prime of the form 4*1 + 1, there
are integers a and m that satisfy the equation a 2 + 1 = mp.
The fact
that a is a quadratic residue of p means that a can be chosen positive and
not greater than {p l)/2.
Consequently, a 2 + 1 < (p 2 /4) + 1, But
Thus,
if r
2

2
(p /4)
<
Thus mp 
If
p.
mi
positive integer
we
1,
such that
b
q2
= mp,
<
(p
l)/2,
being an integer in the
show that we can produce a

sum of two squares.
with |ri| and \r 2 not greater than
shall
rri\p is
m+
true that
it is
there are integers
interval
163
r2
result
= n +
b2
+m
r22
( qi
g2
2
)
2m(r l9l
r 2q 2)
(1)
and
mp =
ri
+ mK
r22
so that
fi
r22
= mim
r22
<
But
ri
for not
both
|ri|
and
\r 2
feY
can be m/2 because
(a, b)
1.
Hence,
mim <
and
<m
mi
Lemma
Applying the identity exhibited just before

(ri
r 2 )fei
g2
?2
(r iqi
r 2q 2 )
10-la,
(nq 2
we have
r 2qi )
or
mmiqi 2
(2)
where
s
ngi
+ r<#
r x qi
and
Recalling Eq.
(1),
we observe
that
mp = mim
+m
(qi
r 2 qi
<?2 )
2ms
and
p
= mi
= mi
m(qi 2
q2
2s
Therefore,
m
and, according to
mim(qi 2
g2
(2),
mp =
x
mi 2
or
mip = (mi
2
-f s)
2mis
2miS
164
Hence,
upon the assumption that m >

rrti < m such that mip is a sum
clear that
it is
structed another integer
We
we have
of
con-
two squares.
must conclude, therefore, that m

1.
The unicity of this representation of the prime p = ^n -f- 1 was proved
at the end of the first proof of Theorem 10-1.
Recalling that the even prime 2 = l 2 + l 2 we shall now proceed to the
second problem stated above by showing in the following steps that any
,
odd prime can be expressed as a sum
of four squares or fewer:
Lemma 10-16 (Euler's Identity). (xi + x -r-z + Z4 )G/i + 2/2 2 +

= {xyx + x y + x y + x y + (xiy x y\ + x y x y
+
2/3
x y
x y x y
+ (xiy
+ (xiy* x y + x y x y
If i has the usual meaning, \J 1, this identity can be proved by find2
2/4 )
4)
-f-
4)
2)
ing the following product of two determinants whose values are X\ 2
x2
Xi
Xz
xz2
+ ix
+ ixi
x 4 2 and
+ v*
y*
x3
+ 1x4
Vi
Xi
y* - iy*
ix 2
--
z)
W2
y^
-2/3
iy*
y\
iyi
A - %B C -iD
-C -- iD A +iB
A + B + C + D'
-
where
Lemma
1
< <
t
Let
lie
Xi,
p,
10-26.
If
A =
B =
xiy 2
C =
x xy z
D =
xiy 4
is
such that tp
where
x{
xzyi
x 4 yi
an odd prime, there
0, 1, 2,
<
in the interval
+ xzyz + x y
+ x y xyz
+xy xy
xzyi
+xy
x 2y 2
x 2 yi
xiyi
<
Xi
(p
x 22
,
Xz
(p
exists
x42
an integer
and no two of these values are congruent modulo

^ 2 (mod p), where i 9^ j and j = 0, 1, 2,
(p - l)/2,
of x^,
(xi
Xi
Xj)(xi + Xj) = 0(mod p).

+ xj would be divisible by p,
than
where
l)/2, represent the integers that
There are then (p
l)/2.
t,
l)/2 values
p, for if
it
x?
follows that
Hence, at least one of Xi x3 and

is impossible because each is less
-
which
p.
same interval and form the

These integers are also incongruent modulo p, for
2
2
y^ = 0(mod p),
if
yj (mod p), where i 7* j, then yf
and we have seen that this congruence is impossible.
Because the Xi 2 and 1 y^ taken together form a set of p + 1
Therefore, some
integers, two of them must be congruent modulo p.
Calling these integers
x^ must be congruent to a particular 1 yj 2
x 2 and 1 y 2 respectively, we have
Again,
let yi
represent the integers in the
1 y
1 yi = 1
numbers
x' as
2
?/
(mod
p)
165
and
x2
so that x
>
and
tp,
<
Moreover, x 2
2p > p2 + 2 and
which shows that p
,
Lemma
10-36.
such that tp
xi
= 0(mod
where
tp,
a positive integer.
is
p)
2
2
2
p /4. Therefore, (p /4) + (p /4) +
2
However, since p > 2, it is evident that
1 > tp.
2
2
2
2
p > (p /2) + 1. Hence, p > (p /2) + 1 > tp,
>
If
<
and y 2
/^,
-\-
(p /2)
As a result, 1 < t < p.

an odd prime and t is the
t.
is
+ x% +
x 22
x 2 then
,
is
least positive integer
odd.
Suppose that t is even and that Xi 2 + x 2 2 + #3 2 + x 2 = p. Then the

xi} where t = 1, 2, 3, 4, are (1) all even, (2) all odd, or (3) two are even
and two are odd. In any of these cases the xi can be grouped in pairs so
Hence, {x\ + x 2 )/2
that x\ + x 2 = 0(mod 2) and x% + x = 0(mod 2).
and (xz + Xi)/2 are integers, and so are (xi x 2 )/2 and (# 3 Xa)/2.
Therefore,
( xi +
2
\
a? 2
S2Y
/ a?i
""
/ ffs
"*"
gj
(x3_-_Xi\
"*"
_ xi
x 22
'
#3
Thus there
Lemma
less
t and such that tp/2 is a sum of

must be odd.
an odd prime and if the least positive integer
Consequently,
10-46.
than
If
p, satisfying
is
2/3*
Choose
Xi
than
s less
x 22
:r 3
+ X4
and such that
st
is
not
yi
1,
2/4
?/,-,
Then y^
the condition tp
then there exists a positive integer

2/2
an integer t/2 smaller than
is
four squares.
t,
2
Xi
where
2
a;*
(mod
=
t),
1, 2, 3, 4,
so that y t
Xi(mod
t),
and let
\yt
<
t/2.
and
2
Vi
X ^ m d
2
Hence,
4
V
Therefore, y x 2
2/2
y%
2
?/,
?A
tp
0(mod
The
s.
integer
s is
not
or each
In this case t 2
2/ would be 0, and then each Xi would be divisible by t.
would divide tp, and t would divide p. Hence, t would be 1. Moreover,
4
since
\y {
<
t/2,
it
follows
that
yi
<
/4 and
V
i=l
<
2
,
and
<
t.
Consequently,
<
<
t.
y- 1
<
2
.
Hence,
166
Theorem
Every odd prime
10-2.
is
sum
most, four positive
of at
squares.
p =
p >
If
If
3,
observe that 3
3,
multiply corresponding members of the equations

xi
yi
+
+
+
+
x22
yi
+x =
+ yi =
x, 2
yz
tp
<t <
st
<
Remember
developed in the preceding lemmas.
We
tive integer satisfying the first equation.
<
that
find
is
the least posi-
upon applying
Euler's
identity that
t
sp
(xtfji
{xiyz
But
since yi
Xiyi
+x
x 2y 2
Xzyi
+ x y + x y^) + (x y - x yi + x y x yz)
+ xy x y^) + {x y x^yi + x yz x y
2
= ^(mod
(3)
2)
t),
+ x y% + xy = Xi + x +
xz2
+ Xzyt
y2
x 2
tp
= 0(mod
t)
Also,
xiy 2
x 2 yi
xy z
Xix 2
It is apparent, therefore, that
right-hand
But
member
of (3)
2
t
Xix 2
x d Xi
sp,
with
<
this conclusion contradicts the fact that
tive integer such that tp
theorem
sum
t),
divides each of the four squares in the
and that
is
x&i = 0(mod
t,
is
sum
of four squares.
was chosen the
of four squares.
Hence,
least posi1,
and the
proved.
is
Theorem
10-3.
Every integer
is
sum
most four positive
of at
squares.
Upon
factoring the given integer into primes, the theorem follows
immediately from Euler's identity.

It may also be added that we must use at least four positive squares to
express some integers as a sum of squares, for we shall show that no integer
that is congruent to 7 for the modulus 8 can be a sum of three squares.
x 2 2 + x z 2 = n. Not all the x i} where i = 1, 2, 3,
Suppose that Xi 2
can be even, nor can just one be, for n = 8k + 7 is odd. But if all the
2
2
2
2
Xi are odd so that Xi = l(mod 8), then xi + x 2 + 3 = 3 (mod 8).
2
are
even
and
one
is
odd,
x 2 2 + Xz 2 =
Finally, if two of the x
Xi
2
2
2
=
Consequently,
sum
of three
x
5(mod
a
8).
l(mod 8) or Xi + x 2 + z
integer
that
modulo
8.
equal
any
is
congruent
7
to
to
squares cannot be
EXERCISES
1.
as a
Prove that integers of the form 4 r (8n
sum
7) with r
and n
>
cannot be expressed
of three squares.
2. Write
prove?
(xi
x22
x 32
rc 4 )
as a
sum
of three squares.
What
does the result
The Equation x 2
10-2.
From
2
.
167
the point of view of the
theory of numbers the solution of the Pythagorean triangle means the

determining of formulas for expressing all triplets of positive integral
2
2
2
When x, y, z is a solution
values that satisfy the equation x
y = z
We shall,
of this equation, it is evident that kx, ky, kz is also a solution.
therefore, confine ourselves to the solutions, called primitive solutions, in
which the values x, y, z are relatively prime. This restriction together

with the equation implies that x, y, z are relatively prime in pairs, for a
common divisor of any two of them divides the third. We infer, then,
that two of these integers cannot be even. But neither can both the
and y be odd, for the square of an odd integer has the form
integers x
and thus the sum
two odd squares is divisible by 2 but not by 4.

is even, and the other is odd.
Let x = 2u.
= \u 2 = z 2 y 2 = (z + y)(z y). However, (z + y) + (z y) = 2z, and therefore any common divisor of z + y
and z y divides 2z. But a divisor greater than 1 of ^ cannot divide
Hence, 2 is the greatest common divisor of z + y and z y.
z + y>
Then
Therefore, let z + y = 2v 2 and z y = 2w 2 where (v, w) = 1.
4u 2 = (2v 2 )(2w 2 ), and x = 2u 2vw. By adding the members of
2
2
2
z + y = 2v 2 and z y = 2w we find, furthermore, that z = v + w and
Sn
1,
of
Consequently, one of x and y

Then y and z are odd, and x 2
2
2
Moreover, because (y, z) = 1, one of the integers v, w is
y = v w
even, while the other is odd.
If (v, w) = 1 and one is even, while the other is odd, and k is an arbi.
trary integer, the substitution of
= 2kvw
k(v 2
2
)
k(v 2
+w
2
)
makes it clear that the equation is actually satisfied

They, therefore, form the general solution of x 2 + y 2 =
2
z except for the fact that the values of x and y may be interchanged.
Hence, we have proved that:
Theorem 10-4. If k is an arbitrary integer, the solutions of x 2
in the given equation
by these
values.
and
2kvw, y = k(v 2 w 2 ), and z = k(v 2

w 2 ) for all integers
so chosen that (v, w) = 1 and one is even, while the other is odd.
are x
EXERCISES
2.
Express 21, 31, and 39 as a sum of squares.

