Jones TheTheoryOfNumbers Text

5T/I.
xroti
THE HARRIS COLLEGE

STREET,
CORPORATION
PRESTON.
All
not
the College Library
Books must be Returned to
below.
shown
date
last
the
than
later
MOV
1974
Jtm.
&
The Theory of Numbers
RINEHART BOOKS IN MATHEMATICS

Ambrose and Morris Lazerowitz
Fundamentals of Symbolic Logic
Ross A. Beaumont and Richard W. Ball
Introduction to Modern Algebra and Matrix Theory
Alice
R. Britton
Calculus
Jack R. Britton
Jack
and L Clifton Snively

Algebra for College Students, revised
College Algebra
Intermediate Algebra, revised
Howard Eves
An Introduction to the History of Mathematics
Howard Eves and Carroll V. Newsom
An Introduction to the Foundations and Fundamental
Concepts of
Mathematics
Casper Goffman
Real Functions
I
$\\.\
Jo J
UL1
i
Richard E. Johnson, Neal H. McCoy, and Anne F. O'Neill

Fundamentals of College Mathematics
Burton W. Jones
Lucien B. Kinney and C. Richard Purdy
Teaching Mathematics in the Secondary School
Harold D. Larsen
liineliart Mathematical Tables, Formulas, and Curves, enlarged edition
Rinehart Mathematical Tables
Horace C. Levinson
The Science of Chance
Neal H. McCoy and Richard E. Johnson
Analytic Geometry
Kenneth
S. Miller
Engineering Mathematics
John
T.
Moore
Fundamental Principles of Mathematics

William K. Morrill
Plane Trigonometry, revised
John A. Northcott
Mathematics of Finance
Plane and Spherical Trigonometry, revised
Lewis M. Reagan, Ellis R. Ott, and Daniel T. Sigley
College Algebra, revised
Paul R. Rider and Carl H. Fischer

Mathematics of Investment
Robert L Swain
Understanding Arithmetic
Gerhard Tintner
Mathematics and Statistics for Economists
Walter
W. Varner
Computing with Desk Calculators
The Theory
of Numbers
BURTON
W.
JONES*
Professor of Mathematics,
The University
of Colorado
and Winston
New York
Holt, Rinehart
CONSTABLE & CO LTD

10-12
ORANGE
STREET,
LONDON
W.C.2
Preface
Though mathematicians
in their publications like to present their subject
were a deductive science, they
will readily
admit that back
postulates and theorems are methods
of discovery
which are largely induc-
made but
are constructed so that
as
if it
Postulates are not arbitrarily
tive.
certain desirable consequences will follow.
of their
Proofs validate conjectures
already made on the basis of experience and experiment. In fact, disguess on the basis
covery of correct results often follows the sequence:
of experiment,
an attempt at proof which
fails
but suggests a modified
which leads to another attempt at proof, and so on.

charms of mathematics is the pervading spirit of discovery.
guess,
these things in
mind that
this
make
of the
was with
book was written.
made
to encourage and equip the
discoveries for himself.
For the most part, reasons
Throughout, a definite
student to
One
It
effort is
for postulates are developed before they are formulated. The statement
and proof of some of the more important theorems are left to the student
March, 1961
Problems are presented and a
after preparation in similar situations.
theory developed to answer the question raised, beginning especially
24711-0115
in the second chapter.

First, this
Copyright, , 1955, by Burton W. Jones

Printed in the United States of America
It
All Rights Reserved
and
Library of Congress Catalog Card
Number 55-6187
:
is,
rather,
book makes no attempt to take the place of the teacher.

his instrument. For instance, at the beginning, postulates
definitions are stated for the natural
that the class will begin rather

definitions and,
tially
on
numbers.
by formulating
its
It is to
own
reference to the text, will find that its
equivalent to those printed.
a method are apparent.
The
text
The
is
limitations of a
be expected
postulates
own
and
are essen-
book
for
such
not supposed to be "read like a
One recommended method is for the student to conceal the proof

a theorem and try to give one himself; failing that, he can reveal the
novel."
of
vu
Preface
viii
first
way
in
two
lines
and try to proceed from
of laying
some
detail at first
but
there.
An
a firm foundation.
less so later
This
attempt
on
in the
is
is
a slow, but excellent,
made
to give material
hope that by that time the
student will have acquired greater facility
In any case, the pace should be very slow throughout the book.
author found, for instance, that the
Most
class periods.
of the
first
book has been
The
Contents
chapter required about twenty

tried out in a classroom at
one
time or another.
first three chapters of the book the author partly had
mind those who are planning to teach in secondary and elementary
schools.
They need discussion of the number system which, after all,
forms the basis for algebra as well as the manipulations of numbers them-
In writing the
in
selves.
wx
Preface
theory contains countless interesting facts, which can
Number
be used to enliven and
clarify
elementary teaching of mathematics.
1 The Development
of the
Number System
For instance, repeating decimals can well be used when the elementary
comes face to face with division by a decimal fraction.
Equations with integer solutions arise from many puzzle problems as well
as from the desire of a teacher to make her problems "come out right."
1.
Introduction
2.
Fundamental Properties
three chapters especially are rich in material useful
4.
Cancellation Properties
5.
Critique
6.
Zero, Subtraction
10
7.
Negative Integers
11
student
first
In short, the
first
for the secondary
This book also gives an adequate introduction to the theory of numbers

and its methods for those students majoring in mathematics or beginning
graduate work.
may be
When
much
of the first chapter
9.
the gap between problem-solving mathematics and mathematics which
10.
develops a theory.
11.
If
taken slowly,
Careful reading of the
Aubrey
J.
Kempner has
it
manuscript by Dr. Carroll V.
resulted in
Newsom and
2
6
7
7
8. Divisibility
should be of assistance in bridging
omitted.
Natural Numbers
3. Inequalities
and elementary school teacher.
used for such students
of
Dr.
12.
28
The Decimal Notation

Rational Numbers
Decimals and Real Numbers
Complex Numbers
27
SI
36
many improvements which the author
acknowledges with deep gratitude.
numbers seems to have a fascination for the layman

and mathematician alike. It is hoped that this book may add to the
The theory
multitude of
of
its disciples.
Zk
Repeating Decimals and Congruences

1.
Repeating Decimals
2.
Number
Burton W. Jones
The University
January, 1955
of Colorado
of
Terms
88
in the
Repetend
of the
Repeating
Decimal
40
3.
Congruence Notation
4.
Properties of Congruences
42
5.
Congruential Equations
4&
ix
Contents
Contents
6.
The Euler
7.
Factorization by
8.
Further Properties of Repeating Decimals
61
9.
Properties of the Euler Phi-function
52
1.
2.
Phi-function
10. Multiplicative
Means
43
of Euler's
Theorem
Functions
Theorem and an Allied Result

Perfect Numbers
Fermat Numbers
The Chinese Remainder Theorem
11. Wilson's
12.
13.
14.
60
O*
Nonlinear Congruences
Introduction
105
106
65
3.
The Remainder Theorem

The Number of Roots of a Congruence
56
4.
Power Residues
109
68
5.
Primitive Roots
60
6.
Primitive Roots
O*
Diophantine Equations
Introduction
2.
A Method
in
3.
of Solution of a Linear
Diophantine Equation
Two Unknowns
68
Linear Diophantine Equations in
More Than Two Un66
4.
Quadratic Diophantine Equations
69
5.
Pythagorean Numbers
72
6.
Other Diophantine Equations
78
Continued Fractions
1.
The Fibonacci Sequence
2.
Properties of Fibonacci-like Sequences
79
3.
Simple Continued Fractions
82
4.
Computation
of the
76
Convergents of a Continued Frac-
tion
5.
Continued-Fraction Expansions
84
86
6.
Closeness of Approximation
89
7.
Expansions of Quadratic Surds

Pure Periodicity and Reduced Quadratic Surds
The Pell Equation
92
8.
9.
10.
Further Information about the Pell Equation
96
98
101
111
Modulo Powers
2.
Summary of Previous Results

The Legendre Symbol
3.
Gauss's
4.
5.
knowns
107
of a
Prime
Quadratic Residues
1.
1.
xi
6.
Lemma
The Quadratic Reciprocity Law

The Jacobi Symbol
Sums of Two Squares
119
120
122
126
183
Bibliography-
187
Index
189
1
The Development of
the
Number System
Introduction
1.1.
The theory
of
numbers concerns
whole numbers; and, because
itself
one hand the intrinsic interest that

puzzles and on the other
restriction to integers.
with the properties of integers or
concerned with integers,
it is
is
it
has on the
associated with the solution of
hand the inherent
difficulty
imposed by the
This explains the fact not only that amateurs
play with the theory of numbers
more than with any other branch
of
mathematics but also that most mathematicians have at various times
worked in
this field.
Some
results in
For example,
This
is
true that
it is
result is divisible
number theory
by
if
6; that
any integer
x3
is,
and easy to prove.

subtracted from its cube, the
are easy to state
is
is divisible
by 6
if
is
an
integer.
true, since
z8
This equation shows that x
(x
is
l)x(x
1).
the product of three consecutive in-
At least one of these integers must be even and at least one

divisible by 3, which shows that the product is divisible by 6.
Some results are easy to state but hard to prove. For example, recall
that a prime number is an integer greater than 1 which is divisible only by
tegers.
itself
and
1,
the
first
ten prime numbers being
2,3,5,7,11, 13,
(An integer greater than
17,
and not a prime
19,23,29.
is
called composite.) Bertrand's
The Development of the Number System
number
postulate states that between any

there
is
prime
greater than
and
double
its
at least one prime number.
between \/2 and 2y/2
5,
proved, but the proof
Some
is
For instance, between 3 and 6 is the

the prime 2. This general fact has been
rather apparent conjectures are easy to state but have not been
proved; in fact, what amounts to the same thing to a mathematician,
do not know whether or not they are

called twin primes
if
they
19 are twin primes.
differ
by
Is there
knows, though there
is
2.
we
For instance, two primes are
true.
Thus
3 and
5,
and
11
13,
and 17 and
No
a largest couple of twin primes?
good evidence to support the
one
belief that there is
none.
Some
propositions are hard to understand and hard to prove, such
Two
as the following:
quadratic forms are congruent in the rational
if
they are congruent in the
is
no point
field of reals
and
in all p-adic fields.
field
There
leading problems in the hope that he can to a large extent under such
much
of the theory himself.
the solutions to exercises in one section
be hoped that the student
will
guage we
will
but
not only
will find his
make
may
Though on many
occasions
be found in the next,
it is
to
not attempt to read in advance but will
the subject
what has gone before. In that way

more interesting because it is his own
is
those of B, that
not on solving numerical exercises
exercises is to illustrate the theory, to find
lished
on the basis
by
proof.
of guessing results
ing the guesses
new
of numerical evidence
The mathematician's
results, or to lead to con-
which
may
be estab-
later
exploration often takes the form
from one kind
by
proof.
of evidence or another and then validatIn this chapter we shall explore the funda-
mental properties of numbers beginning with the integers.

1
.2.
in
way
girls left
the
no great point
so that each object of
if
many
there are just as
call
of the lan-
in trying to define
has one and only one "mate"
this result
If there are
girl.
that the
Furthermore, no matter
of girls.
For instance, a good
boys as girls in a party
number
how
would be the same. In general, then,
if
of
is
to
no boys
or
boys
is
equal to
partners were chosen,
such a correspondence
can be established, the natural number associated with one set will be
the
same as
(that
equal to) that associated with the other.
is,
numbers has four fundamental
Equality of natural
1.
For any natural number
2. If
=
a=
then b
b,
and
=a
3.
If
4.
For any o and
true that
a-
(reflexive).
(symmetric).
then a
c (transitive).
either a
b or a 9* b (determinative).
c,
6,
a, it is
properties:
In each
case a,b,c are assumed to be natural numbers.

If
the number of
call
the number of
Addition of two numbers can be defined as follows:

things in one set
is c
and
in another set is d,
things in both sets together the

result of this definition
la.
4a.
For instance, the number which we
we know
without partners,
number
3a.
natural numbers sometimes called positive integers are
use to count with.
is,
to discover
2a.
Fundamental properties of natural numbers
The
is
and there are no "bachelors" nor "spinsters."
but in exploring the properties of numbers. These results can be stated

most clearly in the form of theorems and corollaries. The function of the
jectures
There
numbers beyond this; we may regard it as an undefined concept

which we assume everyone understands.
Thus with each set of objects we can associate a natural number
which describes how many objects are in the set. The natural numbers
associated in this way with two sets A and B will be said to be the same
if we can set up a one-to-one correspondence between the objects of A and
grounding more sure.
Moreover, the emphasis here
an abstraction independent
It is
it.
natural
try to solve the exercises, using only
he
use to describe
form a procession with each boy taking a
in trying to explain this here.
Instead of presenting a theory to the student in this book, every

effort is made to encourage him to approach the subject from certain
guidance develop
most important thing common to three hats, three houses, and three
is a universal
The system of natural numbers: 1,2,3,4,5,6,
lives.
tally device by which we can tell whether there are more things in one
category than in another.
rather long and difficult.
is
1.2.
sum
of c
and
then we
d,
and write
it c
and those above, addition has four
d.
As a
properties:
The sum of two natural numbers is a natural number

If a = b, then a + c = b + c (addition is well defined).
a + b = b + a (addition is commutative).
(a + 6) + c = a + (6 + c) (addition is associative).
(closure).
what we
three
is
the
The commutative property
holds, since the order in which
we count
the
two
sets
makes no
difference in the sum,
and the closure property la

number in terms of counting. For
follows from our definition of natural
the associative property as shown above,
if
counting as an exercise.
combine
combine
a,b,c,
we denote by A B,C, sets

we on the one hand
while on the other hand we
with the combination of

with the combination of
Definitions: natural
Again our definition shows
Properties
respectively,
A and B,
B and C.
To summarize, we have
If there are 6 sets of
two natural numbers can be defined as
things with c things in each set,
we
define
follows:
ically, if
we
arrange the b sets in rows
we have
ac
(b
a =
Closure
sum
b implies a
of
Multiplication
+c= b+
Throughout the above table the

shall find that the properties are
multiplication
many dots as b rows of c dots. Thirdly, we could count the dots by

columns which would be "c times b" which shows that the commutative
as
each having
rows of glasses with
Then by our
definition, the number of glasses on each tray would
be st. Thus we have
r trays of st glasses, that is, r sets with st in each set, and the total
number
of glasses would be r(st). On the other hand, if we look
at the stack from
the front we will see r rows of glasses with s in each row, that is rs
glasses;
this will
have
be repeated
glasses in each row.
we go back from the

Thus we have r(st) =
times as
(rs)t different glasses.
front.
(rs)t,
Thus we
will
property for multiplication.
Which
property: a (6
+ c) =
ab
ac.
We
leave the proof of this in terms of
ba
ba
ca
letters
=
=
a(bc)
+ c)
+ c)a
a(b
(6
stand for natural numbers.
shared by other kinds of numbers
and multiplication are properly
defined.
(be) ?* (a
+ b) (a +
if
Notice that
in the distributive
c).
of the
above properties are used
in
adding a column of figures?
2.
Prove the distributive property from the definitions given.
3.
What
4.
Which
are
some
of the undefined
of the above properties
terms in the section above?
make
the following expressions
unam-
biguous:
a
5.
+b+
c,
abc?
Using only the above properties, show that (a
+ [6 +
What
-
Finally, there is a property involving both processes, the

distributive
Explain.
(a
the associative
ab
(ab)c
Exercises 1.2
property for multiplication holds.

In order to deal with the associative property of multiplication,
consider a stack of trays of glasses in a cafeteria. Suppose
there are r trays
ab
4- ac
and addition are not interchangeable
property; that is, a
be
numbers a natural number
+b b+a
(a + b) + c = a + (6 + c)
equality, addition,
product of two natural
Distributive
Then "b times c" is understood to be that number obtained by counting

by rows the total number of dots in the rectangular array. In the first
place, the result must be a natural number (the closure property).
Secondly, a = b implies ac = be, since then a rows of c dots will contain
just
b implies ac
Associative
We
two natural
numbers a natural number
Commutative
of multi-
+ ca.
ba
given for natural numbers
c things
6 sets
+ c)a =
number, equality, addition, multiplication.
Well defined
"6 times
c" or "the product of b and c" to be the total number of things in the b
sets. The product of b and c is written b X c, b- c or merely
be.
Schemat-
Addition
that the results are the same.

Multiplication of
Note that the commutative property
plication implies also that ab
numbers
associated with the
1.2.
c])
+ d.
+ b) +
(c 4- d)
Point out at each stage exactly what property
expression does this
What common axiom
is
used.
make unambiguous?
in plane
geometry
is
expressed
by the
third
Property of equality listed above?
7 Which
-
of the following
have the four properties
listed
above for
equality of natural numbers: (a)
same family
in the
(a)
as; (d)
is
is
the son of; (b)
congruent
the reflexive property would hold
the son of
1.3.
to.
if
"A
the brother of;
is
For example,
Explain.
is
the son of
B"
(c) is
in case
implies
"B
is
A."
Suppose we have two sets of objects, namely, A, containing a objects,

and B, containing b objects. We say that a is greater than b (written
a > b) if we have to add some objects to the set B to have just as many
More
things as there are in A.

b
read
is
+x
= a
is
solvable for
greater than
precisely,
we
give the definition: a
some natural number
x.
>
The symbol
2.
If
The second property may be proved as follows: a > b and b > c

imply the existence of natural numbers x and y such that a = b + x and
addition, we may
b = c + y. Then, by use of the well-defined property of
a=
(c
and
have
+ y) + x = c +
equation
by
c
+y
replace b in the first
natural
number
is
a
However, y + x
(y + x) by the associative property.
by the closure property, and
a = c + z and hence a > c.
The
relationship, a greater than
6, is
to say:
and only one must hold:

a > b, a = b, a < b.
>
and
>
then a
3.
If
To
prove the
6,
third property
We shall
is
c,
then a
>
+ c >b + c and ac >
property suppose, for instance, that both a
not
less
than
>
argument used
a form of argument which goes by the
it
may
This
may
+b = a+
reduction to absurdity or
its
This has
its basis in
such arguments outside of the
One example would be
the following: "I
We also have
if
call
then
the can-
c,
then b
for
Q-
loss of generality
some natural number
+c+d=
c,
d,
which, as
lishes the cancellation
c.
If 6 is
not equal to
c,
then,
field of
know John has not
5m.
The proof
If
ab
is left
ac,
by the
either be greater than c or less than
property of inequalities, b must
may without
assume that
+ d.
we have
b is greater
Then a
than
+b=
c,
c.
first
We
that
is,
becomes
This estab-
seen, is impossible.
property of addition.
then b
is
c.
as an exercise.
the axiom of
P is true and P implies Q then Q must be true. It follows that

P implies something which is false then P must itself be false. There
if
of
mean "a
is,
be proved as follows:
be
previous paragraph
in the
name
Latin equivalent reductio ad absurdum.
c to
c."
Similarly, for multiplication the cancellation property
well to consider the type of
many examples
is
c or
be.
Before proceeding with the proof of the second property
are
is,
The well-defined property of addition showed us that

b = a + c. The converse of this statement is what we
be established.
if
"a
Cancellation properties
.4.
5a. If
logic that
This shows that
z.
leave as an exercise.
on occasion the symbols a
also use
cellation property; that
and
a > b. Then, from the definition of inequality, there is a natural number
c, such that a = 6 + c.
But a = b, by hypothesis, and hence by the transitive property of equality, b = b + c. But we cannot set up a one-to-one
correspondence between b + c objects and b objects. Thus we cannot
have both a = b and a > b. The rest of the first property can similarly
first
be designated by
the corresponding symbol with the inequality reversed.
c (the transitive property).
we
may
it
greater than or equal to c," that
of the following
>
he had there would have been tracks in the
and there are none." In order to establish the desired result, it was
supposed false and shown that such a supposition would have false con-
is
b is less than a, that is, b < a. Both > and < are symbols of inequality
and have the following properties expressed in terms of greater than for
the natural numbers a, b, and c:
One
if
s now
6.
Another way to express the
1.
since the snowfall, since
sequences.
Inequalities
means:
Critique
1.5.
mathematics.
left
the house
.5.
Critique
Some readers may be made uneasy by the above formulations, for

there are many undefined terms and concealed assumptions (for instance,
we considered
is
Though
glasses on trays).
not possible to eliminate
all
most rigorous treatment
in the
such things, there
way
another
is
of
We
it
denning
much more
is
reduced to a minimum.
precise in that the assumptions are
may be
His axioms
a natural number
formulated as follows:
1.
There
Every natural number 6 has a unique successor, which is also a

natural number and which we call 6+ also 6 is called the predecessor of b+
a-
We
1.
shall
prove
1.
The
integer
a+
has no predecessor.
= 6+ then a =
If
5.
Addition
6.
Multiplication
7.
(The induction postulate)

clude
a.
b.
all
is
+1=
denned by a
the natural numbers
+ b+ =
a+ and a
1
a,
b+
set of natural
if it
number 1.
of any number
ab
+ b) +
(a
successor
(2)
First,
a=
(q
if (1)
what extent he can do

last
to
be the
set of integers for
must
Then,
of
is
proof
also in A.
If
is
if (2)
holds,
is
1,
If 6
>
if
the case
special attention, for
it
1,
is,
taking the
is true.
that
true for the
ber not greater than
a,
is
number 1.
number of A, then
where
and
is
we assumed
r are natural
the
=
=
1.
which
1)6,
<
and 6
in the above cases that 6
then 6
a,
a,
and the theorem
consequence of the theorem
it
If 6 is
Then,
qb
the
is
first
case of the
Thus our
<
1.
If this is
not
trivially holds.
is
a natural number
qb,
qb
less
than
o,
then there
is
<
a and
The number q of
= [a/6].
(q
1)6
>
o.
this corollary is called the greatest integer in a/6
and
written q
holds for the
another natural num-
then one of the following holds:
and
is
Exercises 1.5
a natural number and b
+r+
(q
contains
which
6,
unique natural number q such that
For example, consider the following important

If
<
the second case of the theorem holds.
successor of that number.
1.5.
or
=1
we have
we have o =
Corollary 1.5.
allows us to
This can be seen by
which a certain theorem
the theorem holds for any
Theorem
that either 6
complete.
An immediate
is
The theorem
-f-
in-
theorem.
show that the theorem is true for all integers, that

all integers, we must show both of the following:
1.
holds,
a = qb
has the following properties:
to
2.
is,
fol-
this.
axiom above deserves
use the method of proof by induction.

set
- 1 - qb,
a 1 = qb + r,r <b.
then a = qb
1.
It follows
a.
numbers
Third,
The
our assumption, one of the
+ 1)6, which is the first case of the theorem,
Second,
From these assumptions all the above properties of natural numbers,

and many others, may be proved. The uneasy reader might be interested
to see to
By
a.
contains the
The
1.
b.
denned by a
is
since
second case.
4.
and q =
1,
true for the predecessor of a, that
is
true for
it
6=1
a implies
o =
if
(1)
3.
<
Now, suppose the theorem
It is certainly true
o.
lowing holds:
2.
is
prove this by induction on
the requirement that b
the positive integers due to Peano, an Italian mathematician of the nineteenth century, which
Critique
1.5.
+ r,
numbers.
<b,
Complete the proof
of the first property of inequalities.
2.
Prove the third property
3.
Prove the cancellation property
4.
Show how
the corollary above follows from the theorem.
5.
Show how
the postulates of
Property of addition.
of inequalities.
of multiplication.
Peano may be used
Hint: For property
to prove the associative
4a, use induction
on
c.
10
Show
6.
that
7.
What
8.
Show by
finite in
a and
if
properties
6 are natural
numbers, then ab
do inequality and equality have
in
We
a.
common?
It is
induction or otherwise that any set of natural numbers,
have discussed
is
less
be a sequence of natural numbers such that each
-,a*
than the one which precedes
that the
first
number
of
numbers
it.
Show
that k
a,i,
in other words,
cannot be greater than the
in the sequence
number
number.
in pasture
is
Thus the fact "No normal cat has two heads," could be
The set of normal cats having two heads is a null set.
there are no cows in pasture A, we could say that the cows
if
form a
It follows that
null set.
if
contains a objects and
a null set then the set obtained by combining
objects, that
b null sets,
Zero, subtraction
.6.
n.
zero.
is,
which
a objects.
way
factory
Also 6
and
would be the number
has a
-+-
of objects in
However, there seems to be no very satisfactory
is zero.
same
interpretation along the

1
having n objects for any natural number
sets
convenient to define a null set as one similarly associated with the
Similarly,
11
stated as follows:
number, contains a greatest natural number.
Let ai,a 2 ,a3,
9.
Negative integers
1.7.
product
lines of the
The most
6.
satis-
merely as a number having the prop-
of thinking of zero is
erties
We
that
>
have defined a
such that a
is
appropriate, since,
would imply
c'
by
definition, say that c is
+ a=a +
number
c.)
by the
Notice that a
if
b is defined only
a number,
the definition of inequality
see that every natural
has the solution x
b.
we
to
6,
gives
a.
(a
b 4-
shall here
number
if
zero,
is
is
greater than
all
With the use

Theorem
r
of zero,
in Section 1.4,
+z=
.6.
we can prove
1.5
exists.
(o
Hence we must have 0-6 =
0.
+ 0)6 =
ab
Thus we
define
0.6 = 6-0 =
0.
b.
However,
can be shown
and multiplication imply
in Section 1.2.
can be stated in a different way:
For any "numbers" a and
there are
6,
Here the word "number"
b\
we exclude
qb
r,
<
numbers q and
6.
shall include all natural
numbers and
zero,
0.
the
Negative integers
1.7.
we
confine ourselves to the natural
has meaning only
If
sion a
unnatural restriction which
if
numbers and
6 is equal to a or less
it is
or zero,
we
define a
number ~6 by
(1)
and we
6
shall say that ~6
if
is
and only
if
0,
c.
a.
This
We
is
an
do so by
a natural number
the equations
+ -6 = "6 + 6 =
~c
zero, the expres-
than
convenient to eliminate.
the introduction of negative integers, as follows: If 6
ab
It
such that
except that
numbers and zero; that is,

zero is the only number which, when added to a number, leaves it unaltered. Hence we may define a a to mean zero. If zero is to be included
in the numbers which have the distributive property, we must define
multiplication by zero in such a way that
cancellation property of addition for natural
0.
lead to a contradiction.
numbers
Theorem
to be extended to include zero,
may
assume that no such contradiction
the properties listed for natural
b.
by the equation
than zero because
0-6 = 6-0 =
always the possibility that several requirements
a.
6 is greater
Furthermore, as
is
a,
that the above definitions of addition of zero
and hence, by the cancellation property,
+ a = a-fO =
Then,
c.
cancellation property of addition, there
when added
c which,
+ c' = b + c,
6;
which we place on a new number
+ c means a b =
It is convenient to define
wc
o "minus"
Of course there
exactly one
is
then,
is,
This
imply the existence of a natural number c
b to
We
c.
Thus
r
.
12
-0 =
(2)
0,
These new numbers exclusive
+ =
since
we
of
where a and b are natural numbers or
0.
negative integers.
call
new numbers.
First, it is necessary to define addition for these
and commutative properties of addition are to

two negative numbers will require the following:
associative
of
If
the
hold, addition
+ -b)(-a) =
(b
Hence we
b(-a)
+ ~b +
On
(a
b)
(a
+ "a) +
+ ~b)
(6
+ b) +
"(a
(a
+ b) =
and,
if
+ ~b +
(a
b)
= ~(a
+ b) +
the cancellation property of addition
(a
The
what we
b).
to hold,
is
-a
To define the sum
+ -b = -(a +
we must
define
+ "6 =
[(a
Similarly, ~b
-6)
+a=
+ "6 =
b.
On
(a
-6)
+ ~b = a +
Thus we
(b
a)
+ ~(b -
(6
+ "6)
the other hand,
define addition so that the following
define addition of a natural
if
(a
<
6)
+ =
we would need
2.
6.
call the integers.
If
a
a
>
<
b,
b,
is
replace a
greater than a, a
~(b
a)
+ b + ~b
number and negative
~(b
=
Thus we
(5)
(b
integer as
by a b on the understanding that when

means the negative of 6 a.
b
+ ~b)a =
to hold for the
ba
(~b)a
The
a(~b)
~(ab)
the properties listed in Section

for the negative integers.
set of positive integers coincides
4.
with the
5.
ba
~(ba),
c.
definition, or
commutative?
can
it
be proved?
Is it associative?
Is ad-
commutative?
properties of inequality listed in Section 1.3 hold for all integers
Do
the cancellation properties of addition and multiplication hold
for integers?
=a
Why or why not?
7.
Does a
8.
Simplify the following expression, pointing out at each step
properties of
imply 6
Here
a, 6, c
c for integers a, b, c?
numbers you are using:

c(a
c.
including zero?
new numbers, and hence
+ -(ba).
Is this
b.
Is subtraction of integers
What
1.5.
b, c.
(~a)c implies b
from Theorem
1.6 follows
+ ~b c + ~b implies a =
define multiplication as follows:
(~b)a
all
that the definitions imposed on ~b and the properties of positive
Theorem
3. Justify ~(~b)
a).
+ ~b = a b = ~b + o.
then a + ~b = ~b + a = ~(b a).
We wish the distributive law
to (6)
(1)
Show
a.
then a
+ ba.
that
b. (~a)b
Thus we may
~{ba)
Show
dition of integers
If
numbers.
integers imply for a,
to
follows:
(4)
(~b){~a)
Exercises 1.7
would be true:
+ ~b -
a)
= 6a=(-o)("6).
and zero follow
of a natural
6]
~(6a)
this definition so that
b).
number and a negative integer we need

to consider two cases. Suppose a > b, then, by the definition of a b,
we have (a b) + b - a, and hence we must define a + ~b so that
a
remains to define the prod-
natural numbers, zero and the negative integers, together form
set of natural
(3)
It
make
We can now say for all integers that a < b if there is a natural
number c such that a + c = b. Thus, for instance, 11 < 1 since
-11 + 10= 1. If an integer is greater than zero, we call it positive.
0.
Thus
-a
these definitions
1.2 for the positive integers
the other hand,
("6)(-o)
(-6)(-a)
From
0.
to
define
(6)
~a
zero.
We want
uct of two negative integers.
13
Negative integers
1.7.
+ 6a)
stand for integers.
ac
+ a abc.
what
14
Integers could be defined as pairs of natural numbers in the following
9.
way:
Theorem
a
(a, b)
b. (a, b)
a.
(a, b)(c,
c.
First,
b or a
d)
Numbers
2.
3.
Third,
(x + y + a, a); (a,
a + a,a)(b,c) = (b,c).
6, b)
a)
(6, c)
(6, c);
(a,
+ a) =
All the properties of negative integers
What number pairs
Prove that a
0.
<b
implies a
may
be derived from the three
correspond to zero?
<
(Ref
8,
pp. 5
ff.)
into three mutually exclusive categories:
but 6 and 15 are not.
and some
by
usually
instance,
it
2,3,5,7,11,13,
of the deepest
delicate analysis,
are irregular in their distribu-
theorems in the theory
where
1,
the number of prime numbers less than
tt(x) is
x.
That is, by
is
a and
b are
a multiple of
two
b, if
integers,
there
is
we say
an integer
that a
is
be.
by 6, or that a
Note that we
b would always
divisible
such that a
introduced negative integers so that the equation a
-f-
be solvable, but note that without introducing fractions we cannot say
that ax
b is solvable in all cases.
the problem of
divisibility is of
talking about rational
problem of
numbers
divisibility, since
When we
deal solely with integers,
fundamental importance.
(ratios of integers), there
If
we were
not hard to prove
Theorem
1.8a.
1.8b.
If
states that there is
no
The greatest integer which divides two integers is called their greatcommon divisor, often abbreviated g.c.d. For instance, the g.c.d. of
6 and 15 is 3. Using the notation (a,b) to denote the g.c.d. of a and 6,
we would have
est
(6,-15) =
Though
g-c.d.,
Define
there
g,
the above
is
another
The
reference
numbers
is
3.
perhaps the most natural
way which
is
way
to define the
sometimes more useful
the g.c.d. of two integers a and
and
b\c,
then
+ canda|6c.
a\c.
1.
g divides both a and
2.
Any number
That these two

'
last
b,
in proofs.
to be that positive integer
which has the following two properties:
a|6anda|c, thena|6
If a\b
prime number theorem.
would be no
number provided the divisor were different from zero.

