C rite ria fo r

D ivisib ility
N. N. V o r o b ’ev
T ra n s la te d and
a d a p te d fro m
th e R u s s ia n by
D a n ie l A. L e v in e an d
T im o th y M c L a rn a n

>sl
Popular Lectures in Mathematics

Survey o f Recent East European Mathematical

Literature
A project conducted by
I zaak W irszup ,
Department of Mathematics,
the University of Chicago,
under a grant from the
National Science Foundation
N. N. Vorob’ev Criteria for

Translated and
adapted from the
Russian edition by
Daniel A. Levine and
Timothy McLarnan

The
University of Chicago
Press
Chicago and
London
The University of Chicago Press, Chicago 60637
The University of Chicago Press, Ltd., London
© 1980 by The University of Chicago
All rights reserved. Published 1980
Printed in the United States of America
84 83 82 81 80 54321

N. N. VOROB’EV is a senior researcher at the

Leningrad Division of the Mathematics Institute
of the USSR Academy of Sciences

Library of Congress Cataloging in Publication Data

Vorob'ev, Nikolai Nikolaevich, 1925-
Criteria for divisibility.
(Survey of recent East European mathematical
literature) (Popular lectures in mathematics)
Translation of Priznaki delimosti.
1. Division. I. Title.
QA242.V6713 513'.2 74-11634
ISBN 0-226-86516-9
Contents

Preface vii
1. Divisibility of Numbers 1
2. The Divisibility of Sums and Products 17
3. Criteria for Congruence and Criteria for Divisibility 22
4. Divisibility of Powers 36
5. Proofs of Theorems 41
6. Solutions to Problems 50
Preface
(to which the author advises the reader
to give his special attention)

Modern mathematics education in the schools is oriented mainly

toward training the student to think in terms of functional relations and
to deal with continuous mathematical objects. The changes planned in
the school curricula [of the USSR] do not affect this orientation. The
last few years have seen the intensive development of new areas of
applied mathematics: computer programming, some aspects of cyber­
netics and operations research, mathematical economics, linguistics,
and so on. A complete understanding of these new sciences and of more
classical studies, however, requires the development of new and fruitful
abstractions for dealing with discrete entities and with combinatorial
problems. There is, therefore, a need for a treatment in the popular
scientific literature of these aspects of mathematics.
Mathematics, like a forest, can be approached and explored by many
paths, and in any one investigation only a small sample of its riches can
be examined. This book may be regarded as a description of one of the
possible walks along the edge of modern mathematics. We shall use
our investigation of criteria for divisibility as a framework within which
we can develop some of the abstract notions of discrete mathematics.
First, ideas arising from the basic theorems of elementary number
theory (such as the fundamental theorem of arithmetic) will be con­
sidered. On a more abstract level, divisibility itself will then be con­
sidered as a relation on the set of integers; that is, as one example of a
more general notion. Finally, criteria for divisibility as such will be
treated as algorithms which enable us to tell whether a given number is
divisible by another. Criteria for congruence, algorithms which act on
any number to produce its remainder on division by a given number,
will also be considered separately.
A number of assertions in this book will be established by two
different methods in order to demonstrate the multitude of interrelations
vii
viii Preface
between individual mathematical facts and the possibility of various
approaches to the same subject.
This book is intended for mathematically inclined students in the
upper grades of secondary school. It presupposes no preliminary
knowledge except the ability to manipulate algebraic identities and a
familiarity with the binomial formula. However, since the logical
structure of the material is fairly complicated, its mastery may require
much attention and patience.
The following plan for the study of this book is recommended. The
first reading may be limited to the basic text of chapters 1-3, and all
problems except numbers 3.3, 3.5, 3.7, 3.13, 3.15, 3.17, and 3.18 may be
ignored. This will provide a general, descriptive acquaintance with the
subject. Most students who are inexperienced in mathematics seem con­
vinced of the validity of the theorem on the unique decomposition of a
natural number into prime factors (apparently considering it almost as
an axiom), and they may therefore simply accept it as true and regard
theorems 1.9-1.13 as its corollaries. On the second reading, however,
the student should provide independent proofs of all theorems in the
order in which they are given. In order that the reader not succumb too
often to the temptation to use prepared proofs, all of the proofs of these
theorems have been placed in a special section. The sole exception is
the proof of theorem 1.7, which has been selected as an example to
illustrate to the reader the desired level of rigor. Chapter 4 should also
be studied on the second reading, and the problems of the basic text
should be solved. Finally, on the third reading, the sections in small
print should be read and the problems relating to them should be
solved.1
The reader who wishes to deepen his knowledge of the theory of
numbers should consult the classic work of Professor I. M. Vinogradov,
Elements o f Number Theory (New York: Dover, 1954), translated from
a later edition as An Introduction to the Theory o f Numbers (London and
New York: Pergamon Press, 1955). For the abstract theory of relations
on a set we recommend Lectures on General Algebra by A. G. Kurosh
(Oxford and New York: Pergamon Press, 1965) or Lattice Theory by
Garrett Birkhoff (Providence, R.I.: American Mathematical Society,
1967). Finally, a more detailed and systematic explanation of the concept
of an algorithm may be found in the booklet by B. A. Trakhtenbrot,
Algorithms and Automatic Computing Machines (Boston: D. C. Heath,
1963); and a rigorous treatment of the theory of algorithms is given in
1. Since the average Soviet student has perhaps had more experience with
rigorous mathematical proof than his western counterpart, these directions are
particularly important to American students.
 Preface ix
a monograph by A. A. Markov, one of the founders of this theory:
The Theory o f Algorithms (Jerusalem: Program for Scientific Transla­
tions, 1961). [Available from the Office of Technical Services, U.S.
Department of Commerce, Washington, D.C.]
1 Divisibility of
Numbers

1.1
The sum, the difference, and the product of two integers (whole
numbers, both negative and nonnegative) is always an integer. We
usually express this by saying that the set of integers is closed under the
operations of addition, subtraction, and multiplication.
It is not true, however, that the set of integers is closed under division.
Division of one integer by another will not always yield an integer. There­
fore, in the study of the division of integers, one of the first questions to
arise is that of the possibility of performing division on two given integers;
that is, of the divisibility of one of these numbers by the other. This
question does not arise with regard to the other three arithmetic opera­
tions, since they may always be performed within the set of integers.
We shall assume in this book that the basic properties of the arith­
metic operations on integers, as well as the simplest properties of
equalities and inequalities, are known. When we use the word “ num­
ber,” we shall mean “ integer,” unless some other meaning is specified.
The nonnegative integers, 0, 1, 2 , . . . , are often called the natural
numbers. When we speak of the natural numbers collectively, we shall
use the term the set o f all natural numbers.
Definition 1.1. The number a is divisible by the number b (or,
equivalently, the number b divides the number a) if there exists a number
c such that a = be. (Recall that “number” here means “integer.”)
If the number b divides the number a, we write b\a. We emphasize
that the symbol b\a does not specify an operation that should be
performed on the numbers a and b, but constitutes an assertion about
these numbers. The statement b\a may be true or false, depending on
the values of a and b. For example, 2|4 is true, but 3|4 is false.
1
2 Chapter One
There are several methods for determining whether b\a is true or not;
that is, for determining the divisibility of a by b. One such method is to
divide a by b, using long division. This process is often long and tire­
some, and one would naturally like to establish the divisibility in ques­
tion without actually performing the division. Besides, we may be
interested only in a “ yes-or-no answer ” concerning the divisibility of the
number a by the number b, but if we perform the division, we obtain
the quotient and remainder as well. These numbers are of no value at
all to us, since at the moment we may wish only to know whether or not
the remainder is zero. So, when we perform the division, we waste some
portion (and, apparently, not a small one) of our work in obtaining
useless by-products. We can only hope that methods for determining
divisibility more directly than by the “ crude” method of long division—
methods that will not give us such abundant by-products—will allow
us to establish divisibility in a more economical way. These hopes are,
in fact, justified, and such methods for establishing divisibility do exist.
They are called criteria for divisibility.
The reader is no doubt already familiar with some criteria for
divisibility. The aim of this book is to examine a number of these
criteria, with an emphasis on principles.
The essence of any criterion for divisibility by a given number b is
to reduce the question of whether b divides a to the question of whether b
divides a', where a' is some number smaller than a. (It is not difficult
to see that the checking of divisibility by ordinary division is, in fact,
based on this idea.)
Thus, a criterion for divisibility is an example of an extremely
important, though perhaps unfamiliar mathematical object. Such a
criterion is not a formula, not a theorem, not a definition, but a process,
just like the multiplication of numbers “ by columns,” or the process
of calculating successive terms of an arithmetic progression.
The concept of a criterion for divisibility will be made more precise
in the next section.

1.2

In the definition of divisibility, nothing is said about how many

different quotients a number a may have on division by b. We must
settle this matter once and for all, so that we need never return to it.
Suppose that

a = be ( 1. 1)
 Divisibility of Numbers 3
and also that

a = b c i.

From these two equalities we obtain

be = bc1 ,

or

b(c — cx) = 0 .

If b 0, then c — c1 = 0; that is, c = cx. If, however, b = 0, then one

can easily see from equation (1.1) that a = 0, and hence that (1.1) is
satisfied by any c.
Thus, if b\a and b = 0, then a = 0 also, and since the “ quotient”
of such a division cannot be uniquely defined, we say that it is impossible
to divide by zero. If the divisor b is distinct from zero, however, division
is permissible, and the quotient is unique.
From now on, when speaking of division, we shall always mean
division by a nonzero number. —•
Let us state some of the simplest properties of divisibility.
T heorem 1.1. a\a (reflexivity).
T heorem 1.2. I f a\b and b\c, then a\c (transitivity).
T heorem 1.3. I f a\b and b\a, then a = b or a = —b (antisymmetry).
T heorem 1.4. Ifb \a and |h| > \a\, then a = 0.
C orollary. Ifb \a and a ^ 0, then \a\ > \b\.
T heorem 1.5. A necessary and sufficient condition that b\a is that
\b\ | \a\.
On the basis of this theorem, we can, in the future, limit ourselves to
consideration of the case in which the divisor and the dividend are
positive numbers.
T heorem 1.6. Ifb \a x, b\a2, . . . , b\an, then b1^ + a2 + • ■• + an.
C orollary. I fb 1^ + a2 and b\au then b\a2} 1

1. 6|ai + a2 is “ shorthand” for 6|(ai + a2). Similarly, m\a — b is another way

of writing m\[a — b).
4 Chapter One
It is by no means correct to suppose that all of these theorems
are obvious and require no proof. One must remember that in mathe­
matics any statement other than an axiom or a definition requires proof.
It is necessary to prove the facts listed above (for example, the reflexive
law of divisibility), because we cannot derive them from the definition
of divisibility alone—we must also use the properties of the integers.

The following example will help to make this clear. The sum, difference,
and product of any two even numbers is always an even number. It is not
always possible, however, to divide a given even number by another. More­
over, even if the division is possible, the quotient need not be even. Therefore,
we can introduce the concept of the “ even divisibility” of even numbers.
D efinition ] .2. The even number a is evenly divisible by the even number
b if there exists an even number c such that a = be.
It is clear that theorem 1.1 is false for even divisibility since for any
nonzero a there exists no even number c such that a = ac.
We shall consider more questions concerning even divisibility later. The
example of even divisibility shows that it is possible to construct theories of
divisibility on sets other than the set of integers (in this case, the set of even
integers) in such a way that theorems which are true for some of these
theories are false for others.

Problems. Prove the following assertions:

Problem 1.1. a|0.
Problem 1.2. 1| a.
Problem 1.3. If a\\, then a = ± 1.
Problem 1.4. For any a / 0, there exists a number b different from
a such that a\b.
Problem 1.5. For any number a there exists a number b different
from a such that from c\b and a\c it follows that c = b or c = a
{a, b, and c are assumed to be natural numbers).
Problem 1.6. Prove the analogues of theorems 1.2,1.3, 1.4, and 1.5 for the
case of even divisibility.
Problem 1.7. Construct a theory of divisibility in which theorems 1.1,1.3,
and 1.4 are true, but not theorems 1.2 and 1.6.

1.3
The most cursory acquaintance with the concrete facts of divisibility
convinces us that the divisibility of numbers is in practice not strongly
related to their magnitude. One may find fairly small numbers that
have a comparatively large number of divisors. For example, 12 is
 Divisibility of Numbers 5
divisible by 1, 2, 3, 4, 6, and 12; and the number 60 has 12 divisors.
These numbers so rich in divisors may be contrasted, however, with
extremely large numbers that have the minimal number of divisors—
two (by theorem 1.1 and problem 1.2 every number distinct from 1 has
at least two different positive divisors). In fact, some laws connecting
the divisibility properties of numbers and their magnitude are known,
but these laws are so complicated and intricate that we shall not touch
upon them here.

1.4
It is perhaps more interesting that divisibility itself allows us to
establish a certain ordering of the integers that is different from the
usual ordering, but has much in common with it.
Indeed, let us consider what we mean by the possibility of ordering
the integers by their magnitude. It is not hard to see that by this
possibility we mean that, for some pairs of numbers a and b, the relation
“ greater than or equal to ” holds:

a > b,

which signifies that a — b is nonnegative (that is, there must exist a

natural number c such that a = b + c). But the notion of divisibility
is nothing more than the idea that some pairs of numbers a and b
satisfy some other mathematical relationship (namely, that there exists
a number c such that a = be). Thus, the relation of divisibility and the
relation “ greater than or equal to ” are concepts of the same nature, and
thus we may speak of their common properties, or, on the other hand,
of their differences.
We note immediately that the relation “ greater than or equal to ” has
more properties in common with the relation of divisibility than does
the relation “ greater than.” This is partly because the relation “ greater
than or equal to,” like the relation of divisibility, is reflexive (that is, the
relation a > a is true for any a), while the relation “ greater than” is not
reflexive (indeed, the inequality a > a is never valid). It is precisely
because of this that the relation “ greater than or equal to ” is considered
to be an order relation on the natural numbers, while the apparently
simpler and more natural relation “ greater than” is not.

1.5
The relation > has the following easily verifiable properties:
Property 1.1. a > a (reflexivity).
6 Chapter One
Property 1.2. If a > b and b > a, then a = b (antisymmetry).
Property 1.3. If a > b and b > c, then a > c (transitivity).
Property 1.4. In any sequence of natural numbers with distinct terms,
fli > a2 > a3 > • ■■> an > • • • ,

there is a “ last number.” This property of the relation is sometimes called

the well-ordering principle for the set of natural numbers.
The well-ordering principle is fairly complicated and might seem at first
to be a bit artificial, but it reveals extremely important features of the
structure of the set of natural numbers ordered by the relation > , and many
other properties of this relation may be derived from it. Moreover, we shall
discover that it is the basis for the method of “ proof by induction,” which
is widely used in mathematical arguments.
As a useful application of this property, we shall prove that there exists a
natural number a such that for any natural number b, if a > b, then a = b.
Indeed, if there were no such number, then for each an we could find an
an+1 such that an > a„ +1 and an # an+1. Starting with an arbitrary ax, we
could then obtain a sequence with distinct terms
ax > a2 > a3 > ■■■> an > an+1 > ■■■

which never ends. But the existence of such a sequence contradicts the well-
ordering principle for the natural numbers.
Thus, the existence of the number a mentioned above has been proved.
This number is called the first or minimal natural number (clearly, it is the
number zero). We note, however, that we have not yet established the
uniqueness of the minimal number. This uniqueness must be shown by a
more roundabout method.
Property 1.5. For any number a there exists a number b distinct from a
such that b > a.
This property of the set of natural numbers is called its unboundedness
under the relation >.
Property 1.6. For any number a that is not minimal, there exists a number
b such that a > b, a ^ b and such that for any number c, a > c > b
implies that c = a or c = b. This formal assertion, when translated into
more descriptive language, means that every natural number except 0 has
an immediate predecessor. (This could be formulated differently by saying
that the set of all numbers less than a given number has a largest element.)
Property 1.7. Either a > b or b > a. This property of the relation > is
called the law of dichotomy. In mathematics the term “ dichotomy” is used
to express the necessity of having one of two possibilities realized. The
word itself is of Greek origin and means “ a division into two parts.”
We emphasize that properties 1.1 through 1.7 are properties of the
relation > on the natural numbers and are not properties of the individual
numbers connected by this relation. Thus, some of the properties 1.1 through
1.7 may not be true of relations other than the magnitude relation >.
 Divisibility o f Numbers 7

Problem 1.8. Using only properties 1.1 through 1.7 of the ordering >
and no other properties of the natural numbers themselves or of the
operations on them,
a. prove the uniqueness of the minimal element;
b. prove the uniqueness of the immediate predecessor;
c. formulate the definition of the immediate successor of a given number
a (that is, of the number a + 1), and prove its existence and uniqueness.
Problem 1.9. Determine which of the assertions (properties) 1.1 through
1.7 remain valid for the relation “ greater than” ( > ) .