Solve each of the equations x 2 + y 2 = 169, x 2
3.
Is there
1.
an
isosceles right triangle
whose three
+y =
2
625,
and x 2
+y
2704.
sides are integers?
4. Find expressions for the sides of all integral right triangles whose hypotenuse is
one greater than a side.
5. Show that if x 2 + y 2 = 2z 2
then x = k(v 2 + 2vw w 2 ), y = (r 2 - 2vw
w 2 ), z = k(v 2 + w 2 ), where k is an arbitrary integer, (v, w) = 1 and one of v and iv
,
is
even, while the other
is
odd.
168
10-3.
Fermat's Last Theorem. About 1637 Fermat stated that there
no solution in positive integers of the equation x n + y n = z n if n > 2.
This theorem is known as Fermat's last theorem, and about it he wrote,
"I have discovered a truly remarkable proof but this margin is too small
To this day mathematicians have been baffled by the
to contain it."
statement, for they have been able neither to prove nor to disprove the
The equations (x m ) 4 + (y m ) 4 = (z m ) 4 and (x m ) p +
general theorem.
m p = (z m p show that the proof can be broken up into the cases in
(y )
)
which n = 4 and n is an odd prime. In 1747 Euler published a proof,
which we shall reproduce below, that there is no solution in the first case.
No proof for every odd prime has been discovered although by 1857
is
Kummer had shown

was Kummer's
It
this
the impossibility of solving the equation
reflection
on
his
own
error
made
if
n <
100.
in attempting to prove
theorem, as well as the misconceptions of Cauchy and Lame, that led

With this new tool he and other
to his invention of the theory of ideals.
mathematicians were able to set up various conditions for the insolvability of the equation.
H. S. Vandiver* has given a complete account of
the present status of the problem, including his own recent contributions
"toward its solution
Theorem
tion x 4
10-5.
and his conjectures as to its truth.

There is no solution in positive integers
of the
equa-
2
.
The method we
shall use in
Fermat's method of
proving this theorem
infinite descent.
is a neat example of
assume that there are
show that under this condition
We
shall
and shall
having a smaller z. It is evident that if
there is a solution of the equation in which the integers x, y, z are not relatively prime in pairs, then there is another in which the restriction holds.
We shall assume, therefore, that x, y, and z are prime each to each. Then
2 2
2
2
2
(x 2 ) 2 + (y ) = z so that according to Theorem 10-4 we have x = a
2
2
2
2
b y = 2ab, and z = a + b with (a, b) = 1, and where we may choose
a odd and b even. Because y 2 = a(2b), it follows that a = u 2 and
Therefore, a = r 2 + s 2 and b = 2rs with
2b = v 2
But a 2 = x 2 + b 2
Hence, 26 = 4rs = v 2 and r = Xi 2 s = yi 2
Thus xi A +
(r, s) = 1.
A
2
and, furthermore, 1 < u < a < z.
Thus there is a set of
yi = u
It is imposintegers with a smaller z that satisfies the given equation.
sible that this always be so, and consequently there is no solution of
integers that satisfy the equation
there
must always be another
set
xA
y = z
Corollary.
x4
2
.
There
is
no solution
in positive integers of the equation
4
.
10-4. The Area

Theorem 10-6.
of
If
an Integral Right Triangle

the sides of a right triangle are integers, the area
cannot be a perfect square.

*
Am. Math. Monthly,
pp. 164-167, 1953.
Vol. 53, No. 10, pp. 555-578, 1946; ibid., Vol. 60, No. 3,
169
We shall develop the proof of this theorem as another example of

Fermat's method of infinite descent, using basically the very method he
outlined in the marginal notes of his copy of Bachet's edition of Diophantus's "Arithmetica."*
If x, y, and z are the sides of a right triangle, again restricting ourselves
to the case where these integers are relatively prime in pairs, let x = 2ab,
2
2
2
b 2 where a and b are relatively prime and one is
y = a b z = a
Then the area of this triangle, which we assume is a
even, the other odd.
Since (a, b) = 1, the integers
perfect square, is A = ab(a 2 b 2 ) = r 2
2
2
a, b, and a b are relatively prime in pairs, and each is, therefore, a
perfect square.
Let a = m 2 b = n 2 a 2 b 2 = k 2
But a
b and
a b are also relatively prime; so let
= m
= m n2 =
n2
u2
and
As a
result
2m = u
2
and
2n 2
Thus
(u, v)
and the
1,
even integers.
= u
last
Hence,
(u
equation shows that
and 2n =
2
even,
is
- -2
y
Therefore, one of the integers
-^ =
2s 2
8q
and u
(u
v)(u
are
+ v).
1_ p
<
even.
is
>
either
+ v)
v){u
and
Accordingly,
t*
^1
2s<
2s 2
let
or
^1 =
and
Then
2n 2
SsH'
n2
4sH'
and
Furthermore, in the
and
is
2s
2s 2
case
and
and
2s 2
m =
2
n2 =
t*
4sH 2
4s 4
4s 2 2
the hypotenuse of a right triangle in which

*
in the second
But
first
Dickson, op.
cit.,
Vol. 2, p. 615.
Thus m
and 2s are the arms. But
2
(t )
(2s 2 ) 2
170
notice that since z
a2
b2
that the hypotenuse of the

(u
new
Moreover, the area
original.
2
and a
and the area
v )/S,
(g.
of the
follows that z
it
triangle
new
>
>
m, so
smaller than that of the
is
triangle
the integer
is
2 2
of the original triangle is
A = a H a* But
6-)
p .)
(**)
>
^^
(^)
(V)
for
>
The assumption of the existence of a triangle whose sides are integers

and whose area is a perfect square has led to the conclusion that a triangle
having these properties but of smaller area and shorter hypotenuse must
Thus there would always be a smaller triangle of this kind,
also exist.
and that is impossible. We must conclude that such a triangle cannot
exist.
On
the basis of the last theorem
statement
Theorem
x2
-\-
If
x2
both p 2
whose
easy to show that the following
There are no integers that
10-7.
,
it is
true:
is
sides are
= w
2p q
and p 2 q 2 = n 2
4
A
p q and p + q*.
consider the right triangle
Hence, the area
p q (p
But we have shown that

perfect square.
satisfy the set of equations
if
= p
is
the sides are integers, the area cannot be a
Consequently, there
is
no solution
in integers of the
given set of equations.

10-5.
The Generalized Wilson Theorem
Theorem 10-8. The product of the positive
integers less than m and

prime to m is congruent to 1 modulo m if m = 4, p n or 2p n with p an
odd prime, but the product is congruent to + 1 modulo m for all other
,
moduli.
product 1 3 = l(mod 4).

t be a quadratic nonresidue of the odd prime p, and let
p
n
=
where
i
1,
2,
Oi,
(f>(p ), be the least positive integers forming a
Then, for each a,-, the congruence
reduced residue system modulo p n
n
n
ctiX = (mod p ) has a solution x = a,- (mod p ) from the set of the a,, and
each integer a is thus paired with an ay distinct from the a^ modulo p n
The integers a are, therefore, separated into 4>(p n )/2
for cti 2 f (mod p).
pairs, and if P is the product of these pairs,
If ra
If
m=
4, the
n
,
let
p =
^(P)/2( mo( J pn)
m l(mod
But t^~ 1)/2
and hence
p),
(J(p-l)/2)p.-i
and
t
171
(_]_
- _X
P-Hp-l)/2
k p )p-*
_|_
-f
Mpn
Therefore,
f*(p-)/2 ==
-l(mod p n )
and
P = -l(mod
m=
2p n
n
)
be a quadratic nonresidue modulo p, and

both of the congruences
x = s(mod p)
x = l(mod 2)
If
let s
let
satisfy
an odd quadratic nonresidue of 2p n for if x 2 = f (mod 2p n )

had a solution, then t = s(mod p) would be a quadratic residue of p. The
congruences ctiX = (mod 2p n ) now pair the positive integers a,-, where
n
n
If P
i = 1, 2,
4>{2p ), that are less than 2p and prime to 2p\
Therefore,
is
represents the product of these pairs,
P=
But
is
(p
~ 1)/2
= l(mod
odd, and
m =
(j>(2p
n
)
n >/ 2
we
find that
(mod 2p n )
= l(mod
P = l(mod
and thus t^ pn)/2
p),
n
Therefore,
<t>{p ).
n
p j. However,
2p n ).
> 2, then 1 is a quadratic nonresidue of 2 U

Hence, the congruences a# = l(mod 2 U ), where the a range through
the positive integers less than 2 U and prime to 2, separate these integers
~
In this case, therefore, if P again represents the product
into 2 U 2 pairs.
If
2U
where u
P = (-iy
of these pairs,
u ~2
= +l(mod
2 U ).
When m =
2, moreover, the above congruence is obviously true.

Finally, suppose that
contains at least two distinct odd primes as
factors or at least one odd prime and the factor 2 U with u > 1.
Let
n
m = 2 u p 1 n ip 2 n
p r *. Let s be a quadratic nonresidue modulo pi,
and let t satisfy both the congruences
x
x
=
=
s(mod pi)
l(mod 2p 2 p 3
'
p r)
'
'
a quadratic nonresidue of m. Again, if the a i} where i = 1, 2,

are the positive integers less than
and prime to m, then the
congruences a x = i(mod m) pair the a* and, as before, the product P of
Then
.
is
4>(m)
the
di is
such that
P =
But
since
i(pi-D/2
ni
(f>{p i
= -l(mod
is
pi),
even and 0(ra)

pc-o/i
*<
and
m>' 2 (mod m)
$*&>i*i>/* ==
<t>(pi'
n
)<f>(p2
l(mod
*)
= +i(modpi0
'
pi ni )"
<f>{p r
),
However,
172
m)/2 =
(1 + 2p 2 ps
p r k, so that t^
2u ~
n2
n3
nr
k)4>(m)/2
anc
j t<i,(m)/2 = -f l (mod p 2
Furthermore,
t
r
)
P2
Vr
p
2
U
M
2
= +l(mod 2 ), and thus t+' = +l(mod 2 ). Therefore, t*' =
+ l(mod m), and P = +l(mod m).
10-6. The Pellian Equation.
The equation x 2 by 2 = 1, in which b
Moreover,
2p 2 p 3
"
'
'
'
a positive integer that is not a square, is known as the Pellian equation,

but it should really be called Fermat's equation, for it was he who proposed the problem of finding its integral solutions. As a matter of fact
the problem is a very old one, having been solved in the Middle Ages by
the Hindus.
We have but an outline of Fermat's proof showing that the
equation is satisfied by an infinite number of pairs of integers all of which
can be obtained from a particular pair, but a proof was also found conBoth
jointly by Wallis and Lord Brouncker and was published in 1658.
Euler and Lagrange contributed to the further development of the
problem. The following argument is based upon that devised by
is
Dirichlet
Lemma
10-lc.
If
is
any
and
< s < m.
Let x have the values
<
such that
m>
number and
real, irrational
integer, then there exist integers r
\r
sa\
is
an
< 1/m
and
0, 1, 2,
There
m+
are, then,
<
integral values of y so that
values of y
m, and determine corresponding

xa < 1 by taking y = [xa] + 1.
xa that lie in the interval from to
,
but including 1. Separate this interval into m equal parts,

up to and including
the first one extending from, but not including,
1/m; the second, extending from, but not including, 1/m up to and
At least two of the m
1 values determined for
including 2/ra; ....
Suppose that these values are
y xa lie in one of these m intervals.
Then
yx X\a and yi xâ with x x\.
1,
excluding
{y x
(x 2
Xi)a\
difference cannot be 0, for x\
x2
1(2/2
x 2 a)
Xia)\
<
ni
or
|(?/2
But the
s
x2
Xi,
we conclude
<
Lemma
an
infinite
inequality
\r
2/i)
sa\
<
a positive integer that
If b is
of pairs of integers r
s b\
<
and
number
\r
Hence,
if
y%
y\
and
that
10-2c.
<
< m
<
2 \/b.
and
is
with
< m
not a square, there

s
>
is
that satisfy the
Lemma
exist
m is a positive integer, integers r
when
10-lc shows that
and
S\
such that
<
In
Vb\ <
si
'
Now
173
<
< m
si
choose a positive integer mi sufficiently large so that
< |fl-lVÎ
mi
Then
r2
and
s2
can be determined so that
<
\r 2
s2
<
'mi
<
s/b\
< mi
s2
Hence,
\r 2
By
Vb\ <
s2
continuing to choose
1, 2, 3,
<
|ri+ i
\n
we obtain
\/b\
Si
This means
we can
<
\/b\
si+1
\A| <
Si
sufficiently large,
m,:
<
mi
for
\n
find integers r -+i, s,+i that give

t
<
s i+1
<
rrii
Tfli
Thus
\ri+1
and
in this
way we can
satisfying the
But
for
y/b\
s i+1
set up an
above condition.
any pair
<
\fi
y/b\
Si
number
infinite
of pairs of integers
r, Si
ri} s i}
<
\n
Si
Vb\ <
< -
rrii-i
Si
Vb
Hence,
<
\n
8i
Vb\ < -
2si
Si
and
<
\n
Therefore, there are infinitely

s
> 0, such that

Lemma 10-3c.
exists
an
|r
an integer k
infinite
number
many
s b\ lies
If 6 is
< + 2-v 5<l+2v
Si b\
/6
pairs of integral values of r
between
and
such that the equation x 2
of pairs of integers x
and
and
with
1+2 y/b.
y.
is
not a square, there
by 2
is
satisfied
by
174
We have seen in lemma

r,
Since b
most
is
2[1
not a square,
pairs of integers
of pairs of integers
s b\ that lie between and 1+2 s/b.

