INTERNATIONAL SERIES IN
epstein
lass
weinstock
weiss
Calculus of Variations
Algebraic
zemanian
Number Theory
ELEMENTARY
THEORY OF NUMBERS
HARRIET GRIFFIN
Associate Professor of Mathematics
Brooklyn College
TORONTO
1954
INC.
LONDON
Number 5312046
6789MP9876
24785
PREFACE
The arithmetic theory of numbers is ever a fascinating subject, the
fundamentals of which can be presented with ease and profit to the average undergraduate student of mathematics and to those who are preparing to teach mathematics provided that these students are stimulated
by a clear and logical treatment of carefully selected topics. It is the
aim of this textbook to offer such a development of the subject. The
facts and methods of proof are old, indeed, but the author believes that
her experience of teaching the theory of numbers for over fifteen years
has enabled her to choose the topics that not only develop the student's
him
is
to be constructed.
at a level
and
in language
some to test his understanding of simple theoretical questions, and others to challenge his originality.
It is to be emphasized that this book is just a text.
It is written for
the student rather than the teacher. It is neither erudite nor exhaustive.
The reader with a good grasp of algebra and the ability to concentrate
will be able to understand it.
It is hoped, moreover, that it will interest
Vi
PREFACE
number
Theorems
and
511,
Theorems
63
through 66 and 69 through 612, Theorems 711 through 716, as well
For the convenience of the reader the
as all of Chap. 8 may be omitted.
theorems just mentioned have been marked with an asterisk. Most
instructors will wish to include as much of Chap. 9, on quadratic residues,
as time permits.
Theorems from Chap. 10 may be selected at pleasure
provided that Theorem 104, on the Pythagorean triangle, is included.
The last two chapters will, doubtless, have to be omitted in a 45hour
course.
Harriet Griffin
CONTENTS
PREFACE
chapter
1.
12.
13.
Definitions
14.
The
11.
2
6
principle of
chapter
Archimedes
2.
24.
The form ax + by
The Diophantine equation ax + by = n
A method for finding a solution of ax + by = n
The solution of the linear Diophantine equation in more than two
25.
21.
22.
23.
chapter
33.
34.
Unique factorization
35.
32.
36.
37.
38.
39.
The
The
least
common
312.
Congruent integers
42.
43.
The
45.
is
53
54
56
58
62
residue classes
<j>
a factor of n!
36
37
40
46
PROPERTIES OF CONGRUENCES
41.
44. Euler's
23
33
34
an integer
Perfect numbers
4.
21
31
multiple
divisors of
chapter
19
variables
25
25
26
28
30
16
PROPERTIES OF INTEGERS
3.
The composite
The sieve of Eratosthenes
The number of primes
31.
14
function
m
vii
CONTENTS
Vlll
chapter
51. Identical
52.
53.
54.
55.
56.
57.
58.
59.
Equivalent congruences
Linear congruences
Division of polynomials
62.
63.
of f(x)
An
7.
71.
The X function
Belonging to an exponent modulo
Another test for a prime
77.
78.
79.
710.
711.
89
92
93
96
ON BELONGING TO AN EXPONENT
72.
76.
88
application of the
chapter
75.
Fermat's theorem
Applications of Fermat's theorem
Wilson's theorem
64.
74.
= 0(mod m)
(mod p 8 )
98
100
104
105
107
108
112
114
116
118
118
Primitive roots
Gauss' method for finding a primitive root modulo p
Primitive roots modulo p n and 2p n
Primitive X roots
Integers belonging to a divisor of X(2 n ) modulo 2 n
Integers belonging to a divisor of <j>(p n ) modulo p n
Integers belonging to a divisor of <f>(2p n ) modulo 2p n
Integers belonging to a divisor of X(m) modulo m
chapter
8.
INDICES
a prime modulus
82. Euler's criterion for the solvability of x n
125
129
chapter
9.
The
92.
Quadratic residues
The Legendre symbol
The prime moduli of which an integer
The Jacobi symbol
The solution of x 2 = a (mod 2 n )
94.
95.
96.
chapter
10.
103.
Fermat's
104.
105.
last
is
a quadratic residue
134
135
139
140
152
156
102.
106.
c(mod m)
101.
QUADRATIC RESIDUES
91.
93.
66
68
70
72
75
79
80
84
85
6.
65.
73.
...
The number
The solution
chapter
61.
5.
z2
theorem
158
167
168
168
170
172
CONTENTS
chapter
111. Integral
domains and
11.
POLYNOMIALS
fields
chapter
121.
of
12.