1.6
The validity of properties of the relation > (and of properties of any
other relation as well) may be established in either of two ways. First, we
may use properties of individual numbers or known peculiarities of the
structure of the set of all natural numbers. This is precisely the way in
which properties 1.1 through 1.7 were verified. Then, having convinced
ourselves that properties 1. 1 through 1.7 are correct, we may begin to reason
abstractly and may derive further properties of this relation solely from
properties 1.1 through 1.7. This is the way in which we proved the existence
o f the minimal element and the assertions o f problem 1.8.
A second approach to this problem is particularly widely used in modern
mathematics. This is the so-called axiomatic approach in which certain
axioms are laid down (in our case, properties 1.1 through 1.7) which reflect
the basic properties of the objects studied and are not subject to proof. All
further assertions, called theorems, are derived from these axioms by strictly
logical means, without any reference to the objects studied.
It may seem to some readers that the consideration of relations in
isolation from the objects connected by these relations (for example,
numbers) is the height of mathematical abstraction, and that so much
abstraction is unnecessary in practical life. We shall make two remarks on
this point.
In the first place, the arguments presented here are not especially
“ abstract” from the standpoint of modern mathematics. Contemporary
mathematicians, moreover, find it necessary to consider many relations
simultaneously, and even to connect pairs of relations by new relations (by
“ relations o f the second order,” so to speak).
The material presented so far will enable us to illustrate the concept o f a
relation between relations by an example:
Let a, j3,. . . be some collection o f relations “ connecting” the natural
numbers. By this we mean that for any pair o f numbers a and b and any
relation y of our collection, we know whether or not the numbers a and b
are related (or connected) by y. If a and b are related by y, we shall write ayb.
We shall say that the relation a is stronger than the relation /3 and write
a => [3 if any pair of numbers connected by the relation /3 is also connected
by the relation a; that is, if afib implies aab.
Thus, denoting the relation o f even divisibility by |e, we may write | => |e.
Furthermore, it is clear that > => > . On the other hand, neither | >
nor > => | h old s: The relations of divisibility and o f “ greater than or equal
8 Chapter One
to ” are not connected by the second-order relation =>. It is precisely this
situation that was described in words in section 1.3.
Further training is, of course, necessary in order to operate freely with
such complicated concepts as relations between relations.
In the second place, similar and even more abstract reasoning is being
employed more and more frequently in the applications of mathematics to
economics, biology, linguistics, and military affairs. Unfortunately, a more
detailed explanation of this reasoning and of the reasons for its use would
take us too far from our basic topic.

1.7
The method of proof known as mathematical induction (sometimes
called complete induction) is intimately tied to the ordering of the
natural numbers by the magnitude relation > . This method is usually
applied in the following form:
Let A(n) be some assertion concerning an arbitrary natural number n.
In order to prove that A(n) is true for any natural number n, we must
establish the infinite sequence of assertions

.4(0), .4 (1 ),..., A (n),. . . .

We shall suppose that:

a. the assertion .4(0) is valid (the “ basis for induction”);2
b. from the validity of the assertion A(n), there follows the validity
of the assertion A(n + 1) (the “ inductive step”).
The principle of mathematical induction asserts that under the
hypotheses (a) and (b), A(n) is valid for any natural number n.
The principle of mathematical induction is not an independent statement,
but may be derived from properties 1.1 through 1.7 of the ordering of the
set of natural numbers under > .
Indeed, let us suppose that conditions (a) and (b) are satisfied for the
assertions A(n), but that the conclusion asserted by induction does not hold.
This means that there must exist numbers m for which the assertion A(m)
is false. Let be one such number. If for all n < mx the assertion A(n)
is true, then m-i is the least of the numbers for which A(n) does not hold.
But if A{n) is true for not all n < mu then there must exist an m2 < m-i
such that A(m2) is false.
As a result, we arrive at a sequence of distinct natural numbers,
mi > m2 > • • • > mT > • • • , (1.2)

2. The assertion A( 1) is often taken as the basis for induction. Clearly, this
difference is not important. It is only important that the basis of induction involve
the first of the numbers we consider.
 Divisibility of Numbers 9
for which A(ri) dees not hold. By the well-ordering principle (property 1.4),
there must be a last element mT in the sequence (1.2). It is clear that the
number mTis the smallest number for which A(n) is false.
Since ,4(0) is true by hypothesis, m, # 0, and so there exists a number
m* immediately preceding mr (in actuality this number is mr — 1). Since
mf < mr, the assertion A{mf) must be true. But then by hypothesis (b) of
the principle of mathematical induction, the assertion A(mf + 1), that is,
the assertion A(mr), must also be true, and we have obtained a contradiction.
This contradiction shows that there can be no natural numbers m for which
A(m) does not hold.
The above argument should not be taken as a proof of the absolute truth
of the principle of mathematical induction. It shows only that it is possible
to derive this one mathematical assertion (the validity of the method of
induction) from others (the properties of the relation >). We took these
properties as axioms and, accordingly, did not prove them but only verified
them as being consistent with our intuitions. Any attempt to give a mathe­
matical proof for them would require the introduction of new assertions
as axioms.
In particular, proofs of the well-ordering principle necessarily involve
some sort of inductive argument (the reader can convince himself of this
independently).
The booklets by I. S. Sominskii, The Method o f Mathematical
Induction, and by L. I. Golovina and I. M. Yaglom, Induction in
Geometry (Boston: D.C. Heath, 1963), contain a large number of
examples of the use of this method. Inductive proofs will be used
frequently in this book.
Problem 1.10. Let pairs of objects of an arbitrary nature (for example,
numbers, points, functions, theorems, and so on) be connected by some
relation £- satisfying properties 1.1 through 1.7. Prove that these objects
(elements) can be enumerated (that is, written in some order) Au A2, ...,
An, . .. in such a way that At £- Aj if and only if i > j.
In essence, the statement above means that a relation having properties
1.1 through 1.7 orders the set on which it is defined in a linear chain of
elements:

Ai £- A2 £- A3 £- • • • .

1.8
Let us return, however, to the relation of divisibility. In the case of the
positive integers, theorems 1.1, 1.2, and 1.3, and problems 1.3, 1.4, and 1.5
show that in properties 1.1 through 1.6 we may replace the relation > by
the relation |. The statement of property 1.7 regarding divisibility would
assert that “ given any two positive integers, at least one is divisible by the
other,” but this is clearly false. Thus, the divisibility relation has all but
one of the properties of the magnitude relation. As a result, the relation of
divisibility orders the natural numbers, but in a more complicated manner
10 Chapter One
than that of a linear chain (see figure 1.1). We note that numbers close to
each other in terms of magnitude often turn out to be fairly “ far ” from each
other in the sense of divisibility. The numbers 4 and 5 or 7 and 8 demon­
strate this clearly.

Let us try to pass from the divisibility of positive integers to the divisi­
bility of the natural numbers; that is, let us try to include zero in our
considerations. Then the scheme in the figure would be enlarged to include
a box lying above all the other boxes in the scheme, for zero is divisible
by any natural number, while no natural number distinct from zero is
divisible by zero.
We leave it to the reader to reformulate and verify assertions 1.1 through
1.7 for the case in which zero is included.
1.9
D efinition 1.3. Any relation £- satisfying the conditions
(1.1') reflexivity (a £- a),
(1.2') antisymmetry (from a £- b and b £- a it follows that a = b), and
(1.3') transitivity (from a £- b and b £- c it follows that a £- c),
is called a relation of partial ordering.
Relations of partial ordering play an important role in situations in
which there is no “ natural” linear ordering, such as situations in which
each object is identified by several indices that are qualitatively not
comparable to one another.
For example, consider the results of a sports competition that includes
several different sports. If one team captures higher places than another in
all of the sports represented, it would be natural to consider that the first
team has attained greater distinction. But if the first team captures higher
places in all events except croquet, in which the second team is stronger,
then the question of the final relative standing of the teams would not be
so clear. Croquet enthusiasts might even insist on giving a higher place
to the second team. In any event, any final distribution of places must be
connected with some agreement on how to evaluate the points.
 Divisibility o f Numbers 11

1.10
It is easy to construct relations satisfying conditions 1.1' through 1.3'
of definition 1.3; that is, relations of partial ordering. Indeed, a great many
types of objects can be partially ordered, often in several different ways.
As a consequence, there is very little that can be said about a general partial
ordering except that it satisfies our three axioms. In particular, we cannot
always apply the method of mathematical induction to the objects on which
a partial ordering is defined.
Suppose, however, that the following conditions are added to conditions
1.1' through 1.3': (1.4') well-ordering; (1.5') unboundedness; (1.6') each
object distinct from the minimal one has at least one immediate predecessor;
and two new conditions:
Property 1.8. Each object has only finitely many predecessors;
Property 1.9. For any a and b £- a (b A a) there exists a c preceding b
and such that c £- a.
It turns out that on the basis of the partial ordering of the set of natural
numbers by a relation that satisfies conditions 1.1' through 1.6', 1.8, and
1.9, it is possible to construct a modification of the method of mathematical
induction. This consists of the following:
Suppose, once again, that A(n) is some assertion about an arbitrary object
n in our given set. We shall assume that
a. the assertion A(a) is valid, where a is any minimal object under the
ordering
b. if n is any object, and all assertions of the form A(m), where n £- m
and n =£ m, are valid, then A(n) is also valid.
The new form of the principle of induction states that if conditions (a)
and (b) are fulfilled, then A(n) is true for any object n in our set.
Problem 1.11. Derive the “ new form” of the principle of induction from
its “ old form.”
Since the relation of divisibility on the natural numbers satisfies condi­
tions 1.1' through 1.6', 1.8, and 1.9 (formulate and check conditions 1.8 and
1.9 for the relation of divisibility), this principle of induction is applicable
to the relation of divisibility.
As applied to divisibility, the new principle of induction may be formulated
as follows: If some assertion A{n) is valid for n = 1, and its validity for all
divisors of a number n distinct from n implies its validity for n, then it
holds for any n.

1.11
Since the division of integers (as we have seen) is not always possible,
it is natural to consider—along with division—another more general
operation that may always be performed and which coincides with
division when the latter is possible. This operation is called division with
remainder.
12 Chapter One
Definition 1.4. To divide a number a by a number b (b > 0) with
remainder means to represent the number a in the form

a = bq + r ,

where q and r are integers and 0 < r < b.

The number q so obtained is called the partial quotient and the
number r the remainder. Clearly, r = 0 if and only if b\a. In this case q
is equal to the quotient of the division of a by b.
We shall show that division with remainder may always be performed
and that the partial quotient and the remainder are completely deter­
mined by the dividend and divisor—that is, that they are unique.
Suppose first that a > 0. We write the numbers

a, a — b, a — 2 b ,. . . , (1.3)

one after another until a negative number appears. Clearly, such a

number must eventually appear.
(In fact, this follows from the well-ordering of the set of natural numbers
by the relation > .)
Suppose that the last of the nonnegative elements of sequence (1.3),
that is, the smallest of them, is the number a - bq. Denoting this
number by r, we have

a = bq + r . (1.4)

Clearly, r < b (otherwise the number r — b, that is, a — (q + 1)b,

would be nonnegative, and this cannot be, since r is the least of the
nonnegative numbers in (1.3)). Thus, (1.4) is the desired representation
of the number a.
Now suppose that a < 0. Reasoning in a manner similar to the above
argument, we write the sequence of numbers

a, a + b, a + 2 b ,. . . ,

until the first nonnegative number r appears (it is easy to check that
r < b). Let

r = a + b q '.

Then, denoting - q ' by q, we obtain

 Divisibility of Numbers 13
a = bq + r ,

as was required.
The possibility of division with remainder is thus proved in all cases.
We shall now prove the uniqueness of the partial quotient and
remainder; that is, we shall prove that if

a = bq + r , (1.5)

and, at the same time,

a = bq1 + rx , ( 1. 6)

with 0 < r < b and 0 < rx < b, then q = qx and r = rx.

In order to prove uniqueness, it does not suffice simply to note that
the numbers q and r obtained by the process above are completely
determined by a and b, since such an argument in no way precludes the
possibility that other values of q and r may be obtained by some
completely different method.
To give a complete proof of uniqueness, therefore, we combine
equations (1.5) and (1.6) to obtain

bq + r = bqx + rx ,

that is,

r - rx = b(qx - q) ;

which says that r — rx is divisible by b. But \r — rx\ < b (since 0 <

r < b and 0 < rx < b), and, according to theorem 1.4, this is possible
only if r — rx = 0; that is, if r = rx. But then

b(<h - q ) = 0 ,

and since b is not zero, qx — q = 0; that is, qx = q.

Thus we have proved the following theorem.
T heorem 1.7 (Division with remainder). Given any numbers a and b
(b > 0), there exist unique numbers r and q such that a = bq + r, and
0 < r < b.
Problem 1.12. Formulate and prove a theorem on division with remainder
for even divisibility.
14 Chapter One

1.12
D efinition 1.5. A positive integer p not equal to 1 is said to be prime
if it is divisible by itself and by 1, and by no other positive integers.
Examples of prime numbers are 2, 3, 5, 7, 11, and 13.
A positive integer distinct from 1 which is not prime is said to be
composite.
T heorem 1.8. There exist infinitely many prime numbers.
Any number that divides both the numbers a and b is called a
common divisor of these numbers. The largest of the common divisors
of a and b is called their greatest common divisor and is denoted by
{a, b).
If the greatest common divisor of the numbers a and b is 1, then these
numbers are said to be relatively prime.
In other words, the numbers a and b are relatively prime if they are
not both divisible by any positive integer except 1.
T heorem 1.9. I f a and p are natural numbers and p is a prime, then
either p\a or the numbers a and p are relatively prime.
Any number that is divisible by both a and b is called a common
multiple of these numbers. The smallest positive common multiple of a
and b is called their least common multiple.
T heorem 1.10. I f M is a common multiple o f a and b and m is their
least common multiple, then m\M.
T heorem 1.11. The least common multiple o f two relatively prime
numbers is their product.
C orollary. A number a is divisible by the relatively prime numbers
b and c if and only i f it is divisible by their product be.
T heorem 1.12. I f c\ab and b and c are relatively prime, then c\a.
T heorem 1.13. I f the product o f severalfactors is divisible by a prime p,
then at least one o f the factors is divisible by p.
C orollary. I f p is a prime and 0 < k < p, then the number

1 2 - (p - 1)p
1-2- • ■(k - \)k \ -2-. (p - k - 1)(p - k)

is divisible by p.
 Divisibility of Numbers 15

T heorem 1.14 (The fundamental theorem of arithmetic). Any positive

integer other than 1 can be represented as a product o f prime factors.
This representation is, moreover, unique (if products that differ only in
the order of factors are not considered distinct).
The fundamental theorem of arithmetic asserts the possibility, in
principle, of decomposing any number into prime factors. The actual
construction of such a decomposition, however, is often extremely
difficult, especially if the numbers involved are very large. Often, large
numbers can be factored or proved to be prime only with the aid of
modern electronic computers. It was not until 1957, for example, that
the number 23217 — 1 was shown to be prime. This number has 969
digits and is the largest known prime at this time (1963). To prove that
it is prime required five and one-half hours of computation by a large
computer.3
Suppose that some number a is decomposed into a product of prime
factors. Collecting equal factors, we may obtain a formula of the form

a = Pia^p2“2- ■-pffi, (1.7)

where p x, p2, . . . , pr are distinct and al5 a2, . . . , aTare positive numbers.
The product on the right-hand side of formula (1.7) is called the
canonical decomposition of the number a.
T heorem 1.15. A necessary and sufficient condition for the numbers
a and b to be relatively prime is that no prime factor appearing in the
canonical decomposition o f a also appears in the canonical decomposition
o f b.
T heorem 1.16. Let (1.7) be the canonical decomposition o f the number
a. Then a\b if and only if

Pf Ab, p 2a*\b,.. .,pffi\b .