b ^ 0, and thus the expression can have at
2 s/b] values so determined.
Consequently, at least one of

determined by an infinite number of the
and these are the values of x, y that satisfy x 2
these integral values, k
by 2
\r
number
10-2c that an infinite
give integral values for
r,
s,
0, is
k.
Theorem
10-9.
is not a perfect square,

has a solution distinct from x = 1, y 0.
Select k ^
so that the equation x 2 by 2 = k is satisfied by an infinite
number of pairs of values of x, y. Now separate the values of x, and in
the equation x 2
manner
If b is
by 2
\k\ residue classes modulo k.

There are, theremodulo k into which the solutions x, y can fit.
Having excluded the finite number of solutions with either x or y zero, we
recognize that the equation shows that for any solution in which either
or both x and y are negative there is one in which both are positive.
Hence, there is an infinite number of solutions having both x and y posi-
like
of y, into the
fore, k 2 pairs of classes
tive.
Accordingly,
if
these positive solutions are distributed
k 2 pairs of residue classes modulo
=
=
X\
2/i
Now
Xi
2/2(mod k)
2/i
2
x2
^
^
x2
Xi,
x2
2/2
2/i,
2/2
=
=
- by, 2
- by 2 2
>
>
k
k
consider the quotient

Xi
x2
From
2 (mod k)
Xi
y/b
2/1
x x x 2 --
2/12/26
O12/2
x*2
Vb
2/2
\/b
X2IJ1)
y* 2h
the above congruences
^1^2
2/i2/2?>
Xi
x 2y x
Xiiji
byi
ss
0(mod
k)
and
xiy 2
xiiji
= 0(mod
k)
Therefore, let
XiX 2
2/12/26
_ u
the
Consequently, two pairs
must contain at least two of these solutions.

2/2 can be so selected that
2/1 and x 2
#i,
among
at least one of these pairs of classes
k,
ano
Xiy 2
x 2 yi
and thus
xi
2/1
Vb =
(u
s/b)(x 2
2/2
y/b)
2/1
V6 =
{u
\A)(>2
2/2
\/b)
Likewise,

Hence,
Xi
by i 2
(u 2
k(u 2
bv 2
bv 2 )(x 2 2
175
by 2 2 )
or
v b)
so that
u2
and the integers u and v satisfy the equation x 2 by 2 = 1.

If v = 0, then u = 1 and then x\ 2/1 y/b = (x 2 y 2 y/b).
Thus Xi = x 2 yi = y 2 But since x h x 2 yi, y 2 are positive, x x = x 2
and yi = y 2j whereas we chose x h 2/1 and x 2 y 2 as distinct solutions of
x 2 by 2 = k. We have, therefore, found a solution x = u, y = v oi the
equation # 2 &2/ 2 = 1 that is distinct from the solutions x = 1, y = 0.
Theorem 10-10. If b is a positive integer that is not a square and if
2
2
Xi, 2/1 is the solution of x by = 1 for which Xi and 2/1 are positive and
#i + 2/i V^> has the least positive value, then any solution x, y of the
equation is determined by the formula x + y s/b = (x + 2/1 y/b) n
for n = 0, 1, 2, ...
Since x 2 by 2 = 1 has a solution in which 2/^0, there must be one,
Xi, 2/1, with Xi and ?/i positive integers, for which x\ + 2/1 V^> is a minimum. Then Xi 2 byi 2 = 1, and if x 2 2 by 2 2 = 1, let
.
and
(xi
2/1
VS)(^2
2/2
V^) =
2/3
V&
Oi
2/1
V&)(#2
2/2
y/b)
^3
2/3
V&
also
Consequently, x 3
mined
(xi
fo/3
1,
and the pair
a solution of the given equation.
is
y/b) n determines a solution of x 2
2/1
of integers z 3
by 2
2/3
so deter-
by induction, that
It follows,
1
for each positive
integral n.
Moreover,
/
Oi
and
X4
x2
if xi, 2/1, is
2/4
fo/
Again,
y/b
1,
2/1
v7rT
fr)
<
Xl
~ ^
)n
a solution of the given equation, then

(xi
2/1
y/b) n for n
2/1
1,
2,
x*, y
is
determined by
also a solution of
1-
n =
if
2/
0,
we
find
(xi
-\A)
= Oi
2/1
*\A)
and
0.
#, 2/ is a solution, x, 2/ and x, y, as well as .t,

Consequently, for n = 0, +1, 2,
all pairs of
.
determined by x
are solutions
2/ -\A =
(#i
2/1 V^)
Furthermore, when
y,
are solutions.
integers x,
of
x2
by 2
2/
71
1.
But the
pairs of integers x, y determined by the formula constitute all

the solutions of x 2
by 2 = 1, for if X, Y with both
and
positive is
176
any solution that cannot be developed by the formula, then X + Y y/b

lies between two consecutive powers of the positive number X\ + y\ Vb',
that
is,
(*i
By
member
multiplying each
using the fact that Xi 2
if
we
yi b
we
1,
2/1
Vb) n+1
(xi
y1
Vb
y\
Vb) n and
see that
Vb) n
Vl
(xi
statement by
of this
Y Vb <
< (X + Y Vb)(x -
But
X+
Vh) n <
2/i
<x +
1
let
(X
+ Y Vb)(xi ~
Vi
Vb) n =
x'
+ y'Vb
- Y Vb)(x +
Vl
Vb) n =
x'
-y'Vb
necessarily giving
(X
then x n
by' 2
and
<
But
x'
y'
Vb =
(x
x'
y'
Vb)~ and
l
<
Adding
<
(4)
and
shows that
(5)
Vb <
y'
+ y\ Vb
X + Y Vb
Xi
Vb
Vb <
x'
>
0,
and subtracting
y'
>
y'
(4)
therefore
y'
impossible that x'
was
Vl
2y Vb, which implies that

it is
x'
however,
x1
(5)
(5)
from
(4) gives
Under these circumstances,
0.
Vb
be
less
than X\
chosen as the smallest of these numbers.
y\
Vb,
for
Therefore,
each solution, not x = 1,

2
2
y = 0, of x by = 1 in which one or both the integers x and y are negative must correspond a solution in which both are positive, we conclude
that all values of x and y that satisfy the equation can be obtained from
the formula x
(xi
Vb) n
2/1
+ y Vb
Since
= (#i
y\
to
Vb) n
EXERCISES
1.
2.
X2
_
3.
Show that all solutions of x 2 2y 2 = 1 can be developed from x = 3, y = 2.

Can you discover a method for developing solutions of both x 2 2y 2 = 1 and
2y 2
= -1
Show
V3 =
from x
1,
1?
that the formula for developing
(2
+ V3)
n
.
all
solutions of x 2
3y 2
is
CHAPTER
11
POLYNOMIALS
11-1. Integral
Domains and
Fields.
Let us
recall that the set
of
rational integers has certain salient properties with respect to the oper-
ation of addition which can be summarized in the following manner:
The sum
1.
the
of
two elements
in a certain order is a
unique element of
set.
Addition is commutative.
Addition is associative.
4. Each element has an inverse with respect to addition.
= a for any a.
such that a
It follows then that there is an element
There is a second law of combination for the set of rational integers
which we called multiplication, and it has the following properties.
1. Multiplication is unique, and the product is in the set.
2. Multiplication is commutative.
2.
3.
3.
Multiplication
is
associative.
4.
Multiplication
is
distributive with respect to addition.
5.
There
6.
an element 1, called the unity element, or unity, such that

= a for any a.
The elements obey the cancellation law, so that if ab = ac and
0,
Any
then
is
c.
set of elements that fulfills these 10 conditions is said to
integral
When
be an
domain.
the set has at least two elements and in addition to the above
properties each element of the set except zero has an inverse with respect
is called a field.
Thus, in a field, ax = 1 with
can be solved for x, and the value is unique. Hence, division by
= 1, then a(bx ) = b and a divides b. The
a t^
is possible, for if ax
rational integers fail to form a field, for division of an integer b by a ^
is not always possible, but the set of the rational numbers (ratios a/b of
two rational integers with b t^ 0) is a field. The real numbers as well as
the complex numbers also form fields.
Now consider the set of all polynomials J(x) a x n + aixtt-1 -j-
n, are selected from
+ a n where the coefficients a with i = 0, 1,
This set includes the field of the coefficients itself, and consea field.
to multiplication, the set
aÔ
t-,
177
178
and 1, which are called the identity

and multiplication, respectively. Conare polynomials of degree zero, whereas
stants not
is said to have no
The laws of elementary algebra show very easily that this set of
degree.
polynomials f(x) has the first nine characteristics enumerated above, but
we must clear up another idea before showing that the cancellation law
quently
it
contains the numbers
elements with respect to addition
also holds.
Two polynomials in x are said to be identically equal if and only if they

have equal values for all values of the variable x. But a polynomial in x
is not reduced to zero by more than n values
of the nth degree with n >
of x, and therefore f(x) vanishes identically if and only if all of its coeffiIf, then, the product of two polynomials with coefficients
cients are zero.
in a field F is identically zero, at least one factor must vanish identically,
for otherwise the product would be a polynomial of at least degree zero
and could not vanish identically. In short, there are no divisors of zero
in the domain of these polynomials, and that is equivalent to saying that
the cancellation law is observed. The polynomials with coefficients in a
field F, therefore, form an integral domain.
After recalling the early definitions and theorems that pertain to the
elements of the integral domain of the rational integers, it is easy to set
up the following analogous definitions and theorems concerning the elements of the domain of the polynomials f(x) with coefficients in a field F:
If two polynomials /(a;) and g(x) have coefficients in a field F,f(x) is said
to divide g(x)
g(x)
if
is
a polynomial q(x) with coefficients in
Then
f(x)q(x) identically.
quotient of g{x)
most that
A
A
there
by f(x).
f(x)
Furthermore,
is
if
such that
a factor of g{x) and q{x)
g(x)
0,
is
the degree of /(re)
the
is
at
of g(x).
polynomial that divides only itself is a null polynomial.

polynomial that divides every polynomial with coefficients in
is
unit polynomial, or a unit.
Theorem
11-1.
If g(x)
mials have coefficients in a
f(x)q(x),
field F,
where
f(x)
the quotient q(x)
and the polynois
unique.
If
g(x)
f(x)qi(x)
f(x)q 2 (x)
then
f(x)[ qi (x)
q 2 (x)}
Hence,
qi(x)
Theorem
q 2 (x)
and
q x (x)
q 2 (x)
11-2.
Zero is the null polynomial of the set of polynomials
with
coefficients
in F.
f(x)
The equation
shows that the quotient of zero by zero exists
f(x) =
but is indeterminate. It also shows that zero divides only itself. Fur-
POLYNOMIALS
179
thermore, every polynomial f(x) divides zero, so that there can be but one
null polynomial.
Theorem 11-3. All the elements except zero of a field F are unit poly-
nomials of the set of polynomials with coefficients in F.