The notation often used for "b divides a" is b\a and "b does not
divide a" is written b H" a.
Theorem
called the
every number would be divisible by every
other
It is
is
talcing
made as close as we please to 1 This

Theorem 1.8g later in this section
prime number. The proof is quite simple.
x large enough, 7r(x)(log x)/x can be

If
For
has been shown that
c.
Divisibility
.8.
numbers, proved
of
have to do with the prime numbers.
lim7r(x)(log a;)/a;=
a
xoo
1
b are integers, then
which are not primes are called composite. Thus
fall
The prime numbers
(a, a).
tion,
definitions.
numbers
relatively prime,
+ a,a) +
where a and
b\a,
Two numbers are called relatively prime (or one is prime to the other) if
For instance, 6 and 35 are
their only common divisors are 1 and 1.
show that
(x
and
The number 1,
The prime numbers,
The composite numbers.
1.
show that addition and multiplication are associative and commutative.

Second, show that
+ a, a) + (y +
a\b
If
1.8c.
~b.
greater than
the natural
basing your proof upon accepted principles for natural numbers,
(x
Every positive integer is divisible by itself and 1. Integers greater

than 1 having no other divisors are called prime numbers or primes.
means a + d = c + b.
means
(a + c, b + d).
(c, d)
d) means (ac + bd,ad-\- be).
(c,
IS
Divisibility
1.8.
refer to the Bibliography in the
back of the book.
follows:
Let
gi
be the
b.
dividing both a
and
b also divides g.
definitions yield the

g.c.d.
same number may be seen as

first definition and g 2 accord-
according to the
16
The
ing to the second.
second that
gx
would then imply g% < g x and the

Hence gi < g i} and it must follow that
To
first definition
would divide g 2
prove
we
since, until
reach the point where one
integers.
can immediately be extended to any number of
definitions
For
instance,
we can
last equality will
rk- 2
35)
1.
and
That any pair of integers has a greatest common divisor can be

shown from the first definition, for any divisor of a positive integer must
be either the number itself or less than it; hence a positive number cannot
have more than a finite number of divisors. Then, to find the greatest
common divisor of two numbers, we need merely list their divisors and
pick out the greatest one common to both. However, this means of finding the g.c.d. is laborious. There is a somewhat shorter method which depends on Theorem 1.6 above and the second definition of g.c.d., which is
ffc-i
divides r*_ 2
imply that rk -x divides r*_ s

the
first
qic-xn-i
Then Theorem
Tk-3
in
must
Thus
1.5.)
the
equation
and
1.8a
qk -iTk-i
We
+ 0,
+ Tk-x
can continue to follow this back
and
rk ~x divides r x
and hence
have demonstrated that rk -x divides a and
until
Therefore we
a.
first
requirement of
then any com-
the
b,
g.c.d.
Since the
mon
equation
first
may be
a and b divides
divisor, d, of
written a
rx.
qb
r%,
Since d divides b and r x
we
from the second equation, divide r 2 and, continuing,
as follows:
be zero,
finally
they form a decreasing
be
write
(6, 21,
r's
is zero,
(See Exercise 9 of Section
sequence of positive integers.
Both
that one of the
this, notice first
17
Divisibility
1.8.
it
must,
see that d divides
Hence the second requirement of the definition is fulfilled, and we

have proved Theorem 1.8d. We have climbed up the ladder and climbed
down again.
There is still a third definition of g.c.d. which we include in the fol/v_i.
Theorem
of
Theorem
.8d.
If
a and b are two positive integers, then the process
1.6 is repeated as follows:
a
b
rx
= qb + i% < ri <b,
= qin + r 2
< r2 < r h
= qtti + r 8
< r 3 < r,,
lowing:
Theorem
Given two nonzero integers a and
1.8e.
positive integer expressible in the

Tk-Z
Tk-z
Eventually one of the
Then
r*_i will
Algorithm.
be the
+ Tt-i, < Th-\ < Tk-t,

qk-xTk-x + rk,0 <rk < Tk-x.
qk-&k-2
r's will
g.c.d. of
(An "algorithm"
is
be zero,
a and
us see
= 228 and
177.
The
and suppose
process
is
r*_i
with x and y integers.

0.
called the Euclid
a term applied to a repetitive process like
the above, used to achieve a desired

First, let
b.
call it r k ,
ax
result.)
how this process works

Then we have
condition
of
a numerical case.
228= 1-177 + 51,

177 = 3-51 +24,
51 = 2-24 +3,
24 = 8-3
+0.
is satisfied,
a and b divides
a.
states, thus, that
is
the g.c.d. of 228 and 177.
+ by =
least
r,
r is the g.c.d. of a
Theorem
To show that
for
1.8a
and
g.c.d.,
b.
we
see that the second
shows that any common divisor
r divides
a and
6,
suppose that r does
Then
Let
a
qr
+ s,
<s
<r.
y such that ax<, + by = r, and

=
s
qax + qby Q However, u
qi- +
im plies qr = a s = qax + qby
Hence s = a(l
qx )
bqy
Thus 8 is a positive integer less than r,
Furthermore, there are integers x
qr
representable in the form ax

result contradicts
The theorem
r.
be the
form
Again, using the second definition of
not divide
in
Then
b, let r
by,
with x
our assumption that
and completes the
proof.
r is
qxo
and y
= ~
qyo-
This
the least integer so expressible
18
We
by the
mx +
and
If
divides ab
and (m,a) =
then
1,
- P^ P2 2
divides b;
above theorem we know
there are integers x and y such that

whence m(bx) + y(ab) = b. However, m divides m(bx)
which by Theorem 1.8a shows that m divides b.
ay
ab,
Then we can express o and
have the following important
Corollary 1.8e.
for
1.8.
p
*n
where no two of
p,
b in the following forms:
VV
r,-,
are equal.
s*
1,
Corollary 1.8e'. If
m is prime to a and to b,
then
it is
prime to ab;
m had a factor in common with ab, it would have a prime factor in

common. Call this common prime factor p. By the first corollary, p
would divide b, which would imply that in and b have a common factor
for, if
Pi
where, for each
These then yield the Fundamental Theorem
of the theory of
numbers:
c is
p,
The
g.c.d. of
a and
b is equal to
e_
P2
the smaller of
P nn V*2 2 -"**
p?
Pn"
a,-
and
,-.
How would one express the
common multiple (l.c.m.) of a and 6?

Any number which is divisible by 2 is called an even number, and
least
others are odd.

to
in contradiction to the hypothesis.
i,
;b
19
Divisibility
be
of the
same
all
two numbers are both even or both odd, they are said
If
parity;
if
one
even and the other odd, they are said
is
number is of the form 2k for an

and
every
integer
odd number of the form 2k + 1, it is not hard to see
that the sum of two numbers of the same parity is even and that the sum
of two numbers of opposite parity is odd.
to be of opposite parity.
Since every even
k,
Theorem
1.8f (The
Fundamental Theorem).
factored into prime factors in
two ways, as
'
P1P2
Pr
qm
If
a positive integer
is
follows:
Euclid proved that there
q
is
more q t may be equal, then r = s and the

Pi are equal to the factors q< in some order. Briefly, we then say
where two or more

factors
p,-
or two or
that decomposition into prime factors
unique except for order
is
to
it
We may
prime numbers.
state
it
is,
there
in the following way,
shall give Euclid's proof.
1.8g.
For any positive integer n there are more than n
primes.
this, notice
that a prime either divides a
number or
it.
Suppose pi is not equal to any of the g.'s.
If
must be prime to each one and hence by Corollary 1.8e',
This
infinity of
Theorem
of the
prime factors.
To prove
an
and we
no greatest prime number, that
is
is
this is true then
orem for multiplication shows that
qtqz
and we may deal with p t as we did with p x

that the theorem is true.
Continuing in this
where now no two
way we
in the
see
form
*... pa ,
Pl p s
n
a,
N=
p.-
are equal.
This decomposition again
only prime numbers dividing both o and b are

Pl>P*, '-,Pn.
list
them as
one
n primes
of
the
1,
which
is
pip 2
divided N,
it
impossible.
Then
+ 1.
would, by Theorem 8.1a, have to

either
is itself
a prime, or
it is
by a prime different from those listed.

In either case there
must be more than n primes.
The least common multiple of two integers a and b is defined to be
the smallest positive integer which is divisible by both a and b. This
statement
f
cept for the order in which the prime factors are written.
n primes and
divisible
This theorem also shows that we can express any integer
suppose that there are only
PhPt, '"iPn>
divide
this,
Then form the number
If
pT
prove
follows:
to their product.
not possible, since pi divides their product. Hence we may renumber the qi if necessary and take p x = q u Then the cancellation theis
PaPs
To
prime
is
a and b
is
equivalent to the following:
if it
m is the least common
has two properties:
unique ex-
Suppose the
1.
2.
m is a multiple of a and b.
m divides every common multiple of a and b.
This bears some similarity to our second definition of
g.c.d.
multiple
20
1.8.
Exercises 1.8
3n
Show
1.
Theorem
that
are plotted in red
on a
1.8e implies the following: If the multiples of
line
and the multiples
are positive integers with greatest

shortest distance from
By
2.
same process
to find integers
3.
Prove Theorems
4.
If (ab, p)
Show
5.
1,
that
x and y so that 576x
and
implies
then g
will
be the
73*/
Use
this
when k =
(a/d,b/d)
Show
1.
and p\a
that
4- b.
a(b,c)
Prove that
if (a,b)
l\b
and
are uniquely determined except for sign; s or
and
may be
rf=
that,
if
ab/d
is
an
where
s\a
the least
is
common
multiple of a and b
if
ka
Theorem
+ r,
for a given integer
Show that
set of
or both
6.
there
Use the
7.
by
will
rk
and
bd
is,
of the
That
be the least
= 0,
number
12.
of
rf.
of integers of the
form
4n -
1.6
1,
4n
odd integers
is
1,
of
4 show that every odd integer is of one

and hence that if the product of an odd
the form 4n + 1 at least one is of the form
with 6
a:
+ 2,
+ a-
,s
1?
a.
In other words, any

a.
show that x 5x s
6
+ Ax
is
of steps, k, in the Euclid
Algorithm show that a
if
1,
rfc _ 2
i
<
= 2,
rt _ 8
k, rv_s
= 3, ri_ = 5,
4
= r*_
r,
-f
in
and
1%-%=
?<_ i
= 1.
8,
Thus a and b
the sequence, the
first
ten
which are
1,2,3,5,8, 13,21,34,55.
is,
if
the Euclid Algorithm process has five steps, the least pair of
is
(a,b)
Why
S and 13.
with b
<
does this imply that the Euclid
34 must contain
less
than 7 steps?
show that a c and r s are either

both divisible by b or both not divisible and hence that the difference of
two numbers is divisible by 6 if and only if their remainders when divided
by 6 are equal.
18.
If
bq
+r
and
bt
s,
19. Prove that if f(x) is a polynomial with integer coefficients and if a is

an integer not zero such that f(a) =
then a divides the constant term.
20. Prove that
if
p and
q are integers
and
p*
= 2^
then p and q are both
1.
even.
Extend the
definition of least
common
multiple to
more than two
integers.
13.
rk -
values of a and b
an odd number
of the follow-
60 for every integer x.
0,
same form.
Using Theorem
4n+
of
Which one
a.
result of the previous exercise to
+ be
ad
^d
d\b; that
Prove that the product
of the forms
1, s
<
if
Prove that the sum of an even number of odd integers is even and the
product of a set of integers is even unless they are all odd.
1 is
by
only one of the set divisible by
is
where for each subscript
terras of
9.
s+
For a given number
6.
and only
a/b and c/d are two fractions in lowest terms, and
integer, then b\d
10.
divisible
is
<
where
a,
can be expressed in the form
a consecutive integers contains exactly one multiple of
Algorithm for
An
shows that every integer
1.6
would have to be two successive numbers

if
its
and
a multiple of a and 6 then d divides both a and
b
is
st
(a,b).
Prove that
8.
5.
and b
Hence show that ab/d

if
ab then r
1.
Prove
7.
r divides
at least one of
+ 2.
Prove that JV in Theorem 1.8g has at least one prime factor of the
form 4n 1 and hence there is no greatest prime of the form 4n 1.
divisible
6.
and
form 3n
s,
bp' only
+2
an integer has the form 3n
if
of the
14.
and
2,
is
ing set of integers
(576, 73).
(ab,ac).
prime factors
576 and 73.
3n
or
1.8c.
show that p kJtl \a-p k
= d
g,
where a and b
point.
1.8c find the g.c.d. of
1.8a, 1.8b,
(a,b)
divisor
any green to any red
Theorem
the process of
common
of b in green,
21
Divisibility
Show
21. Prove that
if
xi
ab where x,a,b are positive integers and (a,6)
then there are integers y and

that every integer not divisible
by
is
of
one of the forms
yz
x,
such that
a =
y*
and
z2
22
22. Prove that
if
common prime
then that
sarily hold
if
common prime
a and 6 have a
factor divides
(b,c)
if
then
(6
conditions will the latter g.c.d. be 1?
24.
all
2c,
Show
c,b
62
c)
Under what
+ by
if
and only
c2
if rb'
b 2 , (c,b)
and
r'b
r'a
a'b
r,
+ b'y = r
^ 0, have
a'x
= d
to! are divisible
by
1
,
where
However, at each step
difference to the result.
it is
im-
portant to begin a row only with a prime divisor.
Prove that for any two integers a and
d.
b there are integers q
a = bq
r,
-6 <
2r
<
6.
This forms the basis for the so-called least remainder algorithm.
and
such that
integer
1, and just one of c,b is even, use Exercise

21 to show that there are integers m and n such that mn = a, c - b = m2
c + b = n2
If
makes no
similar questions for

28.
that the pair of equations, ax
So we continue until a row

The l.c.m. is the product of the numbers in the
3,5,2,7,2. The order in which we divide by these num-
next to obtain the second row.
column, here
bers
1 or 2.
it
consists entirely of ones.

left
we use
fine,
c*
the same result neces-
Answer the
the letters stand for integers and ab'
a2
+ c).
26
solutions
25.
Would
c.
and a2
the word "prime" were omitted?
23. Prove that
(b
factor
23
The decimal notation
1.9.
this
Show
through for 576 and 73.
Carry
that this process always yields the
g.c.d.
26.
If
6)
(c,
and both
and
a
implies that a
is
even.
Show
29. Establish results similar to those of Exercise 17, but use the least
b are odd,
that
(c
show that
6*
6)/2 and
remainder algorithm.
(c
+ 6)/2
are relatively
prime integers and equal respectively to m 2 and n 2 where

and n are
integers whose product is a/2.
Show how this may be used to give all
,
solutions of
is
even, b
greater than
27.
The
is
+b =
2
odd, and no two of
following illustrates for the
35
35
equivalent to b"
31
If (a,m)
if
1,
for
b is divisible
what
integers
Given a fixed integer m,
by m,
will
is
for integers
a and
call
"o
b.
ax be "equivalent to" ab in the

also
an
integer?
2
,
a, 6, c,
have a common factor
15
numbers 35,15,12,6 a device
.9.
Why
does
it
work?
when
properties of integers in the preceding sections are independent
what we
call
them.
We
say, "three divides sue."
If
the Martians
designated three objects by the word "glob" and six objects by the word
"toes" while our "divides" is pronounced by them "sots" then it would
be equally true that "glob sots toes." The corresponding result would
12
~T
hold,
sary,
whatever the words or notation. However, some notation is necesand we are accustomed to the "decimal notation," associated with
the number ten, since we have ten fingers, and say our number system
"has the base ten."

If
not a divisor, we repeat the number above.

a divisor of two numbers of the first row below the horizontal
it is
we had twelve
decimal system," where the

fingers, it
We
we would undoubtedly employ the "duobase would be twelve. If we had just eight
fingers,
first
place the
quotient in the row;
The
for
of
Explanation: Since 3 divides three of the numbers, we use it for

the
row. Where it is a divisor of the number in the line above, we
is
erties of equality listed in Section 1.2:
1.
finding the l.c.m. of a set of numbers.
Since 5
that "equivalence" in the following sense has the four prop-
sense of the previous exercise, where b
where a
Show
30.
would be the "octic system,"
of base eight.
have, in fact, just ten different symbols:
0,1,
-,9,
and make
24
the place where they occur important.
be 3 plus two tens and 576
769301
is six,
0-10
1.9.
For instance: 23
is
understood to
plus seven tens, plus five hundreds.
3-10
+ 9-10 +
6-10
An example
of multiplication
25
would then be
Also
1011010
101
7-10.
1011010
The
zero can be considered to indicate absence of the corresponding
Using
of ten.
letters,
we could
abcde
where
0>12,
a,b,c,d,e are
zero being a
From
lOd
102c
106
111000010.
10*a,
(We
the symbols,
each can be associated with a finger of the hand,
little finger.)
this it follows that
the
tell
from the
system
last digit
whether or not a number written
by 3?
If our number system had the base
symbols 0,1,2. Then abcde would mean
is
Number
2 2 10
two
2*
23
2*
22
1110
number we add the powers
find the decimal value of the first
der which l's occur: 2
2s
2*
10 11
111000010
To
2s 27
1011010
we would need only the
+ 3d 4- 3 c + 3 6 + 3 a,
2
to the base
101
three
much
numbers above would be 90 and 5, and the result, 450. To see this, notice
that the digits in the binary form counting from the right indicate the
occurs) of the corresponding
presence (when 1 occurs) or absence (when
power of 2. Thus we can make a table for the numbers as follows:
the decimal
in
requires
it
In the decimal notation the two
writing for relatively small numbers.
divisible
chief disadvantage of this notation is that
The
call
if the last digit of a number is divisible by

number is also. Why? If the last digit is even, the whole number is
divisible by 2, therefore, even. Why? If the last two digits form a number divisible by 4, why is the number divisible by 4? Why cannot one
5,
0000000
1011010
the digits of the number.
*,9 digits, since
power
write
+8+
16
64 =
90.
The
numbers may be found similarly.

One may convert from the decimal system
of 2
un-
decimal values of the
other
and the number would be

tell
from the
last digit
by 3
divisible
whether or not
In any system of notation the
number
for six,
would be
but
were
would be
number
zero,
but we could not
divisible
by
like
2.
would divide the

by inspection whether or not it is divisible
telling
for three
easier in one notation than in another.
In some respects the
binary system,
is
number system
the simplest.
especially adapted to
machine
It
addition table
is
to the base two, the so-called
has only two symbols:
calculation, since
lack of a certain electric connection
The
it
if
and
and
1.
It is
manner.
Suppose we
highest power of 2 less than 49

2*
1.
Continuing
in this fashion,
0,
+ =
1
1, 1
we have 49 = 2
binary
is
indicated
0,
0-1
0,
1-1
We
write 49, the
number
1.
12
24
17
system to the
49
Then we divide 24 by
0,
below
it.
the binary system.
2,
to be converted, at the right
10,
of every schoolboy:
2*
its existence.
mainder,
0-0 =
Then
17.
The
by the following process:
writing the quotient, 24, to the left
and the multiplication table would be the joy
0-2+1, which, in the binary system, becomes 110001.

A convenient method of converting from the decimal
merely
+ =
26 hence 49 = 2 6
is
can correspond to the
can correspond to
to the binary system in
wish to write 49 in the binary system.
and divide by
2,
of 49, and the remainder below
it.
and the
re-
writing the quotient, 12, to
its left
We continue in this way and see that 49 is
10001 in
26
1.10.
One
application of the binary notation
Peasant method of multiplication.
This
is
is in
the so-called Russian
a method by which the product
may be found by a succession

by two and one addition.
The process for 375 and 39 is as follows:
any pair
of
of integers
of divisions
and
multiplications
3.
him the number he chose. How would you number

pick a number from 1 to 31? Why does it work?
and
cards to
five
27
tell
Suppose, instead of wanting to answer "yes" or "no,"
it
were to be
one of "yes," "no," or "maybe." What number system would you use
instead of the binary system? What assumptions about "yes," "no,"
and "maybe" would have to be made?
375
19
those cards
Rational numbers
750
Translate 57 and 93 into the binary system, multiply them in that

system, and check your result by converting into the decimal system.
4.
9
4*
1500
3000
2*
6000
Four weights are to be chosen so that any object from 1 to 15 pounds

(weighing a whole number of pounds) could be weighed by putting the
object in one pan of a balance scale and balancing it with a proper assortment of the four weights in the other pan. What would they be?
5.
12000
14625
column we divided by 2, disregarding remainders, and in the

column multiplied by 2. In the left column we placed asterisks after
In the
right
left
the even numbers and, in the addition, omitted the numbers to the right of
the stars. The reason for this process may be seen by writing 39 in the
binary system, namely, 100111.
Writing these in a column with the
the weights above could be put in either pan, three

weights would suffice for objects weighing less than 15 pounds. What
would their weights be and how would the proper assortment be found
6.
Show
for
any object?
that
if
first
bottom, it is seen that the asterisks correspond to the two

Thus, to multiply 375 by 1001 11 in the binary system, we multiply
375 by 1, since the last digit for 39 is 1; multiply it by 2, since the next to
digit at the
7. Exercise
zeros.
constructed for a trick using the base three?
the last digit for 39

digit of
but
we
39
is 1
is 1
we do
multiply
it
multiply
not multiply
by
This
32.
it
it
is
by
4, since the third
by 8
from the right
or 16, since those digits are zero,
accomplished by the addition process
8.
Show
that
is
if
Could similar cards be
connected with the base two.
the quotient
ing the remainder,
is
when one
odd then
1 is
divides a
0.
disregard-
the fourth digit from the right in the
binary representation of the number, while

fourth digit will be
number by 28
if
the quotient
is
even, the
Generalize this for any power of 2.
above.
1.10.
Exercises 1.9
Multiply 75 by 423 by the Russian Peasant method. Convert these

numbers to the binary system, multiply and check by converting back
to the decimal system.
1
2.
Four cards are constructed. The first, headed 1, has all the odd num1 and 16; the second, headed 2, hus the numbers:
2,3,6,7,10,
bers between
11,14,15; the third,
fourth,
headed
8,
headed
has
all
4,
has the numbers: 4,5,6,7,12,13,14,15; the
the numbers from 8 to 15 inclusive.
You
ask
someone to choose a number between 1 and 15 and to hand you the cards
on which this number occurs. You then add the numbers at the tops of
Rational numbers
Though the theory of numbers is primarily concerned with integers,

there are a number of situations in which rational numbers are involved,
furthermore, the place of the integers in the number system can better
be appreciated if we see how rational numbers occur.
Just as we introduced negative integers so that we could solve the
equation b + x = a, when a is greater than b, so we determine a new kind
of number, called a rational number, so that bx = a will be solvable (when
o and b are integers with b j& 0), having the property that
(1)
(a/b)b
b(a/b)
o.
28
This we
call
a "rational number," since
it is
written as a "ratio."
we cannot consistently define this symbol,

Whenever we write the symbol (a/b), we assume b
b
two
First, to define equality, consider the
and
where bd
(c/d),
From
0.
as
we
When
shall see later.
which means that to be consistent we must define addition of rational
numbers by
=^ 0.
numbers
rational
(a/b)
29
Rational numbers
1.10.
(a/6)
(4)
(c/d)
(ad
+ cb)/bd.
Furthermore, we wish multiplication of rational numbers to be com-
the definition,
mutative and associative.
(a/b) b
Then,
if
multiplication
is
a,
(c/d)d
to be well defined,
c.
Hence
we have
(a/6)
= a and (c/d)d = c should imply

= (a/6) (c/d)6d = ac.
(a/b)b(c/d)d
(a/b)bd
ad, (c/d)db
cb;
However, by our definition

and,
(a/6)
if
(c/d),
we must have ad =
cancellation property of multiplication

(a/b)
We
(c/d).
to hold,
is
the
An immediate
(a/6)
and
consequence
is
6c,
bd
= a
we would
1,
(a/1)
(3)
five definitions suffice to
(ak/bk),
0,
(ac/bd).
Now we
since
why
can see
(0/6)
(a/6)
by
fraction gives zero,
the exclusion of 6
Also
1.2.
all
definition.
our rational numbers are to have the desired properties.
Since
we want
(a/6)
(1/0)
co,
that
is, 1
0-
co.
a.
This includes a
0.
= 0-
co
numbers we need to have the
(c/d)
bd
=
=
(a/6)
(a/b)b
bd
2(0
following:
Thus 4
(c/d)
(c/d)d
bd
6
= ad + cb,
first equality assumes the distributive property for rational

numbers, the second the associative property for rational numbers and
the commutative property for integers, and the last equality results from
the definition of a rational number. However,
it
bers.
We
a number.
We
(a/6)
(c/d)
-bd = ad
cb,
It is
co
that
is,
much
0) co
co
2.
(2-0) , and hence the associative
This shows that

easier to exclude
is
we were
if
it
it
to call (1/0) a
did not behave like the
from the family
of
In fact (a/0)
not a number.
numis
not
We do not even call it infinity, since that might be misleading.
call
8 are
co)
therefore say that (1/0)
a rational number positive
are positive integers.
and
(2
would be necessary to specify that
posite sign (that

1
2 shows that 2(0
other numbers.
where the
= 2 2 = 4 and
co)
property would not hold.
number
However,
the distributive and associative properties for multi-
if
Suppose we
were to define (1/0) to be a number called "infinity," and wrote
define
the integers are rational numbers.
plication to hold for rational
0.
in (a/6) is necessary
Then, by the cancellation property of fractions, (2/0)
Hence
Hence
ac.
prove that rational numbers satisfy
(0/1) (a/6)
Thus zero multiplied by any
6afc.
These
(a/6)
(a/6)
numbers, (ac/bd)bd =
the properties of integers listed in Section
0.
the following
(a/b) (c/d)
(c/d) are said

all
Cancellation property of fractions:
Since (a/l)l
we
of rational
define multiplication so that
must imply
cb
whenever
ad
ad =
if
(5)
make the following

The rational numbers
(2)
a6/c
the other hand,
therefore
Definition of equality:
to be equal
On
cb.
two
is,
It
is
if it is
called negative
if
equal to a/6, where a and 6
a and
6 are integers of op-
one positive and one negative).
rational numbers,
we say
that r
<
if
Furthermore,
there
is
if
a positive
30
number u such that r + u = s. Subtraction of rational numbers

defined in the same manner as subtraction for integers.
rational
is
31
Decimals and real numbers
1.11.
show that between any two distinct

Is this a property of integers?
third?
a
this
rational
numbers there
always
is
n_1 -\
B
+ a xx + ao whose
Let f(x) be a polynomial a n x + an_ix
Let r and s be two relatively prime integers.
coefficients are integers.
=
then
s divides a and r divides ao. Show that this
Prove that if f(r/s)
12.
Exercises 1.10
Show
that the above definitions and the properties of integers imply
that addition and multiplication of rational numbers are commutative
and
implies that
associative.
There
a.
Show
2.
Section
that equality of fractions has the properties of equality listed in
Would
zero integer.
whenever
If r
5.
d = kb for
k,
a non-
If not,
of the last section,
show that
(a/6), s
and
(c/d),
not zero, define the symbol
s is
4.
(r/s) to
6.
r/s
Show one
be in terms
Use Exercise 20
ber whose square is
7.
8.
What
9.
(c/d)
fail if
of Section 1.8 to
(a
we
defined
+ c)/(b + d).
0.
show that there
integers
that
if
is
no rational num-
and
1.4 also hold for
follow from the properties listed for rational
a/6 and c/d are two rational numbers where
Show
and
that
<
if
a/6
>
mean a =
c,
d.
system has
all
the properties listed at the end of Section
numbers.

all
know, the decimal notation can be extended to represent
102
For instance, we write 176.9304 to mean
+ 7- 10 +
1
6- 10
c/d
if
and only
if
ad
>
where 6
c/d then a/6
<
(a
+ c)/(6 + d) <
9-
10" 1
+ 3- 10" + 0- 10~ + 4- lO"

2
is
a power of 10 and a
is
an
For instance, there
is
no decimal which
is
an integer prime to
3a - 10 e
is
is
and d
posi-
Why
does
equal to 1/3, for

10,
if
1/3
we would have
impossible, since 3 does not divide 10.
the other hand,
if
our number system were to the base
would be equal to a decimal,* whereas 1/10 could
c/d.
integer, are exactly equal to a
were equal to a/10 c where a
decimal.
On
be.
This means that only rational numbers which can be written in the form
which
a/6 and c/d are two rational numbers with
<
the
(a
a,b,c,d are
and d are positive then
a/6
1 1
this
certain fractions.
a/6,
we make the following definitions for
+ c/6 + d).
= (ac- bd/ad + be).
=
and
tive
(a/6) (c/d)
properties of integers listed in Sections 1.2
Show
c.
As we
numbers?
1
(a/6) 4- (c/d)
2.
does (0/6)
a,6,c,d,
(c/d) shall
(a/6)
b.
1.1 1.
numbers?
Why
Compare Ex-
of a,6,c,d?
property which would

(a/b)
rational
a.
1.2 for natural
What must
is 2.
(a/6):
^how that
r.
6how
no rational number whose square
is
Suppose, for integers
symbol
have the property that

(r/s)s
are integers.
Exercise 12.
what
Point out what properties are used.
a/6.
ka,
number?
for k, a rational
Using only the properties and definitions
a(l/6)
the properties of equality then hold?
would your answer be

4.
The number \/2~ + y/Z is not rational. Hint: Show that the
given number is a root of an equation of the type described in
6.
(c/d)
rational roots of f(x)
ercise 7.
1.2.
Suppose we define (a/6) =
3.
1 all
result of Exercise 12 to
Use the
13.
if
This probably would be called a "trimal."
not.
3,
1/3
32
1.11.
However, the decimal may be used to approximate any rational

number to any desired degree of accuracy; more precisely, for any positive
rational
and any small positive number

< r d < e. For instance,
number
d such that
1/3
.3
1/30, 1/3
.33
1/300,
1/3
e,
we can
.33
find
(1/3)10"*,
with the decimal approximation of a rational number.
new kinds
If we write
to be used to define
view
necessary.
is
However,
if it is
numbers, a slightly different point of
of
<
x\
xn
large
enough we
is less
dz,
1/10) J
may make
number which
is
not
less
is
the smallest
Consider the se-
1.415
For instance, 1.414

is
greater than 2.
1.4142 2 1.4143 2
,
list
1.414, 1.4142, 1.41421,
the second.
and
By
please to 2.
number
in the sense that the
more than one number of the sequence (all of which arc

numbers) would have a square closer to 2 than r 2 and hence would
ho greater than r. Second, r 2 could not be greater than 2, for then one of
;i
ional
he decimals
sumption that
impossible
if
r is
and yet greater than

is rational.
the smallest.
r is rational.
is
included because 1.414 2
is less
than 2 and
For the next entry we could compute 1.4141 2
find that 2 lies
r,
(2)
a sequence of rational numbers whose squares approach 2 from
is
below.
rational
Finally,
Thus
all
the decimals of
(1)
each
This would contradict our as-
we have shown
that r2
= 2
is
there are sequences of decimals of the
form
1, 1.4, 1.41,
This
we
2 because
ihaii
would be smaller than
quence of decimals:
(1)
This shows that by choosing n
1/10".
l/lO* < 5/10",
2, 1.5, 1.42, 1.415, 1.4143, 1.41422,
than any dk of the sequence.
<
2x/10
for, as we now show, there is no smallest rational

number greater than all the decimals of this sequence. To show this,
suppose there were, and let it be designated by r. First, r2 cannot be less
3,
be said to define the number 1/3 in the sense that 1/3
aj
x\ as close as
such decimal of the above set
may
decimal defines 1/3,
where there are k threes in the decimal, then the sequence

d2
This sequence does not define
di,
(x H
than 2 and 1/10 2 *
since
dk = .333
their properties
to see that
in the last decimal there are k threes.
we may use
the numbers involved are rational,
all
a decimal
Thus by choosing k to be 6,
-6
we can make the decimal closer to 1/3 than 10 and so forth.
This process of approximation works well when we are concerned
where
Since
33
between the second and
continuing far enough in the sequence,
third,
we can
hence
find
number whose square is as close as we please to 2.