From theorems 1.15 and 1.16 it follows that divisibility by a product

of several relatively prime numbers is equivalent to simultaneous
divisibility by each of them.
Problem 1.13. Determine an upper bound for the least prime divisor
of a composite number a.
Problem 1.14. Let

a = p f i p f z - ■ pffi
3. The prime 219937 — 1 was discovered in 1972. It has 5984 digits.
16 Chapter One
be the canonical decomposition of the number a. Then a necessary and
sufficient condition for b\a is that the canonical decomposition of b
have the form

b = p / W 2' ■'PtB' ,

where 0 < & < cel5 0 < j82 < a2>■• ■>0 < & < aT.
Problem 1.15. Let us denote by r{a) the number of distinct positive
divisors of the number a (including 1 and the number a itself). Show that
for a number a with the canonical decomposition p/ipa®2• • •pT“r,
t (<2) = (<%! + l)(o:2 + 1)- • -(“r + 1) .

Problem 1.16. Find a if it is known that 3|a, 4|a, and r(«) = 14.
Problem 1.17. The canonical decomposition of the number a has the
form pfi-Pz*2, with > 0, az > 0, and t (a2) = 81. What are the possible
values of r(a3) ?
Problem 1.18. What is a if a = 2r(a)?
Problem 1.19. Are the analogues of theorems 1.11-1.14 true for even
divisibility ?
Problem 1.20. Find a method of constructing the canonical decom­
position of the least common multiple and greatest common divisor of
two numbers whose canonical decompositions are given.
2 The Divisibility of
Sums and Products

2.1
In many cases, only the remainder in a problem of division with remain­
der is of interest, and the value of the partial quotient is of no importance.
Suppose that we want to know on what day of the week 1 January
2000 will fall (provided, of course, that we keep the same calendar we
use now). By consulting a calendar, it is easy to find that January 1, 1974,
fell on a Tuesday. The twenty-six years that separate these dates contain
26 -365 + 6 days (the last term being the number of leap years during
this period), that is, 9496 days. These days amount to 1356 weeks with
four days left over. After 1356 weeks pass, it will again be Tuesday, so
that after another four days, on 1 January 2000, it will be Saturday.
It is clear that, for the solution of the problem we have just considered,
it is totally unnecessary to know just how many whole weeks pass in the
26 years, and that all that matters is the number of days left over after
these weeks.
This example is not completely useless, for historians, especially those
who study the Orient, must often consider such problems in comparing
dates given in various calendars.
It might seem at first that the simplest method of finding the re­
mainder on division of one number by another is to perform the division
with remainder directly. In practice, however, performing such a
division is often extremely tedious, especially if the dividend is not
written in the decimal system to which we are so accustomed, but
instead is given by some complicated expression such as 21000 + 31000.
In addition, the lion’s share of the work is wasted in finding the partial
quotient, for which we have no need. It is therefore necessary to look
for a method of determining the remainder directly, without calculating
the partial quotient.
17
18 Chapter Two
We shall demonstrate one such method on the problem that we have
just solved concerning the date January 1, 2000. We may argue in the
following manner: Each year that is not a leap year consists of 365 days,
which fill out 52 weeks with one day left over. A leap year consists of the
same number of weeks with two days left over. This means that the
period from January I, 1974, to January 1, 2000, consists of some
(wholly unimportant) number of complete weeks plus a number of days
that is equal to the number of years in this period, with each leap year
counted twice. This number of days is equal to 26 + 6 = 32. Taking
four complete weeks away from these days, we have four days left,
which we proceed to count off as before, beginning with Tuesday. It
can be seen that this “ replacement” of years by days is an instance of
an extremely general method, which we shall now begin to study.

2.2
D efinition 2.1. We say that the numbers a and b are congruent
modulo m i f the remainders on division o f a and b by m are equal.
We shall now establish some properties of congruent numbers.
T heorem 2.1. The numbers a and b are congruent modulo m i f and only
i f m\a — b.
T heorem 2.2. I f the numbers ax, a2, . . . , an are congruent modulo m to
bu b2, ■■., bn, respectively, then the numbers ax + a2 + ■■■+ an and
bi + b2 + ■■■+ bn, as well as axa2 ■■■an and bxb2 are also
congruent modulo m.
Corollary. I f the numbers a and b are congruent modulo m, then so
are an and bnfor any natural number n.
Theorem 2.2 and its consequences provide fairly rich possibilities for
finding remainders on division. We shall present some examples:
Example 2.1. Find the remainder on division by 3 of the number

A = 1316 - 225-515.

Clearly, the number 13 is congruent to 1, 2 congruent to —1, and 5 also

congruent to —1 modulo 3. So, on the basis of the discussion above,
the number A is congruent on division by 3 to the number

l 16 - ( - 1 ) 25( - 1 ) 15 = 1 - 1 = 0 ;

that is, the desired remainder is equal to 0, and A is divisible by 3.

 The Divisibility of Sums and Products 19
Example 2.2. Find the remainder on division of the same number A
by 37.
For this we represent A in the following form:

A = (132)8 - (25)5•(53)5 .

Since 132 = 169 is congruent to —16, 25 = 32 is congruent to —5, and

53 = 125 is congruent to 14 modulo 37, the entire number A is congruent
to

( _16)8 _ ( —5)5(14)s ,

or, equivalently, to

(162)4 + (70)5 .

But 162, that is, 256, is congruent to —3, and 70 is congruent to —4.
This means that A is congruent to

( —3)4 + ( —4)5

or, equivalently, to

81 - (25)2 ,

and, therefore, to

81 _ ( —5)2 = 81 - 25 = 56.

Finally, 56 is congruent to 19 modulo 37. Since 19 is nonnegative and

smaller than 37, it is the desired remainder.
Problem 2.1. Find the remainder on division
(a) of ^ = (116 + 1717)21 by 8;
(b) of A = 14256 by 17.
Problem 2.2. Prove that for any n:
(a) 6|n3 + lln;
(b) 9 j 4™+ 15« - 1;
(c) 37l+2|10<3") - 1;
(d) for any a.

(a2 - a + 1)|(a2n+1 + (a - l)n+2) .

20 Chapter Two
2.3
If a and b are congruent modulo m, we write
a = 6(mod m)

and say “a is congruent to b modulo m.” This formula itself is called a

congruence.
Congruence of two numbers with regard to some fixed modulus m, or
equivalently, their leaving the same remainder on division by m, is also a
relation that connects integers with other integers.
Some of the properties of the relation of congruence modulo m are:
Property 2.1. Reflexivity: a = <j(mod m). Indeed, m\a — a = 0.
Property 2.2. Symmetry: if a = 6(mod m), then b = a(mod m). Indeed,
if m\a — b, then (by theorem 1.5) m\b — a.
Property 2.3. Transitivity: If a = i(mod m) and b = c(mod m), then
a = c(mod rri).
As proof of this, it is sufficient to note that from m\a — b and m\b — c,
it follows, by theorem 1.6, that m\a — c.
If a relation (which we shall denote by ~ ) has the properties of reflexivity,
symmetry, and transitivity, then it is called a relation of equivalence, or an
equivalence relation. The simplest example of an equivalence relation on a
set of numbers is the relation of equality.
Problem 2.3. Show that an equivalence relation ~ on a set of numbers
breaks that set up into classes (called equivalence classes) which satisfy the
following two conditions: Any two members of the same equivalence class
are connected by the equivalence relation, while no two numbers of different
classes are connected by the relation.
In this problem we are concerned only with equivalence relations over
the integers. This limitation is not essential, however, and the assertion of
the problem is valid for equivalence relations connecting completely
arbitrary objects.
Since the relation of congruence modulo m is an equivalence relation,
it also breaks up, or partitions, the set of integers into classes. These classes
are called residue classes modulo m.
Property 2.4. The number of residue classes modulo m is equal to m.
Indeed, two numbers a and b belong to the same residue class modulo m
if and only if they give the same remainder on division by m. But the
remainder on division by m must be one of exactly m numbers: 0, 1, 2, . . . ,
m — 1. Consequently, the number of residue classes is m.
We note one extraordinarily interesting fact.
In order for each residue class modulo mi to be contained in some
residue class modulo m2, it is necessary and sufficient that m2|/wi.
Indeed, let us consider the residue class Kx modulo mx containing the
number 0. Clearly, the class Kx consists of all numbers that give a remainder
of 0 on division by mu that is, that are divisible by m1. In particular, it
 The Divisibility o f Sums and Products 21
contains the number m-i. Any residue class modulo m2 that contains K-i also
contains 0 and therefore consists of all numbers divisible by m2. Since m-i
is in K-i, it must be the case that m2\mi, which proves the necessity. But the
sufficiency is clear and may be easily proved.
Thus, the relation of divisibility may be defined in terms of residue
classes. This method is useful because it allows us to define divisibility for
objects much more general and complicated than the natural numbers.
In addition, the further development of this idea leads to the theory of
groups, an important branch of modern algebra that has applications to
theoretical physics and crystallography.
We shall consequently continue to enumerate the properties of the
congruence of numbers. From theorem 2.2 we immediately observe:
Property 2.5. If a = £>(mod m) and c = d(mod m), then

a + c = b + r/(mod m) .

C orollary. If a = fi(mod m), then

a + r = b + r(mod rn)

for any integer r.

Property 2.6. If a = fi(mod m ) and c = tf(mod m), then

ac = bd{mod m) .

Properties 2.5 and 2.6 show that, like equalities, congruences may be
added and multiplied termwise.
Problem 2.4. If a given equivalence relation ~ on the set of integers
partitions this set into m classes and is such that from a ~ b and c ~ d it
follows that a + c ~ b + d, then that relation is the relation of congruence
modulo m (that is, a ~ b if and only if a = 6(mod m)).
Problem 2.5. Formulate and prove a rule for replacing any congruence
by an equivalent congruence containing smaller numbers.
Problem 2.6. If the number p is prime and a is not divisible by p, then
no two of the numbers a, 2a, 3a ,.. .,(p — 1)a are congruent to each
other modulop. Therefore, on dividing the numbers a, la, 3a ,. . . , (p — 1)a
by p, we obtain each remainder except 0 exactly once.
Problem 2.7 ( Wilson's theorem). A necessary and sufficient condition for
a number p to be prime is that {p — 1) ! + 1 = 1-2 •••(/> — 1) + 1 be
divisible by p.
Problem 2.8. Formulate and prove the analogue of theorem 1.16 for the
relation of congruence.
3 Criteria for
Congruence and
Criteria for
Divisibility

3.1
An extremely general method for finding the remainder on division
of an arbitrary, but fixed, natural number a by a given natural number
m consists in the following: One constructs a sequence of natural
numbers

a — A o, A ly A 2, .. (3.1)

that are congruent modulo m. One must choose the sequence so that,
after each of its elements that is greater than or equal to m, there is at
least one more element. Then the last member of the sequence (if, of
course, this exists) will clearly be the remainder r on division of a by m.
The simplest example of such a sequence is the sequence (1.3) of
section 1.11. The problems on finding the remainder in examples 2.1
and 2.2 of the preceding chapter also reduce in essence to the construc­
tion of a sequence of this type.
We shall call any method for constructing a sequence of the form
(3.1) a criterion for congruence modulo m.
In particular, one criterion for congruence modulo m is the process
of successive subtraction of the number m until the first number smaller
than m is obtained.

3.2
It is apparent that for a criterion for congruence to be of any use to us,
it must satisfy the following three requirements:
Property 3.1. The criterion must be well defined; that is, the number
a must completely determine all the elements of the sequence (3.1).
22
 Criteria for Congruence and Criteria for Divisibility 23
Property 3.2. The criterion for congruence must be applicable to any
natural number a. This property of the criterion is called its universality.
Property 3.3. Finally, we must have some guarantee that in the
sequence (3.1) there is at least one element that is smaller than m, and
that the sequence terminates at this number. The process of constructing
the sequence should, by this requirement, not continue indefinitely, but
end sooner or later with the appearance of the remainder of the
division of a by m. This property of a criterion for congruence is called
its determinacy.
The requirements we have listed above can be fulfilled in a number
of ways. The most natural of them is the following:
We attempt to find a function f(x ) subject to the following conditions:
a. /(x ) is a natural number for x > m ;
b. /(x ) is not defined for x < m (that is, f(x ) has no meaning for
such an x) ;
(There is nothing surprising in the idea of a function losing its
meaning for some values of the argument. For example, the function
1jx(x — 1) is undefined for a: = 0 and x = 1.)
c. if f(x ) has meaning, f(x ) < x ;
d. if f(x ) has meaning, then the numbers x and f(x ) are congruent
modulo m.
Such functions have already been shown to exist. An example is the
function / 0(x):

x — m if x > m ,
undefined if x < m .

It is precisely this function that is employed in the construction of the

sequence in chapter 1.
To each function /(x ) that satisfies conditions (a) through (d), there
corresponds a unique method of constructing a sequence (3.1); that is,
a unique criterion for congruence modulo m.
Indeed, given an arbitrary natural number a, let us construct the
sequence of numbers

(3.2)

where A 0 = a, and A k +1 = f( A k) for k = 0, 1 ,....

If Ak > m, the number f( A k) is defined, and there is at least one
element after Ak. If, however, A k < m, then f( A k) is not defined, and
Ak is the last element of the sequence (3.2).
24 Chapter Three
Thus, we already have at least one criterion for congruence.

3.3
We shall now show that this criterion has the properties 3.1, 3.2, and
3.3.
The first requirement is fulfilled because each element of the sequence
(3.2) defines the one after it uniquely (if, of course, this next element
actually exists).
Here there is a certain fine point which, although not immediately
apparent, is extremely important. The point is that in defining the
sequence (3.2), we must first determine whether the number f( A k) exists
before we can calculate its value. In other words, we must know
whether the number Ak is larger or smaller than m. If the numbers A k
and m are given in a form that is convenient for comparison, say in their
decimal representation, then this decision can be made without diffi­
culty. But the comparison of the magnitudes of numbers like 220 —
3-52- 11-31-41 and 310 — 78-757 would require considerable effort
even.though the first one is 1 and the second 3.
As a result of this difficulty, we shall in the future apply criteria for
congruence exclusively to positive numbers that are written in the
decimal system.
As to the second requirement, it is sufficient to note the following:
If a > m, we may, in fact, start the construction of the sequence once
we have calculated the value of f(a ), which, by assumption, exists. If
a < m, then f(a) has no meaning. But, in this case, a is its own remainder
on division by m ; that is, the number a comprises the entire sequence
(3.2) .
We pass to the third requirement. By assumption, the function f ( x )
is such that all the elements of sequence (3.2) are nonnegative, and such
that the terms of this sequence decrease strictly. Since the first element
of the sequence is positive, the sequence must terminate. (As can easily
be verified, the index of the last element does not exceed a.) If this last
element (which we shall denote by a) were greater than or equal to m
then there would exist a value /(a), nonnegative as before, and smaller
than a. This means that a would not be the least nonnegative element
of our sequence, which is a contradiction. The process of constructing
the sequence thus must end, and the last element of the sequence is the
remainder on division of a by m.
In summary, we have established that criteria for congruence of the
type that we have described meet the three requirements of being
precisely defined, universal, and determinate. Processes having these
 Criteria for Congruence and Criteria for Divisibility 25
three properties, which are called algorithms, are beginning to play a
very important role in modern mathematics.1 Some simple examples of
algorithms were given at the end of section 1.1 of chapter 1. We shall
become acquainted with other examples below.