It is evident that a constant not zero of F divides every polynomial
whose coefficients are in this field. But a polynomial of degree n >
cannot divide any constant except zero.

A polynomial with coefficients in a field F whose leading coefficient is
unity is a monic polynomial.
The associates of a polynomial with coefficients in a field F are the products of that polynomial by the unit polynomials of the set of polynomials.
A polynomial with coefficients in F that is not a unit and that is divisible only by its associates and the units is a prime polynomial.
A polynomial with coefficients in F that is not zero, a unit, or a prime
polynomial is a composite polynomial.
A common divisor of two or more polynomials with coefficients in F is a
polynomial of the set that divides each of the given polynomials.
A greatest common divisor of two or more polynomials, not all zero, with
coefficients in F is a common divisor that is divisible by every common
When the coefficients of the given polydivisor of the given polynomials.
nomials are in a field F, the monic polynomial that is an associate of a
common
greatest
If
the greatest
cients in a field
Theorem
divisor
is 1,
11-4.
coefficients in
If f(x)
kb
divisor of the set.

coeffi-
and
g(x) are polynomials with coefficients
a unique pair of polynomials q(x) and r(x) with
that satisfy the identity g(x)
f(x)q(x)
r(x)
with
or of lower degree than fix).
Do
is satisfied.
likewise
if
g(x)
is
of lower degree
=
=
0.
and
r(x)
than f(x)
g{x)
and the
~
not of lower degree than f(x), take qi(x) = kx n m where
~
m
n
Then ri(x) = g(x) kx f(x) is lower in degree than g(x)
If g(x)
common
two or more polynomials with
~
a xn
an
aix n 1
~l
m
++&, take q(x)
bix
theorem
a
If g(x)
b xm
called the greatest
divisor of
the polynomials are relatively prime.
in a field F, there exists
either r(x)
is
common
is
and
g(x)
f(x)q 1 {x)
n(x)
or if its degree is lower than that of f(x) } the existence of the

pair of polynomials has been demonstrated, but if neither is the case,
If ri(x)
repeat the operation, using f{x) and r(x).

7*iO)
f(x)q 2 (x)
g(x)
f(x)[q!(x)
Thus we obtain
r 2 (x)
and
Again,
if
r 2 (x)
or
if its
degree
is
q*(x)]
r 2 (x)
lower than that of f(x), the required
180
polynomials are qi(x)
and
#2(2)
number
until after a finite
r 2 (x),
r s _i(z)
f(x)q 8 (x)
g(x)
f(x)[qi(x)
if
not, the process
is
repeated
r s (x)
Hence,
and
but
we obtain
of steps
++
q 2 (x)
q.(x)]
r a (x)
= 0, or it is lower in degree than /(re).

suppose that there are two pairs of polynomials that satisfy the
either rs (x)
Now
Then
stated conditions.
g(x)
f(x)q(x)
r(x)
f(x)Q(x)
R(x)
and
f(x)[q(x)
Q(x)]
r(x)
R(x)
s=
Consequently, q(x) = Q(x), or the degree of the above expression would

be at least that of f(x). Accordingly, r(x) = R{x).
Theorem
The
11-5.
greatest
common
divisor D(x) of
two polyno-
mials f(x) and g(x), not both zero, with coefficients in a field F exists and
can be expressed in the form D(x) = F(x)f{x)
G(x)g(x), where the
polynomials F(x) and G(x) have coefficients in F.

Using Theorem 11-4 we apply the analogue of the Euclidean algorithm
to the polynomials f(x) and g(x) of degrees m and n, respectively, with
<m<
Then
n.
g(x)
f(x)
Ri{x)
Rt-i(x)
= f(x)Qi(x) + Ri(x)
= Ri(x)Q (x) + R (x)
= R (x)Q (x) + R${x)
2
= R
Ri(x)
is
of degree
(x) is of
Rz{x)
(x)Qt+i(x)
is
< m
m < mi
m <m
m\
degree
of degree
+ R t+1
We must finally arrive at a step in which R t+ i is a constant, for the polynomials Ri(x), where i = 1, 2,
Then
t
1, decrease in degree.
by making use of an argument that parallels that in the proof of the
.
Euclidean algorithm, we see that f(x) and g(x) have a greatest common
It is also
divisor R (x) different from a constant if and only if R t+ i = 0.
evident that a necessary and sufficient condition that f(x) and g(x) be
relatively prime is that R +i be a constant different from zero.
If Ri(x) f^ 0, solve for each Ri(x) that is not zero, and substitute the
expression in the succeeding equation of the algorithm.
Thus when
t
R t+1 = 0, we find that

R {x) = g{x) -f(x)Q (x)
l
+ Qi(x)Q
Rz(x)
=
=
= F 1 (x)f(x)
R.{x)
(x)
[1
[1
2 (x)]f(x)
Q2(x)Q 3 (x)]g(x)
G!(x)g(x)
- Q {x)g{x)
- [Q (x)Q 2 (x)Q
2
z (x)
+ Q
1 (x)
Q*(x)]f(x)
POLYNOMIALS
Hence
if
and
f(x)
D(x)
However,
f(x)
if
and
+G
(x)f(x)
(x) is
not a monic
we have
G,(x)g(x)
g(x) are relatively prime,
if
leading coefficient,
its
F,(x)f(x)
R t+1 = F
and dividing by
prime and
g(x) are not relatively
polynomial, by dividing through by
181
we
find
(x)g(x)
we obtain
Rt+i,
F(x)f(x)
+ G(x)g(x)
In the last case the term of highest degree of F(x) comes from the
product Qi(x)Q 2 (x)
Q +i(x). The degree of Qi(x) is n m, of
Q 2 (x) is m mi, and the degree of each Qj(x), where j = 3, 4,
Therefore, when m > 1, the degree of their
t
1, is m/_ 2 W/_i.
(m mi) +
product is less than n, the degree of #(#), for n m
In like manner, when
(m _i m ) = n m
(mi m 2 )
m > 1, the degree of G(x) is determined by the degree of Q 2 (x)Q z (x)
Q t+ i(x) and is less than m, the degree of f(x).
In the special case where R\(x) = 0, fix) is a greatest common divisor,
and then f(x) = f(x)
g(x) and the theorem holds.
Notice that although the method is applicable when the polynomial
f(x) is a constant not zero, the statement about the degrees of F(x) and
G(x) does not hold. For example, if g(x) = 4x
3 and f(x) = 2, we
write 1 = 2 J
If f(x) = 0, the result is obvious.
(4#
3).
'
'
'
t.
The following theorem is now evident:

Theorem 11-6. The polynomials f(x) and
field
are relatively prime
G(x) with coefficients in
Theorem
and
If
Di(x) and
\
1,
= cD
2 (x)
D (x)
D (x).
2
(x).
common
Also
11-8.
is
+ G(x)g(x)
2 (x)
1.
divisor of the polynomials f(x)
unique.
two greatest common
and g(x),
than or equal to the
Hence, di = d 2
d\.
divisors of f(x)
d\ of Di(:r) is less
Di(x), so that d 2
<
(x).
If fi(x)
with coefficients in a
/i(z) divides
g(x) with coefficients in a
there exist polynomials F{x) and
and Z>2(^) can differ only by a constant factor; that is,

But each one is a monic polynomial. Therefore,
and Di(x) =
Theorem
are
if
such that F(x)f(x)
greatest
and the degree
result Di(x)
Di(x)
and only
g(x) with coefficients in a field
then Di(x)
degree d 2 of
As a
The
11-7.
if
field
and f 2 (x) are
F and if /i(x)
relatively
(x).
Since (fi(x),f 2 (x))
1,
fi(x)F 1 (x)
+f
prime polynomials
divides the product f 2 (x )fz(x), then
(x)F 2 (x)
182
Then
MiWW +
Applying the distributive law,
Theorem
11-9.
-/,(*)
evident that f\ (x) divides /3 (x).

g(x) are relatively prime polynomials of
it is
and
If f(x)
f*(x)f t (x)F t (x)
first degree with coefficients in a field F, there exists one and

only one pair of polynomials F(x) and G(x) with coefficients in F satisfying
at least the
and such that the degree of F(x)

than that of /(x).
That the required polynomials F{x) and G(x) exist has been proved in
Theorem 11-5, but if there is a second pair F\(x) and Gi(x) of such polythe condition F(x)f(x)
is less
than that
of g(x)
G(x)g{x)
and the degree
of G(x) is less
nomials, then
F(x)f(x)
G(x)g(x)
= F 1 {x)f{x)
{x)g{x)
Hence,
f(x)[F(x)
But
and
since f(x)
Unless F(x)
F{x)
Fi(x)
F,(x)]
g{x)[G 1 {x)
G(x)]
g(x) are relatively prime, g(x) divides F(x)
F\(x)
is less
this division
than that
would be impossible,
of g(x).
It is
Fi(x).
for the degree of
then obvious that G(x)
It is important to observe the conditions set by the last theorem.

We
admit that when the given polynomials are the constants a and 6, there are
infinitely
many
Theorem
11-9 requires /(x)
solutions of ax
and
by
But
1.
and
places a restriction on the degree of both F(x)
condition
is
it is
to be noticed that
g(x) to be of at least the first degree
removed, we can find
many
G(x).
If
pairs of polynomials F(x)
+ G(x)g(x) =
F(x) + k(x)g(x)
G(x) that will satisfy the equation F(x)f{x)
and G(x)
form
Are there any others?
do, then all pairs of the
k(x)f(x) will also.
nomials
On
is
When
1,
and
the second
for
if
and
Fix)
and G{x)
one of the given poly-
now be obvious.
common divisor of
a constant, the results should
the other hand
when
the greatest
the given poly-
even so simple a case as expressing the greatest common

divisor of x 2 4x + 3 and 2x 6 shows the lack of unicity in the pair of
polynomials F(x) and G(x) having the restriction on degree stated in
nomials
is
Theorem
not
1,
11-9, for
(x 2
4x
3)(
+ l) +
(2.r
6)
(-
lj
= x -
and
(x 2
When
the
field
4x
3)(-l)
(2x
6)
containing the coefficients of f(x)
numbers, on the basis of the fundamental theorem
is
the set of complex
of algebra,
we know
POLYNOMIALS
183
that, except for the order of the factors, f(x) can be factored into linear
factors,
each with leading coefficient unity and absolute term in F, and a

Hence, the identity
constant factor, in exactly one way.
fix)
a xn
where the
aix n
with
Vi,
factorization of f(x)
'
'
1 7 2,
a>n
The reader
tion of f(x) does not exist in
+
+
'
cto(x
r l )(x
r 2)
(x
r)
unique
however, that such a factorizaIn the field of the real numbers
n, are in F, expresses this
realizes,
all fields.
cannot be so factored, but in the field of the complex numbers

Although x 2 - 2 = (x - y/2)(x + y/2) in
i = (x - i)(x + {),
X2
the field of the real numbers, it cannot be factored into linear factors in the
field of the rational numbers.
If a polynomial f(x) of degree n with coefficients in a domain F cannot
be factored into two polynomials of at least the first degree with coefficients in F, then f(x) is said to be irreducible in F.
Otherwise f(x) is
x2
Thus x s 5 is irreducible in the integral domain of the

integers and also in the field of the rational numbers, but it is
reducible in F.
rational
reducible in the field of the real numbers.
The
field of
the rational numbers
tains the rational integers,
and
so
is
we
the smallest infinite
shall consider a
field
that con-
few important char-
polynomials with coefficients in this field. We are especially

concerned with being able to classify the coefficients of the factors of an
integral polynomial.
acteristics of
primitive polynomial
is
an integral polynomial whose
coefficients are
relatively prime.
The primary associate of an integral polynomial is that associate of the

polynomial whose coefficients are relatively prime integers and whose
leading coefficient
Theorem
positive.
is
11-10.
If fi(x)
and f 2 (x) are
integral polynomials, a neces-
sary and sufficient condition that their product be a primitive polynomial

is
that both fi(x) and f 2 (x) be primitive polynomials.