IX
177
186
188
PARTITIONS
numbers
Diagrams
124.
of partitions
190
191
192
193
BIBLIOGRAPHY
197
INDEX
199
CHAPTER
The Development
of the Integers.
The
result of
number
is
elements,
number
itself.
The
and B, and
if
to each element of
If there are
there
is
two
sets of
assigned exactly
recognize the existence of negative quantities even though they did not
admit them as solutions of their problems. It was not until the sixteenth
century that European mathematicians reached this stage of development
1
They compound
basic properties.
the
Thus, although each of the symbols + 1, 1, +5, and 5 represents but one number, that number has
two fundamental attributes. It is evident, moreover, that a onetoone
correspondence can be set up between the positive integers and the natural
numbers
12.
in the following
The System
bers, arithmetica, is
numbers
1, 2, 3,
** 1,
+2
The
+3
 2,
<> 3,
Nevertheless,
1, 2,
rational integers 0,
itself, this
+1
manner:
of Rational Integers.
we
shall
make use
of the
system
of
enable us to develop
classical
an algebraic presentation.
Throughout
word
" integer," unless otherwise stated, will refer to a rational integer, and the
x, y, z, will
Moreover, we
letters a,
b, c,
shall, as
signs in writing
What
shall
through 10, and they will serve to illustrate some salient facts.
It is apparent that the product of three consecutive integers is divisible
by 6, that when two consecutive even integers are chosen, one is divisible
by 4. You will claim you have known all your life that we need only the
10 digits 0,1,
9 to write any integer, but can this be accomplished
in another way?
Have you ever noticed the remarkable fact that, of the
consecutive integers 8 and 9, one is a perfect cube and the other a perfect
square? Surely you have not overlooked the familiar right triangle
from
whose
sides
3, 4,
and
5.
Can you
Again, the
sum
of the positive
having this
Although for its size 6 has many divisors, you notice that
Observe how close the first few integers of the
5 has but 1 and 5.
Would you be interested in examining the law indicated
latter type are.
by the following equations?
divisors of 6
is
double
itself.
property?
5=1+1+3
7=1+1+5=1+3+3
9
+ 1+7
1+3 +
F. Cajori,
"A
10
1+3
6=1+5=3+3
8=1+7=3+5
3
History of Mathematics."
certainly will
some
want
and
why it
so, starting
is so.
That, indeed,
with elementary notions,
is
the attitude
we
we wish
up a
shall set
to foster,
basis for
show
We
is familiar with the concepts of numsum, difference, product, equality, greater than,
absolute value, etc. We shall assume that he understands the fundamentals of algebra and the derivative with respect to x of a rational
We shall also suppose
integral algebraic function of the real variable x.
that he is familiar with the content of the following system of postulates
which the integers obey:
shall
1.
The
+1
and at
Thus we have 1 + 1 = 2, 2 + 1 = 3,
We call all
these numbers positive integers, and 1, 2, 3,
are distinct.
Furthermore, if a + c = b and c is positive, we say that a < b (a is less
than b) or that b > a (b is greater than a)
The set of rational integers is
ordered by this relation "less than."
The fifth postulate says that subtraction exists, and we therefore write
the statement a + x = c in the form c a = x.
We shall show later
that we can derive the fact that there is a single integer
such that
= a for any integer a.
a +
6. The law of closure for multiplication: For each pair of integers a
and b, in that order, the product of a and b exists and is a uniquely determined integer c. Thus a b = c, or ab = c.
7. o( + l) = a for any integer a.
8. The commutative law for multiplication: ab = ba.
9. The associative law for multiplication: (ab)c = a(bc).
10. The distributive law: a(b + c) = ab + ac.
11. The cancellation law for multiplication: If ab = ac and a tA 0, then
obtained.
c.
12.
The
There
last
according to the relation "less than" and then separated into any two
parts without disturbing the array, there is a first integer in one part and
a last integer in the other.
>
Any
13.
fti
n2
14. If
If
15.
>
'
relation.
If
+1
positive integer
method
is
+ +
(a
+ +
+b+
+
+6+
We
based.
shall use
for
m)
c,
a+6+c+d=
+m+n=
n.
3.
Equals
integer
b,
transitive; that
is
The system
of postulates
The
b.
and
But c
this integer Oi
Oi
a.
fifth
is
postulate
shall
show that
exists
such that b
is
unique, for
if
Oi
us that corresponding to
+ a.
= (c + a) + Oi
= c + (a + Oi)
= c +a
=
This integer Oi
tells
an x such that a
b for
a.
any
We
shall call
b.