3.4
One of the most important algorithms in mathematics is the so-called
Euclidean algorithm, which consists of the following:
Let a and b be two natural numbers, with b < a. We divide a by b with
remainder: a = bq0 + rlt where 0 < r3 < b. If r3 ^ 0, then we may divide
b by r-i with remainder: b = r1q1 + r2, with 0 < r2 < rx. Continuing in this
manner, we obtain the equations r1 = r2q2 + r3, r2 = r3q3 + r4, and so on.
We shall show that the process we have described is actually an algorithm;
that is, that it has the properties of being precisely defined, universal, and
determinate.
We note that the process we are considering consists of successive
divisions with remainder. Therefore, the properties of being precisely
defined and universal are consequences of the existence and uniqueness of
the partial quotient and remainder.
It is also fairly simple to show that our process is determinate. The
number b and the remainders on division form a decreasing sequence of
nonnegative numbers:
b, r±, r2, . . . . (3.3)
But there are only b + 1 nonnegative numbers that are not greater than b.
Sequence (3.3), therefore, cannot have more than 6 + 1 elements, so that
the process cannot consist of more than b divisions with remainder.2 Thus,
the process we are considering is indeed an algorithm and fully deserves its
name.
Let us clarify the circumstances under which the process ends. Clearly,
the last division must be such that further division by its remainder is
impossible. But this can happen only if this last remainder is equal to zero;
that is, only if the last division can be performed evenly.
Problem 3.1. a. If Euclid’s algorithm is applied to two numbers a and b,
the last non vanishing remainder rn will equal (a, b).
b. For any natural numbers a and b, there exist integers + and B such
that aA + bB = (a, b).
Problem 3.2. From the result of part (b) of problem 3.1, derive theorems
1.9, 1.12, 1.13, and 1.14. (We emphasize that our arguments about the
Euclidean algorithm were based on the possibility of division with re­
mainder. We did not use theorems 1.9-1.14 in them, nor did we use any
other considerations based on the fundamental theorem of arithmetic.)
1. For a further discussion of algorithms, see the book by B. A. Trakhtenbrot,
Algorithms and Automatic Computing Machines (Boston: D. C. Heath, 1963).
2. In fact, the number of these divisions cannot exceed 5 log b. For a proof,
see pp. 34-35 of the author’s book, The Fibonacci Numbers, in the Topics in
Mathematics series (D. C. Heath and Co., Boston: 1963).
26 Chapter Three
3.5
Of course, the description of an algorithm given in section 3.3 is not
the precise definition of this notion, which is comparatively complicated
and cannot be formulated here.3 The requirements listed do, however,
reflect fairly completely the conditions that algorithms must satisfy.
Algorithms are important because they provide a uniform method for
solving all problems of a given type. For example, the algorithms which
we have just considered allow us to calculate the remainder on division
of any number a by some fixed number m.
Any method of computing using formulas in which numbers are
substituted for variables is also an algorithm. Speaking somewhat more
freely, we can say that all procedures which can be carried out by
machines are algorithms. It is no accident, therefore, that the develop­
ment of the theory of algorithms coincided historically with the appear­
ance and widespread use of computers; but computational problems in
the narrow sense of the word—that is, problems for which a numerical
answer may be obtained from initial data by some type of mechanical
rule—are not the only problems that reduce to algorithms. One may,
for example, attempt to find an algorithm that will provide the proof
of any true proposition in some branch of mathematics. Such an
algorithm must be able to turn the formulation of theorems into their
proofs. Fantastic as this may seem, such algorithms do in fact exist
although not for very wide areas of mathematics. At the same time,
however, the existence of such algorithms may be proved impossible in
some branches of mathematics (such as all branches containing all of
arithmetic).

3.6
Let us now find and examine several criteria for congruence using
the method presented in section 3.1. Here and in the future we shall
assume that all numbers are written in decimal notation.
First we shall find a criterion for congruence modulo 5.
Let A be a natural number. We shall represent A in the form 10a + b
(b is just the last digit in the decimal representation of the number A),
and set

3. For this definition, see the book by Trakhtenbrot referred to in footnote 1

of this chapter.
 Criteria for Congruence and Criteria for Divisibility 27
The reader may check for himself that the function / j satisfies
conditions (a) to (d) of section 3.2.
Thus, to find the remainder on division of some number by 5, it is
sufficient to consider the last digit. If this digit is smaller than 5, then
it is the desired remainder. If not, subtracting 5 from it gives the
remainder. We note that the application of this criterion for congruence
to any number reduces to the construction of a sequence of the type
(3.2) that consists of not more than three elements.
Of course, the object of all these arguments is not to discover this
well-known criterion for divisibility by 5, but to express it in the general
form that was described in section 3.2.
Problem 3.3. State and analyze analogous criteria for congruence
modulo 2, 4, 8, 10, 16, 20, and 25.
Problem 3.4. Suppose some natural number m # 0 is fixed. Let us
represent each natural number A in the form

A = 10ka + b (0 < b < 10fc) ,

and set

(b if A > 10fc ,
f(A ) = <the remainder on division of A by m if m < A < 10te ,
[undefined if A < m .

For what numbers m will such an algorithm be a criterion for

congruence modulo m for some k l

3.7
As a second example, we shall consider a criterion for congruence
modulo 3.
Let us first represent each natural number A in the form

A = \0nan -(- 10" 1an—i -I- • • • -I- lOa^ -I- Qq ,

where 0 < at < 10 (the numbers a0, au . . . , an- u an are the decimal
digits of A). Let us then define
(a0 + a1 + • • • + an. 1 + an, if A > 10 ,
f z(A) = < the remainder on division of A by 3 if 3 < A < 10 ,
undefined if A < 3 .
28 Chapter Three
Problem 3.5. Verify that the function f 2(x) satisfies conditions (a)
through (d) of section 3.2 and thus defines a criterion for congruence
modulo 3.
Problem 3.6. Apply the above criterion for congruence modulo 3.
a. to the numbers 858,733 and 789,988;
b. to the number whose decimal representation consists of 4444 fours.
Problem 3.7. State and analyze analogous criteria for congruence
modulo 1,9, 11, 13, and 37.

3.8

In many cases, it is not only the magnitude of the partial quotient

of two numbers that is nonessential. Often the magnitude of the
remainder is unimportant and all that matters is whether or not it is
zero; that is, whether the first number is divisible by the second. Using
the methods of section 3.1, it is clear how we may approach such
problems.
We shall call the numbers a and b equidivisible by m if either both a
and b are divisible by m or if neither a nor b is divisible by m.

Problem 3.8. Prove that for any number m, all numbers congruent
modulo m are equidivisible by m. Show, by example, that the converse
is false.

Problem 3.9. For what numbers m does equidivisibility of two

numbers by m imply congruence modulo m ?

Problem 3.10. Prove that the relation of equidivisibility by a given

number m is an equivalence relation, and that it partitions the set of
integers into two classes.

Problem 3.11. Does theorem 2.2 hold for equidivisible numbers?

Does its corollary?

3.9
Suppose that it is necessary to determine the divisibility of the
number A by m. We construct a sequence of integers with decreasing
absolute value.

A — A 0, A lt A 2, . . . , (3.4)
 Criteria for Congruence and Criteria for Divisibility 29
that are equidivisible with A on division by m with remainder. The
sequence (3.4) is constructed so that, after every element that is greater
than or equal to m in absolute value, there follows at least one element.
If the last element of this sequence is zero, then A is divisible by m, and
if not, then it is not divisible by m.
We shall call any method for constructing such a sequence (3.4) a
criterion for divisibility by m.
Problem 3.12. Prove that any criterion for congruence modulo m is
a criterion for divisibility by m.
Obviously, criteria for divisibility must satisfy the same requirements
of precise definition, universality, and determinacy that we require of
criteria for congruence.
It is not difficult to verify (and this is left to the reader) that, by using
any function f(x ) that satisfies conditions (a) through (c) of section 3.2
and condition (d*)—if f(x ) has meaning, then the numbers x and f(x )
are equidivisible on division by m—it is possible to construct a criterion
for divisibility by m in the same manner that we constructed a criterion
for congruence modulo m using any function satisfying conditions (a)
through (d).
Let us construct some criteria for divisibility. Because of theorem
1.16, it is sufficient to construct only criteria for divisibility by numbers
of the form pa (that is, by powers of primes).

3.10
A criterion for divisibility by 7. Let A be a natural number. We
represent A in the form 10a + b, where 0 < b < 10. Let us set

(a — 2b if A > 190 ,
f 3(A) = < the remainder on division by 7 for 7 < A < 190 ,
[undefined for A < 7 .

Problem 3.13. Show that the function f f A ) satisfies conditions (a)

through (c) and (d*).
The function f 3(A) gives us a well-known criterion for divisibility by
7: The number 10a + b (0 < b < 10) is divisible by 7 if and only if the
number a — 2b is divisible by 7. The number so obtained is checked
for divisibility by 7 by the same method, and so on.
Problem 3.14. Prove that the criterion for divisibility by 7 that we
have just constructed is not a criterion for congruence modulo 7.
30 Chapter Three
3.11
A criterion for divisibility by 13. Let us represent each natural number
A in the form 10a + b and set

fa + 4b if A > 40,
fi(A ) = < the remainder on division of ^4 by 13 if 13 < A < 40 ,
[not defined if A < 13 .

Problem 3.15. Verify that the function fffx ) satisfies conditions (a)
to (c) and (d*), and formulate the criterion for divisibility by 13 so
obtained.
Problem 3.16. What would be the consequences of changing the
number 40 to a smaller one in the definition of the function / 4?
Problem 3.17. Construct criteria for divisibility by 17, 19, 23, 29, and
31 by analogy with the criteria for 7 and 13.
Problem 3.18. Construct two criteria for divisibility by 49.

3.12
In the preceding sections of this chapter, we became acquainted with
a large number of widely varying criteria for congruence and criteria
for divisibility. The goal in constructing all of these criteria is to obtain
practical algorithms for finding the remainders on division by certain
numbers (criteria for congruence) or for deciding whether these
remainders are equal to zero (criteria for divisibility). To what extent
have we carried out our proposed task ?
Some criteria for congruence, such as those for division by 2, 3, 4,
and 10, actually turned out to be extremely practical and useful. Others
involved clumsy calculations and were therefore less practical.
It is natural, therefore, to seek and use the most convenient and
efficient possible criteria for divisibility and congruence.
One of the difficulties we encounter in such attempts is that we must
know how simple (or how complicated) it would be to use a given
criterion in the case of some particular number. We may, for example,
take as a measure of efficiency the number of arithmetic operations on
single digit numbers that must be performed in using a given criterion
on one number or another.
Unfortunately, any such numerical characterization of the amount of
calculation to be done would depend very much on the individual
properties of the number whose divisibility we want to test.
 Criteria for Congruence and Criteria for Divisibility 31
Thus, it is very easy to decide that the remainder on division of
31,025 by 8 is 1. To do this it is sufficient to find the remainder on
division of 25 by 8. But to find the remainder on division of 30,525 by 8,
it is necessary to divide 525 by 8 with remainder, and this requires a
larger number of calculations (regardless of whether they are done
mentally or on paper).
As another example, we may consider the criterion for divisibility by
37 (see problem 3.7). The remainder on dividing 10,014,023 by 37 can
be found by the addition 10 + 14 + 23 and the subsequent division by
37 of the sum obtained. As can easily be seen, this is equal to 10.
However, there are very few people who could apply this criterion
mentally to the number 782,639,485.
In speaking about the convenience of using various criteria for
divisibility and congruence, we must therefore look beyond the diffi­
culties of applying these tests to particular numbers and evaluate the
effectiveness of each criterion “ on the average.” Only with such an
approach can we hope to formulate precisely the degree of intricacy of a
criterion for divisibility or congruence, or to find the criterion that is
most economical in the majority of cases. Unfortunately, these ques­
tions are very complicated and we will not be able to develop them
here.

3.13
All of the criteria for congruence and divisibility that we have
constructed above appear to be a bit artificial, and at first glance it may
seem that at least some of these criteria were discovered by accident or as
the result of trial and error. In fact, this is not so. Indeed, it may be shown
that there exist methods of constructing criteriafor divisibility and congru­
ence for any previously given number. These methods are called general
criteriafor divisibility and general criteriafor congruence, respectively.
General criteria for divisibility are thus methods for obtaining actual
criteria for divisibility. Actual criteria, therefore, may be considered to
be the results obtained by applying general criteria to particular num­
bers. From this standpoint, the general criteria for divisibility are related
to the actual ones in the same way that an actual criterion is related to
the result of its application to a specific number, that is, to the
remainder on division of a given number a by a given number m.
General criteria for divisibility and for congruence resemble algo­
rithms ; in fact, they seem to be special algorithms whose end products
are themselves algorithms—namely, actual criteria for divisibility or
congruence.
32 Chapter Three
But before we may talk about general criteria for divisibility and
congruence as algorithms, we must convince ourselves that they have
the necessary properties of precise definition, universality, and
determinacy.
Speaking in even more detail, we must verify that the following condi­
tions are satisfied by any general criterion for divisibility or for con­
gruence. First, given any number m, the general criterion must actually
yield a criterion for divisibility (or for congruence) for m. It must, so to
speak, “ convert” each natural number m into the corresponding
criterion. This is merely a precise statement of the property of deter­
minacy. Second, a general criterion must be precisely defined, that is,
when applied to any given number m it must, in a definite manner, lead
to a well-defined criterion for divisibility (or for congruence) for this
number. Third and last, a general criterion should be universal, that is,
truly general, and should give a criterion for divisibility or congruence
for any natural number given in advance.
In this sense, neither the method described in section 3.2 for obtaining
a criterion for congruence nor the method described in section 3.6 of
finding criteria for divisibility are general criteria. In fact, our procedure
for finding functions to satisfy the necessary conditions satisfies none of
the conditions of precise definition, universality, or determinacy.
Indeed, these methods give no guarantee that the necessary function
will be found; hence, they are not determinate. Furthermore, if the
necessary function does exist, it can be found in various ways, to say
nothing of the fact that there may be several functions satisfying our
conditions. This means that the result of our methods is not precisely
defined. Finally, our techniques are not universal enough, since it is
quite possible that we shall be unable to find the required functions for
some numbers. In any event, the method itself gives us no specific
guarantee of success. Thus, for the described process to become an
algorithm, it will have to be supplemented with some precise instruc­
tions that will guarantee the construction of a completely determined
function f m for each number m.
This problem of the “ algorithmization” of the construction of
criteria for divisibility can, in fact, be solved without particular diffi­
culty, and general criteria for divisibility have been known for several
centuries.
We have actually constructed one such general criterion for con­
gruence already in our treatment in Chapter 1, Section 11, about
division with remainder. We may formulate it as follows: To each
positive integer m there corresponds the process of successive subtrac­
tion of m from any other number k until a number is obtained that is
 Criteria for Congruence and Criteria for Divisibility 33
smaller than m (see the last sentence of section 3.1). It is clear that such
a correspondence has the necessary properties of precise definition (we
know exactly what process corresponds to the number m), universality
(the process of successive subtraction can be carried out with any m),
and determinacy (such an attempt inevitably leads to success). But the
practical value of the general criterion for congruence just described is
very small.
A certain improvement of the general criterion for congruence based
on successive subtractions leads to the familiar process of long division,
which is itself a general criterion for congruence. It is worth noting
that most people use precisely this criterion for finding remainders on
division. The reasoning may go according to the scheme outlined in
Table 3.1, in which we give two versions—one in the usual, everyday
language, and one in the language of algorithms.

Table 3.1
In everyday language In the language of algorithms

1. I must find the remainder on The general criterion for congruence

division of a given a by a begins to process the number m;
given m;
2. to do this I must divide by m; the general criterion “ issues” the result
of its processing of the number m:
an actual criterion for congruence on
division by m, which consists in
dividing by m directly;
3. now I begin to divide a by m ... the actual criterion obtained begins to
process the number a;
4. I divide and obtain the remainder. the actual criterion reaches its goal: the
remainder on division of a by m.

Since the first three steps of this reasoning are very simple, we should
not be surprised that the fourth step—the( actual performance of
division—turns out to be so unwieldy. The goal in creating general
methods for congruence and divisibility consists precisely in lightening
the load of the fourth step at the expense of increasing the difficulty of
the second. It is exactly this that we have in mind when we speak of
general criteria for divisibility and congruence.

3.14
Historically, the first useful general criterion for divisibility (in fact,
it is even a general criterion for congruence) is the following method,
proposed early in the seventeenth century by the famous French
34 Chapter Three
mathematician and philosopher Blaise Pascal. Its essence is as follows:
Let m b e a natural number. We form the sequence of numbers

U, r2, rS). . . , (3.5)

by setting

r1 equal to the remainder on dividing 10 by m ,

r2 equal to the remainder on dividing 10^ by m ,
r3 equal to the remainder on dividing 10r2 by m ,

and so on.
Let us now represent an arbitrary number A in the form

10non + 10" + • • • + lOflj + <j0 j

(where 0 < at < 10 for all i), and define the function

a0 + rxa3 + rza2 H-----+ rnan if 10" > m ,

{the remainder on division of A by m if 10" < m < A ,

undefined if A < m .

Problem 3.19. Verify that for any m the function Fm satisfies con­
ditions (a) through (d) of section 3.2.
Thus, we have developed a means for constructing a criterion for
congruence modulo any m; that is, a general criterion for congruence.
Problem 3.20. Formulate the actual criteria for congruence
a. modulo 2, 5, and 10;
b. modulo 4, 20, and 25;
c. modulo 3 and 9;
d. modulo 11;
e. modulo 7;
which are obtained from Pascal’s general criterion for congruence.
Problem 3.21. Suppose that in the sequence (3.5), is the remainder
on division of 100 by m, rz is the remainder on division of 100/-! by m,
r3 is the remainder on division of 100r2 by m, and so on. Use these
values of the rt to derive a general criterion for congruence that is
analogous to that of Pascal.