Now, it is rather clear that the process just described yields a sequence of numbers having the property that by continuing far enough we
can find a number of the sequence as close as we please to V2. (To paraphrase a once popular song: "No matter what you name, I can get
closer.")
This may be formally shown as follows. By our process, for
any positive integer n, we find a number x n of the sequence having the
where r
fine
is
r.oi, r.ai0 2 ,
a natural number or
number
called a real
number as
(x n
1/10") 2
a,- is
We
digit,
which do not de-
then define a new kind of
follows:
With every sequence of decimals of the form (2) we associate a real number R which, by definition, is considered to be so placed
among the rational numbers that
Definition:
1.
is
2.
If
S is any
greater than or equal to each

rational
of the sequence,
Thus
in this sense 1/3
since it is greater
number
then
.O,
[< 2 <
and each
a rational number in the sense above.
property that
x\
r.aia^,
is
S >
number
of the sequence.
greater than or equal to every
number
R.
the real
number defined by the sequence
.DO, .tJOO, .OOOO,
"
"
than each number of the sequence; and
if
is
any
ra-
34
tional
number
S >
number
greater than or equal to each
of the sequence,
1/3.
On
quence
1.12.
real
the other hand, there
no rational number denned by the
is
se-
(1),
number" includes
a rational number defines in
The term
pansion of
"real
rationals, for the decimal ex-
all
way
this
the rational number.
includes integers, since 5 can be written 5.0000
However, as we have seen,
and
also includes
it
2/5, since
It
it is
numbers which
are not rational.
We
have thus denned a
same
to define the
real
number
real
quence of decimals of the form
Thus,
Is it possible for
(2).
The answer
number?
ment
ai/10 and b 2
Many
two sequences
this because
1.12.
Complex numbers
(2)
given in the following
is
where
= 6+1.
it is
value of x2
and thus x 2
is 0,
square plus 1
it is
0, x*
real
number.
0,
and multiplication by
ex-
and
b are real
imaginary;
if
bi
numbers
a
>
1.
some have none.
0,
There
is
since for x real the least
1.
There can be no decimal whose
However, mathematicians are insatia-
bi)(c
di)
o =
if
(c
ac
adi
all
c,
bi)
These numbers have
.&1&2&3-
called a
is
complex number.
di)
If b j 0, it is called
pure imaginary.
?* b, it is called
+ di
(o
(o
pansions to be
R=
than
closer to
is
+1=
and they have created a number whose square is - 1 and call it "\A-I
i.
By definition, & = 1. Any number of the form a + bi where a
dition
Thus we may consider our decimal
real roots,
no decimal expansion of a root of x 2
Then, assuming the cancellation property for addition,

two
the
numbers will be equal if and only if the dissimilar portions of their
ble,
rational ones.)
ir,
x*
rather difficult.
we omit
Repetition of this argu-
9.
numbers can be defined as we defined negative integers.

can be shown that they obey all the laws of the rational numbers, but
define
.Oi0203
num-
the
real
or
S=
equations have one or more real roots:
and so forth.
Negative
It
all
yields the desired result.
To simplify notation, first notice that if we have two decimal expansions, we may multiply or divide both by a large enough power of 10 to
make the dissimilar portions begin immediately after the decimal point.
(This is on the assumption that we may deal with real numbers as with
decimal expansions are equal.
greater than
Since a decimal expansion can be obtained for
Though many equations have
and the other has the sequence 6999
S=
is
both sequences represent the same
if
to correspond with every se-
Theorem 1.11. Two distinct sequences (2) define the same real number if and only if they are identical for the first k digits, k being some nonnegative integer, and from that point on, one has the sequence of digits
Notice that Oi/10
than those for S.
number, we must have
theorem which we shall prove.
a000
less
bers of the sequence for S.
and we designate by \/2 the number denned by that sequence.

number which is not rational we call an irrational number.
.40000
and
35
Complex numbers
d, for a,b,c,d real,
(o
+ bci +
+ c) +
bdP =
(6
(ac
Furthermore,
+ d)i
bd) +
we
and we define ad-
(ad
+ bc)i.
the fundamental properties of the rational
num-
bers.
We
may, without
First, if
will
OiS
loss of generality,
bx
+ 2.
be greater than
However,
c is less
shows that
all
than
less
first
than
all
are zero.
6i/10
9/100
oi
>
all
will
c,
where 10c
<-
= 6j + 1,
Hence
R ^
c.
It is
a remarkable fact that any polynomial in x with complex coequal to a product of factors of the form x - c where c is a
efficients is
complex number, that
any polynomial
is,
decimals in the sequence for S, a fact which
Thus
bi
this case
Then the
1.
the other hand,
be greater than
if 6 2
all
a nxn
cannot happen.
rational
the numbers in the sequence for
On
bi.
decimals in the sequence for R.
cannot be S.
Second, suppose Oi
be
assume
then the rational number
<
8,
number ai/10
unless
all o<
will
dn-ix^
Onix
then the rational number
-2
a-2X n
+ aix + Oo,
o f* 0,
where a< are complex numbers, can be expressed in the form
after the
the numbers in the sequence for
where the
ct are
-
Ci)(x
complex numbers.
Ct)
(X-
C n),
Hence we have reached the end
of the
1.12.
36
road as far as solutions of equations are concerned

of
numbers
numbers
To
to express the roots of
we need no new
any equation whose
kinds
coefficients are
of the kinds already discussed.
We
recapitulate:
greater than
b.
Answer Questions
5.
+x
natural numbers and a

integers
is
Suppose we denned a
4 with "rational " and
bi
If x,y,
>
and
di\f a
a.
x >
b.
>
y,
0,
y
y
>
>
z implies
>
Show
that there
and notice what properties they have

2
is
two thousand years
Therefore,
ago.
this equation
had
real roots.
we invented
Since x2
we invented imaginary and complex numbers, and

as indicated above.
For
this reason, the set of
times called complete, since

and, in fact,
all
the
numbers
it
much
the real numbers and
has no real roots,
the process
is
complete,
complex numbers
is
some-
irrational
of this chapter.
number,
real
number, complex number, and
imaginary number:
-8, -5/4,
5,
V2/V8, V4,
roots of x*
6,
0,
roots of
x2
0,
X, sin 45.
2.
Are the sum and product of two rational numbers necessarily rational?
Are the sum and product of two

3.
Is the
sum
of a rational
irrational
numbers necessarily
irrational?
number and an irrational number ever ranumber and an irrational number ever
Is the product of a rational
tional?
rational?
4.
If
rational,
br
+ dr,
where
show that a =
and
>
with
d,
would hold?
is
VT.
Find the decimal expansions
8.
+ =
for each of the following rational

in
numbers,
common:
and
a,6,c,
1/3, 1/5, 1/7, 1/11, 1/13, 2/7, 3/7, 11/7, 1/17.
Can you prove any
of these properties?
Keep
these results on
hand
and d are rational numbers and
d.
r ir-
for
reference in the following chapter.

9.
If
the symbol (a/b) in Exercise 14 of Section 1.10
4- bi,
is
identified
show that
includes real numbers, rational numbers,
Classify each of the following as being a natural number, integer, ra-
number,
where i
over
Exercises 1.12
tional
not solvable in rational numbers, a highly
unsatisfactory state of affairs, which bothered the Greeks very
saw that
a=
complex number satisfying the equation x2
rational numbers.
c or
complex numbers, which of
xz.
7.
a.
However, x 2
by
b is not a
Thus we invented rational numbers so that every equation

ax - b, where a and b are integers and a ^ 0, is solvable. As it happened,
all such equations where a and b are rational numbers are also solvable in
multiple of
" irrational " replaced
z.
implies xy >
>
z are three
the following properties of inequalities
and then could
However, with integers alone, we could not solve ax = b where
numbers.
a,b,c,d real
such equations where a and b are integers, positive or negative.
all
2, 3,
"real" and "imaginary."

6.
started with the natural numbers and could not
= 6, when a and b are

Hence we invented negative
solve the equation
solve
37
Complex numbers
(c/d) as defined is identified
a.
(a/6)
b.
(a/6) (c/d)
is
identified with (a
with
+ bi)(c
-f-
(a
di).
bi)
(c
+ di).
with
Repeating decimals
2.1.
M = a + b- lOwhere n
is
the
number
"
10" 2 "
10"8n
and
of digits in the repetend
39
b is the repetend.
Then
10"
M-a
M-
10"
+b+6
a+ b
10~"
10~ 2n
10" n
4- b
2"
+
+
lO"
,
.
Subtraction gives
- \)M =
(10
and hence
is
+b -
Thus
We
and
shall begin
by exploring certain
in the process develop certain
fundamentals of the theory of
in
other connections.
which the same sequence
certain point on.
also useful in
repeating or periodic decimal (Ref

of digits recurs
Some examples
numbers
1) is
is
number
is
it is
repeating
one
1/7
2.30000
1/19
where we have underlined the recurring portion which we
Every repeating decimal may be expressed as the quotient of

two integers, that is, a rational number. This may easily be seen from
two examples:
repetend.
Example
1.
To find the rational number n equal to 4.123123123
if,
as above,
we
call
M.
any
.142857
.076923
See Exercise
.05882 35294 11764 7
1/23
.04347 82608 69565 21739 13
.03448 27586 20689 65517 24137 931

.03225 80645 16129
below.
1/17
2,
of repetends
.05263 15789 47368 421
.027
rational
a terminating decimal "repeat-
recurs in the expansion.
= 4123.123123
and hence that lOOOn. - n = 4123.0 - 4
=
Hence
and
n = 4119/999.
4119.
999n
4119,
=
Then lOOOOrc - n = 4123000,
412.34 12341234
Example 2. n
see that lOOOn.
=
=
=
=
=
1/29 =
1/31 =
1/37 =
shall call the
10"
"
A short table
and continues to recur from a
= ~a +
true that the decimal expansion of
ing" in the sense that
are
3.4156156156
fc
1/13
.1212121212
10~ multiplied by the value above for
Conversely,
interesting properties of decimals,
which not only serve to explain these properties but are
many
10"
Repeating decimals
2.1.
number
the rational
10"a
+ b-a,
10"a
Exercises 2.1
n = 4123000/9999. What is the relationship between

by which n is multiplied and the repeating decimal?
Find the rational number equal to 31.2145145
2.
If in
which implies that

the power of 10
To prove
formally that every repeating decimal
the quotient of two integers,

plying
10*A7,
N by
10*, for
let
be expressed as
k an integer (not necessarily positive), we
which we denote by M, have the property that
immediately after the decimal point.
S8
may
N be any repeating decimal.

Thus we may
its
By multimay make
repetend begins
write
M in the form
a process
of division
a remainder appears which appeared before,
why must the expansion repeat from that point on? Why in dividing one
integer by another must eventually a remainder appear which occurred
before? How may this be used to show that any rational number has a
repeating decimal?
3.
Use the development
in Exercise 2
above to show that the number of
40
2.3.
digits in the repeating part of the decimal expansion of a/b
cannot exceed
b-1.
4.
teger
that the repetend of the decimal expansion for 1/wt for any in-
occurs between the
division, if
5.
r,
Show
m is prime to
two times that the remainder
first
less
if 1 is
Compare
10, 11
replaced
There
1/29
There cannot
7.
six different remainders.
a convenient notation due to Gauss (one
is
mathematicians of
all
tion (Ref. 11, pp.
209
when 10
when divided by
Congruence notation
divided by
is
1/3, 1/13, 1/23; 1/7, 1/17, 1/37;
1/19
by
the decimal expansions of each of the following sets:
in the
than m, and hence that any such decimal repeats
1/9
102 11 102 and so forth,
more than
be any
2.3.
that the results of the above exercise hold also
a positive integer
is 1
10.
from the beginning.

6.
of
the successive remainders in the division process are the remainders
11, 11
Show
41
Congruence notation
times),
ff.).
7, 2
which
Instead of saying that 3

2
when 10
2
(mod
is
congruent to b
10
7)
10
of the greatest
is of assistance in the
divided by
is
2 (mod 7)
7,
is
above descripthe remainder
and the
6
(mod
divisible
by m.
10
we
like,
write
7).
In general,
Each of the following letters represents a digit chosen from to 9 inclusive, and no two letters represent the same digit. A fraction, equal to
a repeating decimal, is indicated. Find what values the letters must have
7.
Hence 100
EVE = .TALKTALK
DID
Number
of terms
in
the repetend of the repeating decimal
student's experience with the previous exercises shows that
For instance,
a remainder recurs, the decimal expansion repeats.

division of 11
by 7, the
first
repetend consists of the

in general, since
what
yield at
most
first
first six
follows
and seventh remainders are
when
in the
and the
This must be so,
terms of the expansion.
any point
in the division
2,
depends only on
Furthermore, division by a number
different remainders.
ders are distinct, the next one
Hence,
if
the
must have occurred
first b
before.
61
can
remain-
This shows
that the repetend of the decimal expansion of a/b can contain no
than
by 7 are the numbers congruent to 11 times

7).
In general,
we divide a by
if
remainder
in the
divided by
7,
be seen more clearly
decimal expansion of 1/7
is
divided by
is
when 102
the second remainder
remainder when 10"
5,
etc.
7.
is
if
we
notice that the
the remainder
divided by
7,
when
first
10
b,
10,
102
103 and so forth

,
the remainders in the process are
numbers congruent to a times the powers of 10 (mod b)

Another meaning which could be ascribed to the symbol
a
53 b
(mod m)
are
would be as follows: The remainders when a and 6 are divided by
the same. That this is equivalent to the former meaning can be shown
mq + r, b ms + t, where r and are nonnegative

m; that is, r and t are the remainders when a and b are
divided by m. Then a b m(q s) = r t, and, if a b is divisible
by m, then r I is also; and since r and t are less than m,
Suppose a =
as follows.
integers less than
\r
which *
result can
9, 16, 30,
those
more
digits.
The above
short for "modulo").
and seventh remainders are 4, and the repetend
the remainder at that point.
(mod being
congruent not only to 2 (mod 7) but to
is
begins with the second digit of the expansion and ends with the seventh.
In the expansion of 3/7, the
Thus we could express the results in the last part of Section 2 as

The remainders in the division of 1 by 7 are the numbers congruto 10, 102 103 and so forth (mod 7) the remainders in the division of
(mod
The
mod
follows:
ent
2.2.
read: a
is
II
(mod m) means a bis
have a
shows that r - =
t
= mq
or r
ms, which
is
t.
t\< m,
On
the other hand,
divisible
if r
lt
we would
by m.
is
and the nth
Similarly, in the expansion of 11/7,
* The absolute
value of a number 6 ia written
np gative, and +6 if b L positive or zero.
|6|
and means the number
if
ia
42
In
brief,
by
divisible
we have
shown that if the difference of two numbers is

when divided by to are the same and
just
true, since
is
remainders
to their
visible
conversely.
by
On
Exercises 2.3
Use the congruence notation to

division of 11 by 21.
2.
Prove that
(mod
if
What
7).
find the first three remainders in the
a 10* (mod
o 10 r
7),
where
>
s,
then a
KT =
-
does this show about the repetition in the expansion of
to),
as
is,
to
and the
is, let d be the
is
c)
implies that
b.
(mod
br
one difference
1,
is di-
br
is
does not always imply that
to)
= 27 (mod 21) but 2 ^ 9

by to. Then (o - b)r is divisible
two factors one of which divides r
fact that 6
divisible
of
r;
that
and r. Then r(a

6) divisible by
by m/d. However, r/d and w/d have no common
g.c.d. of in
than
if
Let d be the largest factor of to which divides
since
common factor dq which
a/7?
shown by the
b)(r/d) is divisible
factor greater
(b
is.
must be the product
other a -
(a
= a + c-
Suppose ar
21).
by m\ that
1
the other hand, ar
a ssb (mod
(mod
to the other
43
Properties of congruences
2.4.
if
is
to implies
q were such a factor, to and r would have the
Hence w/d divides a
larger than d.
b,
and
we have proved
3.
If
number
the
result hold?
7 in Exercise 2 were replaced
What
if
by
What
7 were replaced by 15?
11,
would the same
general conclusions
could you draw for 7 replaced by any integer 6?

4.
Which
Theorem
a
of the following properties of congruences hold?
(mod
2.4c.
If ar
br
(mod
to)
and d is the g.c.d.
and r, then
of to
TO/d).
We leave to the student the proof of
Specify any
exceptions.
a.
b.
c.
d.
= b (mod to) implies ac = be (mod to).

ar m br (mod to) implies a = b (mod to). (Compare Exercise 3.)
a m b (mod to) implies a + c = b + c (mod to).
a = b (mod to), c = d (mod to) implies ac = bd (mod m).
a
(mod to), prove that an = b* (mod m). Use this to

remainder when 340 is divided by 10, that is, the last digit in 3 40
5.
If
aT
is
find the
What
the last digit in 7 20 ?
2.4.
Theorem 2.4d. If f(x) is a polynomial in x with integer coefficients,

= b (mod to) implies that /(a) = f(b) (mod to).
Congruence is a kind of equality in that it has the properties of
equality given in Section 1.2 (see Exercise 13 below). Theorems 2.4a and
2.4b show that multiplication and addition are well denned relative to
congruence. From this point of view congruence (mod to) can be thought
then a
the
numbers congruent to each other (mod
ulus
is 3,
In Exercise 4 above you were asked to examine the truth of
This
to
is true,
and hence ac
You were
If
since
(mod
to),
b divisible
by
then ac
also
Theorem 2.4b.
Class of 2:
m implies
(a
other
b)c is divisible
by
number
of that class.
+c=b+
(mod to). This
-7 -4 -1
In the sense of congruence, any
all
our mod-
if
number
From
of
10
11
one class
this point of view,
is
"equal" to any
adding 7 and 4
is
the
and 1 or 5 and 2, since all three sums are in the class

-1. Similarly, 7-8
2 (mod 3).
2 (mod 3), since 1-2
For ease of reference we exhibit below the addition and multiplication
same as adding
(mod m), then a
(mod m).
asked to test the truth of

If
Class of 1:
-..,-9,-6,-3,0,3,6,
-8 -5 -2 1 4 7
be
be is also.
For instance,
to).
the integers are divided into three classes as follows:
Class of 0:
Properties of congruences
Theorem 2.4a.
each class containing
of as a separation of all the integers into classes,
^bles of integers
mod
7.
2.5.
44
of
any number is congruent (mod 9), to the sum of its digits. How
would you express this without the use of the notion of congruence?
that
would you perform the following trick? You ask a friend to

"think of a number " and then form another number by mixing the digits.
He then is to subtract the smaller from the larger and tell you all but one
7)
(mod
Multiplication
Addition (mod 7)
How
8.
11
and mod
properties
Use the
3.
tables to solve each of the following equations,
exist, for b
a.
b.
5x
c.
4x
solutions
=
m
=
(mod
13 (mod
13
is
b.
ax
b).
in order that
10.
6.
If
ab
ab).
(mod
7.
Show
Oi,0o to
and you
8,
+ ilO + Oo
11).
Is there for this result
an-ilO"" 1
which
an analogue to the trick in
is
a two-digit number and
show that N
= 27.
which N
N for
9.)
M the number obtained from N by

-
is
divisible
by
p),
where p
is
9.
Find
all
such
that (x
+ y)
xp
(mod
Prove Theorem
13.
Prove that congruence has the four properties
a),
called
4.
a prime number.
2.4d.
Teams A and B
each of the following

table
=m
(mod
b),
(mod
team B,
5.
2.5.
show that
conclusion be
if
(a,b)
(a,6)
(mod m), show that x

reciprocal
of b (mod m).
the
(mod m) and bx
that a10
is
_1
+
+ aJO
+ a,,-! +
+ 01 +
a_ilO*
congruent to a n
4- a*,
ao
=d >
ac
men
contain six
of equalities given in
each.
How
does the multiplication
a plan by which each man of team

matches being played simultaneously?
7) give
six
Prove that a number which
as its last digit: 0,1,4,5,6,9.
What would your
in a, a*_i,
12.
what
(mod m).
(mod 7/1).
and hence a could be
integers,
If
Show
of
Given x
(mod
For
Section 1.2.
in solving the equations above,
5.
mm
one.
not solvable, give reasons for the lack of solution.
as far as you can in proving your conjecture.
and
digits 0,4,8,
(For other tests of divisibility, see Ref
Exercise 8?
Go
sm
you of the
a10"
congruent (mod
6).
congruences be solvable?
Find an expression
is
is 7.
b).
would you guess to be the condition
him that the other one
numbers
3 (mod
Judging from your experience
a.
tells
interchanging the digits,
7, 11, 12.
Where a congruence
4.
when
He
12.
do the tables mod 7, mod 11, and mod 12 have?

Can you prove any of them for a general modulus m?
What
2.
may
him the missing
9.
mod
he
instance,
tell
choose 57689; he scrambles the digits to get 98567 and
subtracts, getting 40878.
Exercises 2.4
Write out the addition and multiplication tables
You then
the digits in the difference.
inform
45
Congruential equations
is
plays each
man
a square must have one of the following
What
can you say of the
last
two
digits?
implies
It is
1?
(mod w)
where the a,-'s

(mod 9) and hence
are
easy to show that the congruence a
solvable, since
b a =
We
is
a solution.
(mod m)
is
always
We partially prove
= b (mod m) is solvable if the

and then there is only one solution (mod m).
give the proof for m a prime number p and leave to the reader the
Theorem 2.5a.
a and m is
8-c.d. of
The congruence ax
1,
46
proof for
Let a be an integer not divisible by p and write
composite.
To
the numbers
the
a, 2a, 3a,
(1)
No two
of
imply x
(p
them are congruent (mod p), since xa = ya (mod p) would

y (mod p) by Theorem 2.4c. However, there are p of them,
p) to one of the
and only
if
(a,m)
When
1.
ax
by
multiply
(mod m)
is
solvable
if
then b
is
b,
and to solve the congruence
essentially the reciprocal of a,
we need merely
congruence has a solution
this
only
b and have abx
set is then congruent to 1,2,3,-
1)
(mod
p)
The congruence ax
the g.c.d. of a and
divides
a*~\p
Since (p
1)
is
by the
1)!
omitting pa from
(1),
factorial
(mod m),
6c
as
was
of
is
solvable
if
order,
congruent to
(p
(mod
1)!
p).
to get our
we may
desired result.
we have
which means that in the division of
the remainder
(mod m)
is
This theorem of Fermat establishes a property of the repeating part

the decimal expansion of 1/p when p is a prime number not dividing 10,
10"- 1
=x
some
in
not divisible by p, Theorem 2.4c shows us that
(p -
when o
(mod
by
p),
the (p
p,
1)" remainder
1)" remainder in the division of a/p

is
is
is 1.
the same as
divided by p.
and
If it is solvable there are (a,m)
b.
Thus we have
Furthermore, the
is
numbers in
consider the set of
(mod m)
Another consequence of Theorem 2.5a
if
(p
for
seen in Exercise 6 above.
Theorem 2.5b.
The
list.
divide
numbers (1).
Notice that Theorem 2.5a implies that ax
this,
and hence the product of the numbers of the set
l)a, pa.
and thus p classes represented. Hence, since there are only p classes
(mod p), one of them must be the class of b; that is, b must be congruent
(mod
prove
47
2.5.
Exercises 2.5
distinct solutions.
To prove this,
(mod m) then ar bis divisible by m,

m would have to divide b. Suppose
(mod m) if and only if
suppose ar
and any factor common to a and

(a,m)
g and
g\b.
Then ax
{a/g)x
Since (a/g,m/g)
solution
is xo,
1,
b/g (mod m/g).
then the solutions of the congruence ax

xo,
+ 2m/g,
m/g, z
a*
(g
Use the method of proof
5x
2.
Prove that the number
suggested by the proof, but
this
if
(mod m)
are
3.
tion of such a congruence.
does supply a definite
The
way
than that
of finding
a solu-
11).
of
terms in the repetend of the decimal ex-
p does not
p whose every
If (10
(mod 10)?
5.
reader should apply this method to the
solution of the congruences in Exercise
2.5a to solve the congruence
Theorem
divide a,
Prove that for every prime number
multiple
l)m/g.
of solving a congruential equation

it
(mod
of
a divisor of p
is
1,
p being
a prime number.
4.
There are better ways
pansion for a/p, where
has just one solution (mod m/g), and
this
1.
3b and 3c above.
l)/p
If
2 and
5,
there
is
digit is 9.
a positive
is
p, except
3 (mod 10)?
what
7 (mod
integer,
If
is
its last digit if
10)?
If
9 (mod
10)?
Prove Fermat's Theorem by showing that

(1
+1+
l) p
(1
+1+
1)
(mod
p),
A similar method of proof may be used to give the following useful result
sometimes referred to as Fermat's Little Theorem (Ref.
11, pp.
54
ff.).
where the number of

.
Theorem
2.5c.
If
is
not divisible by a prime number p, then

ar~ l
1.
(mod
p).
Prove that a"
l's is less
than
p.
a (mod p) for p a prime number.
Prove that if m p + n p =
(mod p) then
a n odd prime number. (Ref. 3, p. 50.)
7
mp + n" =
(mod p2 )
for
48
State and prove the theorem corresponding to Fermat's Little Theorem
8.
for a composite
1,2,3,-
What
number m instead of p.
by the positive numbers
m+
one value of x
is
common
prime to m.
(m
y (mod m)
<
< m/2 and
Using the same principle as

ax
in Exercise 9,
If
any number prime to m, as
is
see that
aa h aa t
p,
a prime not dividing
1 1
Can
12.
Does a2
is
<
<
Exercise
show that
your answer
if
m is a
we
function denned as follows:
Theorem 2.6b.
less in
absolute value than
a composite number?
p,
which a
(mod m)
or
= -b
first r
(mod m)T
prime or a power of a prime?
prove
number
as Euler's
and,
if
m>
1, <t>(m) is
m and prime to m.
To prove
for
this,
If
we
(a,m)
1,
a (mod m),
t
is
a direct conse-
then a*
(,n)
let
follows:
(mod m).
suppose we write out the powers of o (mod m).
this,
(mod m) and
of a
may
o u (mod m) with
would be positive and
>
leas
The
> n
than
r.
be written in rows as follows:
*\a f
,
a'
contrary to the assumption about
tfrt*
air+ \ a2 ^,
.. -,
.
0'
1,
.,* =
.,*
I,
li
For instance,
and prime
where only the
last
power in each row is congruent to 1 (mod m) If, then,

at the end of one of the rows and hence
.
(mod m), it must occur

must be an integral multiple of
Euler's Theorem we have
a'
Corollary 2.6b.
ai, a,,
s.
a'
if
denned to be
composite numbers analogous to Theorem 2.5c u
Theorem and may be stated as
Theorem 2.6a.
(1)
of positive integers less than
theorem
m and r is the least positive exponent

e 1 (mod m), then r divides
prime to
is
(mod m) and
the so-called Euler Phi-
is
4, since 1,5,7,11 are the positive integers less than 12
powers are distinct (mod m), since
would imply a 1- "

r,
If
a,
theoretic function
<f>(l)
a complete residue system (mod m).
call
o^
An important number
aiaj
generalization of the result in Exercise 2, above,
To
(mod m) imply a
b-
p)
a.
be extended for
by the product of the o's, we have our theorem.

what we call a reduced residue system
The
(mod m); in fact, a reduced residue system is any set of <t>(m) numbers
congruent to the numbers of (1). Similarly, any set of m numbers incongruent (mod m)
ra/2.
known
a's in (1) constitute
The Euler Phi-function
<t>{\2)
,aa
and, dividing both sides
Hence the powers
to 12.
a'(aia 2
for
y/p for
the
Theorem
in the proof of
quence of
(mod
has a nontrivial solution in integers x and y
2.6.
m and having no factors

m and
and no two are congruent (mod m). Hence they are

congruent in some order to the numbers of (1). Taking the product of
both sets, we have
l)/2,
has a solution x,y for which
What
we
2.5a,
than
less
with m, in other words, the positive integers less than
are all prime to
l)/2 values:
congruent to one value of y and hence that
0.
is an odd integer.
If a is an integer, show that no two of the re(m + l)/2 values of x + o are equal. Since (m + l)/2 + (m + l)/2
1 and there are only m different numbers (mod m), show that at
sulting
in
be the positive integers
where
least
<t>(m),
m is a composite number not divisible by 5 or 2?
0, 1, 2, 3,
and prime to m.)
the last restriction necessary?
is
Let x and y each take the following (m
9.
than
are the consequences of this theorem in connection with the decimal
expansion of a/m, where
Why
(In the proof, replace the set

less
where
'
49
The Euler Phi-function
2.6.
(mod w), then
r is
r.
This proves our theorem.
If r is the least positive
a divisor of
</(m).
We
In view of
exponent for which a r
call r
the exponent to which a
50
belongs (mod m)
of a/m,
where
2.8.
This immediately implies that in any decimal expansion
m is prime to
10, the
number
Further properties of repeating decimals
51
Further properties of repeating decimals
2.8.
of digits in the repetend is a
Consider the repetend occurring in the decimal expansion of 1/7,

If we multiply this number by 3 we get 428571, by 2 we
divisor of 0(m).
namely, 142857.
2.7.
Factorization
by means of
Certain types of numbers
Theorem and the

visors of n = 5 7
must divide 7
Now,
that case, 5 7
since 5 7
factored easily by use of Euier's
(5
2 3
)
Suppose we seek the
from
we know
for
51=0
-1 =
different
2.6b
(mod p),
(mod p).
1
would imply
may be
From Theorem
1.
must be
p
by Corollary
7 2, r
if
cyclic
properties of congruences.
exponent of 5 such that 5 r
that
if
r is the least
any prime p dividing
(mod
p)
n,
then
and hence p = 2. In
4 (mod 8). Thus,
-5-1 = 5-1 =
and a
divisor of 7, hence r
2.6b the only prime divisors of
n beside 2
7.
Hence
are primes p for
with the
gives us the
multiplication
4,
hence the term
cyclic.
Similarly,
if
(mod
cyclic
multiply the repetend of the ex-
1 to 16,
we would have
permutations of the digits in the repetend.
Why
is
the sixteen
this so?
Can
same conclusion be drawn for 19,23,13,31,37? If so, why; if not,

what can be said? (For example, if we multiply the repetend for 1/19 by
the numbers from 1 to 18, do we get the eighteen cyclic permutations?)
7);
primes p = 1 (mod 7). However, p

more than an even multiple of 7 that is,
is,
There
is
odd and therefore must be
p
n/4 has a prime factor p
less
(mod
is
another rather remarkable property of such repetends.
instance, the sixteen digits in the repetend of 1/17 are the digits
inclusive and, in addition, the six
The repetend
of
than n/4
it
must have one not greater than
are those less than 139 and congruent to 1
(mod
14).
These primes are
29, 43, 71, 113, 127.
whose
tion shows that
= -64
is
not divisible by 43, 71, or 113, and hence that n/4
-30
= -1
(mod
29). Similar
computais
a prime number.
method enables us to consider n = 57 -+- 1.