We
have
Hence,
b for
Since
6, it
if
6 is positive, 6
is
greater
Moreover,
and conversely.
zero,
and
shows that
+1
_f_
a) =0.
that subtraction
6i
+ a)
62
Since
and
a) +
since
(a)
it
(a
6i)
= (a)
+
+
(a
6 2)
a]
62
follows that
[(a)
a]
6i
[(a)
6i
&i
and that
b2
Hence,
6
tive two.
b.
Theorem
12.
Subtraction
is
Has
it
principles
a^0,
is
so.
Hence,
unique.
We
0*0 =
0.
for
any
that a
ba\b'0 = ba.
Therefore, b
statement includes
Theorem
60 =
We know
stated?
uniquely
tion
we have
However,
c.
and so when
b a = c and
= c, and subtraca,
But
0.
0,
this
Thus we have:
13.
For any integer 6, 6
have seen that c + ( c) =0.
0.
Hence, ac
+ a(c)
0.
But
ac
= ac.
prove
Theorem
14.
+ b).
(
+a
=
=
a
>
member
Theorem
If
15.
b,
= b,
so
= [a+(l)](l)
= a + 1
no integer between a and (a 1) that is, if two
integers are consecutive, as are a 1 and a, so are their negatives.
We
As a
result there
must conclude,
is
EXERCISES
a
a
=
=
7.
On
4.
Show that
Show that
Show that
Show that
5.
If
6.
1.
2.
3.
8.
b,
b.
b, it
c)
ab
ac.
The law
finite
number
when we apply
the opera
We know
into addition
product of a
result
When we
unique.
is
restrict ourselves
it is
number
finite
them always
of
But
exists.
and
any
b are
mean by
An
an
when a
we
it is
integer a
be.
The operation
no integer
If
division.
that a
b.
and therefore
integers,
not always
it is
if
is
we
c exists,
write b
a and
a and read
Jf
is divisible by
"b divides a."
it is
it,
is
read,
and when
of b;
0, c is
the quotient of a by
is
The
b.
complementary to
integer c
is
often
b.
We recall that
\a\
\a\
=
=
6(1
Since b
Theorem
Theorem
Since 6
\b\
>
if
positive or zero
is
18.
If b
19.
If 6
when
When
\b\r.
0, \a\
>
integers d
b,
(ce)d
is
it
0,
thus
evident
that
\b\.
ce.
and therefore c a.
divides a, and b divides c, then b divides a + c.
divides a, and b does not divide c, then b does not
c(ed),
c.
a,
then a
we should have a
b
If b divides a,
17.
a and
Hence, a
divide a
r)
and when
\b\,
Theorem
\a\
tive.
would divide
Theorem
\
= bd and a + c =
= be and 6e = 6c?
6c?
If 6
c.
did divide a
Hence, b(e
c.
d)
c,
c,
and
c.
110.
tive integers 0,
If 6 ?* 0, the difference
1, 2,
6
1 is
>
r2
mb >
b
0.
If ri
But
r2
mb, then
ri
= mb
<
r2
< \b\.
< \b\.)
or
common
divisor, of
two
or
Suppose that
Therefore,
\ (n  r 2 ).
A common factor,
(These state
r2
ri, r 2
more
n >
6,
for
and consequently
integers
is
an integer
greatest
common
divisor of a set of
two
or
more
8
exists,
is, if it
mon
common
divisor of a set of
one of
refer to the
common
greatest
by every com
The symbol d =
cti.
common
greatest
is
(a h
unchanged
a2
if
a r)
15), and 12
(6, 12,
divisor.
divisible
is
any a
is
0,2,
common
by
(36,48).
is
a divisor of every
It is evident that
then
is also.
integer that
common
fits
a ^
Again,
if
is
a least
it is
multiple
is
unchanged
by
integers.
integers,
if
any
common
multiple of
6,
The
multiple.
15, and 9
az
is
least
replaced
is 90.
Since
+1 and 1
Theorem
111.
The
rational integers
is
+1
The
or
1.
null element is
no integer
0,
itself.
itself.
We know
For any
a,
from
hoAv
different
and the
of
* This definition and that of a least common multiple are so worded that they will
apply equally well in a domain of algebraic integers where we cannot say of two disFor instance, the set of algebraic
tinct integers that one must be less than the other.
integers of the form a + bi with a and b rational integers is not ordered by the relation
"less than."
Two
or
more
common
to each,
or relatively
is +1.
The integers 3, 4, and 35 are relatively prime in pairs.
Theorem 113. If d is the greatest common divisor of a and b, the
quotients obtained by dividing them by d are relatively prime.
If d = (a, b), let a = a Q d and b = bod.