Suppose that
n
aix 1
/iO) = a Q x
++
71
and
f 2 (x)
xm
b x x- 1
++&.
with n > m, have integral coefficients that are relatively prime. Their
product necessarily has integral coefficients, but suppose that a prime p
divides each of these coefficients.
Then there is a first a;, say a r and a
,
first bj,
say
a r + s bo
6S
that
is
not divisible by
a r +ib s -i
a rb s
p.
Now
a r-ib s+ i
consider the coefficient of
+ a r+s -m b m
184
of this expression is divisible by p, and thus it is

the coefficients of the product be divisible by p.
Hence, the product is a primitive polynomial.
Each term except a r b s

impossible that
all
On the other hand,

and p
is
if
fi(x)
pg(x),
a prime, then fi(x)f2 (x)
where g(x)
p[g(x)f 2 (x)]
an integral polynomial
and the product is not
is
primitive.
Theorem
11-11.
If f(x) is
a polynomial with integral coefficients and
leading coefficient unity, fix) is factorable into the product of two monic
polynomials in the field of the rational numbers if and only if it is factor-
domain
able in the
Let f(x)
of the rational integers.
where fi(x) and/2 (z) are monic polynomials with

and suppose that not all the coefficients of the
Reduce all fractional coefficients to their lowest
fi(x)f2 (x),
rational coefficients,
factors are integers.
and d 2 be the least common multiples of the denominators

and 2 (#), respectively. Then the coefficients
= difi(x) and h(x) = d 2f2 (x) are relatively prime integers.
terms, and let d\
of the coefficients of fi(x)
of
both g(x)
Consequently, the product g(x)h(x)

nomial.
But then f(x) fi(x)f2 (x)
coefficients unless d\d 2
= did fi(x)f (x) is a primitive poly= g(x)h(x)/did cannot have integral

2
In short, the coefficients of both fiix) and
1.
f2 ix) are integers.

the domain of the integers,
and so the converse is obvious.
If f{x) is factorable in
able rationally
it is,
of course, factor-
a polynomial fix) with coefficients in the field R

is irreducible in R, and if f(x) divides the product
of the polynomials g(x) and h(x) with coefficients in R, then f(x) divides
Theorem
11-12.
of the rational
If
numbers
at least one of g(x)
and
h(x).
Because f(x) is irreducible in R, the greatest common divisor of f(x)

and g(x) either is the monic polynomial that is an associate oif(x), or is 1.
In the first case f(x) g(x). In the second, we have shown that /(a;) h(x).
Theorem 11-13. Any polynomial of at least the first degree with
rational coefficients can be resolved into a product of a rational constant
and one or more monic irreducible polynomials of at least the first degree
with coefficients that are rational. Except for the order of the factors
|
this factorization is unique.

If f(x) is either of
in the field
the
first
monic polynomial, and
c is
degree or of higher degree and irreducible
numbers, then f(x)

a rational number.
of the rational
cg(x),
where g(x)
is
If f(x) is reducible in R, let f{x) = fi(x)f 2 (x).

Each of the new polynomials is of lower degree than f(x). Either fi(x) is irreducible in R, or
it has a factor fz(x) which is lower in degree than fi(x) and is in turn a
Continuing in this manner, we observe that the degree
factor of f{x).
of each factor is lower than that of its predecessor and so the process must
end; that is, there must be an irreducible factor, say p(x),oi f(x). Then
POLYNOMIALS
f(x)
finite
185
We now
operate on q(x) in the same

that
determine
number of steps
p(x)q(x).
f{x)
Cpi(x)p 2 (x)
'
way and
after a
p r (x)
'
where c is a constant and the pj(x) are monic irreducible polynomials.

Then
Suppose that there are two such factorizations of f(x)
.
cpi(x)p 2 (x)
Clearly, c
k, for
p r (x) = kq 1 (x)q 2 (x)
q s (x)
the other factors are monic polynomials.
Pi(x) divides the product of the qi(x),
where
1, 2,
Moreover,
Hence,
s.
But qi(x) is irreducible in R.

and since each is monic, qi(x) = pi(x). Canceling
factors and repeating the argument, supposing that s > r,
Pi(x) divides one of the qi(x), say qi(x).
Thus
qi(x)
the identical
tpi(x),
we have
q r +i(x)
Consequently, each of these factors
is 1,
q s (x)
and the
original factorization
is
unique.
Theorem 11-14. If fi(x) and f2 (x) are integral polynomials, not both
we can choose a greatest common divisor of them so that it is an
zero,
integral polynomial.
Since the coefficients of fi(x) and f 2 (x) are rational, their greatest
divisor D(x) exists and can be written
common
D(x)
Let
and h be the
d, g,
common
least
g(x)/g,
and
2 (x)
h(x)/h.
fl{x)
(x)F 2 (x)
multiples of the denominators of the
and F 2 (x), respectively.
coefficients of D(x), Fi(x),
Fxix)
+f
fi{x)Fi(x)
and
h.
d(x)/d,
common
multiple of
Thus
k,d(x)
where kid(x)
m+ Mx) m
Multiply both members of this equation by the least

d, g,
Let D(x)
Then
is
fi(x)[k 2 g(x)]
an associate
of the
+f (x)[kzh(x)]
2
monic polynomial D(x) and has integral
coefficients.
Very often we use the primary associate of kid(x) in place of the greatest
divisor of fi(x) and f 2 (x) even though we may not be able to
write it in the above form with k 2 g(x) and k z h(x) integral polynomials.
Example. The greatest common divisor x of 3a; 3 2x 2 3x + 2
and 3a; 2 8a; + 4 can be expressed in the form
common
-#=
i(3a;
2a;
3a;
2)
(a
2)(3a; 2
Sx
+ 4)
186
Its associate
9x
6
Instead of 9x
9x
3a;
6 can be written
(3x z
2x 2
Sx
however, we
6,
+ 2) -
may
(3x 2
Sx
4)0
2)
prefer to use its primary associate
2.
EXERCISES
Can
1.
the primary associate of the greatest
2x
5x + 3 be expressed in
polynomials?
2
common
the form developed
in
divisor of 2x 2
Theorem 11-14 that
and
uses integral
Find the greatest common divisor of 6x 3 + I3x 2 + 4z 3 and 2a; 3 + 9x 2 + 13x

and express it in terms of the given polynomials. Write its primary associate.
3. Do the rational integers modulo p, sl prime, form a field?
4. A number r is said to be an algebraic integer if it satisfies a rational integral equa~
tion x n + aix n 1 +
an = 0, where the coefficients a*, for i = 1, 2, .
, n,
are rational integers.
If a is a rational integer, apply this definition to the roots of
the equation x m = a and consider the problem of factoring a.
2.
+ 6,
11-2. Polynomials with Respect to a Prime Modulus.

When the
modulus is a prime p, we pointed out in Chap. 5 that the division of fi(x)
by f2 (x) 7^ 0(mod p), where these are integral polynomials, exists. To
carry out the division, we may use the method of choosing k so that the
leading coefficient of kf 2 {x) is congruent to 1 modulo p, and then dividing
fi(x) by kf2 (x) according to the process of ordinary long division, so that
/i(z)
where r(x)
Hence,
is
kf2 (x)g(x)
r(x)
or an integral polynomial lower in degree than f2 (x).
= f2(x)[kg(x)] +
fi(x)
and
r(x)(mod p)
is the remainder in the division modulo p

This division has already been shown to be unique.
Definitions analogous to those in the first paragraph of this chapter
show that the set of integral polynomials modulo p, a prime, forms an
integral domain.
The unit polynomials, or units, modulo p are the
rational integers that are prime to p, for when (a, p) = 1, the congruence
kg{x)
of fi(x)
is
the quotient while r(x)
by f 2 (x).
ax = b (mod p) has a solution and we can infer that a divides any integral
polynomial modulo p. There can be no other units modulo p, for 1 is not
divisible modulo p by a multiple of p or by any polynomial of the first
degree or higher. Since r is identically congruent to r + kp modulo p,
1, 2,
p 1 represent all the units modulo p. Moreany integer congruent to
modulo p represents the null element.
In like manner, the rational integers congruent to 1 modulo p denote the
unity element of the set of integral polynomials modulo p. The definition
of an identical congruence stated in Chap. 5 is the basis for the statement
the integers
over,
POLYNOMIALS
that
if
f(x)g(x)
0(mod
187
p) identically, at least one of the polynomials
modulo p. There are, theref(x) and g(x) is identically congruent to

modulo p in this set of polynomials.
fore, no divisors of
A monic or primary polynomial modulo p, a prime, is an integral polynomial whose leading coefficient is congruent to 1 modulo p.
The primary polynomial modulo p in a set of associates of f(x) modulo
is
called the primary associate of f(x) modulo p.
p
A prime polynomial modulo p is an integral polynomial that is not a unit
modulo p and which is divisible modulo p by only its associates and the
units modulo p.
An integral polynomial that is not congruent modulo p to zero, a unit,
or a prime polynomial modulo p is a composite polynomial modulo p.
Example. The integers 1, 2, 3, and 4 represent the unit polynomials
modulo 5. The primary polynomials modulo 5 of the first and second
degree can be reduced modulo 5 to one of the following polynomials:
x
x
x
x
+1
+2
+3
+4
x2
x2
x2
x2
z
x2
x2
x2
x2
2
+l
+2
+3
+4
+x
+x
+x
+x
+z
x2
x2
z2
x2
z2
+1
+2
+3
+4
+ 2x
+ 2x
+ 2z
+ 2x
+ 2z
x2
+
+
+
+
z2
x2
2
a;
+ Sx
+ Sx + 1
+ 3x + 2
+ 3x + 3
+ 3z + 4
Of these the following are prime polynomials modulo
The
+3
+4
+2
x
x
z2
z-fl
2
z2
x2
x
x2
z2
+
+ +
-f-
+ 2^+3
+ 2^ + 4
+ 3z + 4
modulo
ic
z2
z2
z2
+ 4z
+ 4z
+ 4x
+ 4z
+ 4z
+1
+2
+3
+4
5:
2
+ 4x +
associates of this set of primary prime polynomials
sent the incongruent prime polynomials
x2
5 of the
modulo 5 reprefirst and second
degree.
common
divisor modulo p, a prime, of a set of integral polycongruent to zero modulo p, is a common divisor of the
set that is divisible modulo p by every common divisor of the set.
We
refer to the primary associate of a greatest common divisor modulo p as
the greatest common divisor modulo p.
greatest
nomials, not
Theorem
all
and f 2 (x) are integral polynomials, not both

modulo p, a prime, then D(x), the greatest common divisor modulo p of fi(x) and fi(x), exists and there are
integral polynomials gi(x) and gi(x) such that D(x) = fi(x)gi(x) +
11-15.
If fi(x)
identically congruent to zero
f 2 (x)g 2 (x)(mod p).