A little computation shows that n = 2 (mod 4), n is divisible by 3 but not
(mod 29). Dividing n by 2-3-29 gives
by 9 and, by the above, n =
slight variation of
449, which is a prime number.
to 9 and
Suppose we have any prime p

and whose repetend N contains p 1 digits. If we
by the numbers from 1 to p - 1 in succession, each digit oc-
last digit is 7
N
N will be transferred once and only once to the right end of the
number. Hence the digits in N will be just those that occur at the right
end of the numbers cN for c = 1,2,
p - 1. The last digit in N will be
7 (see Exercise 4 of Section 2.5). If we multiply N by the numbers from
multiply
curring in
(-4) 1
to
in the repetend of 1/7.
1/47 has 46 digits which consist of four sets of
For
the six which occur in the repetend of 1/7.
because certainly p, a factor, implies that either p or n/4p is not

greater than Vn/2. Thus the only possible primes which need to be tried
Now 5 =
which occur
from
14).
Vn/2
we
pansion of 1/17 by the numbers from
9,
If
by 3
number formed by the six digits beginning

by 2 gives the number composed of the six
digits beginning with the 2, and so on.
One may also accomplish the
same result by writing the digits of the repetend in order around a circle;
multiplication
the
p- 1 =0
1
142857142857,
di-
which
that
and so on. For every multiplier from 1 to 6 we get a so-called

permutation of the digits of the original repetend; that is, if we
write two repetends in succession:
get 285714,
Theorem
Euier's
to 10, inclusive, the last digits will be congruent to 7c
(mod
10) for
10 and hence will be the numbers 0,1,

1,2,
9 in some order.
This will be true for each interval of 10 less than p. This accounts for the
c
sets of
-,
numbers
multiplying
The remaining sLx digits will be obtained by

p
1,
2, p - 3, p - 4, p - 5, p - 6, whose last
respectively; and N = 7 (mod 10) shows that the
to 9 in iV.
by p -
digits are 6,5,4,3,2,1,
products are congruent, respectively, to 2,5,8,1,4,7, which are the digits

occurring in the repetend of 1/7.
52
2.9.

Recall that
(6,c)
1,
we
if
the g.c.d. of 6 and c
and
call 6
we
is
Thus, each positive number
formula for 0(m) for any integer
write
(b,c)
There
c relatively prime.
is
(mod
And
g.
6c) to just
if
m which may be obtained easily once we
than 6c and prime to 6c
is
congruent
and hence, there are exactly
rs
such num-
less
set,
This completes the proof.
bers.
a rather simple
one of the
53
2.9.
Let
be expressed as a product of powers of distinct primes as
follows:
have proved
Theorem
To
and
2.9. If b
prove
let Ci,c 2 ,
and c are
this, let 61,62,
,c
prime integers,
relatively
,6 r
<j>(bc)
4>{b)4>(c).
be a reduced residue system (mod
be a reduced set (mod
6),
Theorem
shows us that
2.9 then
Exercise 4 below,
c).
<p(m)
4>{p\
l
)
0(p2 )
<Kp
which, by
)>
is
The congruence
bx
+ 61 =
=
has a solution x xj for each c 3
Cj
(mod
1,2,
c)
,s,
since
6!
(6,c)
may be written
and which
1.
<fr(m)
= m(l -
l/pi)(l
l/p 2)
(1
l/p f ).
Thus
bxn
(1)
are
61
congruent to
all
c,(mod
6x12
+ 61
(mod
61
bxu
6)
and congruent
bxi,
Exercises 2.9
+ 61
in order to Ci,c 2 ,
The congruence
62
has a solution z 2 for each value of

,
6z 2
(2)
all
(mod
6<
Thus
(mod
2.
Prove that
6)
(mod
c)
and the members
bx 2 ,
and congruent
+6
of the set
in order to Ci,c 2
,c
in order to Ci,c 2
,c,
numbers congruent to
(mod c). Each of these
c);
numbers.
We
by
r.
3.
Prove that
if
have thus shown the existence
4.
Prove that
5.
Show
6.
Show
7.
If
All are
No
3.
is
0(m) = m/2?
by
that 2 P
is
is
last digit is 3
= p"
n_1 .
for
p=
11
and
23,
and
if
the repetend of 1/p contains
said about the digits in the repetend (see Section
the case for
The repetend
(mod m) and
y.
a prime number.
what can be
is
= 1
prime for p = 13 and composite
1 is
bx
(mod m), where

(mod m) implies that y is divisible
a prime number, <(p")
a prime whose
is
What
2.8)?
8.
that 2
16
the least value of x for which a x
if r is
if
digits,
of
set of rs
numbers having the
9.
ha*,
If ai,a,
1.
(mod m) then 2r divides
When
p whose
last digit is 1 or 9?
of 13 contains only 6 digits.
Can
the above results be
extended to such a case?
following properties:
2.
odd?
<f>(m)
prime to 6 and
and (mod
m is
the least value of x for which ax
if r is
for each 6* there will be a set of s
and congruent
6)
c,
62, 6x22 4- 62,
congruent to 6 2 (mod
c).
Cj
c and hence to 6c; no two are congruent both (mod 6)

and hence no two are congruent (mod 6c), and there are rs
of them. Furthermore, every number prime to 6c must be congruent to
one of the 6< (mod 6) and one of the c,- (mod c) and hence is one of the rs
is
For what numbers
a and 6 are prime to m, then av
bx
are
c).
prime to
6c.
two are congruent (mod 6c).

Every number prime to 6c is congruent to one
of the set.
ba
10.
Show
What
(mod m)
is
,a t is
that x 17
is
a reduced residue system (mod w), show that
a reduced system
-x =
the largest
for all x?
(m,b)
if
(mod
number
15- 17-32) for all
ba\,
1.
you can
odd integers
find such that z
19
x.
54
1
2.11.
Show
2.
and
that the
number
terms between
of positive fractions in lowest
and whose denominators are not greater than n
Then F(2)F(3) =
to F(6)
is
if
is
and an
Wilson's Theorem
/(1)/(1) +/(l)/(2) +/(l)/(3) +/(2)/(3),
To
multiplicative.
prove the theorem
55
allied result
which
we assume
is
equal
(6,c)
and have
+ 0(3) + --- + *(n).
*(2)
ZM - /() - E/W/00 -
F(bc)
d\be
1
3.
Write the numbers bx
+b
in the proof of
Theorem
2.9 for b
14.
r\b
6i,6 2 , -,b t
If
-,c,
constitute a reduced residue system
a reduced system (mod
reduced system (mod
(mod
some
c).
Cj
c),
(mod
(mod
6)
Prove that any number d congruent to some
6,-
(mod
c) is in
6c) is
6)
and
prove that every number d in the
congruent to some
6,-
a reduced residue system (mod
6c)
and some C/
(mod 6) and
How may
tion
/.
Now 2>(d) =
this
(mod m)
1
6.
17.
is
divisible
by
m m
show that
the least
numbers
of
power
of a reduced residue
a such that a
of a'
such that
(mod
(a')
p)
2.8.
and
(mod
<t>(p)
p are powers of
*Z,<t>(d)
paragraph of Section
which
(s,r)
1,
+ p-l +
</>(p )
+ 0(pn
),
p being a prime.
p,
system
2.
in the first
power
r is the least
>
if
Answer the questions

If r is
of the
d\p
n
visors of
Prove that the sum
=f(6)-nc).
s|c
The first equality is by definition; the second holds since any divisor of
6c must be the product of a divisor of 6, and a divisor of c, since (6,c) = 1.
The third equality results from the multiplicative character of the func-
be used to prove Theorem 2.9?

15.
r\b
E/wl/(i
3.
Ci,c 2 ,
r\b
5 and
since the only di-
This shows that
-p + p -p +---+pn - p""
2
p",
a particular case of the following theorem.
is
>( d =
Theorem 2.10b.
d\n
p).
From Theorem
we know
2.10a
that ]</>(d)
is
a multiplicative func-
d\n
2.10.
Multiplicative functions
tion, since
The Euler ^-function is called a multiplicative function, since (6,c) - 1

There are various number-theoretic functions
implies <t>(bc) - 0(6) tf>(c).
having this property. Two such functions are <r(n), the sum of the divisors
of n, and d(n), the number of divisors of n (see Exercises 9 and 10 after
the Euler Phi-function
Tm =
d\n
where n = p^p*
!>(<*)
d|p,
is
multiplicative.
E*(d)-
d|p2
EtCOi
rf|p,
r
However, each sum
pr
contained therein, which proves our theorem.
Hence
is
equal to the power of p
Section 2.13).
Recalling that d\n
means "d divides n," we prove a general theorem
on multiplicative functions which we
shall find is
very
There
Theorem 2.10a.
If f(n) is
a multiplicative function and
F(n)
is
First of
right is the
F(fi)
/(I)
of f(d) over all divisors
+/(2) +/(3)
+ /(6),
F(2)
ff.),
is
an interesting theorem called Wilson's Theorem (Ref.
which a trick helps us to prove
11,
easily.
2.1
a.
If
is
the notation should be explained.
summation
pp. 259
Theorem
a multiplicative function.
all,
allied result
a prime, (p 1)! = 1 (mod p).

First notice that the converse of this theorem is easy to prove;
namely, (m 1)!
1 (mod m) implies that is a prime, for otherwise
= /(d),
d\n
then F(n)
and an
Wilson's Theorem
2.1 1.
fruitful.
The quantity on the
d of n.
For instance,
/(l) +/(2), F(3)
= /(l) +/(3).
{m
1)
has a factor in
common with
necessary and sufficient condition that

below,
in.
Hence the theorem gives a
be a prime.
See Exercise
12,
56
To prove
the theorem, write out the set of numbers

1,
2,3, ,?>
aliquot parts or proper divisors.)
Any number which
-2, p-1,
all its
"mate" of 2 be that number a of the set for which 2a 2 = 1

(mod p). (We know there is such a number, since 2 is prime to p.) Conversely, the mate of a t is 2. Then we can associate 3 with its mate, and
so on. If a number b were its own mate, we would have 62 = 1 (mod p)
that is, (b 1)(6 + 1) =
(mod p), which implies that b = 1 or p 1.
Hence omitting these two numbers, we have
and
let
the
(p
2)!
)()
(2o,)(-
()(modp),
to 1
(mod
p)
(p
1)!
1(1)(1)
is
10!
perhaps
is
(l)(p
1)
= -1
(mod
^ M-M-l-(-l)
(mod
11).
A similar trick results in a proof of
It
+ 4 + 7+14.
sum of
has been shown
is,
half the
less
other hand, Euclid developed a formula for
all even perfect numbers.

be found as follows: Let n be a perfect number written in the
form n = 2~ 1wi, where m is odd and q is greater than 1. Then o{n) = 2w
This
may
implies
a(2->(m) - 2m;
is,
2s,
= 2m.
l)tr(m)
must divide
1, it
<r(m),
and we may write
which yields
(2
clarified
l(2-6)(3-4)(5-9)(7-8)(10)
57
than five different

prime factors, but no odd perfect number has ever been found. On the
o-(ra)
p).
number.
no odd perfect number contains
Since 2" does not divide 2
Similarly, 28
of its aliquot parts, that
called a perfect
(2
by an example. Take p= 11. Then

2.r = 1 (mod 1 1) has the solution 6, and hence 2 and 6 are mates.
The
mate of 3 is 4, and so on. Then we can write
This
sum
congruent
is
Hence we have
the
positive divisors,
(Ref. 13) that
that
where each parenthesis contains two numbers whose product
is
numbers
Perfect
2.12.
Certainly
we have
just learned that cr(w)
l)s
= m.
and (2
whose sum is 2% whereas

Hence s and 2 1 are the
only divisors of m, a fact which shows that s = 1 and 2 1 is a prime
p. In that case, n is a perfect number.
Hence the even perfect numbers
has two divisors:
l)a,
= 2's.
are those which can be written in the form
Theorem 2.11b.
p
is
If r
(p
l)/2, then
(r!)
m -1
(-l)'
(mod
p)
if
This is proved by pairing the numbers 1,2,

p - 1 as follows:
and 1, whose product is 1; 2 and p 2, whose product is congnuiit
to 2 3 and p 3, whose product is congruent to 2
(p l)/2
and (p + l)/2, whose product is congruent to 2 since (p + l)/2 =
Hence (p
However, there are r minus
(mod p).
2.12.
1)!
signs,
(-l)(-2 2 )(-3 2 )
which shows that (p
(-r2 ) (mod
1)!
p).
( l) r (r!) 2
o(n) is the
erty that
it is
the sum of
its
a prime number.
since
if
ab, 2"
would have
The numbers
it
with q a prime are called Mersenne numbers, since
was Mersenne who conjectured that
for q
257 only the following
2, 3, 5, 7, 13, 17, 19, 31, 67, 127, 257.
of the positive divisors of n.
(The positive divisors
must itself be a prime number,
values of q yield prime values for 2"
an interesting result connected with the a function, where
sum
with 2
number.
The
is
1)
as a factor.
Perfect numbers
There
However, q may be a prime number without 2 1 being a prime

For example, 2" 1 is composite (see Exercise 5 in Section 2.9).
l)/2.
In fact, q
2a
p-(p-
2-(2
(I)
a prime.
The number
divisors not including 6; that
of a
number
less
is,
6 has the prop6
+ 2 + 3.
than the number are called
its
fact that
Mersenne made
five errors,
namely, q
67,257 yield com-
and q
61,89,107 yield primes, has been shown (Bulletin
American Mathematical Society [1948], Vol. 49, pp. 378-380).
Recently five more Mersenne numbers have been found with the aid of
posite numbers,
of the
high-speed calculating machines, namely, for q
521,607,1279,2203,2281.
58
two
(See
articles
S. Uhler, Vol. 18
[1952] and Vol. 19 [1953]
The term "perfect number" had its origin in the mystical meanings
numbers. Amicable numbers are similarly defined. (Ref. 11, pp.
96-100 and Ref.
If
known
that regular polygons with 2,4,8,16,
constructed with ruler and compasses
draw
straight lines
sides
may
be
when the former is used only to

arcs. The regular pentagon
and the latter only to draw
may
also be constructed with ruler and compasses.

In fact, it can be
proved (Ref. 16) that the regular polygons which can be so constructed
are those of
2r
a prime, show that
1 is
(mod
Prove that
= -1
(mod
Suppose
4.
if
4).
is
2*pip 2
a power of
2.
less
in the proof of
z2
if
m -1
(mod
of 2
2r
1 is
l)/2 and
r!
= -1
(mod
p),
then
2.11a, the
numbers
1,2,
,p- 1
Show
a prime,
it
may
and hence
5.
17
be shown (why?) that
r is
a power
Use the
7.
The
first
(mod
the
first
is
odd or even.
show that
if
x2
= -1
(mod
solva-
four even perfect numbers.
four even perfect numbers end alternately in 6 and
number ends
is
p)
4).
property hold in general?
perfect
n a nonnegative integer are called Fermat numbers. Fermat conjectured that all such numbers are primes. For n = 0,1,2,3,4 we have the
results of Exercise 4 to
Compute
this
= (-ly-^modp);
in the respective cases, r
6.
and the numbers

2*"
p) has a solution, then
(p-l)!
ble then
If
(p
Theorem
than 20 sides which can
those having
3, 4, 5, 6, 8, 10, 12, 15, 16,
sides.
converse of this statement true?
were paired so that the product of each pair is congruent to -1.

that if x 2 = -1 (mod p) has no solution, then
that
r,
Thus the polygons of

be constructed with ruler and compasses are
1 is
a prime, r
Is the
with k a nonnegative integer and with the p's distinct primes such that
2.
(p-1)!^ (-l)'(modp);
n
a power of
If r is
where
sides,
r is
denned as in Theorem 2.11b, prove that (r!) 2

-1 (mod p) if
2
and
(r!)
=
4)
that
1
(mod
m
p
-I (mod 4). The
p) if p
former congruence proves that x 2 m -1 (mod
p) is solvable if p m 1
(mod 4).
2.
3.
It is
59
Exercises 2.13
14, p. 83.)
Fermat numbers
2.13.
Fermat numbers
Mathematica.)
of Scripla
of
by Horace
2.13.
If not, is it at least true that
8.
Does
every even
in either 6 or 8?
for
respective primes
However, Fermat's conjecture
known which
17
257
In
is false.
fact,
Since an equilateral triangle and a regular pentagon may be constructed with ruler and compasses, we can divide
the circumference of a
circle into 3 equal parts and 5 equal parts
by such construction. Show
how
65537.
no other values
of
are
has been shown that Fermat numbers are

composite for the following values of n:
yield primes.
8.
It
this may be used to divide the circumference into 15

equal parts
without using the result on page 58. Generalize this result
to show that if
regular polygons of o and b sides, respectively, may
be constructed, where
(a,b)
9-
(Ref. 6, pp. 14, 15.)
11
12
For example, 2^
18
23
+1=
36
38
73.
tion.
2>(d).
641
6700417.
a In
1,
then a regular polygon of db sides
Prove that
<r(n),
the
sum
may
be constructed.
of the divisors of n, is a multiplicative func-
Find a(p k ) and use these
results to get formulas for <r(n)
and
60
Prove that
0.
number
d(n), the
of divisors of n, is a multiplicative func-
Find d(p n) and use these results

X)d(n), where x is a given integer.
tion.
Is the
product of the divisors of n a multiplicative function?
Show
that
For what numbers

2.14.
1)!
is (ra
Corollary 2.14b.
oi,a 2
,a r are
any
If rai,m 2 ,
,ra r
are relatively prime in pairs
r integers, the congruences
below have a
61
and
common
solution
12.
(ra
and
to get formulas for d(n)
11.
if
2.14.
= 1 (mod m) then m
1) = (mod 7n)?
a prime number.
is
a\
(mod
wii),
o 2 (mod
Corollary 2.14b often goes by the
ra 2),
name
-,
xm
a,
of the Chinese
(mod
r).
Remainder Theo-
rem, since such problems have been found in ancient Chinese literature.
In fact
much mathematical development
in
China
in connection with the
theory of numbers antedated by centuries corresponding mathematics in
There
may
which
a.d.
an old puzzle attributed to Sun-Tsu
is
in the first century
Europe.
(Ref. 4, Vol. 2, p. 57.)
be written as follows: Find the least two positive integers
having the remainders 2,3,2 when divided by 3, 5, and 7, respectively.

Let us see how this puzzle may be solved. All numbers having the rewill be of the form 3n + 2 for some integer
3n + -2 = 3 (mod 5), it will have the remainder 3 when divided by
Then 3w = 1 (mod 5), which implies that n m 2 (mod 5) or n = 2 + 5x;
mainder 2 when divided by 3

n.
If
5.
that
is,
3n
= 8
This gives
4- \bx.
quired remainders (mod 3) and (mod
gruent to 2 (mod 7); that
is,
\hx
numbers which have the rewant 8 + 15a; to be con2 (mod 7), or x = 1 (mod 7).
all
5).
=
+
We
1 + ly and 3n +
105?/.
Thus the two smallestnumbers satisfying the conditions are 23 and 128. Notice that
we began with a solution for 3 and got one for 5, thus one for 15; our last
This gives
x =
2 = 23
positive
mod
step gave a solution
That
3-5-7.
this process
successive moduli are relatively prime is
Theorem
2.1 4.
a number x
is
To prove this,
r
c),
bl for
since b
be.
(mod
and
Theorem
b)
and
notice that x
an integer
t.
if
and
and
(mod
(mod
s are
if
so that r
is
of the
form
(mod
bt
prime.
x and y are two integers which satisfy the conditions
if a; is
a solution, x
by
the is
and by
and hence by
a solution for any
t.
This
less
than
yields the
Corollary 2.14a.
There
is just
one nonnegative integer x
be satisfying the congruences of the theorem,
gruent to x (mod
be).
In
know how
when she counted them two at a time there was
one egg left, and similarly when she counted them three, four, five, and
six at a time; but in sevens there were not any left over.
What is the
smallest number of eggs she could have had?
bill
to the rider's father, she said she did not
2.
any or
If
all of
the following sets of congruences have a solution, find
the solution
b.
=
=
c.
a.
= 1 (mod 7),is2 (mod 18).

(mod 10), x = 7 (mod 15), x m -4 (mod 6).
(mod 4), x = 5 (mod 7), x m 7 (mod 9).
3 (mod
6),
any two numbers, there
and only
t
with a basket of eggs was knocked down by a bicycle.
eggs she had but
3.
b) if
A woman
many
the
c)
2.14, their difference is divisible
Furthermore,
if
presenting her
shown by
Then we may determine
c are relatively
Notice that
of
Tf (b,c)
always works
Exercises 2.14
and
all
solutions are con-
Describe a method of finding solutions of a set of congruences in which
the moduli are not relatively prime.
3.2.
1 if
28
is
Solution of linear Dlophantine Equation
In
positive.
+ Son +
5p =
general solution
is
shows that n
fact, if
if
= 4
we replace c by 4
35n
or
53,
-f-
5w,
5p
p = 5
and p are
solving the congruence 53
5n
7),
63
two unknowns
in the equation
we have
+p=
So our
is,
In
5.
In for integer values
This
of n.
It is also possible to begin
positive.
5p (mod
25; that
in
but this
is
by
longer.
Exercises 3.1
1.
Solve the following Diophantine Equations, indicating
all
solutions
in integers:
3.1.
b.
Introduction
c.
typical puzzle problem is the following:

cows at $70 apiece and pigs at $50 apiece.
much
each did he buy?
of
of pigs,
we have
If c is
if
number
$530.
all,
of
How
cows and p the number
of
50p - 530,
equivalent to the equation
is
7c
Now,
He
bought a number
spent, in
the equation
70c
which
the
A man
we
+ 5p
A man received
2.
a check for a certain amount of money, but on cashing
number of dollars for the number of cents and

Not noticing this, the man then spent 68 cents and discovered
to his amazement that he had twice as much money as the check was
originally drawn for.
Determine the amount of money for which the
check must have been written.
it
the cashier mistook the
conversely.
Under what conditions on a, b, and c will ax + by =

more solutions in integers? Supposing that Xo and yo
solutions, what formulas will include all solutions?
have one or
is
one pair of
3.
53.
could have fractional values of the unknowns,
we
could assign
any value to p and solve for c. However, the values of p and c must be
integers. Any equation whose solutions are restricted to integers is called
a Diophantine Equation, after Diophantos (Ref. 11, p. 179). In this case,
3.2.
a further restriction that the solutions must not only be integers

It can be solved by trial if we notice that we need
to find a c such that 53

7c is divisible by 5. If we try in order c = 1,2,3,4
there
+ 11?/ = 92.
10s + Uy = 39.
20x + I7y = 93.
5x
a.
is
A method of
two unknowns
solution of
a linear Diophantine Equation
in
but cannot be negative.
we
find that the last is the first acceptable value, the values c
not acceptable and any larger value of

the only solution which
There
53
is
is
a better
7c divisible
way
is
-f 5n.
Now
the problem
p =
4,
to solve the problem.
4 (mod
2c
5).
(mod 5); that is, 8

Hence all values of
n must be nonnegative
5,6,7 are
would make p negative.
is
to solve the congruence 53
equivalent to solving 3
has the solution

4
by 5
fits
if c is,
and
it
Thus
5.
Notice that to make
= 7c (mod 5). This

s 2c (mod 5), which
c
must be
of the
form
cannot be as great as
An
equation ax
higher degree than
+
1
by
in
c is called linear, since it
x and
y.
contains no terms of
Solving this equation
equivalent to
is
(mod 6) This is solvable if and only if the

g.c.d. of a and b divides c, from Theorem 2.5a.
If the g.c.d. of a and b
divides c, we can divide all three letters by the g.c.d. and have an equation
in which a and b are relatively prime. Then if x is one solution of the
congruence ax = c (mod b) with (a,b) = 1, all solutions are of the form
solving the congruence ax
#o
the
bn, since ax
ay (mod
equation gives a(x
b)
implies x
+bn) +
by
c;
y (mod b).
that is, by
Putting this in
ax
abn
64
However, axo
implies y
by
yo
c implies that c
axo
by
and by = by
Exercise 14, below).
7x
Theorem
more
solutions
relatively
The Diophantine Equation ax
3.2.
and only
if
if
prime and xQ ,y
by
the g.c.d. of a and b divides
one solution,
is
all
c.
has one or
If
a and b are
Xo
+ bn,y =
yo
The equation ax
by
c represents
and y
a line in analytic geometry.
will correspond to the points
point whose coordinates are integers a lattice point.
3.r -f
lattice point in the first
Qy
co-
of
ways
it
may
3.
212.
We shall
be done by means of continued
frac-
a number congruent to b
(mod wi) which has a factor in common with a and hence which can be
divided into both sides of the congruence. Thus we can reduce the given
congruence to one with smaller modulus and apply the method again and
again to congruences with smaller and smaller moduli. We illustrate the
method by solving
ax
(mod m) aims at each stage
19x
tion
and,
to find
1.
We seek
Since 73
19).
is
is,
residues in absolute value), the last congruence
(mod
2.
(mod
(mod
becomes
Sy
How
is
1?
What is
on the
far apart are the lattice points
a and
b are relatively prime,
by
how
21y
37y
+ by =
53?
7x
ci
Use the results of Exercise 4 to show that

= 1 and ab < c, then the line ax + by =
first
7x
line
far apart will be the lattice
if a,b,
quadrant not on either
and
What does
axis.
c are positive,
one
c has at least
(a,6)
point in the
c?
points on the line ax

5.
15.t
625).
the corresponding general result for the equa-
positive solutions of the Diophantine Equation 23a:
all
If
(mod
this
lattice
imply about
the positive solutions?

6.
If (a,b)
tice points in
with ac
>
If (a,b)
0,
1,
show that the
the
first
quadrant
and none
1,
and
if
a,b,
ab
>
and
if
ax
ab
<
+ by =
has infinitely
0, finitely
with ac
<
many
many
or none
if
ab
lat-
>
0.
c are positive,
show that there
are (a
1)
and on the rectangle whose vertices are

Hence show that within but not on the triangle
1) lattice points within
(0,0), (6,0), (0,a), (b,a).
bounded by the axes and the
(6
line
line
ax
by
ab there are exactly (o
l)/2 lattice points, thus that there are exactly
positive values of c less than ab for which ax
by
(a
l)(b
1)
l)/2
has positive solu-
tions.
19).
We may see by inspection or use the same method to find 3 =

19); that is, Sy m 16 (mod 19), which yields the solution y = 2.
This
y = 2
Sy
19), or
= -3
Find
7.
73, gives a numwe need a y such that 73 + 125?/ s (mod

congruent to 3, and 125 to 8 (taking the smallest
That
the least positive value of c for which the line
+ by =
by
73 (mod 125).
a multiple of 125 which, when added to
ax
4.
(b
ber divisible by 19.
is
lattice points?
dis-
Probably the best method in a numerical case for the solution of
tions.
has
quadrant on
of solving a linear congruence.
What
2.
= 35
21 (mod 96), 145x
Solve each congruence, 69z
any
by = 53.
see in the next chapter that
call
Thus, from the
cussion above, there are no lattice points on the line
from the previous section, only one
whose
We
ordinates are integers and through which the line passes.
There are a number
7) as
shown above.
1.
solutions in integers x
be noticed that a Diophantine Equation
32 can be reduced to a congruence (mod 11) or (mod
an
In this connection, geometrical considerations are of interest and
the line 7x
lly
be used to prove the existence of solutions (see
It should
by
solutions are given
for integer values of n.
The
65
two unknowns
in
Exercises 3.2
x =
value.
may
This method
abn
Thus we have
an.
Solution of linear Diophantine Equation
3.2.
16
8.
Find the
least integer b for
which the equation bx
7y
has ex-
actly six positive solutions.
the end of our "chain" of congruences, and, substituting
into 73
125?/,
we
get 323, which
a solution of the given congruence.
is
19- 17.
This shows that 17
is
9.
A customer buys an article for 48 cents. He has a tl bill and 3 pennies,
while the shopkeeper has 6 dimes and 7 nickels.
In
how many ways can
66
Linear Diophantine Equations
3.3.
the change be arranged?

the same
(Ref. 14, p. 65; see also Exercises 11
two
trains ran 180 miles, one at a speed of 6 miles per
typical problem in elementary algebra
If
d.
t~
Show
1 1
is
for
by a and 180 by
replaced
Show that for
a and
if
the faster train required one hour
Suppose 6
each.
in
Thus the
that gy
10.
other.
and 12
set.)
some
integer
number
that the
a and
integral
is
less
hour
faster
than the
time find the speed of
ad, find the speed in
the speed
d,
the following: Each of
is
an integer
if
may
be used.
12.
How may
However, the
since
any
Show
13.
by
+ 2y + Sz =
may
be illustrated
notice that 6x
+ 3)
Graphical methods
(Ref. 14, p. 66.)
/12.
+ cz =
d has integer solutions
if
the g.c.d. of
How may
the
method used
be used to show the existence

(a,m)
lOy
in the numerical
example
a solution of ax
of
Theorem
part of
first
One might
using
3.2,
Linear Diophantine Equations
Theorem
any
n.
is,
5m.
There
in
1.8d.
2t is solvable for
37,
any I and a solution
3.3.
5v,
if
and only
Suppose
we
aiXi
is
this:
when each
solution
1,
15z
Suppose the
pair of co-
the final solution was
y divided that of
u in 3x
would be considerably longer.

another method of solution whose motivation
In general
4-
5y
it
Then
g.c.d. of a,b,c is 1.
more than two unknowns
anx n
+ 3m = 1 - 3v, z = 1 + 2u.
special, since the coefficient of
an equation in two
we
shall set
if
is
not quite
The fundamental idea

we can find a',b',c and
a'
b'
c'
a"
b"
c"
up the equations
The Diophantine Equation
of 2t
a",b",c" so that the determinant
of the following:
+ by + cz = x',
+ b'y + c'z = y*,
a"x + b"y + c"z =
ax
=d
a'x
z'.
if
the g.c.d. of the
We prove this for n =

teger y
loz
had a common factor greater than
of this section
(mod m) when
Theorem
guess, as a generalization of results for
a&i
for
efficients
is 1,
solvable
lOy
Though the above example showed a
11
1?
unknowns, the truth
is
is
3.3.
may be used to find a solution in a numerical

as follows: To solve
so clear but which gives results very expeditiously.
Prove the
15.
= 11,0= 1. Hence the general solution of the latter is = 11 15w,

Then 6s + lOy = 2t implies 3x + 5y = 11 - 15m or Zx +
z = 1 + 2w.
=
h{y + Su)
11, and the general solution is
is
somewhat
a,b,c divides d.
14.
g.c.d. of g
6a;
Theorem be used to show the existence of a solu= b (mod m) when (a,m) = 1? Would jôu
practical
method
of numerical solution?
a
the g.c.d. of g and a 3
of the proof
Euler's
that ax
if
tion of the congruence ax
consider this
and only
if
divisor of g
of nonnegative solutions of the equation
37
the integer nearest to (n
reduced to finding integers y and x% such
is
t.
x
is
This has a solution
d.
The method
case as
n=
and a 3 is the same as the g.c.d. of

and a 3 divides ai and a 2 while any common
of ai,a 2 and a 3 divides both g and a 3 This establishes our theorem.
a\, at, az,
terms of
and only
a& z =
divides d.
divisor
f or
solution
67
more than two unknowns
in
divides d.
3 and leave to the reader the proof by induction
g is the g.c.d. of a x
can, from
a,-'s
Theorem
a\X\
and a 2
3.2, solve
a-iXi
Then, no matter what
in-
Since the determinant of the coefficients

tegers x,y,
y'
and
z'
and
z in
terms of
difficulty, then, is to find
gy.
This
may
x',y',
arbitrary parameters,
and
is 1,
z';
these
may
setting x'
we have a
be solved for in-
= d and
considering
general solution.
The
chief
the second and third rows of the determinant.
be done as follows: Let g =
(a,b)
and as
br
g.
Then
68
(a,b,c)
gu
cv
implies
and there are integers u and
1,
sa)/g
Let us see
while
u =
such that
last
u(as
how
av/g
bv/g
2, v
1 is
102/
= 2 and s = - 1
sav/g)
5y
+ u(as
+ 2z =
rb)
To
eters y'
and
34,
lOr
2,
2.
Zz',
3z',
= -x'
Along the
tion in three
the general solution in terms of the param-
4.
Find
5.
Show
then (xi
we have
subtract the second from twice the
and 100
19c
(mod
=
=
equations
d,
d'
2
the integer solutions of x
all
that
2,
first,
we have
two
for the solution of
linear equa-
unknowns.
if {xi,y\,Z\)
yi
3/2,
z\
and
- 2)
15.
ax
by
4- cz
0.
(x 2 j/:,z 2 ) are solutions of

is
a solution of ax
and use
4-
by
show that
this to
all
cz
= d
Find
all
solutions of the
former equation can be written in the form
y =
z =
the two equations:
where u and
tions of
the equation
ax
xi
2/1
+ riu + 8\v,
+ r,u + 8aw,
Z\ 4- t%u
v are arbitrary integers
+ by +
cz
-I-
and
s 8v,
(ri,r 2 ,r 3)
and
(si,s 2 ,s 3)
are solu-
0.
100,
= 5u and hence p = 20 equation gives / = 80 + 14w.
5) implies c
stitution in the second given
5p =
shillings?
second method of solution of a single linear equa-
lines of the
the solutions of the latter
3.4.
19c
by + cz
+ b'y + c'z
unknowns, develop one
ions in three
+2z',
+ 3p + / = 100,
c+ p+ f = 100.
we
be solvable in integers.
10c
If
the only
z'.
= -2x'-Sy' +
of chickens,
a'x
of puzzle
and / the number
is
much as two geese one chicken,

together for 25 shillings. What was
sufficient conditions that the
ax
problem leads to two linear equations in three

unknowns: A farmer with $100 goes to market to buy a hundred head of
stock. Prices were as follows: calves, $10 each; pigs, $3 each; chickens,
$.50 each. He gets exactly one hundred head for his $100. How many of
each does he buy? If we let c be the number of calves, p the number of
pigs
Find necessary and
y',
z'
Another type
1.
for as
two ducks, and three geese were sold

the price of each bird in an exact number of
x',
we have
to be at least one of each kind of stock, the second
Three chickens and one duck sold
34,
94 for u
5, 1,
solve
3.
and, putting x'
and
Solving for x,y, and z gives us the equations
2x'+y'
0,
Exercises 3.3
= -lisa solution of 8s 3v = 1. Thus we have
8x+\0y + 2z =
=
-x - y
4x+
is
1.
+ 3z =
a solution of 2w
there
80 for u
answer.
this goes for a numerical example.
(8,10)
If
= cv + gu =
must be nonnegative, and p nonnegative

Hence the only permissible answers are
1.
0, 20,
column, we have c(rbv/g

rb)
c
.
69
value of c must be nonnegative, u

implies
8s
notice that g
Set up the determinant
1.
Expanding by the
cv(rb
(g,c)
3.4.
19w.
Sub-
Since the
arise, for
example, in such prob-
lems as the following: Find an integer b such that the following equation
70
leads to a "factorable" quadratic in the sense of intermediate algebra,

that is, to a quadratic equation whose roots are rational numbers:
71
3.4.
The Diophantine Equation s2 - uhj 2 = d has the solu(mod 2m).