Let e be any common divisor
Then ao = ek, b = em, and a = dek, b = dem. Hence,
of a and b
de is a common divisor of a and b and must divide d; that is, d = des.
Therefore, es = 1.
But this is impossible unless e = +1. Conse
prime in pairs,
if
the greatest
quently, (o
14.
who
The
1.
Among
Principle of Archimedes.
Archimedes
of numbers.
(c.
225
B.C.).
(c.
300
B.C.),
and
little
1)6
< (m
l)b
mb < a < (m
and, for b
<
0,
mb <
Theorem
integers a
114.
and
The theorem
6^0,
Euclid.
of
= bm
<
m
<
Corresponding to
r such that
two
and
6
a
mb and a
mb < b. Thus r = a mb exists and fulfills the
required conditions. Suppose then that a bm\ + n and a = bm 2 + r 2
where
< r 1} r 2 < \b\. Hence, b{m m 2 ) = r 2 r h and b divides
,
10
r2
= r and
m =
Therefore, r\
r\.
2,
We shall
call r
Theorem
An
115.
integer a
b (mi
2)
and that mi
=
=
ra 2
= bm
6^0,
this
is
is
Since
0.
<
<
or
is
is
6^0
not prime to
b.
\b\
and hence an integer d divides a and b if and only if d divides both r and
For example, to find out whether or not 152 is prime to 21, just divide
b.
152 by 21, getting the remainder 5. Since (5, 21) = 1, then (152, 21) = 1.
Theorem 116. All integers take the form 2n or 2n + 1.
According to the theorem of Euclid, any integer a can be expressed in
the form
< r < 2
a = 2n + r
so that r
An
even
either
is
even integer
is
The
odd.
or
is
1.
one that
a multiple of
is
An
2.
integer that
is
not
EXERCISES
Prove that the product of any two consecutive integers is divisible by 2.
integers can be written in the form n(n + 1).
Euclid, n has the form 2k or 2k
1, whence
the product has the form 2k(2k + 1) or (2k + l)(2k + 2) = 2 (2k + 1)(& + 1). In
/
either case the product has the factor 2.
2. Show that the sum of an integer and its square is even.
*
3. Show that all integers take the form Sn, Sn + 1, or Sn 1.
4. Prove that the product of any three consecutive integers is divisible by 3.
5. Prove that the square of an odd integer has the form 8n + 1.
6. Prove by induction: 1 + 3 + 6 +
+ n(n + l)/2 = n(n + l)(n + 2)/6
1.
for
n >
0.
l)/2, is k(k
_o.
1J
1 + 3 +
Hence,
l)/2.
k(k
1)
k(k
l)(fc
We
we
shall build
cated
sum
up the
of the first
series of k
A;
terms.
The
+2)
sum
of the first k
terms of the
terms of
To accomplish
this end,
of the first k
formula.
terms by adding the (k + l)st term to the indi+ l)st term is obtained from the term formula
(k
by substituting k
one
member
n and
for
l)(k
But
2)/2.
to the other
it
Thus we have
equality.
,,,,
1+3+..+ k(k 2+
l)
is (k
we must add
of Eq. (1),
11
+ !)(* +2) +
6
(*
+ l)(k+2)
(k
+ l)(k+2)
6
+
k(k
and
l) (*+2)
+ !)(*
+2)
. n
(t
1WJ
ON
+ 1)(t+2)
.
(k
g+
.
l)(k+2)
2
(2)
simplified, giving
1\
l).
(Jfe
+ l)(*+2)(*+3)
sum
(k
(A;
of the first k
sum
for the
of k
fe
(fc
(A;
correct, for it
Assume that ak
bk
(a
ak+i
=
=
Now
b)F(a, b).
&fc +i
b k+1
(a
divide ak+1
b k+1
by a
b,
getting
Thus
+
+
h)a k
6)a*
a 6
fc
6(a
fc
 6 fc+1
 bk
)
shows that a
b n is divisible
b is a
by a
if
13.
^
or a
positive integer.
If
0,
then
+3(16)
+5 = 54
+ l)st
+3
+5 
4+ 3
is
(10"
4"+ 2
clearly divisible
by
9.
5)
true for
7 2n
b.
7n
3 4 "+ 2
c.
15.
true for
n =
43
1.
4 +2 (3)
=
5
and applying
is
a multiple of
1, it is
likewise
Prove that:
+ 16n
14.
a.
two
all
is
+3
Substituting n
10*(9)
1 is
divisible
by 64
for
n >
0.
0.
is
integers.
To
we
often set
of
12
discovering the law
by
Thus
inspection.