3.15
In section 3.12 we spoke of the comparative advantages of different
criteria for divisibility (or congruence) for a given number. Since a
 Criteria for Congruence and Criteria for Divisibility 35
general criterion for divisibility must give us criteria for divisibility by
any natural number, it is not surprising that the criteria it gives us for
different numbers may differ greatly in efficiency.
For example, Pascal’s general criterion for divisibility gives wholly
acceptable criteria for congruence modulo 3 and 11, together with a
very clumsy and inconvenient criterion for congruence modulo 7 (see
problem 3.20, part e).
In connection with this, we can make observations about general
criteria for divisibility and congruence that are similar to those made
in our discussion in section 3.12 on the practicality of actual criteria for
divisibility. In this sense the best possible general criterion for divisibility
(or congruence) should be the one which, on application to any pre­
viously given positive integer m, gives the best possible criterion for
division (congruence) for this m. The problem of finding the best
general criterion for divisibility is not only far from being solved, but
is also far from being precisely formulated.
4 Divisibility of
Powers

4.1
The question of the divisibility of powers is, in reality, a question
about the divisibility of a certain type of product, namely, the product
of several equal factors. It may be answered, therefore, on the basis of
the results of chapter 2. In the case of large exponents, however,
decreasing the base of a power may not lead immediately to the
remainder on division of the power by some given number, and we must
use some artificial means to obtain this remainder (see the examples in
section 2.2). Moreover, in constructing general criteria for divisibility,
we required calculation of the remainder on division of successive
powers of 10k by m. Although this process in itself is not complicated,
it still does not indicate any regularities in sequence (3.5), nor does it
make it possible to choose the number k so that all of these remainders
are small. At the same time, such a possibility exists; in fact, the number
k may be chosen so that all of these remainders are 1.
These considerations make it desirable to study the divisibility of
powers in more detail.

4.2
Let us begin by establishing several number-theoretic results.
T heorem 4.1 (Fermat’s little theorem). I f the number p is a prime,
then for any positive integer a the difference cP — a is divisible by p.
One should not confuse “ Fermat’s little theorem” with “ Fermat’s
last theorem.” The latter asserts that for any integer n > 2 there do not
exist integers a, b, c satisfying an + bn = cn. Despite numerous attempts,
Fermat’s last theorem has yet to be either proved or refuted.
36
 Divisibility of Powers 37
C orollary. I f p is a prime and a is not divisible by p, then av~x — 1
is divisible by p.
Problem 4.1. Give an example showing that both theorem 4.1 and its
corollary are, in general, not true if p is a composite number.
Problem 4.2. Prove Fermat’s little theorem using the result of
problem 2.6.
Suppose that a natural number m > 1 has the canonical decom­
position
m = p f 'p f z - ■- p f* . (4.1)
We set

<p(m) = p f i \ p x - \)Pz“2 \ P i - !)• • •Pk k KPk ~ 1) (4.2)

and define <p(l) = 1.

Formulas (4.1) and (4.2) assign to each positive integer m a unique
number <p(m). This means that we may speak of 95 as a function on the
positive integers.
D efinition 4.1. The function <p above is called Euler’s function.
Euler’s function plays an exceptionally important role in many
number-theoretic questions. Several of these applications will be
discussed later in this book.
T heorem 4.2. I f mx and m2 are relatively prime, then

<p{mxm2) = <p{mx)<f{mf).

Problem 43. Calculate <p(12), y>(120), <p(1000).

Problem 4.4. Determine all numbers m for which
a. <p(m) — 10;
b. <p(m) = 8.
Problem 4.5. Prove that there exists no number m for which <p(m) =
14.
Problem 4.6. Show that <p(m) is equal to the number of natural
numbers relatively prime to m which are less than or equal to m. This
property of Euler’s function is extraordinarily important. It is often
taken as the definition of this function.
T heorem 4.3 (Euler’s theorem). I f the numbers a and m are relatively
prime, then a’p{m) — 1 is divisible by m.
38 Chapter Four

4.3
We shall now construct several general criteria for divisibility and for
congruence, using the theorems stated above.
Let us fix the natural number m and represent the natural number A
in the form

A = a0 + + a2102®(m) + • • • + ak ,

where

0 < at < 10®(m) for all i from 1 to k ;

that is, the numbers at (i = 0, 1,.. k) have at most <p(m) digits.

The function F defined by

fa0 + ax + • • • + ak if A > 10®<m);

F{A) = < the remainder on division of A by m if m < A < 10®(m);
[undefined if A < m ,

may easily be shown to determine a general criterion for congruence

modulo any m.
Problem 4.7. Verify this.
T heorem 4.4. I f the numbers a and m are relatively prime and the
numbers k x and k z are congruent modulo then the numbers aki and
ak2 are congruent modulo m.
Problem 4.8. Prove that 2730|n13 — n for any n.

4.4

This general criterion for congruence is often not “ sufficiently

economical,” since the number <p(m) frequently turns out to be fairly
large. In applying this criterion, it is often necessary to add large
numbers and to divide cp(m)-digit numbers directly by m (or to make use
of some other criterion for divisibility or congruence). It would thus
be desirable to replace <p{m) by a smaller exponent, and this can in fact
be done in many cases. For example, for m = 37, instead of <p{m) = 36
we may take the exponent 3, since 103 = 1000 leaves a remainder of 1
on division by 37, as must 1036 by Euler’s theorem; for m = 11, instead
of <p(m) = 10 we may take the exponent 2, and so on.
 Divisibility o f Powers 39
D efinition 4.2. The least number S for which a6 leaves a remainder
o f 1 on division by m is called the order o f a on division by m with remainder.

This number is more often called the order o f a modulo m.

Euler’s theorem makes it clear that for any relatively prime numbers
a and m, the order of a modulo m does not exceed <p{m). This order
may be taken in place of <p(m) in the formulation of the general criterion
given in section 4.3.

Problem 4.9. Modify the general criterion for congruence constructed

above by replacing <p(m) with the order of 10 modulo m, and verify
that the modified criterion is itself a general criterion for congruence.

4.5

Applications of Euler’s function and Euler’s theorem are not limited

to criteria for divisibility. Another important use lies in the solution of
equations in integers, often called Diophantine equations.

T heorem 4.5. I f the numbers a and b are relatively prime, then the
equation

ax + by = c (4.3)

is always solvable in integers, and all its solutions in integers are pairs
o f the form (xt, yt), where

xt = c a + bt
1 - a®(W
y t = c ---- u--------a t .

for any integer t.

Problem 4.10. Prove a theorem analogous to theorem 4.5 without

the hypothesis that a and b are relatively prime.

Problem 4.11. Find a method for solving equation (4.3) in integers

on the basis of problem 3.1.

Problem 4.12. Solve the following equations in integers:

a. 5x + l y = 9;
b. 25x + I3y = 8.
40 Chapter Four
4.6
T heorem 4.6. Let m be relatively prime to 10, and let k be congruent
to lO®010” 1 modulo m. Then the numbers 10a + b and a + kb are
equidivisible by m.
We can use this theorem to construct the following general criterion
for divisibility: Let k' denote the remainder on division of 10®’(m)" 1 by m.
We represent an arbitrary number A in the form 10a + b (0 < b < 10)
and set

fa + k'b, for A > a + k 'b ;

F(A) = <the remainder on division of A by m, for m < A < a + k 'b ;
[undefined for A < m .

If k' is large (close to m), we may conveniently replace it by the

negative number k' — m in the formulation of the corresponding
criterion.
Problem 4.13. Verify that conditions (a) through (c) and (d*) are
satisfied for the function F.
Problem 4.14. Use the general criterion just constructed to derive
criteria for divisibility by the numbers 17, 19, 27, 31, and 49.
Problem 4.15. Construct an analogous general criterion for divisi­
bility, using the fact that one may represent any arbitrary natural
number in the form 100a + b (0 < b < 100), and from it derive criteria
for divisibility by 17, 43, 49, 67, 101, and 109.
5 Proofs of
Theorems

T heorem 1.1. It is sufficient to note that a = a -1.

T heorem 1.2. By hypothesis, there exist dx and d2 such that b = adx
and c = bd2. But then c = adxd2, and a\c.
T heorem 1.3. We have a = bc1 and b = ac2, from which it follows
that a = ac!C2, that is, that c1c2 = 1. Since the numbers c1 and c2 are
integers, either cy = c2 = 1 or cx = c2 = —1. In the first case a = b,
and in the second a = —b.
T heorem 1.4. Let a = be. If |c| > 1, then |6| > \a\ implies \bc\ >
|a|, or \a\ > \a\, a contradiction. Hence, |c| < 1, but since c must be
an integer, c = 0, so that a = be is also 0.
T heorem 1.5. Clearly, from a = be it follows that \a\ = \b\\c\, and,
conversely, if \a\ = \b\ |c|, then a = b(±c). Since c is an integer if and
only if |c| is an integer, b\a if and only if |6| | |a|.
T heorem 1.6. We may write

ai — b c i,
a2 = bc2 ,

e^r. bCn 5
where all of the numbers c1; c2, . . . , cn are integers. Adding these
equations term by term, we obtain
fli + a2 + • • • + an = b(c± + cz + ■■■+ cn) .
Since the expression inside the parentheses is an integer, the proof is
complete.
T heorem 1.8. The proof is by contradiction. Suppose that there are
only a finite number of prime numbers, so that they may all be listed:
Ply P2y ***5Pn - (5.1)
41
42 Chapter Five
Let us denote the product of these numbers by P and consider the sum
P + 1. This sum is greater than every prime number given in the listing
(5.1) and therefore cannot be one of these prime numbers. Consequently,
it is divisible by at least one prime pk. But P is also divisible by p k.
Thus, on the basis of the corollary to theorem 1.6, we must have pk|1,
from which it follows that pk = ± 1—contradicting the definition of a
prime number (see definition 1.5 in section 1.12). This proof of the
existence of infinitely many primes was given by Euclid in the fourth
century B.c.
T heorem 1.9. Suppose the numbers a and p are not relatively prime.
Then both are divisible by some positive integer other than 1. Since p
is a prime, the only such number can be p. Thus, in this case, p\a.
So either a and p are relatively prime, or p\a.
T heorem 1.10. Dividing M by m with remainder, we obtain

M = mq + r ,

where 0 < r < m. Since M and m are both divisible by both a and b,
we know by the corollary to theorem 1.6 that the number r must also
be divisible by a and by b, and thus must be a common multiple of these
two numbers. But r < m, and m is the least positive common multiple
of a and b. This means that r cannot be a positive number, so that
r = 0. Therefore, m\M.
T heorem 1.11. Suppose the numbers a and b are relatively prime,
and let m be their least common multiple. Since a\ab and b\ab, we may
conclude by the preceding theorem that m\ab. Let ab = m k, and set
m = ac. Then ab = ack; that is, b = ck, so that k\b. In exactly the
same way, we see that k\a. Since the numbers a and b are relatively
prime, we must have k = 1, which means that m = ab.
T heorem 1.12. Let us denote the least common multiple of the
numbers b and c by m. By the preceding theorem, m = be. Furthermore,
c\ab by hypothesis, and it is clear that b\ab. It follows from theorem 1.10
that bc\ab; that is, that for some number k, ab = bek, which implies at
once that a = ck. But this implies that c\a, the desired result.
T heorem 1.13. The proof is by induction on the number of factors.
If there is only one factor, the theorem is trivial. Let us suppose that
the theorem is true for any product of n factors. Suppose p\aLa2- ■■
anan+1. Then, denoting axa2- ■ an by A, we have p\Aan+1. If p\an+1,
then the theorem is proved. If not, then by theorem 1.9, an+1 and p are
relatively prime. But then, by the preceding theorem, p\A. Since A is a
product of only n factors, the inductive hypothesis implies that one of
these factors must be divisible by p, which proves the theorem.
 Proofs o f Theorems 43
Corollary. The entire fraction is an integer (that is, the numerator
is divisible by the denominator) since, by the binomial theorem, it is
one of the coefficients of the expansion of (x + l)p. We shall therefore
consider the numerator as a product of two factors: p and 1 •2 ■• •
(P - 1) = (P - 1)!
None of the factors in the denominator is divisible by p. Hence, by
the preceding theorem, the entire denominator is not divisible by p,
and is therefore, by theorem 1.9, relatively prime to p. Thus the second
factor of the numerator, (p — 1)!, must be divisible by the denominator.
Denoting the quotient of this division by q, we have = pq, and the
proof is complete.
T heorem 1.14. First we shall show the possibility of decomposing
any number other than 1 into prime factors. Let us suppose that all
numbers smaller than N can be so decomposed. If the number N is
prime, then it may automatically be decomposed into a product of
primes (namely, a product consisting of only one factor—the number N
itself), and the theorem is proved. Suppose now that N is composite
and that N± is some divisor of N that is distinct from both N and 1.
Let N 2 be the quotient on division of N by Nj,. Then N = and
it can easily be verified that 1 < N2 < N. Since Nj and N 2 are both
smaller than N, they are both decomposable into products of prime
factors (by the induction hypothesis). Let = p 1p2- ■ Pk and N 2 =
q^q2*■ qt be these decompositions. Then P\P2- ■TV7i<72• • <7; is the
desired decomposition of the number N, and the possibility of decom­
position is proved.
We now prove the uniqueness of this decomposition. Suppose we are
given two decompositions of the number N into prime factors: N =
PiP2~■ Pk and N = q1q2- ■ qi (where not all of the p t nor all of the q}
are necessarily distinct). Clearly,

PiP2---Pk = qiq2---qi- (5.2)

Since qxq2■■ qt is divisible by p x, we have by the preceding theorem that

at least one of the numbers qx, q2, .. ,,q t is divisible by p2. Suppose that
Pi\qy (we may make this assumption since we have the right to change
the order of the factors so that q1 is the factor which is divisible by px).
Since the number qL is prime, this is possible only if p 1 = qx. Dividing
both sides of equality (5.2) by p u we obtain

P 2P3 ' • 'Pk — 92^3 ' ■<?;• (5.3)

44 Chapter Five
Similarly, we can see that one of the numbers q2, q3, . . . , q, (say q2) is
divisible by p 2, and thus that p 2 = q2. Dividing both sides of equality
(5.3) by p2, we reduce by one more the number of factors on each side.
This process of reduction can clearly be carried out until one of the
products has been completely cancelled. Suppose that the left side of
(5.2) is the first to be cancelled completely. The product on the right
side of (5.2) must then also be cancelled completely, for otherwise we
would obtain an equation of the form

1 — QlC+l ' ' <7l 5

which is impossible since 1 is not divisible by any prime. Therefore, we

have shown that k = /, and that if the qt are suitably reordered,

Pi — Ql, Pi — <fa, ■■■>Pk — <lk ,

so that the two decompositions are identical up to the order of the

factors. The theorem is thus proved in full.
T heorem 1.15. Suppose that Piaip2 z- ■■Pk“k and g / i g / a - . -qfi are
the canonical decompositions of a and b respectively, and that d is some
common divisor of a and b. If d # 1, then d is divisible by some prime
p. Thus, by theorem 1.2, p\a and p\b so that p is among the numbers
Pi, P^, ■■■,Pk as weH as among the numbers qu q2, .. ,,q t by theorem
1.13. Therefore, among the prime factors appearing in the canonical
decomposition of a, there is at least one that appears in the canonical
decomposition of b.
Conversely, if a and b are relatively prime and p appears in the
canonical decomposition for a, then b is not divisible by p, so that p
cannot enter into the canonical decomposition for b.
T heorem 1.16. Suppose a\b. Sincep f^ a (i = 1, 2, . . . , k), we obtain
the desired conclusion from a\b by a simple reference to theorem 1.2.
The converse is proved by induction. We assume Piai\b, all z, and we
suppose that we have already proved p f i- ■-pia‘\b (1 < 1 < k). Then,
since pV+Y^ and since (p V+Y> Pi*1' ' P *0 = we may apply the pre­
ceding theorem and the corollary to theorem 1.11 to obtain p ±ai • • •
p ? r f\Y \b . This completes the inductive step and the proof.

T heorem 2.1. Suppose a and b are congruent modulo m. We write

a = mqi + rx (0 < rx < n i), (5.4)

b = mq2 + r2 (0 < r2 < m ) . (5.5)

 Proofs of Theorems 45
Because of the congruence of a and b, we have r1 = r2. This means that
a - b = m(q1 - q2) ;

that is, m\a — b.