2uy = r - t, where rt = d and r - t =
tion 2x = r +
equation has
the
also
see
that
can
The proof is left to the reader, who
no solutions if d = 2 (mod 4).
Theorem 3.4 shows that such an equation has only a finite number
Theorem
3.4.
t,
-
x -4
On
multiplication
and
collection of terms
x*
3(6
b)
1.
we have
the equation
of solutions.
b 4- 8
It
may
therefore
seem remarkable that the equation
0.
x*
If this is to
have rational
square that
is,
for
some integer
s
Thus we need
solutions, its discriminant
perfect
s,
+ 86 +
b2
must be a
(b
+ 4) 8
12.
this point
At
we
shall
Hence one
factor of 12
must be
(6
+ 4) 2
b -f
s2 .
+s
and the other
+4-
s.
In
symbols,
<
where
rt
12.
Thus
+ =
t
=
=
+ 4)
factor 2 in the right members, these
and s if and only if r and t are both odd or both even (we say briefly,
"r and t are of the same parity"). Since interchanging r and
t does not
change
- Vcy
b,
we may choose r
to be less than
following possible pairs of values for r and
r=
M= 12;r= 2,
in absolute value
and have the
That
is,
is
sum
the
of
all
and
must have the same
the exponent of
Vc
is
odd.
we produce no change;
terms, the sum becomes
if
2,
6 and r
(xo
= 4.
parity,
Taking the product
= -2, < = -6,
The quantity
Only the
fol-
also
and x,y
is
we
we make
replace
the
-Vc
by Vc,
the same replacement in the other
Hence
+ Vcy
= x
+ Vcy.
sides,
two equations and
on the
left side is 1,
of values of
setting it
we have
and hence the right side
a solution of the Pell Equation.
no pair
left side of
x and y
for
Since (xo
+ Vcy
n
)
any n can be the same as
any pair
= 8.
latter value yields a quadratic equation.
- Vcy.
each term of x
If in
in parentheses
increases with n,
and
(xl-cy$ = x i -cy2
is 1
a solu-
1,
of the left sides of these
and hence the only values
which yield
b
is
Vc is even, and Vcy all those for which
-\-Vc-y.
to be considered are
r
terms in the expansion of the
equal to the product of the right
However,
cy*
- Vcy ) n =
equation in which the exponent of
t:
= 6; r = 3,
,2/o
determines integers x and y by means of the
n
Q)
(xo
- t = 2s. Because of the

equations may be solved for integers
and
lowing equation:
+ 4 + 8,
b + 4 - S>
b
2(b
if
with y Q not zero and c not a perfect square, there is an infinite number of
solutions. To this end, let n be any positive integer and see that the expansion of (x
be content to show that
tion of
x2
=d
cy2
1
has either no solution or an infinite number if c is not a square. If d =
fully in the next
this is the so-called Pell Equation, which is discussed more
chapter.
to solve for b the Diophantine Equation
12
The same method may
for another n.
In contrast to the above result for the Pell Equation there
markable theorem of A. Thue, which is as follows: If
be used to prove
/(z)
= a nz n
~l
a n-^ tt
aiz
cio,
n >
3,
is
a re-
77
has integer coefficients and
is
irreducible, that
has no polynomial
is,
fac-
has only a
finite
(Ref. 10, p. 260
+a
number
^ lx K l y
+ aixy n~ +
l
we
y be even and equal to 2m.
let
of solutions in integers x
Ooy n
and
y,
x2
zero being
The equation may then be
ff.)
Exercises 3.4
For what integers
that
since x
x -
-3
Prove Theorem
Show
if
x2
Thus
3.4.
cy*
zJ,
x2
= 4w2
+x
2r, z
2s, rs
which implies
= u2
=
+ 8, x = 8
r.
had a prime factor
rs
2
,
common,
in
m u2 implies that
s = n2 with mn =
,
and
it
would divide
which implies that
relatively prime,
s are
squares which
and
and
may
x.
Now,
How-
s are also.
be written
This gives us
u.
Pythagorean numbers
m +n
2
as the complete solution of x 2
2
,
x = n2
z2
2
,
= 2mn
That is, for each pair of values

n we have a solution, and if x,y,z is a solution in which no two
have a factor in common, there are integers m and n satisfying the above
of
3.5.
4u2 =
written in the form
z are of like parity,
and x are
ever, z
has a solution with y 9

and x2 - cy 2 = d
has a solution, then the latter equation has an infinite number of solutions
if c is not a square.
that
and
and
if r
3.
yield a quadratic equation with rational roots?

2.
to
is,
does
Thus we have reduced our problem
with x odd and no two of x,u,z having a factor in common.

finite.
z2
73
finding solutions of
tors with integer coefficients, then the equation
ax"
3.6.
m and
equations.
There
is
a particular kind of quadratic Diophantine Equation which

connected with the right triangle. It is well known that there is a right
It is
is
triangle with sides 3,4,
Pythagorean
To
set.
and
5.
This collection of numbers
find all right triangles
has no solutions in positive integers x,y,z (Ref. 14, pp. 408

it is
is
not hard to show that the Diophantine Equation x s
called a
ff.).
y*
z*
In fact,
conjectured that
whose sides are integers we
xn
_|_
yn
is
zn
must solve the equation

has no solution in positive integers
x*
is
if
if
x,y,z is
any two
a solution, then bx,by,bz

of x,y,
and
is also
have a prime factor
it
tions
we
call primitive solutions; all solutions will
these.
Furthermore,
2 (mod
4),
which
is
ruled out the case
if
greater than
if
then be multiples of
x and y were both odd, z would be congruent to

They cannot both be even, since we have
2
factor in
common.
efforts of
many
integers n.
in
impossible.
when they have a
all
This conjecture
2.
Fermat's Last Theorem, and, in spite of the
a solu-
commust divide the third, and, after division, we get another solution.
Hence we may assume that no two have a factor in common. Such solumon,
called
mathematicians, has never been proved for
First, notice that
tion; and, conversely,
+y
Then suppose
3.6.

for solving many Diophantine
many solutions depend on the ingenuity and observation of
To show how such solutions may be effected, we consider the
While there are systematic methods

Equations,
the solver.
following problem
74
3.6.
For the fraction 16/64 one might erroneously cancel the

arrive at 1/4, a correct result.
10a
and 5c
6,
for 106
To
is
a two-digit
what
for
c,
to a/c where the numerator
denominator
If ah stands for the two-digit

is
which shows that k divides 9 and hence

cases
we
number with digits a and b, the

number with the digits b and c, and b > 0?
10ac
6c
10a&
55/55, 26/65, 19/95.
Thus the only
solutions of our problem are
6,
and a
a
c'
Exercises 3.6
ca, or
10a(c
(1)
6)
c(a
side,
and
it is
not a divisor of either
since a,6,c are positive integers less than 10.
or
bis divisible by
5, or
b is
even and
Hence
either c
c i6 divisible
by
is
(mod
5. In the first case
5)
(1)
and
shows that c =
6,
c'.
Then a -
left side is
6 is
odd or even.
even, since the
the plus sign the
left
having the same property?
Show
>
0,
Will the
and we have
(6-5)(2c'-6) = -c ,6>5,
/
if c' =* 4,
9,
62
- 6(5 + 2c') + 9c' - 0. Thus,

= 4, and we have the fractions
16/64
2.
2u, c
5.
divisor of
6.
6)
(1)
ka,
6-
6,
a-
1;
gives
6.
must divide the right side, and hence it is a

and substitute in the last equation to get
left side, it
Write 6
2,
49/98
Then equation
2a(5
Since a divides the
if c'
= k(2a -
1),
c)/(6
+ d).
that
if
c.
the numbers 30 and 25,
result squared (55 2 ) the result is the original
number.
other numbers are there composed of four different digits
that
if
a/6
<
c/d with 6 and d positive integers then
even according as
integers?
which becomes
and the
c'
Then a =
negative.
The number 3025 has the property
must then be not greater than 4 and since with

side would be greater than 4, the ambiguous sign
Since
formed from the digits in order of the two halves of the number, are added
'.
is
(a
(giving 55)
or second factor
first
solutions in integers of this equation.
all
and our fraction is aa/aa. In
c/d
6 and d are positive integers and prime respectively to a and
3.
The
Find
How many
(65)(2c'-6) =
to write
Assume
2.
an integer
2c' for
is
We
even
5.
the second (1) reduces to
must be
mistake often made by beginners
nor
consider these two cases separately.

1.
b).
a/6
a divisor of the right
is
In the respective
to find positive integers less than 10
+6
+c
106
Now, 10
9.
aa/aa, 16/64, 26/65, 19/95, 49/98.
10a
is,
3 or
1,
the quotient equal
such that
that
is
75
get the fractions
a two-digit
is
we need
solve this problem
fractions ab/JJc
and
6's
number
same
result hold
if
+c
+d
d'
we remove the
restriction that the letters are
77
4.1.
the quotients of successive terms are formed, their

differences are alternately negative and positive and each is less in abIn other words, if we call u n the nth
solute value than the preceding.
This means that
if
term of the Fibonacci Sequence where
and
are the
first
two terms, the
Continued Fractions
8
^
4.1.
There
is
"^
a very interesting sequence of integers which bears the name of

Fibonacci, an Italian mathematician of the twelfth century. The first
few numbers of this sequence are
8
ratios
u n /u n+ 1 for n
>
form a decreasing sequence
and an increasing sequence
1,1,2,3,5,8,13,21,34,-...
for
even values
being greater than each of the latter.
Each number
in
the
is
sum
of the
two preceding
nature in the following manner.
starting with one bud, one's
If
ones.
These numbers occur
a twig or branch
hand moves
selected, and,
is
bud and so conreaches a bud just above
to the nearest
tinues in spiral fashion around the twig until
it
number of intervening buds will be one of the

Fibonacci numbers, and the number of revolutions around the stalk will
the starting one, the
be another such number, which ones depending on the particular plant.

pine cone gives a similar result.
The
what the Greeks
the ratio of whose sides
There
is
shown later.)
call k, as-
The
Golden Section.
called the
of
rectangle
l/k was supposed to be the most beautiful.
is
an interesting trick by which one can give a closed expression
similarly determined.
First of
notice that the nth term
means
(This will be rigorously
nth term of the Fibonacci Sequence and
for the
odd values
for
each of the former ratios
approaches a limiting value which we shall
ratio n/w + i
sociated with
of n,
is
many
other sequences
consider the Fibonacci Sequence and
all,
given in terms of the two preceding ones by
of the formula
We
are concerned here, however, with the mathematical properties

of the sequence. Notice that 1 3 - 2 2 = -1,2-5 - 3 2 = 1,3-8 - 5 2 = -1,
U n = U n -\
(1)
5- 13
1,
and so
deception which
may be cut up
as
shown
as
on.
This property
shown
in the figure below
2/3
is
and the pieces
1/15, 3/5
trick is to set
(2)
= ca" -f
d/3'
fitted together
and show that

(1) holds.
c,d,a,
It will
and
/S
may be
then follow that
determined so that ui
(2) gives
1,
uz =
and
a formula for the terms of the
Fibonacci Sequence.
equivalent to the sequence of equations
= -1/6, 2/3 - 3/5 =
The
square 8 units by 8
of the Fibonacci Sequence.
The property noted above
76
connected with a geometrical
by 13. The area is thus increased by 1

Notice that the numbers appearing are those of four suc-
numbers
1/2
is
be familiar to the student.
to give a rectangle 5
square unit.
cessive
may
+ W_J.
5/8 = -1/40,
First,
If (1) is
we
find
how a and
/S
can be chosen so that
to hold, the following equation
must be
(1) will
true:
etc.
Ca
d/3
ca"-
dP
n~
+ ca
n~2
-2
d/S"
be
satisfied.
78
Continued Fractions
The
coefficients of c
and d
in this equation are
a -2( a2 _ a
which Wi =
equation will be satisfied
last
u2 =
1,
if
a and
/3
are the
If
>
roots of the equation
the equation x
a,
and
and u n = 2u_i
1,
for the Fibonacci Sequence, find a
1),
5.
Hence the
respectively.
and jSÔS2
i)
in
ax
u_ 2
Using the methods used
formula for the nth term.
has two distinct roots, a and
the sequence of integers u
if
79
4.2.
defined
is
by
its first
/3, where
two terms
and the equation
38
(3)
For
we
= ca
= cc?
?*
/3
U n = aUn-l
0.
choose c and d so that U\

1
Since
whatever the values of c and
this choice (1) will hold
Second,
and n 2 =
show that a formula
d.
that
+ d/3
un =
is,
where q =
+ d/3,
2
.
6.
these equations can be solved for c
and d with the following
result:
Then,
we
if
we denote by a
see that
/3
is
the root
equal to
00 -
'
a)
(1
- V5)/2
V5, and a
and
/q
- au^^/q,
the sequence defined in the previous exercise
If in
Exercise
If in
(wi
-f 46.
a =
5,
/3
Is there
infinite.
show that the nth term

u
the other root,
manipulation shows that
little
is
- u 2)a n~
(/3wi
btln-i,
|a| >|/3|
a limit
show that
if a = /3?
of the sequence
is
given
by the formula
a)
of (3)
Va
un
ujun-x approaches a as n becomes

If so, find it; if not, show why not.
7.
a(0
for
where a
2(2n 2
c=-l/V5\d=l/\/5.
is
the root of x
au\)/a
can
ax
Note that a =
2
.
+ ndan
=
0,
4{aui
)/a2 and d
In this case will u n /u n -i ap-
a/2.
proach a limit as n becomes infinite?
If so, find it; if not,
show why
not.
Hence
Suppose we have two sequences whose nth terms are p n and q, respectively, where p = p_i + p_ 2 o = g_i + g_ 2 and pi = 1, p 2 = 2,
8.
wn -
(4)
03*
- a-)/V5.
while gi
4, q 2
Prove
Exercises 4.1
p n q n-i
Using the numbers 5,8,13,21, dissect the square 13 inches on a side in
2.
Using the formula
(1) for
un
of the Fibonacci
As we found from
Sequence given above,
having
many
may combine
same
3.
result using
left side is
formula
Using previous
equal to
n*).
(1)
results, find
lim
Pk
the exercises above, there are
p*.-,g*,
sequences
In
fact,
= a k pk-\
where
Pk-i, qk
for the Fibonacci Sequence.
Pk
P*g*-i
Consider the sequence
qkPk-i
akPk-\
Pk-i
a-kqk-i
fffc- 2 .
Pk~2
Pk-\
=
qk
many
Now
u n+ i/u
1, 1, 3, 7, 17,
(-1)".
we
the types mentioned in Exercises 5 and 6 above and con-
Achieve the
(4).
n
4.
-(w n-iw n +i -
of the properties of the Fibonacci Sequence.
sider the pair of sequences
Hint: Show that the
q n pn-i
4.2.
a fashion similar to that in the figure above.
show that
qk-i
Pk-2
Pk-i
qk-t
Cfc-l
=
a k qk-i
g*- 2
qk-\
80
4.2. Properties of Fibonacci-like
Continued Fractions
The second
equality follows from
and the third is obtained by subtracting a k times the second column from the first, an operation which wo
know
leaves unaltered the value of the determinant.
if
Pi=l, Po=l, 9i=l,
Thus, noticing that
qk-iPk-z),
in the expression p*g*_i
qkPk-i
we
(pkQk-i
qkPk-i)
(-l)*-*(p!gi
qiPi)
(-l) (p*?i
fc
if
Hence equation
Pk/qk
(3)
that
is,
Pk-x/qk-i
two consecutive quotients
the differences alternate in sign.

Pk
Pkqk-i - qkPk-i
13
21
34
55
89
13
21
34
55
89
144
range over the Fibonacci Sequence with pk=qk+i.
Qk
u*
is
(3)
the
qk-i
qk
results
the determinant from the
first.
O-kpk-l
Pk-t
Pk-2
ak qk-i
qk-i
q k -r
- WjfcM-l =
empirically and have
as small as
we
/(UkUk -l).
above pî q*P\
= 1)
u = ( 1)*,
2
now
established.
A;
Theorem
than
ratio is greater
Fur-
than every odd-numbered
A;
of
4.2a
form a decreasing sequence
proved in terms
that the quotients Pk/qk form a de-
even, and an increasing sequence for k odd.
P2k/q2k> Plrlqir> Pit- l/<?2r- 1-
satisfy the conditions of
ak is positive, then the fractions ffe/gi
- pM
We have seen
please.
thermore, each even-numbered
>i
just
l)
it
creasing sequence for
Thus we have
now what we have
(-
was shown that the ratios uk + i/w* approach a limit.

position
to show that under the conditions of Theorem
We are now in a
By "approaching a limit" we
approach
a limit.
4.2a the ratios pk /qk
mean that there is some number L so that, by choosing k large enough,
In an exercise
\L
for
k even and an increasing sequence for k odd.
Let us see
= qk = Pk-\.
we can make
from subtracting the second column
p k and qk
term of the Fibonacci Sequence, we have
becomes
ratio, since if r is greater
where the third equality
and
value by l/ty^-i, and
-a(-l)-\
qk-i
If
we found
Pk-i
Pk-\
- ak
Theorem 4.2b.
(-l)7fa*0*-i);
=
qk
fcth
states (notice that for the choice
Theorem 4.2a
and Theorem 4.2b states that uk+ i/u k form a decreasing sequence for k
even and an increasing sequence for k odd. These are the results which
is
Furthermore,
Pk-i
11
U*+ iUk- i
differ in absolute
10
Pkqk-i- qkPk-i - (-1)*.

consequence of the theorem
Uk+l/Uk
If
g*:
Thus the pk and

Furthermore,
and
Uk
p k and q k are the fcth terms of two sequences defined by the equations (1) and if p^qi q 7 pi = 1, then
An immediate
for every k,
112
1
q ipx).
Thus we have proved

Theorem 4.2a.
0123456
Pk
(2)
= 0.
of values for p*
qk
decrease the subscripts
?o
have the following table
will
we merely change its sign. Hence, if we decrease the subscripts by

k 2, we change the sign k 2 times; k diminished by A; 2 is 2, and
we have
by
ak
(1) let
and choose
Then we
(pk-iqk-z
see that
In equations
(1),
the last determinant is equal to
we
Sequence with which we started.
Sequences
However, these properties alone do not guarantee the existence

as is
-1,
of the Fibonacci
of a limit,
shown by the following sequence:

4, 0,
-r-
1/2, 1
1/2, 3
+ 1/4, - 1/4, 3 + 1/8, 1 - 1/8,

1
82
Continued Fractions
Here the even-numbered terms (second, fourth, and so on) form

a decreasing sequence, the odd-numbered terms form an increasing
sequence, and
each of the former
is
greater than
all of
we must prove
approaches zero.
This
is
much more compactly
that the difference between successive terms

easily done for the ratios p k /q k for, from
(3), the
differences between successive quotients of the sequences

(1) are in absolute value equal to the following:
Vfogi),
may be
written
(a h a t , ),
knowing the a t % we can easily write the longer form. Such a fracmay or may not have a last term. In the latter case we call it an in-
since,
tion
finite
continued fraction.
If /is
(ai,Oi)
The
the continued fraction (ai,a 2 ,
second convergent,
its
fcth
convergent
relatively
1/Cmj),
l/(fcff,),
fractions
form
in the
/=
(4)
Such
instead of "simple continued fraction."
83
the latter.
However, the terms of

the sequence do not approach a limit since the even-numbered
terms approach 3 and the odd-numbered terms approach 1. The difference
between successive terms approaches 3-1, which is 2. For the existence
of a limit
Simple continued fractions
4.3.
may
we call a
),
(ai,Oi,
,a k)
its first
its
be written in the form pk /qk
fcth
convergent,
convergent.
where p k and
q k are
prime positive integers.
in
simplest continued fraction in a certain sense would be that
are
convergents
successive
which the a's are all equal to 1. Its
The
where the products
in the
denominators form an increasing sequence of
Hence the
positive integers.
rapidly than 1/2, 1/3, 1/4,
Suppose, then,
we
fractions in (4) certainly decrease

,
which approach
qtpi
Theorem
1.
(1)
Then we may summarize
4.2c.
If
p k and qk are the
quences, they satisfy the conditions of
p k /qk approaches a limiting value as
A;
with ak positive integers and

the results so far obtained in
A;th
terms of a regular pair of se-
Theorem
4.2a,
1, 2,
(1)
define a regular pair of sequences to be sequences
p k and qk denned by the relationship

p.qi
more
zero.
of the quotients of successive
which are the reciprocals

Fibonacci Sequence.
3/2, 5/3, 8/5,
We know
So we may
tinued fraction approaches a limit.
and the quotient

1
increases without bound.
Regular pairs of sequences are connected with so-called continued

which have many uses and interesting properties. A continued
fraction is a number of the form
x =
or x t
a*
a, except
perhaps the
first is zero.
d/a positive integers, except that the

continued fraction, and
be considered here.
it is
first
o.
Use
of the
we
If
the 6/s are
may be
zero,
all 1
and the
we call it a
where the positive sign
is
l/x,
+ V5
1.618034
chosen, since x
is
positive,
.618034
simple
is
to
use the term "continued fraction"
1/x-
V5-1
start with the
formula for the solution of a quadratic
as
only this kind of continued fraction that
From now on we
bt
a,+
where no
if
equation yields
Since this continued fraction has the same value

second term as if we started with the first, we have
fractions,
a,+
let
Simple continued fractions
of the
of the conthese quotients converge to a limit and that hence the value
4.3.
numbers
from the previous section (why?) that
and
84
Continued Fractions
4.4.
Exercises 4.3
We
Find the limit approached by the quotients
sequence
of Exercise
4 in Section
un =
2.
Find the limiting value
What would be
4.1.
terms of the
of successive
for k
have already shown

r
and show that
the limit for
Pr/q r
+ w_ ?
aun-i
we make
this
pk and
qk in terms of
p*_ 2 and
pk -\,
qu-i, qk-2, respectively.
the
p k/q k is
fcth
qupk-z
(ata*_i
l)(p k^q k _ 3
gt_4p _ 8), where

fc
convergent of a continued fraction.
this
by replacing a T by a r
(a,p r_i
+ Pr-2) + ffr-1
+ qr-z) + q -i
a,+l(arPr-l
a r+1 (arq r-i
qr +i
of proof suffice to evaluate 1
p r-i and a r+iq r
1/2
is
1/4
valid?
1/8
Does
method
this
"r+lffr
by Theorem
column
4.2a,
of the
we need
Much of the labor of computation

may be avoided by establishing
of convergents of
formulas for p k and qk

putation shows that for the simple continued fraction (a h a 2
+ Pr-l
^ q -i
,ak )
which
is
holds for
is
l)/a 2 [a 8 (aia 2
,
1)
ai]/(aa 2
factor of the elements of the first
= a z pi + p lt
a k p k -i
this
shows that
pr+i
(1)
in Section 4.2
and
If
p k /qk
is
pitik-i
the
fcth
convergent of a continued fraction,
QkPk-i
(-1)*,
The
conditions of
Theorem 4.2b
1).
is
a systematic way to compute the convergents of a given conTo do this, notice that p\ = a\, p 2 = aia 2 + 1 would
tinued fraction.
q3
(1) in
= a 3q 2 +
permit us with consistency to define p
q lf
Section 4.2 for k
we would have
o 2 = 2, a% = 3,
fashion,
3.
That
this
ai
1,
p*_ 2 , qh
= a k qk-i
g,-
1
1
and p_i =
0.
In a similar
and g_i = 1. Then suppose, for example,

and we form the following table:
12
Pi
qo
the statement of
Pk
also hold.
Theorem 4.4a. If pk /qk is the kth convergent of the continued fraction

), where p k and q k are positive, relatively prime integers, then
prime and completes the proof of the theorem.
a,-
(ai,a 2 ,
we
and the continued fraction converges.
exactly the relationship

all
qr
and any common
the
be seen that
Pz
+ q -i)
Pr
gvfi are relatively
There
may
l/o r+ i.
only to show that a r+ip r
determinant would have to divide
Theorem 4.4b.
Com-
three convergents are
this it
q 2pi =1, the sequences p k ,qk are regular sequences and we can
conclude from Theorems 4.2a and 4.2c the truth of
P2?i
a continued
fraction
From
this will
However,
Since these relationships are the same as
+ ftr-l
+ q -i
a T+l q r
qr -\ are relatively prime.
dr+lPr
and
Computation of the convergents of a continued fraction
Oi, (aia 2
pr-2)/(arq r-i
Qr+lPr
positive integers, to establish our result
'
proof invalid while the similar proof above
first
".Or),
(Oi,02,
replacement in p T/qT
'
4.4.
holds
Since the numerator and denominator of the fraction on the right are
What is wrong with the following proof " that 1 + 3 + 9H

=-1/2?
Letz= l + 3 + 3 2 + = 1 + 3(1 + 3 + 3 2 +--.). Hencea;= l+3z,
which gives x = -1/2. What is the essential difference that makes this
4
it
get
Pr+l
Prove that p k qk-z
k =
Now
and thus p T+ i/qr+i will be obtained from
(1,2,1,2,1,2,..-).
3.
Let us assume that
3.
85
of the periodic continued fraction
this continued fraction express
this for k
this implies that it holds for
then prove the theorem by induction.
If
For
Computation of the convergents of a continued fraction
10
43
225
1393
30
157
972
where the number 3 in the second
line is 2- 1
forth, the third line being similarly
computed.
g*_,.
1,
10
is
3-3
1,
and so
86
4.5.
Continued Fractions
if
1.
Find the
first
convergents of each of the following continued
five
fractions, the second
and
is
the continued fraction (ai,a 2 ,
x; that
b.
2.
third being periodic (see Section 4.7)
(1,3,5,7, )
119/108
1
Example b above converges
that the continued fraction of
to
1/ast,
tained in this
4.
5.
Find the continued-fraction expansion
If pk/qk
(ai,fl2,
,o-k),
show that
Pk/pk-i and hence, from the result for

Pk/pk-i
6.
If qk
pansion Of
7.
three convergents.
of \/5,
Pfc-i,
a*
compute the
three
that
all k,
Notice that
in
the greatest integer less than
is
we might
1
,
x/17
below the
-,Oj,ai).
ex-
).
To
48
53
260
573
19
21
103
227.
have chosen
Oj
1,
ob-
a.'s
a*
in place of an
2,
convergent would have been 313/124.