=
=
=
=
23
33
43
p _
Q2
32
32
62
10*
l2
62
V n(n
Show
Show
Show
16.
17.
18.
or 5n
1) 1
that
if
that
if
7z
2x
that
if
an integer
(n
l)rc
n 2 (n 2
a multiple of
is
2^i
n2
2w
i;
3,
5,
is
not a multiple of 2 or 3
is
of the
form
1.
20.
a multiple of
is
1 is
4.
Show
19.
24A;
square, n 2
through In
is
a perfect
21. Prove that every odd cube, n 3 is the sum of n consecutive odd integers.
Find
a corresponding law for an even cube.
22. We have assumed the principle of Archimedes, but just as some of our other
postulates are not independent, it is possible to prove the principle of Archimedes on
the basis of the assumptions already made.
= 0. If a = b,
and 6 > 0. If a < b, then
< a < b and
Case 1. Let a >
= 1, If a > b, there is a positive multiple of b that is less
then b = a < 26 and
,
than
But
a.
<
Thus
(a
there
<
(a
>
ab
>
1)6.
1)6
Let a
2.
1)6
is
>
greater than
a.
Then
mb < a <
Case
6,
is
be (m
<
and
>
(in
1)6
Then nonnegative
0.
cb
<
\a\
<
(c
c exists so
that
1)6
and
cb > a
If
> (c 
1)6
include
all
26. If
n(n
odd primes.
is
l)/2.
Find by
trial
some
and
triangular.
13
What
lack?
Peano stated the following postulates, together with the principle of finite
numbers
a. There is a number 1.
+
b. Every number n has a unique successor n
c. The number 1 is not the successor of any number.
d. If n + = m + then n = m.
Define addition and multiplication, and derive the commutative, associative, and
distributive laws for the natural numbers on the basis of these postulates.
28.
CHAPTER
The Form ax
by.
A polynomial in the variables Xi, x 2
xT
a rational integral algebraic expression in these variables.
Thus
~
a n where the exponents of the variable x are
a Qx n
aix n l
21.
is
bixy
bzxy
is
is
a polynomial in
a polynomial in x and
The
y.
x.
The
expression b\X 2
degree of a polynomial
is
the
degree in all its variables of its term or terms of highest degree. The first
polynomial is of degree n, and the second is of the fourth degree. The
second, however, is of only the second degree in x, but it is of the third
degree in y. If all the coefficients of a polynomial are integers, it is said
to be an integral polynomial.
x r is a homogeneous polynomial
A form* in the variables x h x 2
in these variables; that is, each term of the polynomial is of the same
The degree of a form is the degree in all its variables of any term
degree.
The polynomial Sx 2 y + 5xy 2 y z is a form of the third
of the form.
All the forms with which we shall be concerned will have
degree.
We shall make use of the form ax + by to show
integers as coefficients.
that the greatest common divisor of two rational integers (not both zero)
,
exists
and
is
a rational integer.
Theorem 21. The least positive integer in the set of integers defined
by ax + by, where a and b are not both zero, is the greatest common
divisor of the set.
by when
Consider the set of integers defined by the linear form ax
a and b are constants and x and y are variables whose values are all the
Since there is but a finite number of integers between zero and
integers.
by contains a positive integer,
any positive integer, and since the set ax
this set
Let
L =
This integer
ax
it
be represented by
by
This technical use of the word "form" is not to be confused with the ordinary
we have made use of the term. When we say, for instance, that an
integer has the form Qk + 1, the word is synonymous with "mold" or "structure"
and in this case designates that the given integer is always a multiple of six, plus one.
Whenever "form " is used to mean a homogeneous polynomial, the implication will be
*
sense in which
clear
from the
text.
14
theorem
and
n = ax
by\
and L
such that
n =
15
mh +
<
<L
Hence,
ax\
byi
a(zi
mxo)
= m(ax
by
m?/
and
Therefore, r
is
an integer
6(?/i
than L,
must be
0.
The
least positive integer of the set thus divides every integer of the set
and
is
necessarily a
greatest
common
any common
therefore
common
But L
is
in the set,
Hence,
is
and
the
the stated
definition
set
ax
f
by.
The
integers a
the greatest
If
(a,
common
b),
common
Theorem
by
by
divisor of a
ax
d,
is
The
and
and
b d,
+
b,
we
b y).
On
for both a
d.
by.
see that d
is
common
and
b is a divisor of d.
by necessarily divides
common
24.
d(a x
so that a
The
Thus a common
integer d
is,
there
greatest
common
unique.