If there is
any
difference in degree,
assume that the degree of fi(x) is

is evident unless both polyno-
higher than that of f 2 (x), for the theorem

mials are nonconstants.
Then
fi(x)
=f
(x)q!(x)
riO)(mod
p)
188
= 0(mod p) or ri(x) is of lower degree than/2 (x). Repeat

by dividing the remainder into the previous divisor whenever
k.
Vi(x) ^ 0(mod p), where i = 1, 2,
After a finite number of
steps we shall arrive at a remainder which is congruent to
modulo p,
for the degree of the remainder continually decreases.
Thus
and
either r\(x)
this process
/2O)
n(x)
r k - 2 (x)
r k -i(x)
Then every common
=
=
r 2 (x)q 3 (x)
+r
+r
+
r k -i(x)q k (x)
2
3
(x)(mod p)
(x)(mod p)
r k (x)(mod p)
r k (x)q k+1 (x)(mod p)
divisor
modulo p
of fi(x)
and/2 (V)
divides r k (x),
common divisor modulo p of these polynomials.

is a greatest common divisor modulo p offi(x) and
Or).
r k (x) is
r k (x)
s= ri(x)q 2 (x)
successively for the n{x) in terms of fi(x)
nix)
= fi(x)hi(x) + f
and f 2 (x), we
and
Therefore,
By solving
find
(x)h 2 (x)(mod p)
the leading coefficient c of r k (x) is not congruent to 1 modulo p, we

determine d so that cd = 1 (mod p) and multiply each member of the congruence by d, thereby obtaining
If
D(x)
^f^g^x) + f
(x)g 2 (x)(modp)
Consider the problem of determining when the degree of g 2 (x) will be

than the degree of fi(x) and that of gi(x) will be less than the degree
less
oif2 (x).
It will now be easy for the reader to prove the following theorems:
Theorem 11-16. If the integral polynomials fi(x) and f 2 (x) are relatively prime modulo p, a prime, and if fi(x) divides f 2 (x)fz(x) modulo p,
then fi(x) divides f^(x) modulo p.
Theorem 11-17. If p is a prime, the integral polynomials fi(x) and
f2 (x) are relatively prime modulo p if and only if there exist integral polynomials gi(x) and g 2 (x) such that fi(x)gi(x) + f2 (x)g 2 (x) = l(mod p).
Consider the problem of the unicity of #1(2) and g 2 (x).

11-18.
A composite integral polynomial modulo p, a prime,
can be factored into prime polynomials modulo p, and except for the
order of the factors and associated polynomials modulo p the factorization
Theorem
is
unique for the modulus
A Method
p.
Congruence Modulo p, a Prime. If p

modulo p of f(x) = 0(mod p) are among
the solutions of x p x = 0(mod p). Therefore, to solve f(x) = 0(mod p),
find D(x), the greatest common divisor modulo p of f(x) and x p x.
Then the solutions of D(x) = 0(mod p) are the distinct solutions of
Of course,
f(x) = 0(mod p), and their number is the degree of D(x).
11-3.
is
a prime,
all
for Solving a
distinct solutions
POLYNOMIALS
f(x)
0(mod
p)
may have
189
a repeated solution, but the existence of a
r(mod p) can be determined by removing the

factor x r modulo p from fix) and substituting r for x in the quotient.
Example. We can exhibit the usefulness of this device by finding the
2
The application of the
solutions of x 5 + x 3 + x x + 3 = 0(mod 5).
5
x and a; 5 + x z + x 2 x + 3 shows that D(x) = x 2
algorithm to a;
3x + 2 (mod 5) and therefore that the only distinct solutions of the given
congruence are x = 1, x = 2(mod 5). But x b + x z + a; 2 x + 3 =
2
3
2 The congruence z 3 + 3a; 2
3a; + 2)(x + 3a; + Sx + 4) (mod 5).
(a;
+ 3a; + 4 = 0(mod 5) has the solution x = 2 (mod 5), and a; 3 + 3a; 2 +
2
- 2) (mod 5). However, a; 2 - 2 = 0(mod 5) is
3a; + 4 = (a; - 2) (a;
Hence, the solutions of the
satisfied by neither x = 1 nor x = 2 (mod 5).
original congruence are x = 1, x = 2, x = 2 (mod 5).
multiple solution x
EXERCISES
Write the primary prime polynomials modulo 3 of degrees 0, 1, and 2. Then
all the prime polynomials modulo 3 of degree 2 that are incongruent modulo 3.
2. Factor 2x 5 + x 3 + 2z 2 + 2x + 2 into prime factors modulo 3.
2
4
3. Find the solutions of x - 2x - 3 = 0(mod 5).
4. Find the solutions of x 5 4x 3
Sx = 0(mod 7).
5. Find the solutions of x 4 x
1 m 0(mod 7).
4
6. Find the solutions of 2x - 10z - 27 = 0(mod 35).
7. If f(x) is an integral polynomial and p is a prime, develop a method for solving
the congruence /(x) = 0(mod p) by using the derivatives oif(x) with respect to x and
1.
write
certain greatest
common
divisors
modulo
p.
CHAPTER
12
PARTITIONS
12-1.
The Additive Theory
Numbers.
of
Leibnitz (1646-1716) was
mathematicians who paid particular attention to developing the theory concerned with the separation of an integer into all possible
summands selected from a given set, for example, the representation of 4
by 4, 3 + 1, 2 + 2, 2 + 1 + 1, 1 + 1 + 1 + 1, where selections are
made from 1, 2, 3, and 4. To Euler, however, is due a large part of the
This additive theory of numbers is a difficult subject. We
basic theory.
The student can refer to the
shall develop only the fundamental ideas.
work of G. H. Hardy, one of the modern experts in this field, for an
extensive treatment of this topic.
If from any set of positive integers a, where i = 1, 2, 3, . . . , finite
among
the
or infinite,
first
we
select
numbers
representation, whether or not

integer n.
of
in
it
so that
Thus the representation
which the selection
this chapter
we
is
n = ax
of
n as a sum
made from
a2
1,
2,
q or
is
In
addends from the
the set of positive primes.
shall confine ourselves to selections of
set of positive integers
a m the
a partition of the
of primes is a partition
contains repetitions,
from
all
the positive integers,
considering both the case where repetitions are permitted and where they
are prohibited,
the order of the
summands
in each situation
being
irrelevant.
We shall represent the number of partitions of n containing m selections

with repetitions allowed by P(n, m, <g).
shall use the symbol Q(n, m, <q).
If the
representations are to have at most m parts selected without repetition
from 1, 2,
q, their number will be expressed by Q(n, <m, <g).
from the integers 1,2,

If repetitions
When
are prohibited,
we
made from the set of all positive integers, we

employ P(n, m) and Q(n, m) for the number of partitions of n, with
repetitions and without repetitions, respectively, that have exactly m
the selections are
shall
parts.
If
the selections are unrestricted as to the
number
of parts,
we
<q) to mean the number of partitions of n, with

repetitions permitted, into any number of parts all of which are less than
or equal to q, and Q(n, U) to mean the number of partitions of n into any
number of parts selected without repetition from the positive integers.
shall write P(n,
U,
190
PARTITIONS
If
the selection of the unrestricted
1,
2,
Q(n, U,
and
191
number of parts is to be from among

q, we shall write P(n, U, q) and
always to include
is
q).
remaining n
Thus when n
separate n into m parts with

one unit, and then distribute the
To
12-2. Partitions with Repetitions.

repetitions let each of the
m parts have
m units to
> m and the
one part, to two parts,
from the positive
selections are
to
parts.
we
integers,
have
= P(n - m,
P(n, m)
1)
+ P(n -
m, 2)
+
+
P(n
m, m)
But then
P(n
1,
m-
= P(n -
1)
m,
1)
P(n
m,
2)
+
+
P(n
m,
1)
Hence,
= P(n
P(n, m)
1,
1)
P(n
m, m)
and we have proved:
Theorem
equal to
When
12-1.
integers, the
P(n
the addends are selected from the positive
number
of partitions of
P(n
the number
1,
To
Example.
find P(7, 3)
find
P(6, 2)
recursion formula of
1)
P(4, 3)
P(3, 2)
=
=
=
=
P(5,
1)
1)
P(3, 2)
P(2, 1)
into
parts with repetitions
of partitions of 7 into three parts,
12-1,
P(3,
is
m, m).
Repeating the application
P(4, 3).
Theorem
P(6, 2)
P(4, 2)
we
we
of the
obtain
+ P(4,
+ P(2,
+ P(l,
+ P(l,
2)
2)
3)
2)
=
=
=
=
1
1
+
+
P(4, 2)
1
P(3, 2)
1
Hence,
P(7, 3)
= P(n 1, m 1).
and P(n m, m) = 0.
Corollary 2.
P(2n, 2) = n, and P(2n + 1, 2) = n.
The partitions of an even integer 2n into two parts are
Corollary
If
m>
If
1.
n/
2,
m > n/2,
m < m
P(n, m)
(2w
2)
(2n
n)
an odd integer 2n
+
+
-f-
The
are
partitions of
(2n
1)
greater than
into
two parts
192
+
+
2n
(n
By making
1,
(2n
1)
1)
use of the formulas P(n, n)
1,
P(n, n
P(n,
1)
P(n - 3, 2)
+ P(n - 3,
3)
1)
1,
as well as
P(n, 2)
P(w, 3)
= P(n = P(w -
=
P(n, k)
1, 1)
1,
2)
+ P(n -
+ P(n - 2, 2) =
+ P(n - 3, 3) =
3, 2)
= P(n - 1, k = 1 + P(n - *,
P(n
+ P(w + P(w -
1)
2)
+ P( -
P(n
2, 2)
2, 1)
3, 3)
ft)
fc,
*, 3)
P(n
we can set up a table of the number of partitions of n into

from the positive integers with repetitions permitted.
k, k)
m parts selected
Values of n
m, the
number
of parts
10
12
11
15
10
13
11
2
3
5
6
7
8
9
10
11
12
P(n 1) = 1. The formulas P(n, n) 1

account for the two diagonals of l's. Passing to
the second row, to find P(3, 2), we merely add the numbers in the column
= 2,
under
To find P(4, 2), add the numbers under
etc.
To write the third row, sum the numbers under 4 3 = 1 for
P(4, 3), the numbers under 5 3 = 2 for P(5, 3), the numbers in the
first three rows under 6 3 = 3 for P(6, 3), the numbers in the first three
rows under 7 3 = 4 for P(7, 3), etc.
From the way the table has been set up it is evident that we can find
P(n, <m) by merely looking up P(n + m, m).
The
row
first
and P(n, n
of the table uses
1)
4-2
3-2=1.
12-3.
exactly
Diagrams
of Partitions.
parts selected from
If
1, 2,
we
.
set
.
all the partitions of n into

with repetitions permitted,
up
q,
193
PARTITIONS
and always employ at

partitions,
least
one
and then remove one
q,
from each
of the
partitions with repetitions permitted of
we obviously have the
P(n, m, q) =
m 1 parts selected from 1,2,
q; that
_ q m i <g). If repetitions are prohibited, Q(n, m, q) =
q, m 1, <q 1).
We can diagram a partition of n into m parts that include q as a greatest
n
P( n
Q(n
q into
is,
integer in the following
thus graphically exhibit the statement
way and
made above:
1
1
...
... 1
(q units)
(m rows)
Moreover,
we read
if
the diagram by columns,
we have a
partition of
Such partitions are said to be

into q parts of which the greatest is m.
The diagram shows that the following statement is valid:
conjugate.
There are as many partitions of n into m parts selected
and always including q as there are partitions of n into
parts such that one is m and the others are less than or equal to m, repe-
Theorem
from
1, 2,
12-2.
.
q
titions being permitted in each case.
Similarly the diagrams
Theorem
of
12-3.
n into at most
parts
repetitions permitted of
P(n,
<m) =
P(n,
<m) =
show the following theorem, due to Euler:
The number
n
is
of partitions
the same as the
into parts
P(n, U, <m).
Furthermore, by subtracting P(n,
The diagrams also make it clear

Theorem 12-4. The number of
n into
repetitions
which do not exceed m; that
<m
1)
P(n, U,
<m
1)
is,
from
P(n, U, <m), we have:

P(n, m) = P(n, U, m).
Corollary.
of
with repetitions permitted

number of partitions with
that:
partitions with repetitions permitted
more parts is the same as the number of partitions with

permitted of n into parts containing an element that is greater
or
than or equal to m.
12-4. Generating Functions for the Number of Partitions.
Consider
the addends that are summed to determine the exponents of x in the
product (1
z 2 )(l + x*) = 1
x 1+2 + x 3 + x 1+ *
x)(l
x + x2
i+2+3
j j s evident that these exponents are the results arising
a-2+3 _[_ C
from all possible selections of one, two, and three distinct summands from
the set 1, 2, 3.
Consequently, the coefficient 2 of x 3 gives the number of
ways 3 can be produced by adding together distinct integers selected from
Similarly the coefficient of x n in the expansion of (1 +
1, 2, and 3.
x)(l -f- x 2 )
(1 + x q ) is the number of partitions of n into dis-
<
194
from the
tinct integers
set
Q(n, U, <q).
Now take the product
2,
1,
+ zx){\ +
(1
and
zx 2 )
therefore, the value of
is,
(1
zx q ).
The
pres-
ence of the z in each factor enables us to count the number of addends

used to produce the exponent of x. Thus the coefficient of z m x n is the
number of ways in which n can be represented by exactly m distinct
addends from among

mine Q(n, m, <q).
The
1, 2, 3,
effect of multiplying
~
z qiq 1)/2 is
(1
therefore, enables us to deter-
It,
q.
zx){\
zx 2 )
(1
zx q )
to collect the terms representing the
by 1 +
number
ways n can be produced by one, two,

m distinct addends, for
m ~ l to produce z mx n z 2x n by z m ~ 2 etc. Thus the
is multiplied by z
coefficient of z mx n is the number of partitions of n that can be obtained by
choosing at most m distinct integers from the set 1, 2,
It is the
, q.
of
zx n
<m,
value of Q(n,
<q).
The fact that the series 1
xm
x 2m
developed from the
m
quotient 1/(1
x ), is absolutely convergent for
< x < 1 enabled
Kronecker* to prove that the coefficients of the expansion of the generating function 1/(1 x)(l x 2 )(l x z )
give the number of partitions of n with repetitions permitted.
The discovery of this theorem is
due to Euler. We shall merely indicate the truth of the statement by the
),
first
,
To produce any integer n, we need at most n addends,
following argument:
and hence but the
of the factors (1
+x +x
n
(1
2n
-f-
+x+x +
2
determined by
),
(1
x2
the generating
term of the (n + l)st factor in the product of the

would merely reproduce the product of the first n
factors, and the next term would add n + 1 to each of the exponents of x
already produced so that the resulting exponents would exceed n. It is
evident also that all succeeding exponents so derived would exceed n.
Moreover, any term developed from the product of the first n factors is
the result of selecting exactly one term from each of these factors. The
The
function.
first
first
1 factors
choice can be represented as the selection of one of each of the factors

x a x 2b x 3c
x nk where the values of the integers a, 26, Be, . . . , rik
are among 0, 1,
n, and a is to be interpreted as the sum of a units,
2b as the sum of b 2 s, 3c as the sum of c 3's, etc. Each time the sum of
,
the exponents,
a, 26, 3c,
...
the coefficient of x n gives the
nk,
is n,
number
we have
a partition of n.
of partitions of
n with
Thus
repetitions
permitted.
Let us illustrate the use of the generating function by finding the partiof 5.
We need but the factors (1 x + x 2 x 3 + x 4 x b ),
2
4
x
x
x 3 ), (1
x 4 ), and (1
z 5 ). The product is to be
+
(1
), (1
tions
*L. E. Dickson, " History of the Theory of Numbers," Vol.
2, p. 104.
PARTITIONS
195
interpreted in the form

(1
x l+l
x l+l+1
x 1+1+1+1
a;i+i+i+i+i)(l
+
+
+
+
1+1+1
1+2
+
+
+
+
x 1+1+1+1
z 1+1 + 2
partitions of
1, 2, 3, 4,
_|_
^ 1+1+ l+i+l
^1+1+1+2
x 2+ 2
_J_
x l+2+2
1+3
_|_
^1+1+3
+
The
Z 5)
is
1+1
(1
Then the expansion
+ x + x 2+2
+ x )(l + z )(l +
+
+
+
x*
as well as
5,
X 2+ 3
Z 1+4
5
+
+
+
+
+
+
+
are thereby enumerated,
and the
coefficient 7 of x 5 is the value of P(5, C/).
enumerates
Similarly we can see that 1/(1 x)(l x 3 )(l x h )
the partitions of n into odd integers with repetitions permitted and that
1/(1
x 2 )(l
x A )(l
x 6)
does the same
when the
parts are even.
EXERCISES
1.
2.
3.
4.
Show
Show
that P(n,
<m,
q)
P(n,
<
q,
in).
that P(n, U, <q) = P(n

q, q).
Find the number of partitions of n into parts selected from 1, 2, 2 2 2 3
Show that the number of partitions of n in terms of odd integers with repetitions
equal to Q(n, U).

5. Write a generating function which will enumerate the partitions of n into parts
that are odd and unequal.
6. Show that x r /(l x 2 )(l x 4 )
(1 x 2q) enumerates the partitions of
n r into even parts that do not exceed 2q with repetitions permitted. Show also
that when n r is even, the same function enumerates the partitions of (n r)/2
into parts not larger than
with repetitions.
7. Find a method for listing all the partitions of n into m parts by starting with
1 units and the integer n
1.
is
m+
BIBLIOGRAPHY
P., "Die Lehre von der Kreistheilung," B. G. Teubner, Leipzig, 1921.
"Grundlehren der neueren Zahlentheorie," Walter De Gruyter & Com-
Bachmann,
,
pany, Berlin, 1931.

Cahen, E., "The*orie des nombres," Gauthier-Villars & Cie, Paris, 1900.
Cajori, F., "A History of Mathematics,' Macmillan & Co., Ltd., London, 1906.
Carmichael, R. D., "The Theory of Numbers," John Wiley & Sons, Inc., New
'
York, 1914.
Chrystal, G., "Algebra," A.
&
C. Black, Ltd., London, Vol.
I,
1931, Vol. II,
1932.
Dickson, L. E., "History of the Theory of Numbers," Carnegie Institution of

Washington, Washington, D.C., 1920.
,
"Introduction to the Theory of Numbers," University of Chicago Press,
Chicago, 1929.
,
"Studies in the Theory of Numbers," University of Chicago Press,
Chicago, 1930.
,
"Modern Elementary Theory
of
Numbers," University
of
Chicago
Press, Chicago, 1939.
Hancock, H., "Foundations of the Theory of Algebraic Numbers," The Macmillan Company, New York, Vol. I, 1931, Vol. II, 1932.
Hardy, G. H., "Some Famous Problems of the Theory of Numbers," Oxford
University Press, New York, 1920.
and E. M. Wright, "The Theory of Numbers," Oxford University Press,
New York, 1938.
Hecke, E., "Theorie der algebraischen Zahlen," Akademische Verlagsgesellschaft m.b.H., Leipzig, 1923.
Kraitchik, M., "Theorie des nombres," Gauthier-Villars
&
Cie, Paris, Vol.
I,
1922, Vol. II, 1926.
Landau, E., "Vorlesungen iiber Zahlentheorie," S. Hirzel, Leipzig, 1927.

Mathews, G. B., "Theory of Numbers," G. E. Stechert & Company, New York,
1927.
Nagell, T., "Introduction to
Number Theory," John Wiley &
Sons, Inc.,
New
York, 1951.
Ore, Oystein,
"Number Theory and
Its History,"
New
McGraw-Hill Book Company,
York, 1948.
Reid, L. W., "The Elements of the Theory of Algebraic Numbers," The Macmillan Company, New York, 1910.
Smith, D. E., "History of Mathematics," Ginn & Company, Boston, Vol. I,
Inc.,
1923, Vol. II, 1925.

197
198
"A Source Book

New York, 1929.
,
Stewart, N. M.,
in
"Theory
of
Mathematics," McGraw-Hill Book Company,
Numbers," The Macmillan Company,
New
Inc.,
York,
1952.
J. V., and M. A. Heaslet, "Elementary Number Theory," McGrawBook Company, Inc., New York, 1939.
Uspensky,
Hill
Wright, H. N., "First Course in the Theory of Numbers," John Wiley

Inc.,
New
York, 1939.
&
Sons,
INDEX
Chrystal, G., 197
Abacus, 39
Absolute value, 7
Addition, 3, 54, 177
Additive theory of numbers, 190
Ahmes papyrus, 1
Algebra, 1, 16, 67
Algorithm, 31, 180
Amicable integers, 37
Archibald, R. C, 37n.
Archimedes, 9
principle of, 9, 12, 28
an integral right
Arithmetic function, 58
Area
of
Closure,
3,
Commutative
Composite,
179, 187
Conditional congruence, 67
Congruence, 53
conditional, 67
triangle,
168
degree
of,
68
identical, 66
properties
solution
of,
54-55
67
of, 66,
Congruences,
integer, 8
of a polynomial, 179
Associative law, for addition,
177
25
9,
Composite polynomial,
"Arithmetica," 169
an
3,
Complementary factors, 7
Complex numbers, 47, 182
Arithmetica, 2
Associates, of
law, for addition,
for multiplication, 3, 177
54jf.
equivalent, 68-70, 91
3,
linear, 70-71,
177
89
simultaneous, 79-83
for multiplication, 3, 177
operations on, 68
quadratic, 134-135
Bachet,
C, 169
Bachmann,
solution
P., 145n.,
of, 66,
67
Congruent
197
integers, 53ff.
Consecutive integers, 3, 6
Base, 38, 125
Belonging to an exponent, 100

Bernhard, H. A., 36n.
Bibliography, 197-198
Binary scale, 39
Bocher, M., 67n.
Consecutive squares, 51
Crelle, A. L., 100, 106
Criterion for solvability, of ax
&(mod m), 71
of x n = c(mod m), 129
Cunningham, A., 127
Brauer, A., 36n., 50n.
Brouncker, W., 172

Buck, R. C, 46ri.
by induction, 3, 4
Denary system of notation, 39
Definition
Cahen, R, 197
Derivative of f(x) with respect to
Cajori, F., 2n., 197
Descartes, R.,
Calculating machines, 39
Cancellation law, 3, 177
Dickson, L. E., 30n., 36, 88n.,

159n., 194n., 197
Carmichael, R.
Cauchy,
C,
A., 145,
99, 105, 112,
197
168
Chinese, 88
Chinese remainder theorem, 79-80
2,
2, 39-40
Diophantine equations, 16
linear, 17, 19-23
simultaneous, 23-24
Digits,
199
x, 77,
86
88
127/i.,
ELEMENTAKY THEORY OF NUMBERS
200
Diophantine equations, solutions

17, 19,
of,
16,
21
Fermat, P., 30, 37, 88,

last theorem, 168
method of descent,
Fermat numbers, 32
Diophantus, 1, 16, 88
"Arithmetica," 169
Dirichlet, L., 27, 172
Field, 177
Discrete set, 3
" Disquisitiones arithmeticae," 136
Form, 14
Finite induction,
degree
Distribution of primes, 50, 51

Distributive law, 3, 177
Division,
30,
168
10-11
14
linear, 14r-16
quadratic, 46
Franqui, B., 37n.
178
7,
of,
4,
89, 105
by 3, 55
by 8, 9, and 11, 56
modulo ra, 55, 57, 72
Fundamental theorem
of arithmetic, 29
Divisor, 7
common, 7
Garcia, M., 37n.
greatest, 7, 14-15, 18, 62, 179, 180,

185, 187, 188
an
Divisors, of
number
integer,
34
34-35
sum of, 34-35
of zero, 57, 178
Egyptians,
Goldbach, C., 51
Goldberg, B., 127
Greater than, 3
Greatest common divisor,
of,
1,
14-15, 18,
188
used for solving congruences, 188
Greeks, 1, 16
Gupta, H., 37n.
39
9,
7,
62, 179, 180, 185, 187,
Eisenstein, R. G., 144
"Elements,"
Gauss, C., 50, 53, 107, 136

law of quadratic reciprocity, 145-147
lemma of, 141
139
Equals, 4
Equivalent congruences, 68-70, 91