1/xi,
six places the
value of y/Yl
is
where x x
the expansion of
We
4.123106.
four convergents, their decimal values to six places,
first
11/9
Now \/l7 = 4 +
Example 2. Expand V17.
=
=
Since x a = x
\/l7 + 4
8 + I/**
4)
1/(V17
where x 3 =
Using the sequence of
last
l/x t where x 6
119 14
also
of 573/227.
l/x z
x,
what can you conclude about the continued-fraction
of Exercise 5,
2
(-l)*
prove that pk-\
= 0(modp
list
and
from the true value.
their discrepancies
>
qk
and sufficient condition that a

same forward as backward.
imply p k~\ =
finite
qk-
continued
2177/528
Convergent
33/8
268/65
Decimal value
4.125000
4.123077
4.123106
+0.001894
-0.000029
+0.000000
.123106
Discrepancy
state a necessary
fraction read the
qk implies
t ).
Conversely, prove that the congruence and pk
we know
Actually, since
that the true value of the
number
lies
between
any two successive convergents, we can tell from the expansion that
Notice that the
4.1231 is accurate to four places and 4.123106 to six.
continued fraction
4.5.
227/119 =
which case the next to the
is (4,8,8,
(a,k,a,k-i,
illustrate
Pfc/^jt?
Using the notations
Hence
first
2.236068.
1/2.
convergents, and compare your values with \/5
where x 2
where x 4 = 108/11 = 9
Find the continued-fraction expansion of 355/113, and compute the
first
= 9/2 = 4 4manner, we have
where
1/xa,
l/x 2
y/2 and that of Example c to \/33.
We shall now
1.
Now, 573/227 = 2
(1,2,2,2,-..).
Show
the greatest integer in
is
x.
the method of expansion, by three examples.
(1,1,2,1,2,1,2, ).
c.
) then a x
the largest integer not greater than
is,
Example
a.
from the form of the continued-fraction expansion that
It is apparent
Exercises 4.4
87
Continued-fraction expansions
is
periodic.
We shall prove later that this is true of the
continued-fraction expansion of any real quadratic surd (real root of a
Continued -fraction expansions
quadratic equation with integer coefficients).
We
have so
far
which continued fractions converge.

point of view,
it is
Example
been chiefly concerned in finding the numbers to

of
Actually, from a computational
more importance
pansions of given numbers.
This
to seek continued-fraction ex-
is useful, since,
as
we have
various convergents give approximations to the number.
one of the convergents of w
is
seen, the
For instance,
22/7, the familiar approximation, and no
fraction with denominator smaller than 7
is
a better approximation.
3.
what along the
Find the expansion \/2.
lines of the
This could be done some-
above example, but there
is
a simpler way
a method which can be applied to get a positive solution of any polynomial equation. We seek, then, the real root of x* 2 = 0. Since the
root
is
between
One may
but
it is
and
2,
we have x =
find the equation in X\
\/x\,
where
X\ is greater
than
1.
by substitution in the given equation,

by 1 and reverse the order of the
simpler to diminish the roots
88
4.6.
Continued Fractions
Since the process usually employed in the theory of equa-
coefficients.
may be unfamiliar to the reader, we
tions
him to any
referring
mechanical process
illustrate the
college algebra text for the reasoning involved.
89
Closeness of approximation
tinued fraction, of course, represents a rational number.
See Example
of this section.
Notice in this example that p/g = 260/103 and a solution of 573x

227y = lis x = q6 ,y = p6 The next to the last convergent of the alterfl
Thus,
-2
2
1
1
-1
x =
if
+ y, y satisfies the
+ Zy + Sy - 1 = 0.
equation y*
Then y =
\/x\ implies
+ 3zi + Zx\ - x\ =
x\ - 3x\ - 3xi - 1 =
1
0, or
0.
Notice that the last equation
and changing the
efficients
=
-
Xt
xi
x2
+
1 +
The computation
pk
- 6xf Zx\ - 12x* 1 < z < 2.
l/x 2 where 10x|

,
where
where
1/zs,
l/au,
Pk/Qk
6x 2
1.
Theorem 4.5b. If a/b is a positive rational number in lowest terms

and if a/b is expanded into a continued fraction with k terms, then
by = ( l)
qk-i, Pk-i is a solution of ax
This theorem gives a means of solving Diophantine Equations, linear
in two unknowns, but it is not as good computationally as methods given
fc
previously in this book.
24x 3
and accuracy
Exercises 4.5
0,
10
0,
1
of results is as follows:
29
34
23
27
1.000
1.3333
1.2500
1.2609
1.2592
-0.2599
Discrepancy
Further computation yields
signs.
of convergents
ak
obtained by reversing the order of co-
is
native expansion, 313/124, similarly gives a solution of 573x 227y

Why is this so? The following is left to the reader to prove.
+.0734 ] -.0099
+.0010
-.0007
2.
Using continued
3.
Find the
fractions, find a solution in integers of IZx
3.14159 and of 3.14160.
4.
23y
1.
convergents in the continued-fraction expansion of
first five
Since
ir
lies
between these two numbers, what
does this show about the continued-fraction expansion of
of
Va +
2
1,
7r?
Va + 2
2
for posi-
tive integers a.
5.
Notice that the larger the a, the greater the gain in accuracy from one
Prove Theorem 4.56 above.
Approximate by continued fractions accurate to three decimal places
s
the positive root of x
7x 2
0.
convergent to the next.
With the above examples
in
mind
it is
not hard to prove
Closeness of approximation
4.6.
Theorem 4.5a. The continued-fraction expansion of any real positive

number is unique except that every rational number has two expansions:
(Ol, d2,'
where o = a
To prove this,
with
Xf+i
", Ctn-l,
and the
let
x,-
first
O n )>
iflh
1
a 2>' "t -l>
imation for this constant by the Babylonians.
is
was used as an approxfraction with smaller
Similarly, 22/7
of the a's are the same.
denominator gives a better approximation.

approximation for the size of its denominator.
In
is
the best
fact,
be defined as that part of the expansion beginning
and x t+ i that beginning with ai+i. Then x = a,- + l/x,-+i, where

1.
If Xi+i > 1, this shows that a,- is the greatest integer in Xi and
uniquely determined.
may
-k
No
n> *)>
pansion for a real
p/q is a convergent of the continued-fraction exnumber x there is no rational number a/b with b ^ q
which
Theorem 4.6a.
a,-,
hence
tion
in the expansion of
The convergent 355/113
terminate with
(a,-
However,
if
1) or with a<
Xi+\
1/1.
1,
the continued fac-
terminating con-
is
The
closer to
If
proof of this theorem will follow immediately from the
90
4.6. Closeness of approximation
Continued Fractions
Lemma:
than b and
If
a/b
<
<
c/d
and
a'/b'
ab'
= -1, then d
a'b
is
greater
(All letters represent positive integers.)
b'.
us see why the theorem will follow from the lemma. Supthe kth convergent in the expansion of x and a/b is closer to
First, let
pose pk/qk
is
and k - 1st convergent,

since it is then either between x and k k or between x and
p /q
pk-i/q k-\.
However, p k /q k and pk-i/qk-i satisfy the conditions of the lemma, and
xo
than
hence b
To
is pk/qk-
is
Then a/b
greater than qk
prove the lemma, see that
<
It
-ad) <
b'(cb
<
therefore b'
{cb
d.
placed
by Xk+iva
kth.
we
will
Pk+i
m ak+\Pk
?k+i
a,k+iqic
have an expression
for
+
+
From
Pk-i
qk-i
this it readily follows that
c/d
Similarly,
<
may
it
<
a/b
ad)/db
conditions imply that
its
However,
d.
obtain limits for the closeness of approximation of a convergent

be the
to the number, xo, which the continued fraction represents, let x k +i
rea
portion of the expansion of x beginning with ak +i. Then, if k+ i is
To
<
Hence
between the
is
91
a'/b'
(a'b
cb
a/b,
ab')/b'b
ad
a positive integer, and
is
<
we know from
x k+ iqk
holding in the
be shown that b
should be noticed that, though
Furthermore,
= \/W.
first
qk-i
q k -i)
+ qk-i =
qk+i
qk(xk+iqk
qk
ak+ iqk
instance unless
qk-i
the above theorem
if pk/qk and pk-i/qk-i are two successive convergents of

x there is
then no rational number closer to x than p k/qk whose denominator does
not exceed q k yet there may be rational numbers between the two con-
inequality
Furthermore,
terminates with ak+i.
d.
that
the
qk,
a rational number whose expansion
is
a;
>
qk+i
Ofc+i9*
implies
Xk+iq k
+ qk-i =
(Xk+i
a k+i)qk
+ qk+i <q + <?*+i <

k
2qk+i-
vergents which are closer to x
than pk-i/qk-i and whose denominators

Suppose k is odd. Since x - qk y = p k is one
are between qk and qk-\.

solution of pk+ix - qk+^y
solutions are of the form
x - qk
Then,
0,1,2,
Pk
1,
as
tqk+h
in the previous chapter, all the
= pk
we form the sequence of values

a*+2 we get the sequence of fractions
tpk+ i for integers
all
these results,
Theorem 4.6b.
we saw
if
expansion of
x,
If
we have
p k /q k are convergents of the continued-fraction
then
t.
of the ratio x/y for
<
Pk
qk
qkqk+i
<
T
E
Pk
+ pk+i
qk
Combining
q k +qk+i
pk
+ 2pk+
qk
Jt2q
k+ i
pk
'
qk
(ak+2
(a*+ 2
l)p
1)^+1
p k+i
'
It is
not hard to see that the equalities cannot hold
'
qk+2
It follows that k odd implies that

pk /qk is less than x, and any rational
a/b between p k /qk and x must satisfy just one of the following three
conditions: First, be one of the fractions in the sequence above;
second, lie
between one fraction and its successor; or, third, he between the last
fraction and xq. In the third case, its denominator must be
greater than
there are values of x
x -
qkqk+i
the intermediate convergents.
Notice that, if we consider the numerator

the ^/-coordinate of a point and the denominator the x-coordinate,
all
these fractions represent points on the line
p k+1 x
qk+iy
l.
is
irrational,
but
the following differences arbitrarily small:

Pk
"
Ik
pk
qk
qkiqk
qk+i)
Furthermore, the larger the q k the closer the approximation.

,
q^,
since it lies between pk+2/q k+3 and k +i/qk+i. In the second case,
p
its
denominator must be greater than the denominator of the largest fraction
in the above set to the left of which it lies. We call the
sequence above
making
if
Exercises 4.6
In the manner described above, plot on a graph the

vergents of VT7 and the intermediate convergents.
1
2. Similarly, plot the first three convergents
vergents of x.
first
three con-
and the intermediate con-
92
Continued Fractions
the details of the proof of Theorem 4.6b.
3.
Fill in
4.
Find values of x and k
Section 4.6
differences
5.
1
6.
7.
is less
than
would be
for
Would
.01.
less
which one
than
From Theorem 4.6b how
A;
beyond a certain value?
must the
close
end of
third convergents in Exercises
and 2 be to the true value?

Prove that
(x
p k /qk )/{x
Since the proof
Pk-i/qk-i)
-qk-i/qkXk+i.
not
pansion beginning with
If
and
r,
with
a*.
= xT
a*.
give here merely an outline leaving
the expansion
the portion beginning with a r
is
This shows that, for some k and
Take
< k and
denote that part of the ex-
let Xk
let p'Jq'i
some k
periodic, then for
is
equal to the portion beginning

the following equality holds:
r,
denote the convergents of x T
Then
we have
If
and
tegers r
with
<d and cs - rd
a/b
9.
coefficients.
we
As above,
a/b and c/d are two fractions in lowest terms for which a,b,c,d are
positive integers and be - ad > 1, show that there are two positive in8.
difficult,
the details to the student.
Xk
Using continued fractions, find -\ZA accurate to three decimal places.
is
periodic continued fraction represents a root
with integer
of a quadratic equation
there be a value of x for which both
.01 for all
Any
Theorem 4.7a.
of the differences at the
93
Expansions of quadratic surds
4.7.
<
The so-called Farey Sequence
in order
which are between
not greater than
5,
that
and
r/s
<
that this implies
c/d.
of order
1
Show
1.
QLr-X
which shows that Xk
the set of rational numbers

and whose denominators are
is
in value
that
if
V'k-r-1
= xk
X k-%-r-2
a quadratic equation with integer
satisfies
cients.
Then x =
satisfies
a quadratic equation with integer coefficients.
It is
is,
{pk-\x k
somewhat harder
Theorem 4.7b.
0, 1/5, 1/4, 1/3, 2/5, 1/2, 3/5, 2/3, 3/4, 4/5, 1.
Show
V'k-r-X k
x,
If
Pk-t) I\qk-\Xk
qk-z)
coeffi-
shows that x likewise
to prove the converse of this theorem, that
an
is
a/b and c/d are any two successive fractions of a Farey Se- ad = 1. In this connection the results of
is
irrational root of a quadratic equation,
ax 2
bx
+c=
0,
quence of any order then cb

Exercise 8 above are useful.
(Ref. 6, pp. 23
where
ff.)
a,b,
and
c are integers, then its continued-fraction expansion is
periodic.
As above, we
4.7.
We
saw
Chapter II that every rational number has a periodic

decimal expansion; that is, from a certain point on, the same sequence of
and again without
ceasing.
This includes the case
the expansion ends with a succession of zeros.
continued fraction
there
is
is
This definition excludes
represent rational numbers.
finite
To
continued fractions, which
Pk-J/iWk-l
qk-i).
spectively,
by
z,r$f,,u, re-
and have
X
(i;
rz
+s
tz
+u
ts\
y/vr,
|e|
\ru
1.
Then, by Theorem 4.6b, we have
It is
root of a quadratic equation with integer coefficients and, conversely,

coefficients
simplify our notation, replace x k pu-i, Pk-i, qt-i, qk-i
a remarkable property of continued

fractions that every periodic continued fraction represents an irrational
every irrational (and hence
(XkPk-\
Similarly a periodic
defined to be one in which, from a certain point on,

a sequence of values of the a,- which repeats again and again with-
out ceasing.
X =
in
digits recurs again
when
see that
real) root of
a quadratic equation with integer
has a periodic continued-fraction expansion.
We
first
have
x =
since
it
if
must be
r/t
Pk-i/qk-i (that
l/gj-j (that
absolute value.
is,
e/i2
is,
= s/u
r/t) is less in
1/t2 ) multiplied
<
>
|if|,
absolute value than l/ql_x
by a number,
A similar statement is true for x
e,
less
Pk-i/qk-*
than
If
in
we sub-
94
ax 2 4- bx
where
stitute in
C=
0,
+c=
the value for x given in
we
(1),
Az2
get
+ Bz +
However, x
and hence
A=ar + brt + ct
B = 2ars + b(ru + ts) + 2du,
C = as + bsu + cw
2
=-
r/t
\eu/t
t/P
rjt/u\
s/u
\ru
\B\
4-
\2ax
and
is
+ 8t)/tu =
v/u* implies (-ru
8t\
95
4.7.
Continued Fractions
1,
e/&
r,/u2 ,
which shows
b\
|2o|,
therefore bounded.
Now,
less in
if
we can show
absolute value than
depending only on
number
some
fixed
x, then, since
of different equations
As
the same.
number independent
which
must be
are bounded, that
can
identical
of z*,
the details of the proof of Theorem 4.7a.
However, the con-
2.
Verify
by computation that 4AC
This fact would show
3.
Making use
is
only a
finite
and hence two values of
z are
soon as a value of z recurs, the continued fraction repeats.
Therefore the rest of the proof consists in showing that A, B, and
are
Now,
Thus, since ax2
+ bx + c =
0,
\A\
be/P.
5.
Using s/u in place of
bounded.
Second,
the same
it
may
is
r/t,
we
see that
C is
l)
VD,
common
Show
to
be shown directly, as follows, that
method as that above, the reader may

Tjt/u)
(2ax
As a matter
terms.
(The
maximum
x,
e,
and
t\
is
d(2d
1),
+ f\t/u\ <
that
is
bounded.
By
easily find that
c'
noticing the element in the
determinants
first
aa'
bd
ab'
bd'
d!
ca'
+ dtf
cb'
dd'
is
row
products of corresponding elements in the

determinant and first column of the right determinant.
sum
the
of the left
of the product of the
row and column
first
of the
row and second column of the product is

obtained by multiplying elements of the first row of the left determinant
and the second column of the right determinant. Show that the correct
Similarly, the element in the first
values for A,B, and
+ b) + 2aerj/tu;
\eu/t
d being the
Ref. 15, p. 34.)
'/
are of opposite signs, a fact which shows
\eu/t
-D=
of fact, this estimate has
satisfy
2a
2A
2c
2C
Since r/t and s/u are successive convergents of the continued fraction,
expansion of
x2
these expansions?
all
bounded.
determined except for sign.
B = - (eu/t +
if
is
prove that
|6|,
To complete the proof, one may proceed in either of two directions.

First, it may be shown by direct computation, in virtue of the values of
This is straightA,B, and C given above, that AAC B2 = ac b 2
forward but rather long. However, it does show that, once A and C are
determined, then
bl in the last proof.
VD
erties
\a\
4.
we have
-2aex
4ac
Find the continued-fraction expansions of \Z7, VTl, \/l3. Compare

them with those already found for \/3, V5, Vl7. Can you find any prop-
4- c,
+ ae /P - 0e,
< \2ax\ + +
A =
is
greatest integer in
ae2 /fl
a positive integer, the number of equations which the accom2

panying z may satisfy is not more than 2(2VD + l) and hence that the
contains no
repeating portion of the continued-fraction expansion of
where
been considerably improved.
A/e - a(r/ty + b{r/t) + c,

= a(x - e/t2 ) 2 + b(x - t/f2)
= ax2 + bx + c - 2aex/t2 +
-B =
of the fact verified in Exercise 2,
more than 2(2y/D
bounded.
which
Exercises 4.7
must be
or z, and
Fill in
satisfy.
is irrational.
is,
they are integers, there
tinued fraction does not stop, since x

that two such equations
that A,B, and
nt/u\.
Use
this result to
show that iac
b2
- 4AC - B2
96
Continued Fractions
Show
6.
that whenever the continued-fraction expansion for x
periodic," that
if
is
is, if
the repeating portion begins with the
the length of the period
xHence show that

pansion and
x =
if
(pkX
VD
first
is
"pure
term, then
we have
satisfies, in brief,
then p k
qk-i, pk-i/qk
D,
D in
show that
if
a\
Pure periodicity and reduced quadratic surds

of this chapter will
be primarily concerned with de-
where
D is a positive integer,
- Dy 2 =
ai
f(<H +
veloping a method of solution of the Pell Equation

x*
lies
for
l/s/2,
l/xz),
if
f(x)
and f(y) =
bears his name, Pell apparently did not contribute to
the equation
Exabove gave a preview of the kind of connection to be expected.

However, as was shown, it does not apply for D an integer. To find a
its solution.
we
If
we may
First,
is also, if
we need
x%
is
de-
y be the conjugate of x,
is defined by the equation
let
call y 2 ,
the equation for
x,
the equation for x 2

for y 2 is /(ai
is
l/y 2);
hence y 2 and x 2 satisfy the same quadratic equation. However, y - l/y 2 =

implies 1 <y 2 < 0, which shows that x 2 is a reduced
ai > 1 and y <
quadratic surd. Hence by continuing this process we see that if x is a
Let
Though
is
l/x 2
shows that the equation
reduced quadratic surd, so
1,
not a perfect square.
between
if" part of the following
a reduced quadratic surd, then x 2
is
fined by the equation x
Exercise 6 cannot be an integer.
The remainder
conjugate
its
have thus proved the "only
then the conjugate of x 2 which

4.8.
if
Theorem 4.8. A number x has a pure periodic continued-fraction

expansion if and only if it is a reduced quadratic surd.
To complete the proof we must show that any reduced quadratic
to
that
and
surd x has a pure periodic continued-fraction expansion.
PS-DgJ-sbi.
Show
call
theorem.
and
7.
We then
and 1.
between
lies
greater than
if it is
We
and -1.
has a pure periodic continued-fraction ex-
D is a positive rational number,
the conjugate of x,
a positive root of a quadratic equation with integer coefficients a reduced

quadratic surd
+ pt-i)/(q& + qt-i).
97
Pure periodicity and reduced quadratic surds
4.8.
portion
x =
(ai,a 2 ,
is <ir+ h
<
and, since
-y,
a r+k
<
,a r ,
is
Xk for every
a r+ i,
Now y r =
1,
a r+fc
ar
k.
Qr+k+i,
1/l/rfi
), where the repeating
implies
-l/y r+i = Or-yr,
we have
ercise 6
an integer we
need more information about the form

of the continued-fraction expansion of \/T).
We shall first need to char-
solution for
numbers x whose continued-fraction expansions are pure
acterize the
periodic; that
Or-C-VlfHjJ
first
the square bracket denoting the greatest integer in l/yr+i.

yk+r
is,
(ai,a 2 ,
in Exercise 6 above,
-,^,01,08, ,*,
ar
we then have
q&*
(gjî
pk)x
p*_i
0.
when x =
and
being positive and the
Pk
qk-i
Pk-i
when x = 1, the latter

we see that the equation has a
former being negative,
and 1.
That
is,
a,k+ r
= [-l/W+i].
Similarly,
However,
x r+l
Or+k.
Similarly,
we may show
that a r_i
Or+k-i
and
so on back to

We leave to the student the proof of the
Corollary 4.8.
Since the quadratic expression has the value pk-i
1/yk+r+i
implies
ai.
qk
ak+r +
integers in
Xr+k+i implies yr+i
Ut+k+i, which implies that the greatest
equal,
that is,
are
the negative reciprocals of the last two quantities
As
The continued-fraction expansion of VZ?, where D is
a positive integer, not a perfect square,
is
of the
form
root between
the other root of the quadratic equation which x
where the repeating part
is
that portion from a% to a*+i inclusive.
98
Continued Fractions
4.9.
The
4.9.
Equation
Pell
Now,
x2
(1)
a solution
is
(xi+iPk+i
However, x =
VD
p*)/(zfc+?*+i
Qk)
implies that x t
a convergent
+ qVD)q <
l/(p
\/2<f,
above theorem
2,3, the
will
if
D >
3.
be an immediate con-
(-1)*.
Theorem
*+2 implies
is
implies
-VD\ =
Hence, except for
of
99
Equation
sequence of
-Dy*=
this notice that x%
\p/q
VD
portion of k terms, then p kf qk
Dg =
p*
Theorem 4.9a. If D is a positive integer which is not a perfect square

and such that the continued-fraction expansion of
has a repeating
Pell
with p and q relatively prime positive integers, then p/q

in the continued-fraction expansion of VD.
We can now prove
To prove
The
If \r/s
l/2s2 and r and s are relatively prime
<
x\
positive integers, then r/s is a convergent of the continued-fraction ex-
(xzPk+i
= \/{VD
4.9c.
pk)/(x 2 qk+i
ai),
pansion of
?*)
To
and substitution in
odd be
the above displayed equation gives us
x.
prove
(oi,a 2
let
this,
,a k ),
the continued-fraction expansion of r/s with k
and
let Xk+i
be defined by the
finite
continued frac-
tion
VD
Pk+i
Qk+i
which
+ pk(^ D - a{)
+ qk(VD ~
qkD
Equating the rational and
If Xk+i is
we
gfc+i
Pk+i
pt+i
aiqk
aipk
Eliminating ai by multiplying the
by qk and adding,
a x qk)
+ p k (VD -
o x ).
first
will
not
xic+i
Pk
0,
qk D
0.
than
less
is the
so denned
not
than
less
(pkXk+i
-p k
and the second
VD
shall
now show
X
is
r/s
the con-
Now, k odd
"
pre-
which
is
p/q
is
Pk/qk
less
<
than
x2
- Dy = 1,
n
-
if
l/2g and qk-i
a solution of one of the equations
which proves our
result.
of our theorem tells us that
<
pk/qk and hence we may write

,
l/2g 2
x, substituting in (1)
pk
l/2q\.
not
e
<
greater than
Xk+i
VD. More
Pk/qk\
is
We
However,
If
for x*+i,
implies that x
x,y positive,
we have
Theorem 4.9b.
be solved
-*>*-*
and the hypothesis
Solving the equation for
given by a convergent of the continued fraction for
cisely,
of
gives a solution of equation (1).
Dy% = 1
Hence we need only show that
x.
-qtx
I*
that every solution of

s
and
of x, which, in virtue of its
+ qk-i) may
= fcg
Xk+1
0,
have just shown that a properly chosen convergent
tinued fraction expansion of
into a continued fraction
1.
pt^/iqkXk+i
(1)
However, p k/qk
Dql =
it
giving
equation by
+ qkPk+i + Pl~
we can expand
1,
which results in our theorem.
We
jCLkjXk+i).
continued fraction for
is
Now, x =
gives
-pflk+i
have a continued-fraction expansion
uniqueness,
irrational parts gives the equations
(2)
(ai,o 2 ,
<*i)
results in the equation
+ VD(qk+i -
(1
and simplifying, we get
eqkqk-i)/eql,
eqk q k -i
<
qk implies
eq\,
that
is,
e(q*
+ qflk-i) ^
1-
100
Continued Fractions
Further information about the Pell Equation
4.10.
Exercises 4.9
from the solution
4,1 of the corresponding equation,
1 on
with
101
the right
side.
1
Prove Corollary
Have we shown that the least positive solution of
4.8.
10.
x2
- Dy* =
In Exercise 8
Would
1 1
1,
that
is,
terms
the solution in positive integers for which y is the least, is given

convergent in the expansion of VD, where k is the number of
/cth
in
3.
Would Theorem
"less than or equal to"?
4.
Using expansions found in the previous set of exercises, find solutions
4.9c necessarily hold
if
Though we have
Dy =
2
choosing
choosing
(that
is,
5.
Show
that
Theorem 4.9b holds
6.
Show
that
if
to be 1)
also for
D=
2,
D=
is
tion
is
4),
the equation
term.
DyZ= -i
the
D=
of
3 (mod 4) the
VD must be even.
kth.
it
Hence we
would
D an integer, not a perfect square,

(r
then the integers
+ sy/D) = rn + VD 8n
then r
of
is
the number of terms
point of view, this informafirst
convergent which gave
free us
from trying any until we reach the kth
If the rth
convergent of
VD gives a solution of
is
divisible
by
k,
the number of terms in the period of the expansion
VD.
r, s,
defined
by
of the
such that
x*-Dy*=> (1)",
= x2
= x 2 whenever
Then,
Equation.
P DQ =
2
VD by Theorem
expansion of
the rth convergent of

if
1, P/Q
4.9b.
Call
it
P,Q
will
is
the least pair
be a convergent
the Rth. convergent,
Then k =
R, since if k were less than R, Theorem

show that the Pell Equation would have a smaller solution
than P,Q. Then xr+2 = x 2 would imply r divisible by k. It thus remains to
show that Xr+2 = Xt, whenever the rth convergent of
gives a solution
of the Pell Equation. Notice that 1 = ( l) r since p r /Qr is less than
VD if r is odd and greater if r is even.
and have z +2
integer, is a solution of
4.9a would
VD
where the ambiguous signs correspond.
Use the
the
prove
shall
of positive integers
9.
is
a solution for this choice, otherwise
not necessary, since one would use the
VD gives a solution of the Pell
x*-Dy = 1,
n an
is
convergent, where k
Suppose we show that x r+i
a solution of
2
with
there
x*-Dy*= 1,
if
number of terms in the periodic part of the expansion

if r,s is
that the smallest positive solution of
From a computational
Theorem 4.10a.
use of exercise 6 or other means, show that
that
the machinery for dealing with the Pell
all
if
has no solution.
Show
If
3.
a solution, but
3 (mod
xz-
8.
were a negative integer?
the solution in positive integers for which y
to be
in the periodic part.
3,5,17,7,11,13.
Dm
hand
at
we have not shown
First,
x*
x*-Dy*= 1
By
if
be?
Equation, there are two additional refinements which are of interest.
least,
7.
were a perfect square?
the result of Exercise 8 hold

it
positive in-
4.10.
the inequality were replaced
of the equations
fori)
if
be
all of r,s,n to
the period?
by
the result hold
what help would
so, of
by the
Would
tegers?
necessary for any or
is it
result of Exercise 8 to find a solution of
17y
NOW,
VD - X
+ Pr-l)/(grZr+l + fr-l) shoWS

(Pr - VD~q )Xr+l = -Pr-l + Vltyr-l.
-
(p&r+l
that
102
4.10.
Continued Fractions
Multiplying both sides by p,

r
2
%
p - Dq = (-l) we get
VDq
4-
and making use
While the computation of the
of the fact that
difficult to
accomplish, there
convergent of
fcth
way
a quicker
is
VD
is
103
not very
to find other solutions of the
Xr+l
=
=
(~l) r {Dqrqr-l
VD + (-iy(Dq qr-i
tfrPr-l)
We
ctr+i
Now, x
VD -
a r+i
+i
where
ai,
+ l/xr+2,
l/zr+2
thus Xt
= x r+t, which was
There
is
served that the
first
So we
the middle.
- p pr-i) +
f
ai.
VD
quite
is
Theorem
- ai=
l/xt,
If p,q is the smallest positive integral solution of the
positive integral solutions
p n ,q n
by the
are given
To
prove
this,
+ qVD) n =
suppose that
r,s is
+ VDq n
symmetry about the continued-fraction

easy to find. The reader may have ob-
(p
Then
a positive solution.
between two successive powers of p + qy/D.

positive integer m, the following inequalities hold:
will lie
+ qVD) m < r + sVD <
(p
That
+ qVD)^
is,
for
sy/D
some
1
.
terms of the periodic part are symmetric about

Letting (p
demonstrate
The
4.1 Ob.
all
(p
tity,
is a positive integer, not
section.
formula
to be proved.
shall
(-l) r (Z>g,g,_i
4. 1 Oc.
Equation,
Pell
where, from the line above,
a certain interesting
expansion of y/D, which
dr+i
8 of the previous
VrVr-l),
Theorem
in Exercise
have
which implies that

Xr+i
was suggested
Pell Equation, as
+ VZ)(prr-l ~
PrPr-1
continued-fraction expansion of
a square,
VD,
where
pm
+ qm VD
and dividing through by
this
quan-
we have
(1)
form
of the
is
+ qVB) m =
< + uVD <p + qVD,
where
(oi,
a 2 a 3 a4
,
a*,
o3 a2
,
2ai,
),
t
where the periodic part consists
From
the
first of
of all the
equations
However,
Exercise
5,
qt+i
aji+i?*
+ qk-i
section 4.4).
t=
a*+ i
= ak +i
+ qn-i/qk
[qk+i/qk], since g*_i
<
(compare
qt.
Replacing
rpm
at+i-r-
Con-
a*
- uVD =
p_
and hence
Suppose
qk+\/qk
(ajt+i,
ak
and u
,as).
Ot_x,
(r
+ 8>/D)(pm - qn VD),
u = pm s- qm r.
VD by -VD in the above, we get
Multiplying corresponding
%*-*
+ qk/qk-i
qk
VD) =
sqm D,
- sVD)(pm +
(r
we have
tinuing this process,
qm
the second equality following from Exercise 8 of Section 4.9, and
+ Pt/9*.
a\
implies qk+i/qk
Hence
+ uVD =(r + sVD)/(pm +
first.
the proof of Theorem 4.9a we have
(2) in
qk+i/qk
terms exhibited except the
sides,
]) U m
2
(r
we have
_ Ds )(pl - qW) =
2
1.
Then
are positive.
&m
qm VD).
+ Du
2
,
p*
+ Dq
However,
ai
and the equality
+ pk/qk =
of the
ak+i
Thus the theorem
is
would show that either

a3 a8
(2ai,
two quantities on the
= 2d, a = ak a t =
proved.
ak ),
left
a*_i,
shows that
equalities (1) then
tion that p,q
proof
(2)
we need
is
> p and
imply that the
>g
t < p
and u < q. Inwhich denies our supposi-
or else
latter holds,
Thus, to complete our
the smallest positive solution.
only show that

p*
and u are
- q*D =
positive.
2
- Ds
Now,
104
Continued Fractions
implies r
> sVD, p m >
sign as (p m s
qmr)(p m s
2 2
(2) implies p s
same
qm
VD
and hence
qn r), which
2
2
q s D,
2
q r
>
0.
Also u has the same
equal to ps 2
is
2
+ Ds q
2
'
2
,
qr 2
However,
and hence u has the
sign as
pV - #* = s - = (r - p )/Z).
p m + qm VD <r + sVD implies either pm <
2
(3)
<?
However,
the former case, the second equality of (3) shows that

latter case the first equality of (3)
shows that u
Using this theorem, we can find
all
r or qm
is
<
In
s.
positive; in the
positive.
solutions of the Pell Equation
by continued
finding the least positive one
is
fractions
by
and Theorem 4.10c
to find all the rest.
If
the length of the period of the expansion of
be the simplest way to find the