Suppose that both d\ and d 2 are greatest common divisors of the set
+ by and that they are positive. Then, according to the definition,
Consequently, d\ < d 2 and d 2 < d\, so that di = d 2
d\ di and d 2 d\.
then,
that the greatest common divisor of a and b is the least
It follows,
positive integer in the set ax + by and that d = (a, b) can be expressed
as a linear function of a and b with integral coefficients.
Thus 4 =
=
=
12(18) + 20(11).
(12, 20) can be written 4
12(2) + 20(l)
The fact that the greatest common divisor of any two rational integers
a and b, where not both are zero, can be written in the form ax + by with
x and y rational integers is an important characteristic of the set of
ax
16
Not
rational integers.
by
a(kx
b(ky
).
Therefore,
by ax
all
of
d are
mem
by.
the associative and commutative laws for addition and if the equation
a + b = c is satisfied by an element of the set whenever two of the three
elements a, b, c are chosen from the set. It is now apparent that the
integers ax + by form a modul.
EXERCISES
Describe the set of integers 3z + 6y.
Use the form ax + by to define a set of integers all of which are even.
3. Use the form ax + by to define a set of integers that are multiples of 5.
Are
all multiples of 5 included?
4. Use the linear form in two variables to determine a set of even integers that are
multiples of 5.
by include all the integers?
5. When will the set ax
1.
2.
22.
of note
by
n.
Diophantus
who devoted
(c. 275)
himself to algebra.
and twodollar
bills
Theorem
paid in fivedollar
bills?
The
26.
integers.
17
linear
Diophantine equation ax
\
by
= n
has a
ax
if
by
1.
if (a, b) = d and ax
by = n
then this very pair satisfies a Q x
b y = n'
where a = a Q d, b = b d, n = n'd. Again, if (a, b) = 1, the equation
kby =
by = n has a solution and that same solution holds for kax
ax
The problem of solving the equation ax
by = n is, therefore,
kn.
reduced to finding a general method for solving the equation in the case
where the coefficients of the variables are relatively prime.
Theorem 27. If m divides ab, and m and a are relatively prime, then
has a solution x
m divides
b.
Since (m, a)
y
1,
the equation
mx
ay
Xi,
(m, a)
has a solution x
Then
z/i.
b(mxi
ayi)
mbxi
abyi
and
But
mb, and
m\
ab.
it is
necessary that a
sequently, 6
0,
Therefore,
= +1.
But
b.
when
0, for
then
ab implies that ab
if
0.
Con
and m\b.
need not be relatively prime; for example, 6 \ 15, but (6, 15) = 3.
Corollary 1.
If m is prime to both a and b, then m is prime to ab.
Corollary 2.
If m is prime to a h a 2
, a^, then m is prime to their
,
product.
EXERCISES
*
1. Prove that the product of three consecutive integers is divisible by 6.
Let P = n(n + l)(n + 2). Then P is divisible by both 2 and 3. Hence, P
= 3m. But (2, 3) = 1, and 2 3m. Therefore, 2 m. Thus m = 2s, and

3 (2s)
6s.
2k
P =
18
4.
6.
Prove that
2.
3.
A number
rational
is
4.
V2 is irrational.
if
and only
if it
27,
7.
When m is
= +d
9.
=
=
If (a, b)
2.
Show
integer,
b)
1,
if (a, 6)
1,
6,
the greatest
6) is
common
common
2.
62
and a
divisor of a 3
+
+
6 is either 1 or 2.
63
and a 2
6 2 is
6.
The
common
greatest
for
Xi,
1,2,
a 2 2
.
"
nX
n, are integers.
oi,
Then
(a h a 2 ).
exist so that
Any common
divisor of a h a 2
a divisor of
divisor of a 2
the greatest
aiyi
of di, a 2
c.
\
(ac, be)
(a
the greatest
ai^i
where the
then
0,
1,
28.
>
(a
Theorem
and
that
a divisor of a
and
irrational.
an
Thus d =
11. If (a, 6)
?/i
is
Notice that (a
(2a, 26)
0i, as,
it is
or d.
10. If (a, 6)
12.
sum
the
8. If
CL2V2
di
three integers.
Since integers
Z\
is
the greatest
and
z 2 exist so
common
divisor of these
that
diZi
a 32 2
d2
a 2 y 2 )zi
a 3z 2
d2
a 3 3
d2
we have
(i2/i
and
finally
diXx
a 2 2
raise the
number
set to n.
common
divisor of d\
and
a^+i.
Theorem
29.
If
19
axo
by
= n
ax'
by'
xo)
and
so that
a(x'
b(y
and
a{x'
b{y'
a/
xo, we obtain
akb = b(y f y )
Since
(a,
b)
that
1,
That the
all
and
kb
for
integers x
we find
that
y'
ka.