Eratosthenes, 25
Erdos, P., 27, 50
Erlerus, H. G., 103
Euclid, 9, 36
"Elements," 9, 139
formula for perfect numbers, 36
theorem, 9, 14
Euclidean algorithm, 31
Hancock, H., 197

Hardy, G. H., 158n., 190, 197
Heaslet, M., 198
Hecke, E., 197

Hilbert, D., 158
Hindu- Arabic system, 38
Hindus, 1, 172
Euler, L., 36, 46, 58, 159

criterion for solvability of x n sb
c(mod m),
Ideals, 168
129, 136
Identical congruence, 66
identity, 164
<t>
Identically equal polynomials, 67, 178
function, 58-60, 96, 106
Even
Identity element, 178
integer, 10
Exponent, of a prime contained in

40-43
to which an integer belongs, 100
178
common, 7
greatest, 7
Factor,
7,
(See also Divisor)

of a polynomial, 74, 178
Factorization, 28-30, 183-185
n!,
Incongruent integers, 53
Index of n, 125
Indicator, 58
Indices, 125-131
used in solving congruences, 127-129
Advanced Study, 40
Institute for
Integer, associate of, 8

divisors of, 34-35
expressed as a sum, 190

of 4 squares, 158,
in
164-166
terms of a base, 37
INDEX
Mathematical induction,
Mathews, G., 197
Integer, indicator of, 58
unique factorization
of,
201
28-29
4,
10-11
Meissel, E., 49
Integers, algebraic, 8n.
amicable, 37
Mersenne, M., 37
belonging to a divisor of X(2 n ) modulo

2, 114-115
Method, for finding the exponent to

which a belongs, 100
belonging to an exponent modulo

103, 106
congruent, 53
consecutive,
3, 6,
p,
for finding a primitive root of p, 107
for finding a quadratic residue of p, 138
of infinite descent, 30, 168

for solving congruences
51
W.
by
indices, 127
H., 46n.
even, 10
Mills,
negative, 5
Mobius function, 93
Mobius inversion formula, 95-96
odd, 10
Integral domain, 177, 186
Modul, 16
Modulo m, 53
Modulus, 53
Multinomial theorem, 44
Integral logarithm of
Multiple,
positive, 3
rational,
square, 51
x,
50
7,
common, 8
Inverse, 5
Inversion formula, 95-96

Irreducible polynomials, 183
least, 8,
33
modulo m, 72
Multiplication, 3, 54, 177
Jacobi,
C,
Multiplicative function, 58, 94

Multiplicity of solutions of
127, 145
0(mod
Jacobi symbol, 152
p), 76,
f(x)
77
Multiply perfect numbers, 37

Kraitchik, M., 127, 197
Kronecker,
L., 50, 145,
Kummer,
E., 145,
Lagrange,
J.,
194
Nagell, T., 197
Nonresidue,^quadratic, 135
Null element, 8, 186
168
Null polynomial, 178
Lambda
75, 136,
158
function, 99, 100, 112
Number,
to d
Lattice points, 144
to d
Legendre symbol, 139

Lehmer, D. H., 37, 88n.
Lehmer, D. N., 26n.

Leibnitz, G., 190
rule for derivatives of a product, 78
Less than, 3
Linear congruences, 70-71, 79-83, 89

Liouville, J., 145, 158
Littlewood,
J.,
159
Logarithm, 50, 126

Lucas, E., 99, 105
an
integer, 34
d modulo
p,
102
Lame, G., 168

Landau, E., 197
Lawther, H. P., 113
Least common multiple, 8, 33
Legendre, A., 47, 50, 136
of divisors of
of integers that belong, to
modulo p n 117
modulo 2p n 118
,
of integers that belong to a divisor of
X(w) modulo m, 118-124

not exceeding \m\,
and prime to m, 58
and prime to factors of m, 60-62
of positive primes not exceeding n, 47
of primitive roots of p, 106
of solutions, of a linear congruence, 71
of /(re) Ô(modp), 75
of x d = l(modp), 90, 103
of/(x) = 0(modp"), 86
of f(x) m 0(mod m), 84-85
of x 2 = a(mod 2 n ), 156-157
of a; 2 " = l(mod2), 116
of positive integers
Numbers,
202
Numbers, additive theory

complex, 47, 182
Fermat, 32
of the form, 2*> 2 2 " + 1, 32, 46
Mersenne, 37
natural, 1, 2
perfect, 36-37
of,
Polynomials, domain
190
178
of,
divisor of, 179, 180,
185, 187
37
1,
identically equal, 67, 178

relatively prime, 181-183
Position in writing numbers, 1

Postulates,
13
3,
Power residues
multiply, 37
of a
modulo m, 101
Prime, 8
Mersenne, 37
rational, 2, 177, 183

real,
common
greatest
relatively (to each other), 9
183
signed, 2
in pairs (each to each), 9
test for, 25, 93, 104
Prime
Odd
integer, 10
One-to-one reciprocal correspondence,
Ordering relation, 3
Ore, O., SOn., 197
factor, of a composite,
Prime number theorem, 50

Prime-representing function, 46-47
Primes, contained in
distribution
notation
number
for,
for,
of,
of, 50,
n\,
41-43
51
factorization into, 28-29
Partitions, 190
diagrams
25
40
of n!,
1
formula
of form, 2 n
192
190-191
for,
generating functions
191-192
193-195
for,
2 2"
An
An
1,
27, 141
1,
30, 58, 141,
number
table
not exceeding
of,
of,
n,
47-50
52
Primitive X roots, 112, 118-124

Primitive polynomial, 183
Primitive roots, of m, 105
of
n
,
108-110
of 2p
111
and quadratic
integral, 14
159-164
many, 26-27
form An - 1, 27
Peano's postulates, 13
Pellian equation, 172-176
Perfect numbers, 36-37
composite, 179, 187

factor of, 74
37
infinitely
multiply, 37
Phi function, 58, 88, 96, 100
Polynomial, 14
with coefficients in a field, 178-183
28,
46, 151
Pascal, B., 88
of
1,
1,
residues, 138
Primitive solutions, 167
not a prime-representing function, 46

irreducible, 183
Pythagorean
triangle, 167,
168
Pythagoreans, 10
monic, 179, 184, 187

null, 178
primary associate
of,
183, 187
prime, 179, 187

primitive, 183
Quadratic forms, 46
reducible, 183
with
with
183-185
prime modulus,
rational coefficients,
respect
Quadratic character, 135, 140

Quadratic congruence, 134-135
to
186-189
unique factorization
unit, 178
of,
184-185
vanishes identically, 178

Polynomials, division of, 72-73, 178
quotient in, modulo p, 186
remainder in, modulo p, 186
Quadratic nonresidue, 135
Quadratic reciprocity law, 144-147

Quadratic residue, 135
of p, test for, 137
Quadratic residues and primitive roots,

138
Quotient,
of
7,
10
two polynomials, 178

modulo p, 186
203
INDEX
Sussman, B., 118
Symbol, definition
Radix, 38
Rational integers, 2, 3, 18, 183
Rational numbers, 177, 183
Real numbers, 183
Reciprocity law, 144-147
si,
66
a\b,
Ep {m),
Reducible polynomials, 183

Reflexive property, 4, 55
Regular function, 58
Reid, L., 197
Relatively prime, 9, 179
g(k),
(a, 6)
41
158
d,
40
139
in pairs, 9
Remainder in division, of a by
of g{x) by f(x), 72, 178
b,
10
(see
A(m), 99
93
58
tt(x), 47
<r(m), 34
r(m), 34
fi(m),
Residue classes, 53, 56-57

Residue system, 54, 62-65
complete, 54
least numerical values, 63
reduced, 54
Riemann, G., 50
Roots, primitive
53
of,
<t>(m),
Symmetric property,
Primitive roots)
4,
55
Table, of indices for prime 193, 132-133

of primes, 52
Taylor's theorem, 77, 86

Tchebysheff, P., 50
Sarrus, F., 88
Scale of notation, 37-40
Telephone cables, 113
American, 40n.
Selberg, A., 27, 50
Sequence, 4
Shapiro, H. N., 27
Sieve of Eratosthenes, 25
Simultaneous congruences, 79-83
Single-valued function, 58
Smith, D. E., 197
Smith, H. J. S., 21n.
Transitivity, 4, 55
Scientific
Solution, of f(x)
= 0(mod
of
= b (mod
of x n
+ by = n,
+ yn = zn
ax
of ax
p),
85-86
Uhler, H.
Unique
S.,
37
factorization, 28-30, 183-185
8, 178, 186
Unit polynomial, 178, 179
Unity, 177
Unity element, 177
Uspensky, J., 198
Unit,
m), 89
19-20
168
Vandiver, H. S., 168

Vinogradov, I., 51
Solutions, of Diophantine equations, 16,

17, 19,
J., 172
Waring, E., 158
Wertheim, G., 127
Wilson's theorem, 92-93, 136
generalization of, 170-172
Wright, E. M., 46n., 197
Wright, H. N., 198
Wallis,
21
= 0(mod
p), 75, 91, 92, 188

multiplicity of, 76-78
2
2
of x 2
y = z 167
of x 2 - by 2 = 1, 175
of f(x)
Squares, integers that are, 12
primes between, 51
Stewart, N., 198
Substitution, 4, 55
Subtraction, 3, 5, 54
of the divisors of an integer, 34-35
over the divisors of an integer, 60, 62
of 2 squares, 159-164
of 4 squares, 158, 164-166
Sum,
Zassenhaus, H., 27
Zeller, C., 145
Zero,
1,
4-5, 178
divisors of, 57, 178
no degree, 178
as null element, 8

Lingua

Benvenuto in Scribd!

Elementary Theory of Numbers, Harriet Griffin

Caricato da

Informazioni sul documento

Descrizione:

Data di caricamento

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Descrizione:

Copyright:

Formati disponibili

Elementary Theory of Numbers, Harriet Griffin

Caricato da

Descrizione:

Copyright:

Formati disponibili

Titoli correlati

ELEMENTARY THEORY OF NUMBERS

PURE AND APPLIED MATHEMATICS

ahlfors Complex Analysis

Partial Differential Equations

golomb and shanks Elements of Ordinary Differential Equations

Elements of Pure and Applied Mathematics

Distribution Theory and Transform Analysis

McGRAW-HILL BOOK COMPANY,

ELEMENTARY THEORY OF NUMBERS

an ample basis for

Consequently, care has been taken

to develop the student's power of inspection,

him and induce him to pursue his inquiries further.

by their lively reactions the parts to be retained,

It is apparent that there is more material in the text than can be

of primes not exceeding the positive integer n, it is suggested

5-7, 5-8, 5-9,

THE FUNDAMENTAL LAWS

The development of the integers

THE LINEAR DIOPHANTINE EQUATION

Simultaneous linear Diophantine equations

Fermat's method of infinite descent

The highest power of a prime that

Basic properties of congruences

3-10. Scales of notation

Residue systems modulo

The number and multiplicity of

The Mobius function

Mobius inversion formula

8-1. Indices for

The area of an integral right triangle

SOME FAMOUS PROBLEMS

general quadratic congruence

The Waring problem

THE THEOREMS OF FERMAT AND WILSON AND

the solutions of a congruence

Other simultaneous linear congruences

THE SOLUTION OF CONGRUENCES

and conditional congruences

Polynomials with respect to a prime modulus

The additive theory

12-2. Partitions with repetitions

Generating functions for the number of partitions

THE FUNDAMENTAL LAWS

rational integers are the

many centuries of development of the concept of number.

abstracting the idea of

basic notion in the concept of

that of one-to-one correspondence.

one element of B, while each element of B is thereby related to a single

ELEMENTARY THEORY OF NUMBERS

of the idea of number, and it remained for Descartes (1596-1650) to grasp

The signed numbers have two

idea of quantity with that of quality.

theory of numproperly concerned with the study of only the natural

because, besides being of interest in

set has properties not possessed