2.
The equation x 2 Dy z =
y/D
is
odd,
least positive solution of the Pell
represents an hyperbola.
what
will
Equation?
Draw
perbola.
Show
then
all
that
if
x2
Dy = 1
2
is
solvable
and p,q
positive integral solutions p,q n of \x

(p
and that p*
- Dq\ =
+ qVD) n =
(-1)".
pn
is its
Dy =
+ VDqn
smallest solution,
1
Just as in algebra
we
are given
by
consider roots of equations f(x)
0,
here
we
are
concerned with the congruence

f(x)
a graph
showing the relationship of the solutions of the Pell Equation to the hy-
3.
Introduction
5.1.
Exercises 4.10
(mod m),
Sometimes we might
is a polynomial with integral coefficients.
would
times
we
be satisfied to know
know
roots
and
at
other
want to
the
Of course, in the latter instance we mean the
the number of roots.
number of incongruent roots (mod m)
where f(x)
We
have already found that some
some
linear congruences
have no
roots,
and some have many. However, if the modulus is a prime,

is more satisfactory, since each linear congruence, where the
have one,
the situation
coefficient of
is
shall find that for
theory
is
prime to the modulus, has exactly one solution. We

congruences of higher degree and prime modulus, the
also quite satisfactory.
Exercises 5.1
= a (mod p) has two solutions or none if p

Find integers a and m for which x 2 = a
(mod
p).
^
(mod m) has more than two solutions.
1
is
Show
that a congruence x 2
a prime and a
important theorems in the theory of equations, guess

whether or not they hold for congruences, test your guesses by examples,
105
2.
List
some
of the
106
and then
see
if
The number of roots of a congruence
5.3.
One such theorem would be the
you can prove them.
remainder theorem.
are not congruent

since in the latter case corresponding coefficients
write
should
also
we
(mod 5). Strictly speaking, perhaps,
6ml
The remainder theorem
5.2.
reader will recall that in algebra conditional equations and
The
identical equations
identically equal to zero
polynomial f(x) was denned to be
after terms
if,
each power of x was zero.

is
were discussed.
identically equal to zero
were collected, the
second definition reads:
if f(x)
for all values of x.
tions are equivalent for polynomials
whose
coefficient of
coefficients are
defini-
complex num-
no such polynomial equation can have more roots than its

However, the definitions are
all coefficients of x are zero.
seen
by the example
not equivalent for congruence, as can be
bers, since
(mod m)
(mod
5),
for each value of
x (mod
5),
x*
which
not
is satisfied
all
x and
the powers of x are divisible

5x*
5x
by
5.
In
we have used
coefficients of
Thus we
(mod
25)
R=
and
/(a).
shall
to emphasize identity.
(x
a),
f(x) is a
the quotient
is
= {x-
a)q(x)
-t-
is
and the
The
correspond-
re-
expressed in
(The remainder theorem.)
5.2.
q(x)
R,
This was called the remainder theorem.
ing result for congruences
polynomial in
be written in the form
f(x)
If f(x)
is
a polynomial
such that
=3 (x
f(x)
/(a)
To
prove
then
we
may
by
(mod m).
if
gruence
(2)
a)q(x)
+R
(mod m),
(mod m).
that the identity (1) implies the identical conhold for all integer values of x. If we then replace
this, notice
which must
x by o we get the desired
result.
write
each coefficient of f(x)
The number of roots of a congruence
(mod m)
divisible
is
f(x )
when
it
(1)
5.3.
if
R, then
is
divided
it is
(2)
when we wish
f(x)
only
when
in the theory of equations that
fact,
three horizontal lines for congruence,
use four for identical congruence
if,
mainder
has 25 roots.
Since
showed
Theorem
though the
5),
o does not imply f(x)
for all integers
We
degree unless
(mod
are involved,
but we customarily omit the fourth line when only numbers
our notation, f(x)
since such omission will cause no confusion. Using
integers
o, but /(a) =
for
all
(mod
m)
/(a)
that
implies
(mod m)
polynomial f(x)
These two
107
m g(x)
by m, and,
similarly,
(mod m)
two polynomials with the property that the coefficient of each power of x in f(x) is congruent (mod m) to the corresponding coefficient in g(x). For example, we would write
f(x)
and
g(x) are
5x*
x*
+ 2x -
m, ifa
+ 4)
their
have seen examples of congruences which have more roots than

2
Howdegree, for example, x = 1 (mod 8) whose roots are 1,3,5,7.
ever,
if
We
the modulus
Theorem 5.3a.
(mod
5),
a prime number, we have
If f(x) is
a polynomial of degree n, the congruence
f(x)
(mod
p)
p being a prime number.

(mod
Suppose the congruence has a root a. Then /(a) =
has no more than

2
is
distinct roots,
the remainder theorem shows that
but
x*
^ x (mod 5),
f(x)
^{x-
a)q(x)
(mod
p).
p),
and
108
5A. Power residues
(mod p) implies (6 - a)q(b) m

(mod p).
Then b f a (mod p) implies q(b) m (mod p). Thus, by the remainder
theorem, q(x)
(x - b)r(z) (mod p). Continuing in this fashion, we see
that each root leads to a linear factor, and, since there can be no more
If 6 is
another root, /(&)
than
NRh
are the quadratic nonresidues.
a^-1-
of degree
exercise the proof of the following theorem.
Wilson's Theorem.
for
Theorem 5.3b. Let f(x) be a polynomial

an odd prime, and suppose
of degree
n less than
that
(mod
if iV*i
(x-l)(x-2)
are any two
2 has p
by
1 roots
(x-p+1) =0(modp)
and hence, by Theorem
p being an odd prime.

See Sections 2.5 and 2.11.
coefficients divisible
Show
p,
8.
Prove Theorem
9.
In a manner analogous to the definition of
5.3a,
has
all its
that this implies
5.3b.
where
p,
solvable.
is
p)
and
Using Fermat's Theorem, show that
7.
moduli which are powers of primes (Ref. 10,

pp. 85-90), but they are too complex for inclusion here. We leave as an
is
linear factors, our proof is complete.
There are theorems
Show
N& = N
nonresidues, the congruence
109
v l,
define for a
number j such that,?' 1 = N.

bx + c =
(mod p) has a
given prime p and a quadratic nonresidue N, a
x*
Then
f(x)
If /(0)
p).
=
^
m f(x)q(x) + r(x),
(mod
p) has
(mod
p), x*
r{x) of degree
distinct roots
x may
if
<
and only
if
be replaced by x*~
r(x)
+
+ sj, where r and s are integers. Hint: Write
x* + bx + c = (x + b/2) + c - & /4 and see that -c + b /* m r (mod p)
for some integer r, or else, using Exercise 6, c + 6 /4 = Nr2 m fr
Show
n.
solution of the form r
(mod
above.
Exercises 5.3
10.
t
Is the following true?
(mod m) has one
Suppose a
root, it
"no," give an example.
it; if
(mod
has two?
If
Would your
ra).
Then
your answer
is
if
ax 1
bx
"yes/' prove
conclusions be the
same
(mod
1 .
that every quadratic congruence x 2
the
if
modulus were a prime?
p) for
Show
and
some
that x
r.
s are integers
Show how
1 1
2 (mod 7) has no solution of the form r
and j
is
defined so that j2
3 (mod
the solutions of the congruence fix)
+ sj,
where
7).
(mod m) could
be obtained from the solutions for the prime power factors of m.

12.
Show
that ax 2
+ by = c
2
(mod p)
is
solvable
if
(ab,p)
1.
See Ex-
ercise 9 of Section 2.5.

2.
A polynomial
.
equation of odd degree with real coefficients must have
at least one real root.
Give an example
of
a cubic congruence with no
Power residues
5.4.
roots.
Where does the proof of Theorem 5.3a break down when p

prime number? Does the theorem hold for a power of a prime?
3.
not a
is
The above
exercises
Show that x 2 = a (mod
ber theory.
is
a matter of concern
called a quadratic residue of p; if not, a nonresidue.
x2
= a
5.
If
that
6.
RN
a quadratic residue of p and

a quadratic nonresidue, show
quadratic
is a
nonresidue for p an odd prime.
is
Let Ri,
odd prime
,Rk, k- (p l)/2, be the quadratic residues of an

and N a quadratic nonresidue. Show that NRi, NR t
R
p,
of the usual theory of
ginning that the problem of divisibility which
in
in the theory of
complex numbers
is
numbers.
for real
So
5.4.
If
is
a prime and
xn
numbers
the solution of
definite criterion is given
by
(p,a)
t,
is trivial
always possible, but the corresponding
congruence (mod p) often has no solution.
Theorem
num-
Perhaps this was to be expected, for we noticed in the be-
p) has two roots or none if o is not divisible by

being
an
odd
prime),
and hence there are (p l)/2 values of a for
p (p
2
which x = a (mod p) has a solution. If the congruence is solvable, o is
4.
show that not much
equations carries over to the solution of polynomial congruences in
a (mod p)
the congruence
10
is
solvable
(n,
p-
if
1).
We
5.5.
(mod
where
and only if
a'
When
solvable there are (p
it is
p),
e is
postpone until later the proof for
defined
by
(p
l)/e
l)/e solutions.
all n,
2.
being content for the
1 or when a - 1. In the
- 1.
1), implies (n, p - 1) = n and ne = p
first case n, a divisor of (p
n - a, we can write
ne a is divisible by x
Then, if we notice that x
present with the cases
when n
a divisor of p
is
x where
q(x) is a
x^- 1
xn
a*
polynomial with integer
x ne
+ a* - I =
a'
-a=
(x
o)g(xo)
a)q(x)
= sp -
p,
a*
it
in the
(mod
3),
(mod
a (mod p) has exactly one solution.
a fixed integer independent of
r is
are congruent to
By
many
the
(mod
Why
8).
primes congruent to
method
there are infinitely
of Exercise 3,
many
= -1
(mod
p) for
odd prime
all
a.
an odd prime
divisors of x*
does this show that there are in-
(mod 8)?
prove that for any positive integer r
primes of the form
1.
p= 1 + 2^'N,N=
is
N=
p).
1, 2,
3,--'.
a famous theorem of Dirichlet which generalized the above re-
It states, in fact, that for
sults.
p),
and hence
1 1
one solution of the congruence.
Xo is
x*
factor greater than
ae
show that
form or where
Theorem and the
(mod
if
4.
Hence
independent of x and
Assuming Theorem 5.4, prove that x*

and only if p = 1 (mod 8) that is,
3.
finitely
coefficients.
(x n
If
v is
Express
There
a'
a)q(x),
solution Xo, Fermat's
(mod p) has any

above identity show that
If
(x n
where
Primitive roots
0,1,2,
b without a common
many primes of the form a + Nb,
(a,b) = 1 there are infinitely many
any integers a and
there are infinitely
In other words,
if
primes in the arithmetic progression

Conversely,
if
the last congruence holds, Theorem 5.3b shows that

x
(mod
a,
We can also prove the theorem when a =

p-
To do
1.
this, let
are integers r and s such that nr
and sj
(mod
(mod
p).
there are
of x n
k =
(p
(n,
without restricting
l)s
We know that there

1).
= k. Then x = 1 (mod p)
= 1 (mod p).
s 1 (mod p), it
x*
follows that a s 1
divides n and re*
then shows that
paragraph
preceding
in
the
proved
we
What
k = 1 (mod
number
of solutions
Hence
the
p).
k solutions of x
if
(mod
+ 26, a + 3&,
A:
p)
is
A:,
which
is
equal to (p
l)/e.
number.
of
We call
p if p
If
(p,a)
1.
1 is
a number
a,
a primitive root (mod p) or a primitive root

is congruent to 1 (mod p)
the least positive power of a which
and prove
Exercises 5.4
is
if
a (mod p) has 3 or no solutions for

1 (mod 3),
Where there are solutions, express them in the form xo&v&i?,
show that x8
tf>(p
1)
primitive roots, where
<f>
the Euler ^-function.
In
1
There is another question connected with power residues suggested

by Fermat's Theorem. Certainly there are often smaller powers of a than
-1
a"
which are congruent to 1 (mod p), for example, 2 3 = 1 (mod 7), but
for each prime p is there an integer a such that no power of a smaller
than p 1 is congruent to 1 (mod p)? For instance, the successive powers of 3 (mod 7) are congruent to 3,2,6,4,5,1 (mod 7), and 3 is such a
Theorem 5.5a. Every prime p has
n to be
p-
p) imply
x"^^ ^ =
Conversely,
b,
Primitive roots
5.5.
p-1.
-1
This completes the proof for the case when n divides
has n solutions.
a divisor of
p)
is
that
fact,
suppose we say that "a belongs to the exponent k (mod p)"
the least positive power of a which
Theorem
5.5a
is
included in
is
congruent to
(mod
p)
and
see
112
Theorem 5.5b.
If
is
fc
any
divisor of
1,
there are
(mod p) which belong to k (mod p).

Theorem 2.6b implies that k must be a divisor of
we have
17 in Exercises 2.9,
also belongs to k
k of p
1,
(mod
there are
seen that
p-
By
1.
if
0(fc)
Exercise
a belongs to k (mod p) then a'

k) = 1. Hence, for each divisor
if
then
If
p).
we
stand for the number of integers belonging to k (mod p),
from
to
number
every
1
Since
or 0(fc).
see that for each k, p(k) is either
p - 1 belongs to some k (mod p), we have
we
let p(k)
p
where
h,---,K
1, fet,
p(l)
+ p(fci) +
are the divisors of
a primitive root (mod
0(1)
0(fci)
However we showed
in
Example
5
(mod
is
N
Belongs (mod 13) to
(mod
13)?
Example
= 2*
(mod
10
11
12
12
12
12
12
12; 2
0(3) to 3;
The
0(2) to 2; and 1
0(1) to
1.
The
following examples
0(6) to 6; 2
0(4) to 4;
in that
primitive roots of a prime p serve a very useful purpose
conexpedites
which
logarithm
made a basis for a kind of
they can be
Notice that from the table above, 2 is a primitive

is
table of the residues of the powers of 2 (mod 13)
illus-
Solve 3x
(mod
Table
13).
shows that 3
II
24
=7
Solve x*
(mod
13); that
is,
x*
2 11 (mod 13).
=2
=2
If
and then
r
,
which would imply (why?) that 4r = 11

an impossibility, showing that there can be no solution."* We
12),
(mod
11
13),
Theorem
5.4, since
12/4
3 and 7 3
13).
= 2r and
3.
Solve x*
7 (mod 13); that
is,
x6
2" (mod
13).
Let
we must have 5r = 11 (mod 12). The only solution

congruence is r = 7, and hence the only solution of our given
congruence
have a log to
13) will thus
Hence our congruence may be written 2 4 x = 2s (mod 13)

2B = 6 (mod 13). How would you similarly solve 5x = 3
2.
and 2 ir
Example
proved.
Note that 4 = 0(12) numbers belong to
(mod
13).
m 2 ~* =
of the last
TABLE
1.
and x
(mod
*(*;).
Hence p(fc.) = 0(fc<) for every fc,, and our theorem

For example, we have the following:
except
trate the use of a table of indices or logs.
could have deduced this from
For instance, we would have from Table
p).
Any number
9.
the base 2 (or any other primitive root).
x*
Theorem 2.10b that
there were a solution, x would be a power of 2, say x
+ P(ku),
1.
for g,
II, log 2 ,ia
and only if (r,

or no numbers belonging to k (mod
p)
numbers
4>(k)
113
Primitive roots
5.5.
see that
is
27
11
(mod
13).
Exercises 5.5
Use Table
2.
What
II to find all solutions of x*
9 (mod
13).
other primitive roots does 13 have beside 2?
such as Table
II, for
each such root.
Form a
table,
Use one table to solve the congru-
ence in the previous exercise.
gruential computation.
root
(mod
13).
below:
TABLE
3.
Show how Table
4.
Using primitive roots, prove Theorem
II
5.
n
2n
12
11
The values
of
primitive root
10
We
shall
i?
10
are usually called the indices for the
2.
11
by
for the
adopt the nomenclature of logarithms and
logfl.p
6 to
mean
a (mod p)
Show that x* = a (mod p) is solvable

(e, p 1) for every primitive root g.
II.
5.4.
if
and only
if
log B>Pa
is
divisible
How
a primitive root?
7.
write
could be constructed from Table
would you define a primitive root (mod p*) where p is a prime?

Using your definition, find whether 9 has a primitive root. Does 8 have
6.
numbers 2"
8.
Find
Find
all
all
the numbers that belong to 4>{m) for

the numbers which
belong to 2"-*
m=
(mod
3,9,27.
2*) for
3,4,5,6.
114
Primitive roots
5.6.
where u
modulo powers of a prime
an
is
115
Primitive roots
5.6.
Raise both sides of the equation to the pth power
integer.
and obtain
It is natural to call
power
of
a which
is
of p*
a a primitive root
congruent to
(mod
if </>(p*) is
the smallest
aip
+ kp* + up* )" = 1 + p(kp' + up* +

(p/2)(p-i)(kp* + up+ y+--m 1 + kp* (mod p*
1
(1
Exercise 6 above shows that
p).
8 has no primitive root but the following theorem holds:
Theorem 5.6a.
positive integer
If
is
an odd prime, p has a primitive root for every
r.
If it is the
we
find primitive roots
former
since all terms but the first
+ hp)*" m 1 (mod p
(a + kp) p = a + kp (mod
2
(a
that
We
shall
show below that
all
<f>(p
primitive roots,
1)
a, of p.
2
p are obtained
primitive roots of
in this
we
consider primitive roots of higher powers of p, notice that

is
not
However,
8.
and a
roots of 27,
all
3.
The
the numbers congruent to 2 (mod 9) are primitive
similar statement
may
be made for numbers congruent
This suggests that any primitive root (mod p 2 ) may be a

primitive root for all higher powers of p. We now show that this is the
to 5
(mod
9)
+ kp* -*
1
we have a =
1
p")
it
where k
is
we need
Lemma:
8
>
the following auxiliary result:
Let p be an odd prime and
s 1 + kp'
a'
(mod
p*
with
then
First,
pothesis
is
1
t
tpT
== 1
+ &p*+T
we prove the lemma
for r
(mod
1.
is
equivalent to the equation

a'
kp*
is
a primitive root of
kp*
From what we have just

(mod p), and hence a"-1 = 1 + kp,
lemma with = p 1 and s = 1, we
a primitive root of p2
îp*1 m
(mod p
is
if
by
+ kprH
(mod
pr+2)
_ 1+
=
*(p
kp 2 (mod p 3 ), which shows that a
3
2
however, a belongs to t (mod p ), a = 1 (mod p ) implies
,#(p*)
1,
If,
).
divisible
+ up*
is a
and thus a
a
of the hy-
4>{p
),
and hence
primitive root of
must be
<(p
2
)
or <(p 8 ).
t
The former
must be
3
.
Similarly, (1) implies a*
"
z
), and
(mod p 4 ),
4>{p
a primitive root of p4 Continuing in this fashion, we see

that every primitive root of p2 is a primitive root of p r for all r, and every
odd prime
1,
The congruence
+r
p** ),
would contradict the previous incongruence. Hence
p** ).
have
Thus,
1
by
p*.
First,
must be a primitive root

prime to p. Then, by the
(1)
case.
Third, suppose a
proved
(mod
a primitive root of
primitive roots of 9 are 2 and 5 but
"
p*4 1").
p*, for if o
and hence, by the lemma,
T
and if a is a primitive root of p* T tp r =
(mod
)
However, 4>{p* T ) = p T 4>(p'), and hence t =
(mod <f>(p')), and a
tf>(p* )).
the only primitive root of
implies
Hence, by the conclusion of the lemma for
last congruence.
(mod
(mod p*
belongs to
by
true for r replaced
lemma
what we wished to prove.
is
Before
is
(mod
Second, every primitive root of p*'
a tp
it
of the
lemma, we replace t by tp^ 1 and
(o*""V
which
follows that
1, it
manner.
2
= l + kp
in the hypothesis of the
"
by p*4 2
1.
assume that the hypothesis

1
).
Thus a p a m kp (mod p2 ), which holds if and only if k m (ap a)/p

(mod p). Hence, f or p 1 values of k the last congruence fails to hold,
and for these values a + kp is a primitive root; that is, p 2 has (p 1)
primitive roots, since there are
is,
we have the
is,
that
a*'"
we have
1
in the expansion are divisible
prove the lemma by induction assume
To
r
If,
<(p ~~ 1)
two
This completes the proof for r
(mod p2)
Let o be a primitive root
(mod p), and suppose a + kp belongs to t (mod p2 ). Now, (a + kpY =
(mod p 1), and since from Theorems 2.6a and
1 (mod p) implies t =
2.6b, t also divides p(p 1) = tf>(p2 ) it must be either p 1 or p(p 1).
First,
is
p.
116
The number 2 seems
Euler phi-function of a power of p

other respect in which
For
to be a peculiar prime.
it differs
when p =
a power of p only
is
in behavior
instance, the
no power of 2 beyond 4 has a primitive root.

2 (mod 8), 4 (mod 16), 8 (mod 32), and so on.
is
that
Lemma:
If
>
In
+ 2n)
(1
'" 3
2w4) (mod
This lemma can be shown by expanding the
If
if
elect to
left side
by the bino-
prove
+ 2n) =
by induction as
it
may
1 4-
4n +
4?i
),
5,
gives the
for all n.
assume the lemma for
(1
Exercises 5.6
Are there values of
<
r is
an
+2"
t
(1
since 2{t
Squaring both
integer.
2)
+ 2n)
> +
t
'"2
2
= 1+
when
over the positive integers not greater than 2 _2 ?
every odd number a there
> 5 and
seek to prove
it
for
+L
(n
rc
same be true
(mod
sides,
Suppose
0o
m=
is
= 5
2p*p
a positive integer
(mod 2
such that, for the proper
<
),
<
<
"2
.
we
get
the ambiguous sign
- n2 + 2n (mod
>
This completes the induction and
-pf,
5 (mod
'" 3
or 3
(mod
m/2'o),
so that
and, for each positive

if
8), g
2***),
a
5.
Would
where the pi are distinct odd primes.
(mod
4),
n n2
i,
= 1
choose
gi
</,
(mod
4),
a primitive root (mod
(mod w/p.
p .) such
that
).
4)
Why is such a choice possible? Show that for every integer a,

there are positive integers r<
not greater than
tQ
2,
<
If
>
4,
m belongs
to 2*~ 2
(mod
2') if
(mod 8) otherwise m belongs to a smaller power.

and 7 all belong to 2 (mod 8).
;
and only
If
3,
if
the
for the
above choice
of
<(?>>)
and an integer
prime to m,
which
is
zero or
such that
r
3,5,
(mod 2'), why does it

(mod 2') as r ranges
Hence show that for
be chosen so that
2'~ 1 (n
the theorem, notice that
Corollary 5.6b.
numbers
2*
m m 3
<
5 were replaced by 3?
if
+ 2n + 2%
= 1 (mod 2<) and 1 + 2n belongs

and hence (1 + 2n) 2
~
t 2
power
less
than
2
(mod
2').
to a
On the other
hand, if n = 1 or 2 (mod
'" 3
2
2
= 1 + 2 ? 1 (mod 2')n + 2n* = 2_ (mod 4) and (1 + 2n)
4), n
However, (1 + 2n) 2
= l (mod 2'). Hence we have
2n*
choice of sign
the
proves the lemma.
To prove
<
(mod 2 8 ), where
24?
where
3*
follow that 5 r takes on exactly 2*" 2 different values
Let
'" 3
<
and r such that 5 r
Since Corollary 5.6b shows that 5 belongs to 2*~ 2
2.
3.
2n) 2
the proof as an exercise.
Also show that r (with these restrictions) and the sign are unique.
we may assume
is,
We leave
an odd prime.
We now
to be a
That
is
be
follows.
+ 4(n + n which
of the lemma, since n + n = n n + 2n* (mod 4)
(1 + 2n) = 1 + 4-2n + 6-4n + 4-8n + 16n
H 1 + 8(n + 3n + 2n (mod 32)
bb 1 + 8(n - n + 2n
(mod 32).
=[4, (1
formula
t
we
(mod m)
2').
mial theorem, but this leads to certain complications which
avoided
where p
and
- n2 +
(n
define a primitive root
2,4,p',2p',
= 1 + 2<~
If
we have
fact,
then
3,
we
number
a such that <p(m) is the least power of a which is congruent to 1 (mod m),
then m has a primitive root if and only if m is one of the numbers
5.6c.
Notice that 3 belongs to
~
Theorem 5.6b. If t > 3, an odd number 1 + 2n belongs to 2 t 2 (mod
or a lesser power of 2 according as n = 1,2 (mod 4) or 0,3 (mod 4).
This depends on the following auxiliary result, which we prove first.
2')
If
One
2.
from the other primes
Theorem
117
Primitive roots
5.6.
= g g
r
l
*'
ambiguous
-g]'
sign.
(mod m),
Hint: Show that the
gruence holds for each prime power dividing m.
last con-
Furthermore, show that
118
the gi and a uniquely determine the
r,-
and,
if
m is divisible by 4, the am-
biguous sign.
4.
to
Show
that
then ak
if
m is factored as in Exercise 3 and o is any number prime

1
(mod m) where k
is
the least
common
multiple of
Quadratic Residues
number o such that this fc is the least power congruent
(mod m)? Under what conditions is k - </>(m)?
Is there a
1
5.
Prove Theorem
6.
Find the number
7.
How many
to
5.6c.
of primitive roots of
numbers
will there
r
,
for
p an odd prime.
be belonging to 2 t_2 (mod 2')?
Summary
6.1.
Recall that
of previous results
we have
a a quadratic residue (mod m)
called
x*
has a solution (see Exercises

not solvable,
we
call
a (mod
(a,m)
and
wi)
If (a,m)
5.3).
if
a a quadratic nonresidue
and the congruence
or,
simply, nonresidue.
is-
If
(a,m) ft 1, a is neither a quadratic residue nor a nonresidue.
we showed
In the same exercises

(p
l)/2 quadratic residues and the
that each odd prime number p has

same number of nonresidues. This
could also be seen from the fact that the even powers of a primitive root
of
p are quadratic residues and the odd powers are the nonresidues.
Furthermore, an immediate consequence of Theorem 5.4 is
Theorem
then a
is
6.1
(Euler's Criterion.)
a quadratic residue (mod
and a nonresidue
p;
if
s +1
(mod
a P -D/2
= _!
( rao
must hold
Corollary 6.1.
= 1
The
and p
is
an odd prime,
p),
if
of these congruences
if
a (r-/j
One
p)
If (a,p)
Why?
(mod
If
4), it is
(mod
dp).
if (a,
4),
p)
then
1,
is
a quadratic residue of
a nonresidue.
proof of the theorem and corollary
is left
as an exercise.
119
120
Show why Theorem
from Theorem
6.1 follows
5.4,
and why the
corol-
lary follows from the theorem.
Show that the even powers of a primitive root of p are
(mod
Show
3.
r
,
4.
is
if
is
p,
why must
one of the congruences of Theorem 6.1
an odd prime p if and only if it is

There are at least two methods of
b is a quadratic residue of
all
powers of
p.
a.
Use Exercise 3, above, and the

a primitive root of p.
b.
Show
Show
(mod
a.
b.
8).
that 6
a root of x 2
+ kpY =
so that (x
6.
Xo is
is
any primitive root
fact that
residue, it is
of
is
Euler's Criterion gives a
a (mod p 2 ), and so on for higher powers of
t>
if
and only
if
+ 4fe) =
1
2.
these results as
It is a quadratic residue of
3.
=
=
1
1
(o|p)
for
(mod 8) if p = 2, r =
(mod 2 r) if p = 2 and
of p,
i13
below.
Theorem
or
of
if
and only
if
The Legendre Symbol has the
6.2c.
1, (a|p)
1.
If (a, p)
2.
The symbol
is
a*"-1 )' 2
2.
(mod p) by
multiplicative, that
(a\p){b\p)
3. If
(mod
as b
4. If (r, p)
p may
1,
p),
then
then (a|p)
(ar'lp)
following properties:
Euler's Criterion.
is,
(ab|p).
(b|p).
(a|p).
In words, a square factor prime
be deleted from the left portion of any symbol without
altering its value.
The proof is left as an exercise.

The Legendre Symbol is a
p odd,
3,
the
coming into vogue and is much more popular with the printers. There
is a certain elegance about this symbol, since it has the properties listed
to
A number a is a quadratic residue
= 1 if a is a quadratic residue of p,
= 1 if a is a quadratic nonresidue
= if a = (mod p).
is
follows:
ThiB symbol was originally written (-), but the notation given above
a (mod 16), and so on for higher powers of
2.
p an odd prime as
The results of Exercises 5 and 6, above, show that whether or not a

number is a quadratic residue of a power of a prime depends on the same
information for the prime itself. For purposes of easy reference we list
Theorem 6.2a.
Law which we shall prove
step on the road to this result
first
The Legendre Symbol
6.2.
The
(a|p), defined for
(a|p)
of proof are
that the even powers of 5
so that (xq
8).
of determining the quadratic
in the Quadratic Reciprocity
(a|p)
p.
(mod 2 ) are congruent to the numbers congruent to 1 (mod 8) and less than 2', in other words, that
2x =
b (mod 2') is solvable for x.
b m 1 (mod 8) implies 5
2
=
a (mod 8) then k may be determined
Show that if x is a root of x
Show
way
4),
number (mod p), that is, whether or not it is a quadratic

cumbersome in practice. There is a much more elegant
Legendre Symbol
= a (mod p), then k may be determined
a quadratic residue of 2 l
Two methods
(mod
(mod
character of a
later in this chapter.
if
a quadratic residue of each odd
m and, in addition,
6 = 1 (mod 4)
m=
b s 1 (mod 8) if m
method embodied
proof, suggested as follows:
that
b is a quadratic residue of a composite

if it is
if
Though
a quadratic residue of
as an exercise the proof of
Theorem 6.2 b. The number

number m prime to 6 if and only
an odd prime.
is
prime to
Show that
We leave
factors.
by
replaced
hold?
5.
power
prime factor of
that the statement in Exercise 2 remains true
where now p
If
quadratic residues
and the odd powers nonresidues.
p)
121
The Chinese Remainder Theorem shows that the quadratic residues of a

composite number may be found from the quadratic residues of its prime
Exercises 6.1
The Legendre Symbol
6.2.
Quadratic Residues
particular
example
of
what
is
called a
character, which could be defined as follows: Let g be a primitive root of
122
p,
6.3. Gauss's
Quadratic Residues
and
^1
p be a number for which p 7
let
Then
unity.
that
is,
is
a p
root of
symbol
define the
that
the
is,
first
tive residues
(p
(mod
l)/2 multiples of
recall that
and
(a,p) ?* 1,
if
log iPa
than p for which a
p*
(a,p)
if
the value of the character
is
Their least posi-
17.
7, 14, 4, 11, 1, 8, 15, 5.
where
1,
log iPa.
You must
independent of
g.
Notice that each residue

if
was denned to be that unique positive integer less

g' (mod p).
For the Legendre Symbol, p is 1, and
character depends not only on p but on
123
17) are in order
Mp)
to be zero
where p =
7,
Lemma
necessary, reducing
derived from
predecessor by adding 7 and,
its
Now,
17).
three of these are greater than
affirms that (7|17) = (-1) 8 = -1; that is, 7 is

In a similar fashion, 13 can be shown to be a quad-
Lemma
and Gauss's
17/2,
a nonresidue of
is
(mod
17.
Usually the value of the

ratic residue of 17.
g.
Theorem 6.3a. (Gauss's Lemma.) If p

is the number of least positive residues
an odd prime,
is
and u
(q,p)
1,
of the set
Exercises 6.2
q,2q,Zq,---,q(p-l)/2
(1)
1.
2.
Prove Theorem 6.2b.