If b
0,
clearly trivial.
is
kb,
+ by = n for
by substituting
ka satisfy ax
them
Xo)
x'
A Method
23.
by
n.
If (a, b)
and d\n, we have shown that, by dividing each term of the equation
ax + by = n by d, we obtain an equivalent equation, that is, one which
is satisfied by all and only the solutions of the original equation.
Consequently, it will be sufficient to solve the equation ax + by = n when
(a, b)
If
1.
n, so that
n n
a, let
and x = n
is
satisfied.
If
Jf
n,
then a
1 and we may
5*
suppose that
qia
+n
<
ri
<
a
w =
#20
7*2
<
r2
<
a
(oia
q 2a
r2
by
= n
exists, riy
<
\a\
<
\b\.
Then
and
Therefore,
ax
Since a solution oi ax
ri?/
az
r2
by substituting
T\
Jf
r 2,
If
\
r2
choose
ri)i/
0,
r% is
and y
is
divisor since r\
<
z as
Therefore, where
\a\.
qsri
r3
<
rz
<n
But
if
as the
20
and
#471
7*4
ny =
q*ri
r4
T2
<
<
r4
ri
we have
(gs^i
+ r )
3
and, as above,
r 3z
riw
r4
r 2s _i
2fc
fc
substitution.
23a;
S7y
Example.
Since
Solution.
111)
(69,
3,
we
14)y
(130) (23)
solve
Illy
9000.
the equivalent
equation
Thus
3000.
2Sx
(23
10
Hence,
Uy +
2Sz
10
9)2
10
9z
+ Uw
10
(9
9v
4)t>
5s
1)
IHl
'
and
14t/
(14
so that
Again,
9z
S)w
and
bw
Therefore,
bw
(5
and
4^
Finally,
4z;
so that
(4
a^
1, w =
+
j
= 2,
l/
and
124.
Therefore,
quently,
all
2,
x
y
=
=
124
4
?/
37k
23/c
4,
a;
Conse
21
124
37/c
>
23/c
>
v
and
so that
4
124
0.
EXERCISES
number
of solutions for
y are positive.
16x + 7y = 601.
2. Ux  45y = 11.
75z + 91y = 320.
4. 56x  50y = 74.
40x  63?/ = 135.
6. 123a; + 57y = 393.
77z + 165?/ = 3553.
Separate 1591 into two parts such that one part is a multiple of 23 and the other
1.
3.
5.
7.
8.
a multiple of 34.
More than
Diophantine equation in more than two
variables or a simultaneous set of such equations can be solved very
24.
Two
The
Variables.
single linear
solutions.
a 2 a 3)
,
d,
there
by
d.
is
a 2y
a zz
= m
no solution unless d
we should
first
(1)
Supposing that d m,
of the equation
m.
divide each
member
(a h a 2 ), it is necessary
atf
H.
New
diw,
dit
t
But
(a 3 , di)
a z w, where
1,
Then
1.
if
d\
m in order
=
1,
that
and therefore
to is
one
22
Hence,
solution.
gives
all
Now
diw, for
is
a multiple of
d\.
=
=
x
y
We,
all
which a 3 z
of z for
of course,
wish u and
+bv
+cv
biu
c\U
(2)
to be integers,
c 2x
bic 2
biy
bic 2
b 2y
b 2 c\
Cix
b 2 Ci
bic 2
b 2 Ci
61, b 2 , Ci, c 2
that
makes
will force
u and
Applying
stituting z
v to
ai(biu
b 2 v)
a 2 (ciu
c 2 v)
and sub
is
a 3 (^o
diw)
= m
or
(01&1
a 2 Ci)u
(ai&2
a 2 c 2 )v
a 3 diw
dit
(3)
since
a dz
= m
^1^0
v,
we
are per
mitted to impose a second condition upon the coefficients of the transformation. We, therefore, set
aib 2
Since
(a,i,
a 2)
di, let a\
doidi,
dQidib 2
a>iC 2
a2
aoarfi,
= a 02 dic
and then
or
01?>2
Let
us, therefore,
6iC 2
6 2 Ci
choose
b2
CLo 2 C 2
= a 02 and
c2
02Ci
condition
this equation
if
Then the
of values so
Now
floi
becomes
aoi&i
and
a x bi
a%c x
= d
d\U
(3)
a^diw
23
dito
and
u =
which
Eqs.
a>zw
is
(2) for
of the
to
x and
y,
we
Then by
t.
eliminating u from
form
x
y
z
=
=
=
bit
a 02 v
crfo
a 01 v
20
+ dzbiw
+ a ciw
3
diw
where v and w are the parameters. That all these values of x, y, and z
determined by integral values of the parameters satisfy the equation is
easily verified.