Conversely, suppose that m\a — b. Dividing a and b by m with
remainder, we obtain (5.4) and (5.5). Here

a - b = m(^i - q2) + r1 - r2;

that is,

(a - b) - miq-L - q2) = r1 - r2 .

By theorem 2.6, m\r1 — r2. Since \r± — r2\ < m, however, it follows
from theorem 1.4 that r± — r2 = 0, or rx = r2, the required result.
T heorem 2.2. By hypothesis and theorem 2.1 we may write

«i = & i+ mqx ,'

a2 = b2 + mq2 , (5 6)

an = bn + mqn

Adding these equations term by term, we obtain

(<2j + a2 + • • • + an) —(Z>! + b2 + ■■■+ bn) = mfq^ + q2 + • • ■+ qf) ,
which, by virtue of theorem 2.1, shows the congruence of the two sums.
To prove the congruence of the two products, we make use of the
following identity:

(dx + s1m)(d2 + s2m) = dxd2 + (dxs2 + d2Sx + s^n^m .

Arguing inductively, we can easily see that this implies

(d-x + s1m){d2 + s2m)- ■-{dk + skm) = dxd2- ■-dk + tm ,

where t is some integer.

If we now multiply all the equalities (5.6) term by term, and apply
the arguments just given to the right-hand side, we obtain

ciiP-2**■ = bi&2 ***bn + m t ,

where t is some integer. The congruence of the products modulo m is

thus proved.
46 Chapter Five

Theorem 4.1. The proof is by induction on a. For a = 1,

ap — a = 1 — 1 = 0 ,

and p|0.
Let us suppose that a? — a is divisible by p. We shall show that
(a + 1)p - (a + 1) is also divisible by p. Indeed, expanding (a + l)p
by the binomial theorem, we have

(a + l)p - (a + 1) = ap +

+ [p - , ) » + 1 - « - 1

= ap - a + + ( y ^ " 2 + '''

+ {/-,)“■ <5 7 >

ap — a is divisible by p by hypothesis. By the corollary to theorem 1.13,

is also divisible by p for 1 < k < p — 1. Hence, each term of the
I,
sum (5.7) is divisible by p and therefore (by theorem 1.6), so is the
entire sum.
This justifies the inductive transition, and thus the entire theorem is
proved.
C orollary. By Fermat’s theorem,

p \a p — a = a(ap 1 — 1).

If a is not divisible by p here, then by theorem 1.13, p must divide ap~1 — 1.

Theorem 4.2. Let m 1 = p ^ - ■■p ka* and let m 2 = q / '- ■ qiBl- By
theorem 1.15, each of the numbers P i , . . . , p k is distinct from each of
the numbers qu . . . , qt. This means that the canonical decomposition
of m 1m 2 is p ^ i- ■-pjA g/1• ■-qfr. Therefore,

<p("*im2) = pV- K p i - l)- • -P kk K pk - I )??1 Kqi - !)• • q?1 \ q t - i)

= 9(.m1)<p(m2) .
 Proofs of Theorems 47
T heorem 4.3. We shall first prove by induction on a that for any
prime p, a<-pa~1{p~1)) — 1 is divisible by pa. For a = 1 the assertion is
simply a corollary of Fermat’s little theorem, the validity of which has
already been established. Thus, the basis for induction has been
proved.
Let us suppose now that — 1, and let us consider the
expression apHp~1) — 1. We must prove that it is divisible by p a+1.
However,
aP“(P- 1) _ J = (QPa- VP- l))p _ 1

Since aPa- 1<p-D — 1 is by hypothesis divisible by pa, the number

/ ■ 1(p'i ) has the form Npa + 1. This means that

avHv-V) _ i = + l)p - 1 ;

that is, by the binomial formula,

/ " ’- 11 - 1 = N pp ap + ^ N p~fpaip~v + . . . + ^ ^ ^ N p a + 1 - 1.

In the last sum, the first term is divisible by p a+1, since it is divisible by
pap, and ap > a + 1. In each of the followingp — 1 terms, the exponent
of p is at least a; moreover, by the corollary to theorem 1.13, the
binomial coefficient is divisible by p. This means that each of these
terms is divisible by p a+1. Finally, the difference 1 — 1 = 0 may be
discarded. By theorem 1.6, therefore, p a+x\apHp~Vl — 1, completing the
inductive step. What we now have is Euler’s theorem in the case that
m = pa for some a.
Let us now suppose that Euler’s theorem has been proved for the
numbers and m2, with m1 and m2 relatively prime. We shall prove
Euler’s theorem for the number m = m1m2. Once we have done this,
we may set mx = p f i ■■-pka^ and m2 = p'kW1 to obtain the inductive
transition needed to establish the theorem in general. Let us therefore
proceed with this proof.
Since a is relatively prime to m = m1m2 by the hypothesis of the
theorem, a must also be relatively prime to mu as is av i m Therefore,
by hypothesis,
(^<j® ( m 2))«>(m 1) _ J _ <j«>(m 1)<i>(m2 ) __ J _ a <p(m 1m 2 ) _ J _ a <p(m) _ |

is divisible by In exactly the same way we see that a®(m) — 1 is

divisible by m2. And since the numbers m± and m2 are relatively prime,
48 Chapter Five
_ j js divisibie by their product, that is, by m. Euler’s theorem is
thus proved.
T heorem 4.4. Let

*1 = + r,
k 2 = <p(m)q2 + r (0 < r < <p(m)) .

Then

aki = a0™9i +T = (a®(m))9l«r -

On the basis of Euler’s theorem and theorem 2.2, a0(m)9w Tis congruent
modulo m to aT. Similarly, aki is congruent modulo m to a1, which
means that the numbers aki and ak2 are congruent modulo m.
T heorem 4.5. Let us first find one solution (x', y') of this equation.
To do this, it is clearly sufficient to find a number x' such that b\ax' — c.
By Euler’s theorem, b\a'Mt>) — 1. This means that b\ca,pib> — c, and we
may take the number ca"'m ~1 for x '.
Now suppose that (x", y") is some other solution of the equation
ax + by = c. We shall show that the numbers x' and x" are congruent
modulo b. Indeed, let

ax' + by' = c ;
ax" + by" = c .

Subtracting the second equation from the first, term by term, we obtain

a(x' - x") + b(y' - y") = 0 ,

from which b\a{x' — x"). Since a and b are relatively prime by hypothe­
sis, b\x' — x" by theorem 1.12; thus, by theorem 2.1, x' and x" are
congruent modulo b.
Thus, all the desired values of x are to be found among the numbers

x t = ca0™-1 + b t .

But if xt = ca,Hb) ~1 + bt, it is a simple matter to check that y t =

c[(l — avm)lb] — at is the unique number satisfying axt + byt = c;
that is, the pairs {xt, y t) are exactly the solutions of our equation.
 Proofs of Theorems 49
T heorem 4.6. Since 10 and m are relatively prime, the numbers
10a + b and (10a + b)I0®(m)-1 are—by theorem 1.15—equidivisible by
m. But

(10a + b) 10<p(m)- 1 = 10®(m)a + lO ® '^"^,

so that 10a + b and a + kb are equidivisible by m, since by Euler’s

theorem, 10®<m) and 1 are congruent modulo m and by hypothesis,
j0«>(m)-i and k are congruent modulo m.
6 Solutions to
Problems

Problem 1.1. 0 = a-0 for any a.

Problem 1.2. a = 1 a for any a.
Problem 1.3. Suppose that a\\. This means that 1 = ac for some
integer c. From this it follows that \a\ < 1. And since a # 0, we must
have a = ± 1.
Problem 1.4. It is sufficient to take any c > 1 and set b = ac.
Problem 1.5. For b we may take, for example, 2a. Suppose that for
some c, c\2a and a\c. This means that there exist dx and d2 such that
2a = dYc and c = d2a. From this it follows that 2a = dxd2a, or,
cancelling a, that

2 = did2 ■

But such an equality is possible only if one of the numbers and d2

is equal to 1, and the other is equal to 2. If d1 = 1, then c = 2a = b\
if d2 = 1, then c = a. The same proof clearly holds if 2 is replaced by
any prime p.

Problem 1.6. The proofs differ in no substantial way from the proofs for
ordinary divisibility.
Problem 1.7. Let n be some fixed number greater than 1. We shall say
that (b\a)n if there exists an integer c such that a = be and c < n. The
correctness of the theorems analogous to theorems 1.1, 1.3, and 1.4 may be
verified without difficulty. If we take b = na and c = nb, however, then
(a\b)n and (fi|c)n, but c = n2a, and, since n2 > n, the relation (a\c)n does
not hold. In exactly the same way, the divisibility (b\a + a)n does not hold.
Problem 1.8. a. Suppose that there are two minimal elements, a1 and a2.
By the law of dichotomy, either a1 > a2 or a2 > ax. If ax > a2, then, by
50
 Solutions to Problems 51
the minimality of au it follows that ax = a2. If, however, a2 > <?i, then
Oi = a2 follows from the minimality of a2.
b. Let a be some number, and let bi and b2 be two numbers immediately
preceding it. By the law of dichotomy, either b± > b2 or b2 > b1. We may
suppose without loss of generality that bi > b2. We then have a > bx > b2,
and since the number b2 immediately precedes the number a, either b-i = a
or &i = b2. But, by hypothesis, 61 ^ a, so that bx = b2, proving the desired
uniqueness.
c. An immediate successor of a number a is a number b such that b > a,
b ay and from b > c > a it follows that either c = b or c = a.
Let us suppose that some number a has no immediate successor. This
means that, for any an > a, A a, there is an an+1 distinct from an and
from a so that an > an+1 > a. Let us now take an arbitrary ax > a and
distinct from a (by property 1.2 such a number exists), and, beginning
with it, construct an infinite sequence of distinct numbers
fli > a2 > • • • > an > an+1 > • • • > a .
The existence of this sequence contradicts property 1.4. Consequently, an
immediate successor exists. Its uniqueness is established by using the law
of dichotomy exactly as in part (b).
Problem 1.9. Transitivity (property 1.3), unboundedness of the set of
numbers (property 1.5), well-ordering (property 1.4), and the existence of an
immediate predecessor (property 1.6) remain valid. The law of dichotomy
(property 1.7) may be replaced by the law of trichotomy (either a > b, or
a < b, or a = b).
The property of reflexivity (1.1) does not hold, for a > a is always false.
Finally, property 1.2 remains formally valid, for, strictly speaking, this
assertion reads as follows: “ For any two natural numbers a and b, from
a > b and b > a it follows that a = b.” Since the premise {a > b and
b > a) never holds, the implication is formally true.
Problem 1.10. Suppose that a set is ordered by a relation £- having
properties 1.1 through 1.7. As has already been shown, it has a minimal
element. Let us denote this minimal element by a0. It follows from the
results of problem 1.8 that each element has an immediate successor. Let us
denote the immediate successor of a0 by alt the immediate successor of ax
by a2, and so on. As a result, we obtain a sequence
flo> fli> , (6-1)
in which for any n, an+1 &-On■By reflexivity and transitivity of the relation
it follows that at £- a, if and only if i > j. We now have only to show
the sequence (6.1) contains all the objects of our set. This is accomplished
by means of a fairly subtle argument by induction.
Let b0 be an element of our ordered set. We shall construct inductively a
decreasing sequence of objects all smaller than b0 as follows: Suppose
has already been chosen, and compare it with a0. If bn- x = a0, the sequence
is complete. If 6„_i + a0, then take bn to be the immediate predecessor of
bn- 1. This process results in a sequence of distinct elements:
bo &- bi £- b2 • • • £- bn E—• ■■•
52 Chapter Six
On the basis of property 1.4, this sequence must have a last element. But
from the very rule of construction of this sequence, its last element can only
be fl0- We may suppose without loss of generality that bn = a0.
It is not difficult to verify that if some number a immediately precedes b,
then b immediately follows a. This means that bn-i. = au bn- 2 = a2, .. . ,
bo dn.
This last equality shows that any element b0 of our set belongs to sequence
( 6 . 1).

Problem 1.11. Let a be some number. We shall call any sequence of

distinct numbers a0 = a, alt a2, . . ., an, for which

do £- Qi £- d2 £- • • • £- an , (6.2)

where an is minimal under the ordering £-, a chain of predecessors of a0.

The number n will be called the length of this chain.
We shall show first that, under the conditions we have imposed on the
order no number can have arbitrarily long chains of predecessors.
Indeed, let a be some number, and let bu b2, ■.., ; be numbers that
immediately precede it.
If a1 does not immediately precede a0, we may, on the basis of property
1.9, insert in (6.2) a number immediately preceding a (recall a = a0).
Consequently, if there are arbitrarily long chains of predecessors of a, there
must also be arbitrarily long chains of its predecessors that begin with
numbers immediately preceding a. In the future, we shall consider only such
chains.
Each of these chains of predecessors of a is longer by exactly one than
some chain of predecessors of one of the numbers immediately preceding a.
If each of them had chains of predecessors of bounded length, then a itself
could not have arbitrarily long chains of predecessors.
This means that, under our assumption, at least one of the numbers
immediately preceding a0 has arbitrarily long chains of prececessors. Let us
denote this number by ai and apply to it all the arguments just given with
regard to a0. This gives us some number a2 that immediately precedes
and has arbitrarily long chains of predecessors. Repeating this process, we
arrive at a sequence

do £- i £- d2 £- ■• • >

which, by virtue of property 1.4, must break off sooner or later. This means
that the sequence will have an element to which our arguments will no
longer apply. But we have already established the applicability of these
arguments to each element of the sequence. This contradiction shows that
no number has arbitrarily long chains of predecessors.
For each number a, therefore, it is possible to choose a longest from
among its chains of predecessors. Let us denote its length by n{a). If b
immediately precedes d, then clearly n(b) < n(a) — 1, and, for all minimal a,
n(a) = 0.
Suppose, finally, that A(a) is an assertion depending on a. Let us denote
by B(n) the assertion “ A(a) is valid for all objects a for which n(a) = n.”
 Solutions to Problems 53
Then, as may easily be seen, the formulation of the principle of induction in
the new form for the assertions A(a) coincides with the formulation of this
principle in the old form for the assertions Bin).
Problem 1.12. For any even integers a and b, there exist even numbers q
and r such that
a = bq + r (0 < r < 2b) . (6.3)
These numbers q and r are unique.
Proof. By theorem 1.7, there exist q0 and r0 such that

a = bq0 + r0 (0 < r0 < b) .

Because of theorem 1.6, the number r0 must be even. If q0 is also even,
we may set q = q0 and r — r0. But if the number q0 is odd, we take q =
qo — 1 (which is even), and r = r0 + b. In either case, we obtain relation
(6.3).
Let us suppose that, in addition to (6.3), there is another representation
a = bq' + r' (0 < r' < 2b) ,
in which the numbers q' and r' are even. Then we have
b(q' - q) = r - r' .
Since
0 < \r - r'\ < 2b ,
we must have \q' — q\ <2. But q' — q is an even number, which means
that q' — q = 0, from which uniqueness follows at once.

Problem 1.13. Let p be the smallest prime divisor of the number a.

From this it follows that a = pb. Any prime divisor q of the number b is,
at the same time, a divisor of a. Therefore, q > p, and so b > p, so that
a > p 2, and, finally, p < No better bound than this is possible,
as may be seen by considering the number a = p2.
Problem 1.14. Necessity. Suppose that b\a. From theorem 1.13, it
follows that each prime divisor of b is a prime divisor of a. Thus, b has
the form

/ ’/ W 2- • -p/ ’ , (6-4)

where 0 < j31( 0 < j82, . . . , 0 < j8r. Let us suppose that some pt > at.
Without loss of generality, we may assume > a±. Set a = be. Then

P-L'Pff*- • P t*’ = P i01P20*- ■- P f ’c •

54 Chapter Six
Cancelling i on both sides of this equation, we have

p 2a2- ■.Pr«T = p { i - aip202- ■-p/rC .

Since f t > implies f t — ax > 0, p 1 must divide the right side of the
new equation; but p 1 does not divide the left side. Since this is a
contradiction, we are forced to conclude that ft < a, for each i = 1,
2 completing the proof of necessity.
Sufficiency. To prove the sufficiency we need only note that if b has
the indicated form, then

a = bp*! Bip%2 ■-Prr 6r ■

Problem 1.15. As we have already shown, each divisor of the number a

with canonical decomposition piaip2“2- ■■pf* must have form (6.4), where
ft assumes <*i + 1 values: 0, 1, 2, . . . , f t assumes a2 + 1 values, and
so on. Since any combination of these values is possible and the set of all
such combinations gives us all the divisors of a, with each occurring exactly
once (if any divisor were repeated more than one time, then it would have
more than one canonical decomposition), the number of divisors of a is

(“1 + 1)(“2 + I)' ' '(“r + 1) •

Problem 1.16. Suppose that the canonical decomposition of a is p\a^p2 2
• • ■pit?*. Clearly, we may set Pi = 2, ^ > 2 and p2 = 3, a2 > 1. Further­
more, we have
(“1 + l)(a2 + I) - • '(«)£ + 1) — 14.