If
(mod
which are greater than p/2, then
and p =
3)
but that
(alp)*
if
w, an imaginary cube root of unity, show
Let
Then u
p/2.
3.
Prove Theorem
4.
Prove that properties 2 and 3
&i, 6j,
which are
primitive root.
less
5.
,b be those least positive residues

t
6.2c hold for the
b\, hi,
bt
symbol
all
and
positive
(mod
p) of the set
than p/2 and c h d,

,c u those which are greater than
= (p - l)/2, and the numbers
less
Ci,
c 2,
than p/2.
Prove that
6.3.
Theorem
of
are
same
6.2c.
(2)
is
(1)
(a\p)
= (-D u
(q\p)
not independent of the primitive root of p

for one primitive root it is equal to 1 for every
that the character (a\p) a
result
(l|p)
may
is
(2|p)
p is a prime.
the character denned above.
(p
be shown to hold for
l|p)
0, if
The
lemma
of
it is
portance: First,
it
Gauss which gives a rather curious way
gives a
number (mod
means
a quadratic residue; and second,
of
it is
Law whose
it
1, 2,
q(p
l)/2
p).
(p
were
(2)
some
l)/2 in
distinct
all
order,
(mod
p),
and we would
bfi xc 7
u
(-l) 6x6 2
b t (p
cu
d)(p
c 2)
(p
cu )
(p-h
(-D-^y^Kmodp).
That
is,
,(P-l)/2
is
the goal of this chapter.
Since
paris
by
it
is
prime to
p,
Use
_iW2Lli
1
(
(-tytp-J )*
we may divide both
(>dp).
sides of the congruence
and have
.{p-D/2
Write
1-7, 2-7, 3-7, 4-7, 5-7, 6-7, 7-7, 8-7,
/pjz1\)l
s
V
an important step on the road to the

is
From a computational
determining the primes of which 2
proof
bj) t
of find-
lemma formally, let us see how it works in a

Suppose we wish to determine whether or not 7
quadratic residue of 17.
they would have to be
have
has at least two claims to im-
Before stating the

ticular example.
could show that the numbers of
not very useful, but
Quadratic Reciprocity
we
q(2q)(Zq)
ing the quadratic character of a
point of view
Lemma
Gauss's
There
If
of Euler's Criterion then
(-l) u (mod p).
would complete the proof.
124
Quadratic Residues
It
remains to show that the numbers of
Suppose
r 9* s
bi =
b,-
(mod
would imply
p
and
a^p-
with
p),
rq
impossible, since r
ilarly,
6.3. Gauss's
p),
Then
bi
and hence
s are positive, distinct,
(mod
Cj
^ j.
(mod
sq
p).
Finally,
b<
if
(mod p).
(mod p) with
are distinct
(2)
rq, bj
and
less
sq
(mod
p).
This
where
bi
rq,
is,
if
sum
the
is
Theorem
of the greatest integers in the
by
6.3a, divided
numbers
of set
then
p,
u = M+(p -
is
= sq (mod p), we would have rg -sq (mod p) that r m s (mod p),

or r + s =
(mod p) This is impossible, since both r and s are positive
Cj
is,
(1) of
125
l)(q
l)/8
(mod
2),
Sim-
than p/2.
cjt
that
Lemma
where u
(q\p)
is
Theorem
defined in
and p
6.3a,
is
an odd prime.
Recall
(-1)".
First notice that
and
less
The
iq
Thus the proof of the theorem is complete.

use we make of this theorem is to show
than p/2.
first
Theorem 6.3 b.
If
is
To show
r,-
this, let
p-[iq/p~]
of
Theorem
6.3a.
- 1)/8
iq=pM + rt + rs+
pM + & + b
= pM + bi +
=
(mod 8), and (2|p) = 1 if p = 3 (mod 8).

q = 2 in Theorem 6.3a, and the set (1) becomes
+r
all less than p and hence are all least positive residues
Hence
we
must determine how many are greater than p/2.
p).
if p = 1 (mod 4), then
2(p
l)/4
<
p/2 and 2- (p
+ 3)/4 >
3)/4
< p/2 and
2(p
l)/4
>
it is
odd.
This proves the theorem.
this theorem, see Ref. 14, pp.
There
is
if
4),
then
way
6.3c.
is
2)
of (2) are 1, 2,
X>g =
(mod
(p
2).
l)/2 in
some order and
M + u + Yl (mod
M + *" - 52* (mod
2),
ss
g-
2),
sums being over from 1 to (p l)/2. Each of the sums in the last
is, by the formula for the sum of an arithmetic progression,
2
equal to (p
l)/8. Hence
all
congruence
u
and the theorem
is
M+
(1
proved, since
g)(p*
1
l)/8
= q
(mod
1
2),
(mod
2).
Exercises 6.3
(For other proofs of

1
is
Show
even
if
that
Theorem
and only
if
It is
2.
Of what primes
3.
Use Theorem
If
M = iq/pl + [2q/pl +
[(p-l)<Z/2p],
ff.)
to our ultimate goal.
odd,
p/2.
another criterion for quadratic character which takes us
further along the
Theorem
278
Then, since the numbers
Hence u = (p - l)/2 - (p 3)/4 = (p + l)/4; thus u is even if p =

1 (mod 8), and odd if p = 3 (mod 8). Hence u is even if p = 1 (mod 8),
and odd if p = 3 (mod 8). In the former case (p2 l)/8 is even, and
in the latter
-(p-c)
p/2.
Hence u = (p - l)/2 (p l)/4 = (p l)/4; thus u is even

1 (mod 8), and odd if p = 3 (mod 8).
Second, if p = 1 (mod
2(p
+ c +--- +cM
(p - ci) - (p - c -
Ci
+ b + up
= pM + wp + & + (P - d)
t-i
i-i
These numbers are

First,
one of th6
+u
+--- +&, +
(mod
is
Thus
(P-D/2
2,2-2,3-2, ...,2(p-l)/2.
(3)
+ ri,
the least positive residue of iq (mod p) and hence
is
c,-'s
= 1
if
= (-l) (p2
or
bi's
an odd prime,
(2|p)
that is, (2|p)
where
to 17.
is
m=
Theorem
1 (mod 8).
a quadratic residue?
6.3c implies
6.3b.
Show
that
(m1 - l)/8
6.3c to find the quadratic character of 13 with respect
126
Quadratic Residues
Show
4.
that
according as p
(p
= 3
if
(mod
l)(3p
if
and only
3) or
M = (p - l)/6 or (p + l)/6
6.3c then
-1 (mod
+ 5)/12 or
(p
3),
is
(mod
(35|71).
hence that
l)(3p
This shows that 3
in the respective cases.
prime p
Theorem
in
1)/12 (mod 2)
1 or 11
(mod
6.
We
many
showed
is
The Quadratic Reciprocity Law and property 3
(5|71)
the prime must be congruent to

of only these primes; not
If
1.
x2
(mod
8), since
the prime factors of x 2
all
- 2 m -1 (mod 8). Hence,

a prime factor p = 1 (mod 8). Why does
infinitely many primes of the form 8N 1?
the forms 8N + 3, 8N + 5.
since for x odd, x2
7.
Of what primes
by an odd prime,
divisible
is
is
- 2 can be =
for
each odd
this
x,
(mod
To prove
x2
8),
(mod
2)
by Gauss
imply that there are
been
many
Thus,
if
it
Theorem 6.4a.
distinct
we
shall first
show how
proved
is
not at
first
it
shall
l)(g
remains to prove that
use Eisenstein's geometric proof, which
q).
Let
let
the line
is
denote the region bounded by the lines x
p=
11, q
5.)
If
p and
fM)
5
q are
(-i)',
is,
2
(g|p)
(q\p)
= -
(p|g) if
The second form
(3>k)
tfp
(mod
of the law,
4) or q
(mod
4).
which
is
(mod
4),
easier to use in practice, fol-
*
)
lows from the
first,
since
is
q/2 and the x and y axes, not including any of the bounding
(See the illustration for
even unless both p and q are congruent to
j[
I1
<)
Ii
>
tJ
On
as follows:
connect the origin with
That
l)/4.
the point (p,
ap-
odd primes, then
(p -
- (-D"
used.
(The Quadratic Reciprocity Law.)
(q\p)(p\q)
where
it is
first
There have since
several proofs.
we have
M + N= (p-l)(q- l)/4.
We
and
q,
4-".
Hence
in 1783
(-1)*.
N is that sum obtained from M by interchanging p and

(q\p)(j>\q)
Because the elegance of the result
proofs.
parent, after stating
two
who gave
(q\p)
Devise similar proofs for
was stated by Euler
thirteen years later,
-1,
and hence
2 has
This law has been characterized as one of the most beautiful results
It
the law, notice that q odd implies from Theorem 6.3c that
a rectangular coordinate system,

numbers.
1,
are primes.
a quadratic residue
in the theory of
-(1|7)
shall later
(4)
6.4.
(1|5)
-(71|7)
of the two values shows that 35 is a nonresidue of 71.

remove some of the restrictions on p and g with the result
that computation can be shortened to (35|71) = (71|35) = (1|35) =
1, but we first shall prove our result when both numbers in the symbol
6 a quadratic residue?
is
(71|5)
of the Legendre
and the product
12).
a quadratic residue?
SN +
=
=
127
equal to (5|71)
is
gives the following:
a quadratic residue of an odd
in Exercise 3 of Exercises 5.4 that there are infinitely
primes of the form
the law applies to the evaluation of
(7|71).
(7|71)
how
Symbol
We
Of what primes
By
us see
property 2 of the Legendre Symbol this
if
p
5.
First, let
4).
6.4.
i)
10
11
p/2,
lines.
128
The
6.5.
Quadratic Residues
ways the
proof consists in counting in two
lattice points (that
is,
the points with integral coordinates) within the region R.
The
simplest
way
them
to count
vertical dimension is (q
Hence
l)/2.
is
(p
contains in
l)/2 and whose
its interior
3-1
2
We then shall establish
lattice points.
(4)
No
2.
The connection with the geometric argument may be seen,

L is py - qx = 0, and if (x, y) is any point above L, y > qx/p,
qx >
and, for points below L, py qx< 0.
that is, py
Example 1. Find (71|73). Now (71|73) = (73|71) since 73 m 1
(mod 4) and (73|71) = (2|71) by property 3 of the Legendre Symbol.
Hence
Finally, (2|71) = 1 by Theorem 6.3b, since 71 = 7 (mod 8).
=
could
also
evaluate
of
73.
One
quadratic
residue
and
71
is
a
(71|73)
1,
the symbol as follows: (71|73) = (2173) = (-1|73)(2|73) = 1-1 = 1, by
properties 3 and 2 of the Legendre Symbol, Corollary 6.1, and Theorem
(4) holds.
by showing that
in
there are
3.
N lattice points above L.
on L,
Example
below L,
lattice points
properties 3
M + N lattice points in R and M + N =
Hence it will follow that there are
<p-l)(g-l)/4.
Example
Second, on the line x
below
low
are 1,2,
L and onx=
L and
L is
on x =
This point
is
outside R.
Therefore
[g/p]; that
Hence the
2.
Hence
total
number
+ [2<?/p] ++[(?-
of lattice points in
as
is left
the number of lattice points in
is
R and
exercise.
Notice that an algebraic method of proof would be to consider the
(2|301)(3|301)
= -1,
(3|301)
(301|3)
(31J301)
(301|31)
(186|301)
1,
since 301
(113)
5 (mod
1,
(22J31)
and 186
is
8),
= ( 9|31) =
(-1)31)
= -1.
a quadratic residue of 301.
Compute the
following quadratic characters:
Use the Quadratic Reciprocity Law to do Exercises 4 and 5
of the
previous section.
a quadratic residue of 65?
3.
Is 7
4.
In the proof of (4) show that
Is 14 a quadratic residue of 65?
N is
the
number
of lattice points in
and above L.
py-qx,
(5)
as y takes on the values
The number
It
(17|89), (73)89), (10|137), (29|101).

2.
an
(2(301)
l)2/2p],
numbers
are distinct
(1861301)
Exercises 6.4
R and
M.
Third, the proof that
above
Now
is,
just
( 2|73)
1.
(186|301).
there are [_q/p] lattice points in R,

Similarly, there are [2q/p~\ lattice points in R, be-
1.
Lq/pI
is
Find
(-2|73) by
We have just shown
Legendre Symbol.
(55J73)
= ( 18|73) =
the y coordinates of the lattice points be-
below
which
3.
of the
(55|73)
is
contains no lattice points within R.
in
and 4
Here
(55|73).
(31|301).
to the origin than the point (q,p).
Find
2.
and hence
to be 1
py = qx and p and q are relatively

prime, p must divide the a>coordinate of any lattice point on L, and q
must divide the y-coordinate. Hence there is no lattice point on L closer
First, since the equation of
low
29
6.3b.
1.
lattice points
since the line
to notice that the lattice points
is
form a rectangle whose horizontal dimension
The Jacobi Symbol
of
is
numbers
1, 2,
of (5) is
5.
- l)/2, and x = 1, 2,
(p l)(q l)/4 (whether
(q
(p
Complete the algebraic proof
of (4).
l)/2.
or not they
As above, we can show that none are zero.

negative and hence that
(5) are positive and
6.5.
The Jacobi Symbol
immaterial).
can be shown that
N of
There
is
an extension of the Legendre Symbol which shortens the
130
Quadratic Residues
P\Pi
Jacobi Symbol
The
computation.
"
where the
Pk,
is
since
any product
of p<
and
p,-
(-1
i.m\Pk)-
'
be noted that (m\P) =
For instance,
quadratic residue of P.
(7|5)
(7|13)
-1.
But the
latter
1 does not
(7|65)
shows that 7
imply that
(7|5)(7|13)
1,
(m\P)
m is
must be 1, and hence
(ra|pi)
= 1
implies that
is
a nonresidue
is
necessarily a quadratic residue of
The
4.
Hence
and property 4
only if
but
is
(m|P)
1,
l)/2
(mod
2),
is
To prove
then
proved.
we
proved similarly, and this
Property
that
6, notice
(P|Q)
the product being over
all
each repeated as often as
7T
(pfa),
prime factors
it
leave as an exercise.
occurs in
piOlP and prime factors Qj of
Q,
Hence
or Q.
an odd prime.
Symbol depends on
usefulness of the Jacobi
if
is
resi-
Thus
a nonresidue of P.
of P,
(Pi
Property 5
not a quadratic
is
is
since
due of 5 and hence cannot be one of 65, since x % = 7 (mod 65) solvable
would imply x 2 = 7 (mod 5) solvable. For m to be a quadratic residue of
P it must be one of each factor of P. However, if (m|P) = 1, one of the
symbols
by
1 is divisible
+ Pi)/2 m
i-i
First, it should
p's are primes not necessarily distinct:
= (m|pi)Mp)
(m|P)
numbers
defined as follows for odd
131
The Jacobi Symbol
6.5.
its
having the
we
properties of manipulation of the Legendre Symbol, which
six
{P\Q)(Q\P)=v(p i \q i)(q i \p<) -(-1)S
list in
where
Theorem
P has
1.
(a|P)(o|P)
2.
If
Jacobi Symbol (a|P) for odd positive integers
3.
If (r,P)
(-1|P)
5.
(2|P)
(ab|P).
(mod P),
4.
6.
The
6.5.
the following properties:
(a|P)
- 1, (ar*|P) =
= (-1) <P_1)/2
to
(a|P).
(P -
just
= (-l) (p? 1)/8

(P\Q)(Q\P) = (-1)', where
The
shown above that the first sum on the right is congruent

l)/2 (mod 2) and the second sum to (Q - l)/2 (mod 2). Hence
But we have
(b|P).
(P\Q)(Q\P)
proofs of the
To prove
first
Property
= (P -
1)(Q
three properties are
left
l)/4.
and property 6
Two
as exercises.
Symbol would
_1| P)
(
We
_i| pj)
( _i| pi)(
=,
have to show that

(P
where
to the
ir
is
l)/2
r is of
( _i| pjfc)
(_!),,
r-Z(Pi- 1)A
i-l
the same parity as (P
= (-1 +
pip 2
p k)/2 = (-1
(P
1.
show the advantages
will
TPi =
t-1
Tr(p i
require evaluating (5|97)
l)
Symbol.
Using the Legendre
would be Been
(17|97); each
is 1.
Using the Jacobi
Symbol, we have
(85|97)
+ fc p ),
(97|85)
(12|85)
(3|85)
1.
and hence that 85
is
(85|3)
(1|3)
Notice that
p's,
Now,
we have shown
that (85|97)
(mod
does not imply that 12
nonresidue, since (12|5)

4),
analogous
= l + Y,(Pi-V
and
to have the value 1, and hence their product
-l +
2),
of the Jacobi
Is 85 a quadratic residue of 97?
ratic residue of 97, since the latter is
(mod
l)/4
However,
l)/2.
the abbreviated notation for the product of the
notation for their sum.
1)(Q
holds.
examples
Example
that
4, see
- -
Example 2. Find
is
a prime.
On
a quad-
the other hand, (12|85)
a quadratic residue of 85.
In
fact,
12
is
= 1.
(171|173). It
is
equal to (173|171)
(2|171)
= -1.
132
Exercises 6,5
Find
10.
all
m for the following pairs of values:
respect to
535
877*
93
461*
51
103*
m
The values of m with asterisks are primes.
and 5
1303
2305
2631
Theorem
sion.
3.
Define an extension of the Jacobi Symbol as follows: If a
define
(a\
P)
to be equal to (a|P).
erties listed in
Theorem
ties negative.
Why is this
6.5,
Show
6.5.
Sums of two squares
6.6.
is positive,
'
that this symbol has the prop-
provided that in no symbol are both quanti-
we
In this section
that
is,
for
what
what numbers are the sums
find
integers
Use the
restriction necessary?
results of Exercise 3 to find the primes of
which
is
x2
a quadis
ratic residue.
solvable in integers x and y.
by the
ation of m, a prime,
5.
+ bx + c m
Lemma
(mod
p)
Show
Show
that x2
a = b c, where
If
if
8.
ma
that x*
and only
if
7.
The
question
may
be reduced to consider-
following:
is
(mod
p') is solvable for all
is
primes
p and
all integers
is
if
and only
if
of the Jacobi
x
Suppose p
z cm
if
may
this
be done?)
it
follows that
MP) =
Does
this
+ <P =
c2
If
is
(mod
quadratic
and x 2
E(*1p) =o,
all
positive
i less
than
(a2
+ 6s
(1),
then
of
if
all
a quadratic nonresidue
is
4), since
then
primes congruent to
it
(mod 4), in fact,

would have such a
(mod
4) are the
sums
two squares, as is shown by
sum
of
If
is
two squares in
aside from the order
(mod 4), it can be exone and only one way, that
a prime congruent to
and signs
The fact that such a prime

p.
(mod p), with y prime to p.

and (y'x) z + {y'y) 2 =
for y',
m cannot be a prime congruent to 3
the other hand,
pressed as a
is,
4).
factor.
Theorem 6.6a.
Using the method suggested in the previous exercise, show that
(o|p) p is defined as in Section 2 and is not 1 for all o, then
9.
over
m in
(mod
+ y"s0
cannot even be congruent to 3 (mod
On
many
4)
p) or
would be solvable
This shows that
of p.
prime
just as
(mod
1 (mod p)
Then y'y
(mod p) implies (y'x) 2 = -1 (modp), whereas -1
o.
mean that any composite number P has
is
Then m-n =
n.
an odd prime dividing
residues as nonresidues?
where the sum
is
also useful:
m is any integer prime to P and the sum is over all positive values
m can be chosen so that (m\P) ^ 1
less than P. Hence show that
(when
is
6.6b.
Symbol im-
where
of i
following
Lemma
plies that
MP)
two squares, so
of
a quadratic residue of m.
P is an odd number, show that Property

MF)
and n are each sums
The
solvable for (a,ra)
has no square factors and
If
To prove this, let a2 + b2 = m,

2
2
2
(c + d?) = ific + bd) + (ad - be)
the square of an integer.
a (mod m)
6.6a.
m-n.
be solvable.
two squares,
Find the conditions that the congruence

ax2
6.
of
the equation
fit,
(1)
4.
and y positive integers. Record how many times each value occurs,
and guess what numbers in general are represented by this expres-
Prove Properties
of
+y
x2
for x
781
2.
1, 2, 3,
by the expression
the values less than 40 taken on
Use the Jacobi Symbol to compute the quadratic character of a with
133
Sums of two squares
6.6.
Quadratic Residues
essentially
of x
is
and
sum
y.
of
two squares
mediately once the truth of the following statement
is
will follow
im-
established: If s
is
134
Quadratic Residues
(mod
m 1
s2
an integer such that
( l|p) =
4) implies
6.6.
(2)
has a solution xo
To prove
yo
with
(mod
s since
z* 1
Thus we have shown that
X1X1
(See Exercise 9 in Section 2.5.)
\yo\-
However,
(3) implies
xi
0,1,2, .--^Vp].
Then x
values, that
is,
more than p
Hence two
different values.
of
them must be
- syi = x 2 -sy* (mod p).

x yo = y\ yz, we see that
only
Xi
Thus,
we
if
set xo
x\
2,
solution of the congruence (2)
Theorem 6.6b.
we have
congruent (mod p), and
=
That
(io- sy
and x and y are both

2
if
congruence
sy
x\
x2
and
.To
less
yo
every prime factor of
than
Proof: Suppose p
form a
p.
This
m x\ + y\ m
(mod
y*.
(Ref 10, pp. 188

.
ff.)
the uniqueness of the solution, suppose

x\
(3)
+ y\ =
p =
= x2 + y* implies
y%\
v(.V\
ytXi
Vf)
Now,
(3)
implies that Xi,X2,y hy 2 are
y&x =
If
= yix
yiXi or
first
Xi
2/ 2
(x\
is
Cnj/2
solvable, all the
above
results
may be
m is the sum of two squares
(mod
(mod
and divides m. Then by

py\, and m = p*mi, where
4)
x\
if
and
4) occurs to
Lemma
6.6b,
y\.
divisible
sum
of
two squares.
than
+ y-iXt =
yiy 3
Exercises 6.6
circles
Thus
(4)
shows that
+ y = m have lattice points on

2
0, for
3.
them?
Find for what integers m, the following equation
gers:
p.
+ x iVl Y +
x2
2. Prove that if the lengths of the sides of a right triangle are integers
whose g.c.d. is 1, all the prime factors of the length of the hypotenuse
are congruent to 1 (mod 4).
p.
What
p).
x1
multiplication of
How many
=
= ys
(mod
(3) gives
+ y\)(xl + y*)
and hence
m which is congruent to 3
x = px h y =
equation by
(modp).
all less
the second equation holds, then X\X%
the equations in
xt,yi
1,
+ y\,
x\
Hence
(4)
mi
sentable as the
where Xi,yifr$% are all positive integers. Multiply the

y\ and the second by y\, and subtract to get
y\x\
by p, it must be divisible by p2 and so we can continue.

If p occurred to an odd power in w, we would eventually have the impos2
2
1
On
sible equation m = x + y to solve where p, but not p divides m
the other hand, application of Lemma 6.6a shows that all numbers in
which the primes congruent to 3 (mod 4) occur to an even power are repreIf
p).
To show
y x or xi
we have
(2),
s*y\
{xhy)
1,
mi =
of
an even power in w.
2
y = fcp. Furthermore, the above statement implies X* < p,
p and hence x 2 + y* < 2p. Thus k = 1 and p is the sura of the two
is.
y <
squares x 2 and
2
)(x
y 2 xt
,
y&i = y&t.
positive integer
completes the proof of the statement.
For such a solution
im-
xi,xt,yi,yt all positive integers,
yiy* or
and our proof is complete.

Since x2 + y% = 2 is obviously
combined in the "if " part of
sy takes on
{[Vp]+l|
that (x h yi)
and y range over the values
this statement, let x
with
(3),
plies
sy (mod p)
< Vp >
|zo|
such an
p), (there is
then the congruence
1),
135
Sums of two squares
fax*
yi y 2 )\
solutions are there
+ 2y =
when
m.
m is a prime?
is
solvable in inte-
Bibliography
1.
Numbers,"
F. E. Andrews, "Revolving
Atlantic Monthly, 155 (1935):
459-466.
2.
W. W.
R. Ball, Mathematical Recreations and Essays, 11th
Macmillan Company,
3.
New
ed.,
The
York, 1939.
Harold Davenport, The Higher Arithmetic, Hutchinson's University

Library, London, 1952.
4.
L. E. Dickson, History of the Theory of
Washington, D.C., 1919-1923, Vols.

5.
Numbers, Carnegie Institute,
I, II,
III.
Introduction to the Theory of Numbers, University of Chicago
Press, Chicago, 1929.

6.
G. H. Hardy and E.
Numbers, 2nd
7.
D. H. Lehmer,
Academy
8.
Wright,
An
Introduction to the Theory of
Clarendon Press, Oxford, 1945.

Guide
Tables in the Theory of Numbers, National
to
of Sciences, 1941.
C. C. MacDuffee,
Sons, Inc.,
9.
ed.,
M.
New
An
Introduction to Abstract Algebra,
John Wiley
&
York, 1940.
H. A. Merrill, Mathematical Excursions, Bruce Humphries,
Inc.,
Boston, 1934.
10.
Trygve Nagell, Introduction

Inc.,
11.
New
Number
Theory, John Wiley
&
Sons,
York, 1951.
Oystein Ore, Number Theory and Its History, McGraw-Hill Book

Co., Inc.,
12.
to
B.
M.
New
York, 1948.
Stewart, Theory of Numbers,
The Macmillan Company, New
York, 1952.
13. J. J. Sylvester, Collected
Press,
14. J.
Mathematical Papers, Cambridge University
London, 1912, Vol.
V. Uspensky and
M.
McGraw-Hill Book Co.,
4,
pp. 607-610.
A.
Inc.,
Heaslet, Elementary
New York,
Number
Theory,
1939.
137
138
15.
Bibliography
&
Sons,
A. Youngs, Monographs in Modern Mathematics, see
"On
H. N. Wright,
Inc.,
16. J.
New
W.
First Course in Theory of
Numbers, John Wiley
York, 1939.
Construction of Regular Polygons," by L. E. Dickson, Longmans,
Green
&
Co.,
New
York, 1911, pp. 374ff.
Index
Index
Continued
fractions, infinite,
83
simple, 82
Absolute value, 41
Convergents of continued fractions, 83
Addition, natural numbers, 3
intermediate, 90
negative numbers, 12
rational numbers, 29
Aliquot parts, 57
Approximations by
tions,
continued
frac-
Decimals, 31
periodic (repeating), 38
89
Associative property, addition, 3

multiplication, 4
Determinative property of equality, 3

Diophantine Equation, 62
linear,
63
quadratic, 69
Belongs to an exponent, 49
Distributive property,
5,
12
Binary system, 24
Eisenstein, 127
Cancellation property, fractions, 28
Equality, natural numbers, 3

natural numbers, 7
Character, defined, 121
Euclid algorithm, 16
quadratic, 121
Chinese remainder theorem, 60
Closure property, addition, 3
Euler Phi-function, 48
Euler's Theorem, 48
multiplication, 4
Commutative property,
properties of, 52
Euler's Criterion, 119
addition, 3
multiplication, 4
Congruence, properties of, 42

Congruence notation, 41
Congruential equations, 45
Conjugate, 97
Farey Sequence, 92
Fermat's Last Theorem, 73
Fermat's Little Theorem, 46
Fibonacci Sequence, 76
142
Index
Index
Fundamental theorem
of arithmetic,
18
Transitive property, equality, 3
natural, 2
perfect,
prime,
inequality, 6
56
1,
Twin
15
pure imaginary, 35
Gauss's
Lemma, 122
common
Greatest integer
Well-defined property, addition, 3
divisor (g.c.d.), 15
in,
One-to-one correspondence, 3
"How many,"
Parity, 19
Peano's axioms, 8
Pell Equation, 71,
96
Phi-function, 48
Indices, 112
Prime number theorem, 15
Integers, 13
Primitive root, 111
negative, 11
Primitive solutions, 72
positive, 13
Pythagorean
72
set,
Q
Jacobi Symbol, 129
Quadratic character, 121

Quadratic nonresidue, 108
Quadratic Reciprocity Law, 126
Lattice point, 64, 128

Iieast
common
Quadratic residue, 108
multiple (l.c.m.), 19
Least remainder algorithm, 23
Legendre Symbol, 120

properties
of,
121
Reduced quadratic surd, 97

Reductio ad absurdum (reduction
Reflexive property of equality, 3
Multiplication, natural numbers, 4
negative integers, 13
Remainder theorem, 106

Residue systems,
Multiplicative function, 54
complete
and
duced, 49
Russian Peasant multiplication, 26
N
Number, complex, 35
composite,
to
absurdity), 6
1,
Fermat, 58
15
Subtraction, 10
Sums
of two squares, 133

Symmetric property of equality, 3
imaginary, 35
34
Mersenne, 57
irrational,
Thue, A., 71
re-
multiplication, 5
Wilson's
primes, 2
27
33
real,
Golden Section, 77
"Greater than," 6
Greatest
rational,
143
Theorem and an
suit,
Z
Zero, 10
55
allied
re
^+\-

Jones TheTheoryOfNumbers Text

Uploaded by

Copyright:

Available Formats

Jones TheTheoryOfNumbers Text

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Jones TheTheoryOfNumbers Text

Uploaded by

Copyright:

Available Formats

5T/I.

THE HARRIS COLLEGE

The Theory of Numbers

RINEHART BOOKS IN MATHEMATICS

and L Clifton Snively

Richard E. Johnson, Neal H. McCoy, and Anne F. O'Neill

Fundamental Principles of Mathematics

Paul R. Rider and Carl H. Fischer

Computing with Desk Calculators

CONSTABLE & CO LTD

in their publications like to present their subject

were a deductive science, they

admit that back

postulates and theorems are methods

which are largely induc-

are constructed so that

Postulates are not arbitrarily

certain desirable consequences will follow.

Proofs validate conjectures

an attempt at proof which

but suggests a modified

which leads to another attempt at proof, and so on.

book was written.

to encourage and equip the

discoveries for himself.

For the most part, reasons

Problems are presented and a

after preparation in similar situations.

theory developed to answer the question raised, beginning especially

in the second chapter.

Copyright, , 1955, by Burton W. Jones

All Rights Reserved

Library of Congress Catalog Card

book makes no attempt to take the place of the teacher.

definitions are stated for the natural

that the class will begin rather

reference to the text, will find that its

equivalent to those printed.

a method are apparent.

not supposed to be "read like a

One recommended method is for the student to conceal the proof

and try to proceed from

a slow, but excellent,

hope that by that time the

student will have acquired greater facility

book has been

chapter required about twenty

theory contains countless interesting facts, which can

be used to enliven and

elementary teaching of mathematics.

three chapters especially are rich in material useful

for the secondary

This book also gives an adequate introduction to the theory of numbers

of the first chapter

the gap between problem-solving mathematics and mathematics which

Careful reading of the

manuscript by Dr. Carroll V.

should be of assistance in bridging

and elementary school teacher.

used for such students