Hence, k = 2, <*i + 1 = 7, and a2 + 1 = 2 (since ^ + 1 = 2 implies that

4fa). Thus, a = 26-3 = 192.
Problem 1.17. We have
r(a2) = T(Pl2“ip22°2) = (2a! + l)(2a2 + 1) = 81 ,
so that (2aj + l)(2a2 + 1) is a decomposition of 81 into two factors. Since
the numbering of the prime divisors of a is arbitrary, we need only consider
the following possibilities:
2ai + 1 = 1, 2a2 + 1 = 81;
2oti + 1 = 3 , 2a2 + 1 = 27;
2ai + 1 = 9 , 2a2 + 1 = 9 .
In the first of these cases, ai = 0, which contradicts the assumption that
a > 0. The remaining cases give us
ai = 1 , «2 = 13;
“i = 4, «2 = 4 .
 Solutions to Problems 55
This means that either

r(a3) = T(piaip2a2) = rip^pz39) = (3 + 1)(39 + 1) = 160,

or

T(a3) = t (pl°ip23a*) = t (pi12pz12) = 13-13 = 169.

Problem 1.18. Let pf'pz**- ■-pif* be the canonical decomposition of the
number a. The condition of the problem gives us

P i aiP z 2 - ■■P z k = 2(«1 + l ) ( a 2 + 1)- • -(a * + 1) ,

or

Pi*1 P2a* P k*
(6.5)
ai + 1 a2 + 1 ait + 1
We note that
21 22 23 2a
1< (« > 4) ,
1+ 1 2+ 1 ' 3+ 1 2< a+ 1
31 3“
1< (a >
1 + \ <2< a + 1

Pa ( p > 5, a > i).

a+ 1
Therefore, each fraction on the left side of (6.5) is not smaller than 1, and,
consequently, none of the fractions may be larger than 2. This means that
only fractions from the following collection can appear on the left side of
(6.5):
21 22 23 31
1 + 1’ 2 + 1’ 3 + 1’ 1 + 1 ’
and the product of the fractions must be 2. But this can be so in only two
cases: when the only fraction on the left side of (6.5) is 23/(3 + 1), or when
only the two fractions 22/(2 + 1) and 33/(l + 1) appear there. The two
answers to the problem—8 and 12—correspond to these two cases.
Problem 1.19. The analogues of theorems 1.11-1.14 are false for even
divisibility.
Indeed, the numbers 30 and 42 are even primes (that is, they are not
products of any two even numbers). Their least common even multiple is
420 and their product is 1260.
Furthermore, 60 = 6-10 is evenly divisible by the even prime 30; 6 and
30 are evenly relatively prime, but 10 is not evenly divisible by 30.
Finally, the decompositions 60 = 6-10 and 60 = 30-2 are two distinct
factorizations of the number 60 into even primes.
56 Chapter Six
Problem 1.20. Let p lyp 2, . . be a complete list of all the primes
entering into the canonical decomposition of at least one of a and b.
Let us set

a = Pia'P2a2- ■-Pkk ,
b = Pi^Pz1*2- ■' Phk ■

(If a is not divisible by p h then a* = 0; if b is not divisible by pu then

ft = 0.) Let yx be the larger of the two numbers ax and ft for each
i = 1 ,2 ,...,/ :, and let ft be the smaller of them.
Then, because of our result in problem 1.14, the greatest common
divisor of a and b is Px^p2 *- ■■Puk, and their least common multiple is
PiyW 2- ■■Pk k■

Problem 2.1. a. 116 is congruent to 4 and 17 is congruent to 1 modulo

8. This means that A is congruent to 521 = (52)10-5. But 52 = 25 is
congruent to 1 modulo 8. Consequently, A gives a remainder of 5 on
division by 8.
b. It is clear that 14256 = 196128 = 9128 = 8164 = (~ 4 )64 = 464 =
1632 = ( - 1 ) 32 = l(mod 17).
Problem 2.2. a. Let n1 be the remainder on division of n by 6. Then
n1 may assume the values 0, 1,2, 3,4, and 5, and + 11«! is congruent
modulo 6 to n3 + 1In. This means that we need test only the numbers
0, 12, 30, 60, 108, and 180 for divisibility by 6. But all of these numbers
are divisible by 6, and so n3 + 1In will be divisible by 6 for any n.
b. For n > 2, we have (using the binomial theorem):

4" + 15n - 1 = (3 + l)n + 15n - 1

+ 18w,

which is clearly divisible by 9

For n = 1, the expression is equal to 41 + 15 -1 — 1 = 18, which is
also divisible by 9, completing the proof.
 Solutions to Problems 57
c. The proof is by induction.
For n = 0,

10<3°> - 1 = 101 - 1 = 9 and 30+2 = 9 ,

so 3ii +2|10(3") - 1.

Suppose now that the divisibility

3« +21 J0<3n> _ i

is known to be valid. Then

10(3"+1) - 1 = (10<3">)3 - l 3
= (10(3"> - l)(102'(3n) + 10(3"> + 1).

The first factor on the right is divisible by 3" +2 by the induction hypothe­
sis. In the second factor, we may replace the tens by ones, since 10 is
congruent to 1 modulo 3. The resulting sum, 3, is divisible by 3, and,
therefore, so is the second factor itself. Consequently, the whole product
is divisible by 3n+3 = 3<7l+1) +2, which is what is required.
d. Clearly, a2 is congruent to a — 1 modulo a2 — a + 1. This means
that a2n+1 + (a — l)n+2 is congruent to

a2n+1 + (a2)n+2 = a2n+1 + a2n+i

= a2n+\ 1 + a3)
= a2n+1(l + a)(l —a + a2) ,

so a2 — a + 1|a2n+1 + (a — l)n+2, as was to be shown.

Problem 2.3. Let ~ be an equivalence relation on a set of numbers. Let

us take an arbitrary number a and consider all numbers equivalent to a.
At least one such number, namely a, is known to exist because the relation
~ is reflexive. Any two of these numbers are equivalent to one another by
virtue of the transitivity of the relation ~ . We shall denote the class of these
numbers by K.
Now let us consider an arbitrary number b not belonging to K. If b ~ c
were to hold for some number c in K, then b ~ a would also hold, but this
cannot be, by the choice of b. This means that none of the numbers lying
outside K is equivalent to any of the numbers in K. Consequently, K is an
equivalence class containing a.
Finally, since the number a was completely arbitrary, the arguments
given show that each number belongs to some equivalence class, and so the
proof is complete.
58 Chapter Six
Problem 2.4. Clearly, among the m + 1 numbers 0, 1, . . m, there are
two that belong to the same class, since ~ splits the numbers into only m
classes. Suppose that these numbers are k and /: k ~ /. Since, generally
speaking, there may be more than one such pair of numbers, let us choose
that pair for which the quantity |k — I \ is greatest. Since —/ -----/, we
obtain:

k - l ~ I - I = 0.
Furthermore, we find that, for any integer rt,
n(k — ~ 0.
Finally, for any r.
n(k — I) + r ~ r;

that is, from a = 6(mod k — I) it follows that a ~ b. Thus, each equivalence

class of the relation ~ contains some residue class modulo k — I.
In order to have m equivalence classes under —, it is therefore necessary
to have k — I > m. But by our construction, k — I < m. Thus, k — I = m
and each equivalence class under ~ contains exactly 1 residue class
modulo m.
Problem 2.5. Two such rules are the following:
a. Both sides of a congruence and the modulus may be divided by the
same number (distinct from zero, of course).
Indeed,
ad = bd(mod md)

means that
md\ad — bd = (a — b)d;
that is, that m\a — b, and hence, a = 6(mod m).
b. Both sides of a congruence may be divided by a number that is
relatively prime to the modulus.
Indeed, if d and m are relatively prime, then from
ad = bd(mod m) ,

that is, from m\(a - b)d, it follows on the basis of theorem 1.12 that m\a -
b, as was required.
Problem 2.6. Let us suppose, on the contrary, that
ka = /a(mod p) , I < k < I < p — 1.
This means thatp\(l —k)a. Sincep does not divide a, we must havep\l — k.
But this is impossible, since 0 < I — k < p.
 Solutions to Problems 59
Problem 2.7. Necessity. Suppose that the number p is prime. Let us
choose any q such that 0 < q < p. Among the numbers q, 2q, . . ., (p — 1)q,
there is exactly one that leaves a remainder of 1 on division by p. Suppose
that this number is qq:
qq = l(mod p) . (6.6)
On the other hand, among the numbers q, 2q, .. . , (p — 1)q there is also
exactly one that leaves a remainder of 1 on division by p. This, as has already
been established, is the number qq.
Let us clarify the circumstances under which q = q. In all such cases, the
congruence (6.6) may be rewritten as
q2 = l(mod p) ,

or, equivalently, as

q2 — 1 = 0(mod p ) .

This means that

p\q2 - 1 = {q + 1)(q - 1) .

Since p is prime, theorem 1.13 shows that either p\q + 1 or p\q — 1. Since
the number q lies between 0 and p, the first case is possible only for q =
p — 1 and the second for q = 1.
Consequently, given any q different from 1 and p — 1, q A q\ so the
remaining numbers 2 ,..., p — 2 may be paired in such a way that the
product of the numbers forming each pair leaves a remainder of 1 on
division by p. Let us write out the congruences of the form (6.6) for all
such pairs and add to this list the congruence

p — 1 = p — l(modp);

and then let us multiply all these (p — l)/2 congruences term by term.
As the result of this multiplication, we obtain on the left side the product
of all the numbers from 2 to p — 1 and, on the right side, p — 1:

2-3- • -(p - 1) = p - l(m odp),

or

1-2-3- • -{p - 1) + 1 = 0(modp ) ,

which means simply that

p\l-2 - ■-{p - 1) + 1,

as was required.
60 Chapter Six
Sufficiency. If the number p is not prime, then it may be decomposed
into two smaller factors: p — pip2.
If Pi 9* p2, then both pi and p2 occur as factors in the product 1-2-••
(p — 1), which is therefore divisible by pip2, that is, by p. If, however,
Pi = Pi = q, then p = q2 (that is, p is a perfect square). If q > 2, then
p > 2q and both the numbers g and 2q enter into the product 1 -2- •• (p — 1),
so that this product is divisible by q2 = p. In either case, 1 ■2 (p — 1) + 1
cannot be divisible by p. Finally, if q = 2, so that p = 4, then 1-2-3 + 1 =
7, which is not divisible by 4.
Problem 2.8. Theorem. Let m = p i ,'p2a'>■■■pkak be the canonical decom­
position of m. Then for the numbers A and B to be congruent modulo m, it is
necessary and sufficient that they be congruent modulo pia‘ (i = 1, 2, . . k).
Proof. For A and B to be congruent modulo m, it is necessary and suffi­
cient that m\A — B. But by theorem 1.16, this is equivalent to

Pi^\A - B ,. . .,p kai‘\A - B ,

that is, to

A = B(mod pfi) , i = 1, 2,.. ., k .

Problem 3.1. a. Application of the Euclidean algorithm to the numbers a

and b yields a series of equations of the form:

a = bq0 + f i ,
b = r ^ + r2 ,
fi = r2q2 + r3 ,
(6.7)
rn- 2 = rn- t f n-! + rn ,
rn-i = rnqn .

We have rn\rn-!. This, together with rn_2 = rn- t f n- x + rn, gives us

rn\rn-.2- Arguing similarly, we may move up the system of equations (6.7),
until we finally conclude that rn\a and rn\b. This means that rn is a common
divisor of a and b.
Let d be any common divisor of a and b. This fact, together with the
equation a = bq0 + ru gives us d\rx. Moving in this way down the system
of equalities (6.7), we find successively that d\r2, d\r3, . . . , d\rn. This means
that rn is divisible by any common divisor of a and b and is thus the greatest
common divisor of these numbers.
b. The proof is by induction and uses the Euclidean algorithm applied
to a and b. Setting A0 = 0, B0 = 1, A-! = 1, = —q0, we have r0 = b =
aAo + bB0 and r± = aAi + bBx. Suppose now that

rk-1 — Ak- t f + Bjc-ffi,

rk = Aka + Bkb .
 Solutions to Problems 61
Then
rk +1 = rh- - rkqk = (Ak- - qkAk)a + [Bk- 1 - qkBk)b ,
1 1

and we have only to set

Ak+i — Ak-i —qkAk ,

Bic+1 = Bk - 1 —qkBk ,
in order to have rk +1 = /ffc+i(z + Bk+1b. Then clearly, by this construction,
the numbers An and Bn will be the desired A and B.
Problem 3.2. We will consider only theorem 1.12 as an illustration. The
proofs of the other theorems use similar techniques.
Suppose that c\ab. If b and c are relatively prime, then, by the preceding
problem, we may find integers B and C such ihat

bB + cC = 1 ,

or, after multiplication by a,

abB T acC = a ;

but c\ab by hypothesis and since c|ac, we have c\a, as desired.

Problem 3.3. We shall limit ourselves to considering a criterion for

congruence modulo 8.
Suppose that an arbitrary natural number A is represented in the
form lOOOn + b, where 0 < b < 1000 (that is, b is simply the last three
digits of A). Let

/ (A) = < the remainder on division of A by 8 if 8 < A < 1000,

undefined if A < 8 .

The one may easily verify that the function/is the required criterion for
congruence.
Problem 3.4. For those having a canonical decomposition of the form
2a5B for some a and /?.
Problem 3.5. Conditions (a) and (b) are automatically satisfied. Since
the numbers 10 and 1 (and, therefore, 102 and 1, 103 and 1, and so on)
are congruent modulo 3, so are the numbers A and f(A). Finally, the
fact that f ( A) < A for A > 3, is established by a simple calculation.
62 Chapter Six
Problem 3.6. a. ^,(858,773) = 3 8 ;/2(38) = 11 ; / 2(Il) = 2 ./ 2(789,988)
= 4 9 ;/2(49) = 1 3 ;/2(13) = 4 ; / a(4) = 1.
b . U A ) = 4444-4 = 17,776; / a(17,776) = 2 8 ;/2(28) = 1 0 ;/2(10) =

Problem 3.7. The criterion for congruence modulo 9 is exactly

analogous to the criterion for congruence modulo 3.
To obtain a criterion for congruence on division by 11, we represent
the number A in the form

A = 102nan + I02n~2an_1 + • • • + \02ai + a0 ,

where 0 < at < 100. Such a representation clearly corresponds to

simply breaking up the number into two-place “ blocks” (from right to
left). Let

fa0 + Ui -i-----+ an if A > 100 ,

f (A) = < the remainder on division of A by 11 if 11 < A < 100 ,
[not defined if ^4 < 11 .

It is not hard to show that the numbers A and f ( A) are indeed

congruent modulo 11 and that, in addition, f (A) < A.
Another criterion for congruence on division by 11 is obtained by
representing the number A in the form

A = 10nan + \0n~1an^ 1 + ■• • + 10^! + a0

and using the fact that 10 is congruent to —1 and 100 is congruent to 1

modulo 11. Then A is congruent to the number a0 — + a2 — a3 H----
± an, and one may easily formulate a criterion for congruence on this
basis.
Finally, we may break up the number A into 3-place “ blocks” and
represent it in the form

103nan + 103n~3an_1 H-------h lO3^ + a0

(0 < Uj < 1000). Then, modulo 37, A is congruent to the sum a0 + ai +

■■■+ an, and, modulo 7, 11, and 13, to the alternating sum a0 — ^ +
U2 it ^71-
Problem 3.8. If the numbers a and b are congruent, then m\a — b.
Therefore, by virtue of theorem 1.6, the numbers a and b are simultane­
ously divisible or not divisible by m.
 Solutions to Problems 63
Four and 5 are equidivisible on division by 3, but not congruent
modulo 3.
Problem 3.9. Suppose that congruence modulo m implies equi-
divisibility by m. Then all numbers not divisible by m leave the same
remainder on division by m. Therefore, either there are no numbers
which are not divisible by m (that is, m = 1), or there is only one non­
zero remainder on division by m (that is, m = 2).

Problem 3.10. The relation of equidivisibility is clearly reflexive (any

number is equidivisible by m with itself), symmetric (if a is equidivisible
with b, then b is equidivisible with a), and transitive (if a is equidivisible
with b and b is equidivisible with c, then a is equidivisible with c).
Consequently, equidivisibility is an equivalence relation. All numbers
that are divisible by m fall into one class, and all numbers that are not
divisible by m fall into another class.

Problem 3.11. It is not difficult to verify that for m > 2, equi­

divisibility of sums does not follow from the equidivisibility of their
terms.
A necessary and sufficient condition for the equidivisibility of products
to follow from the equidivisibility of their terms is that the number m be
prime.
Indeed, if one of the products is divisible by a prime p, then, by
theorem 1.13, at least one of the factors of this product must be divisible
by p\ but then p would divide the factor in the other product that is
equidivisible with this factor in the first product, and hence, p would
divide the second product. So p divides one product if and only if p
divides both products; that is, the two products are equidivisible by p.
Conversely, if the number p is composite, then products of equi­
divisible factors may fail to be equidivisible. It is sufficient to set p =
P1P2 (Pi ^ !)• Then the numbers 1 and p x, as well as the
numbers 1 and p2, are equidivisible by p, while their products clearly
are not.
The corollary does not hold, as 2 is equidivisible with 4 on division
by 8, but 22 and 42 are not equidivisible by 8.
Problem 3.12. This is an immediate corollary of the first part of
problem 3.8.
Problem 3.13. a. Clearly, it is enough to check that a — 2b > 0
whenever A > 190. Now if a — 2b < 0, then a < 2b. Since the maxi­
mum possible value for b is 9, this cannot happen if a > 18. But a > 18
whenever A > 190.
b. is clear.
64 Chapter Six
c. If A > 190, we must show that 10a + b > a — 2b, or 9a + 3b >
0. But this is always true, since a is positive and b is nonnegative.
If 7 < A < 190, (c) is clear.
Finally, 10a + b and 50a + 5b are equidivisible by 7 (since the
numbers 5 and 7 are relatively prime) and so, therefore, are 10a + b
and 50a + 5b — 7(7a + b) = a — 2b.
Problem 3.14. 191 is congruent to 2 modulo 7, while/3( 191) = 19 —
2 1 = 17 is congruent to 3 modulo 7.
Problem 3.15. Condition (c). To verify that f( A ) < A, it is sufficient
to show that a + 4b < 10a + b\ that is 3b < 9a. Therefore, for a > 4
(i.e., A > 40), the necessary condition is satisfied.
Condition (d*). Since 4 and 13 are relatively prime, 10a + b is
equidivisible by 13 with 40a + 4b, and this last number is equidivisible
with a + 4b.
Problem 3.16. The criterion for divisibility would no longer be
determinate, as / 4(39) would be 39.
Problem 3.17. Suppose that we are to construct a criterion for
divisibility by some m. We choose the smallest number s satisfying
m| 10jt + 1 (for m = 1; s turned out to be 2) or m\\Qs — 1 (for m = 13,
s = 4).
In the first of these cases, A = 10a + b is equidivisible by m with

lOas + bs = (10j + l)a — a + bs

(since s and m must be relatively prime); that is, with a — bs, and in
the second case, with

lOaj1+ bs = (10j — 1)a + a + bs;

that is, with a + bs.

In connection with the above, the number 10a + b

is equidivisible with a —5b on divisionby17,

is equidivisible with a + 2b on divisionby19,
is equidivisible with a + lb on divisionby23,
is equidivisible with a + 3b on divisionby29,
is equidivisible with a —3b on divisionby31.

The precise formulation of these criteria for divisibility is left to the

reader.
 Solutions to Problems 65
Problem 3.18. a. Since 100 is congruent to 2 modulo 49, any number
of the form

102nan + 102" -2(2n_! + • ■• + lO2^ + a0 (0 < at < 100)

is in fact congruent, modulo 49, to

2nctn + 2n 1an-1 + • • • + 2cii + a3 .

b. 10a + b is equidivisible with a + 5b by 49.

Problem 3.19. Conditions (a) and (b) are automatically satisfied.
Conditions (c) and (d) hold because in order to obtain f (A) from A for
10" > m, one need only replace certain numbers whose sum is equal to
A by their remainders on division by m (which are smaller than but
congruent to the numbers themselves), thus obtaining as a new sum a
smaller number which is congruent to A by theorem 1.6.
Problem 3.20.a. r1 = r2 = ■■• = rn = 0; that is, rk = 0for k > 1;
b. r2 = r3= • ■■= rn = 0; that is, rk = 0 for k > 2;
c. r1 = r2= • = rn = 1; that is, rk = 1 for all k ;
d. rY = r3= ■ ■■= r2t-i = - 1; r2 = r4 = • • • = r2t = 1; that is,
rk = ( —1)* for all k;
e- ret +i = 3,ret +2 = 2, ret +3 = 6,r et +i = 4, ret +3 = 5,r et = 1, all /.

Problem 3.21. This is left to the reader.

Problem 4.1. Neither 2i — 2 nor 23 — 1 is divisible by 4.

Problem 4.2. If p\a, then p\ap — a, and the theorem is proved. If,
however, p does not divide a, then a and p are relatively prime, and the
congruence in the theorem is equivalent to the congruence

ap_1 = l(m odp ) .

To prove that this last congruence holds, we divide each number of

the form ta (t = 1, 2 , . . . , p — 1) by p with remainder:

ta = qtp + rt .

This may be rewritten as

a ~ /^(mod p) ,
2a = r2(mod p) ,
( 6 . 8)

(P - 1)a = rp_1(modp) .
66 Chapter Six
From the result of problem 2.6, it follows that each of the numbers
1 , 2 — 1 occurs exactly once among the numbers rlt . . . , rp_1.
Multiplying together all the congruences in (6.8) term by term, we
obtain

1 -2 - • (p — 1)ap 1 = 1-2■■ (p — l)(m odp ) .

The proof is completed cancelling 1-2■■ (p — 1) on both sides of the

congruence.
Problem 4.3.

9^(12) = <p(22•3) = 22-1(3 - 1) = 2-2 = 4 ,

9(120) = <p(23-3-5) = 23“ 1(3 - 1)(5 - 1) = 4-2-4 = 32,
9(1000) = 9(23-53) = 23-153“ 1(5 - 1) = 4-25-4 = 400.
Problem 4.4. Let m = p xaip 2 2■■■P k k- Then

a- PttP ~ \ P i - 1)Pz3~1(P2 - 1 ) - ■pl*~\P)c - 1) = 10 .

The product on the left must be divisible by 5. This means that either
one of the numbers p u p 2, . ■., p k is 5 (to be definite, let us suppose
that p 1 = 5) or one of the differences Pi — 1, p 2 — 1 ,..., p k — 1 is
divisible by 5 (in this case, suppose that 5|/>! — 1). In the first of these
cases, Pi — 1 = 4, which cannot be, since 10 is not divisible by 4. Thus
we are reduced to the second case, which is possible only for p 1 = 11,
since p1 must be prime, and (px — 1)| 10. But then = 1, and from
theorem 4.2 it follows that 9(m /ll) = 1; that is, either mj\ 1 = 1 or
mj 11 = 2; that is, either m = 11 or m = 22. It is easy to verify that
9 ( 11) = 9 (22) = 10.

b. Pi1-1(Pi - 1)P%*~\P2 ~ 1)' • -p I*~\P k - 1) = 8 .

If m is odd, then 0^ = «2 = • ■■= = 1 (since the right side of the

equality above is a power of two):

(.Pi ~ O O 2 - 1) • ■ (pfc - 1) = 8 .

This is possible only for k = 2, p1 = 3, andp2 = 5; that is, for m = 15.

Suppose now that the number m is even. In order to be definite, let
us set p 1 = 2. Clearly, as before, a2 = - - • = ak = 1, and we have

2a~ \ p 2 ~ 1)•■■(/>* - 1) = 8 .
 Solutions to Problems 67
Clearly, a < 4. If a = 1, then the case is exactly like the one just
considered: The equality above is possible only for k = 3, p2 = 3, and
p3 = 5, that is, for m = 30. If a = 2, then k = 2, p2 = 5, and m = 20.
If a = 3, then k = 2, p2 = 3, and m = 24. Finally, if a = 4, then k = 1
and m = 16.
Thus, the solutions to our problem are: m1 = 15, m2 = 30, m3 = 20,
= 24, mk = 16.
Problem 4.5. Let us suppose that

PaP ~KPi - 1 K 2-1(T2 - !)• • P> ~\Pk - 1) = 14.

Every number of the form pt — 1 is either 1 or an even number, and

therefore cannot be 7. In addition, pt — 1 ^ 14, since 15 is not a prime.
This means that one of the numbers p f " 1 is 7. But then p { — 1 = 6 ,
and 14 is not divisible by 6, so <p(m) is not 14 for any m.
Problem 4.6. The problem is easy to verify for m = 1. The next case
to be considered is that in which m is some power of a prime; say,
m = pa. For a number to be relatively prime to m, it is necessary and
sufficient that it not be divisible by p. But among the numbers 0, 1,
2 , . . . , m — 1, there are in all mjp numbers that are divisible by p.
Consequently, there are in this list

m ~ = m{\ - ^ = p °^ l - = p p - \ p - 1) = <p(m)

numbers that are relatively prime to m.

We now note that for a to be relatively prime to m, it is necessary
and sufficient that the remainder on division of a by m be relatively
prime to m.
By what was established above, the number of possible remainders on
division by p"' (that is, the number of numbers less than />,“■) that are
relatively prime to p* ‘ is equal to <p(plai). But, as was shown in problem
2.8, the congruence of two numbers modulo m follows from their con­
gruence modulo ptai for all i, and conversely. In particular, for a number
to be relatively prime to m, it is necessary and sufficient that it be
relatively prime to each of the numbers p
Consequently, to each combination of remainders on division by
Pi“q p2a2,. . . , pk k that are relatively prime to the corresponding
divisors, there corresponds exactly one remainder on division by m that
is relatively prime to m. We now have only to note that the number of
such combinations is equal to <p(piai)<p(p2a2) • •<p(Pkak) = 9(m)-
68 Chapter Six
Problem 4.7. This is left to the reader.
Problem 4.8. n13 — n = n(n12 — 1). Butn12 = «<“(13) = n2v,a) = n3'‘>i5)
_ n6®(3) _ ni2®(2) Therefore, either p\n or p\n12 — 1 for p = 2, 3, 5, 7,
13 (by Euler’s theorem and the fact that in general, x — 1 divides x m — 1
for any m); so, by theorem 1.6, n13 — n is divisible by 2-3-5-7-13 =
2730.
Problem 4.9. This is left for the reader to do independently.
Problem 4.10. Let d be the greatest common divisor of a and b. If c
is not divisible by d, then it follows from theorem 1.6 that the equation
ax + by = c cannot be solved in integers. If c is divisible by d, however,
then both sides of the equation may be divided by d, and we arrive at
the case already considered; so the equation does have a solution in
integers.
Problem 4.11. Let A and B be such that aA + bB = 1. Let us set

. » 1 — aA
x t = cA + b t , y t = c — r------- a t .

Then

axt + byt = a(cA + bt) + b{c -— — at j

= caA + abt + c(l — aA) — abt — c ,

and the pairs (x t, yt) are indeed solutions to our equation.

Problem 4.12. a. x t = 9 ■5s + I t = 28,125 + It,

yt = 9 - - 51 = -20,088 - S t .

Since the numbers 28,125 and 20,088 and the coefficients 7 and 5 are,
so to speak, “ approximately proportional,” we may hope to obtain a
representation of our solution using smaller numbers. Indeed, we may
write

x t = 6 + 7(t + 4017) , y t = —3 — 5(f + 4017) ,

or, setting t + 4017 = t ’,

Xt> — 6 + 7r , JV = - 3 - 5f .
 Solutions to Problems 69
b. Let us use the fact that the order of 25 modulo 13 is 2. We may
write

xt = 8-25 + 13? = 200 + 13/,

2

y t = 8 1 ~ 325 - 25t = -3 8 4 - 25?;

or, after simplifications,

xv = 5 + 13/', y v = - 9 - 25?'.
Problem 4.13. Conditions (a), (b), and (c) are guaranteed auto­
matically, and condition (d*) follows from theorem 4.6.
Problem 4.14. The criteria can be easily constructed using the general
criterion and the following table:

m 17 19 27 29 31 49
k' 12 (or —5) 2 19 (or - 8 ) 3 28 (or - 3 ) 5

Problem 4.15. This is left for the reader.

 Popular
Lectures
Mathema e d fro m
series
Iz a a k W irs zu p , D a n ie l A . L e v in e a n d
E d ito r T im o th y M c L a rn a n

T h e P o p u l a r L e c t u r e s in N . N . V o r o b ' e v ’s C r ite r ia fo r
M a th e m a t ic s s e rie s , tra n s la te d D iv is ib ilit y i n t r o d u c e s t h e h i g h
an d a d a p t e d fro m th e R u s s ia n , s c h o o l o r e a rly c o lle g e s tu d e n t
m a k e s a v a i l a b l e to E n g lis h - to a s p e c i f i c n u m b e r - t h e o r e t i c
s p e a k in g te a c h e rs an d stu d e n ts to p ic a n d e x p la in s th e g e n e ra l
s o m e of th e b e st m a th e m a tic a l
m a th e m a tic a l s tru c tu re s w h ic h
lite ra tu re of th e S o v ie t U n io n .
u n d e rlie th e p a rtic u la r c o n c e p ts
T h e le c tu r e s a r e in te n d e d to
d i s c u s s e d . V o r o b ’e v d i s c u s s e s
in tro d u c e v a rio u s a s p e c ts of
m a th e m a tic a l th o u g h t a n d to th e id e a s o f w e l l - o r d e r e d sets,

e n g a g e t h e s t u d e n t in p a rtial a n d lin e a r o rd e rin g s ,

m a th e m a tic a l a c tiv ity w h ic h e q u iv a le n c e re la tio n s , e q u iv a ­
w ill f o s te r i n d e p e n d e n t w o r k . le n c e c la sse s, a lg o rith m s , and
S o m e of th e le c tu re s p ro v id e an th e re la tio n s h ip b e tw e e n th e
e l e m e n t a r y in tro d u c tio n to d e te rm in a b ility o f a lg o rith m s
c e rta in n o n e le m e n t a r y to p ic s ,
d e fin e d on th e in te g e rs an d th e
a n d o th ers p re s e n t m a th e m a tic a l
w e ll-o r d e rin g p rin c ip le .
c o n c e p t s a n d i d e a s in g r e a t e r
A l l t h i s is d o n e c o m p r e h e n ­
d e p th . T h e b o o k s c o n ta in m a n y
siv e ly w ith th e h e lp of a u n iq u e
in g e n io u s p r o b le m s w ith h in ts,
s o lu tio n s , a n d a n s w e r s . A p la n fo r s tu d y w h ic h e n c o u ra g e s
s i g n i f i c a n t f e a t u r e is t h e a u t h o r s ’ t h e s t u d e n t to s k i p l a r g e s e c ­
u s e of n e w a p p r o a c h e s to m a k e tio n s o f th e b o o k o n first r e a d in g
c o m p le x m a th e m a tic a l to p ic s a n d r e tu rn to th e m la te r. T h e
a c c e s s ib le to b o th h ig h s c h o o l m o re g e n e ra l a n d c o n c e p tu a lly
a n d c o lle g e s tu d e n ts .
c h a l l e n g i n g m a t e r i a l a p p e a r s in
s m a ll p rin t, so th a t th e s tu d e n t
m u s t h a v e a g o o d g ra s p o f th e
n u m b e r-th e o re tic co n c e p ts on
w h ic h th e g e n e ra liz a tio n s are
b a s e d b e fo re m a k in g th e step
to g e n e r a l i z a t i o n . T h e b o o k le t
p ro v id e s b o th s p e c ific k n o w l­
e d g e in a p a r t i c u l a r fie ld o f
m a th e m a tic a l in v e s tig a tio n an d
a fin e b a s is on w h ic h to c o n tin u e
s t u d i e s in m a t h e m a t i c s .

N. N. Vorob ’ev is a s e n i o r
r e s e a r c h e r at th e L e n in g r a d
D iv is io n o f th e M a th e m a t ic s
In s titu te of th e U .S .S .R .
A c a d e m y of S c ie n c e s .

Paper ISBN: 0-226-86516-9