It
is
now be made
=
=
= t, it is easy to compute the answer in the
1.
Since
C\
u
bi
3,
form x = 3(22  41ti>)  4v = 66 + 123w  4v, y = 22  41w +
variables can
variables in the
v,
25.
to
same manner
as
Qw.
Simultaneous Linear Diophantine Equations. A set of two equamay or may not have a solution in integers. Con
aix
bix
If (ai,
2,
a2
there
CJ3)
is,
+ay+az =
+by+bz =
2
mi
ra 2
mi
d 2 fails to divide
no solution for the set. But even when these conthere need not be a common solution.
Take, for
or
of course,
2x
2x
If
is
+

When
together with an
4y
+z=
+ 3z =
Sy
(4, 2)
2,
2s
first,
the result
is
= 
x, satisfy
24
common
exist,
the
solutions,
they
if
equation
aix
a 2y
ri
=
=
= mi
a 3z
(5)
are
with parameters
&2?/
>32
= m we
2,
r2
r3
+
+
+
(5),
+
+
t\W
t2
(6)
s 3^
common
Wo
v
in bix
Bv = C
solution
solvable.
w =
v =
When
s 2v
Aw +
tion
SiV
\
exist
if
Bit
Ait
first
equation
the
x
y
z
where there
is
Example.
= X
= Y
= Z
+ Kit
+K
+ K^
2t
t.
2x
24?/
Sy
412
7z
91,
2.
first
lift)
(77
 4(77 +
=
327/c)
= 4 + 45/c; y =
124/c; 2 = 1 6(2 +
327/c)
27
 41(2 + 11/c) +
11/c) = 13  66/c.
22
EXERCISES
1.
Solve: 2x
3.
Solve: 10a;
5.
7.
A room has
to realize exactly
the children,
8.
$100,
If
5y
16?/
3s
17.
2.
4z
4.
2y
4y
7x 100 seats.
$10
if
the
Solve: 3z
6y
2z
11.
= 521.
= 22,
= 11,
6. Solve the set: x + 3y  hz = 29.
5a:  5y  3z = 33.
How many men, women, and children should be admitted
men will pay 50 cents each; the women, 20 cents each; and
48.
Solve: 127a:
319?/
43z
7z
cent each?
100 pieces of
how many
money
of each
in denominations of 50 cents,
denomination must there be?
$5,
to
amount
to
CHAPTER
PROPERTIES OF INTEGERS
31.
The Composite.
most familiar
proved it?
is
Theorem
31.
Every composite has a prime factor.
Because any negative integer can be expressed as the product of its
positive associate and the unit 1, we shall assume that the composite m
Since m is neither a unit nor a prime, it has a factor that is
is positive.
not a unit or an associate of m. Therefore, let m = /i/ 2 where both
Then /i < m, and /i is either a prime or another
factors are positive.
composite.
If /i is a prime, the theorem is complete, but if /i is a comThen
posite, it, in turn, has a factor other than an associate or a unit.
< /3 < /i. If /3 is not a prime, the line of reasoning
/i = fzfA, where
continues in the above manner, but only for a finite number of steps, for
,
since
we must
ciates
way
>
arrive at a positive factor f<m\ that is divisible only by its assoThe integer 2n i is, therefore, a prime, and by subunits.
and the
stitution
32.
it is
The Sieve
of Eratosthenes
(c.
230
B.C.).
It is evident that
is
one
a prime would be to
write all the integers from 1 through m; then to leave 2, and strike out
every second integer thereafter; next to leave 3, and strike out every third
integer thereafter; generally, to leave the next unstruck integer p, and
Each
integer except
that
is
not
by
Theorem
32.
where I
is
than or equal to m.
Suppose that m is not a prime but is a composite. Then m has a prime
This prime factor p must be greater than / according to the
factor.
is less
25
26
that
m >
(I
l) 2
(2
p)
But
if
If
p,
5,
p, or
N
N
is
a prime,
is
a composite,
of the
it
it is
greater than
would divide
1,
which
is
primes
Consequently,
it is
is infinite.
'
'
to 10,006,721,"
PROPERTIES OF INTEGERS
Theorem
27
34.
Suppose the
(4
p)
AT"
is
a prime,
it is
1)(4
1,
1)
4/c
l)(At
1)
Ak
28
Theorem
If 2 n
35.
a prime, n
1 is
is
a prime.
Suppose that n
is
when p =
for
11,
2 11
EXERCISES
1.
6n
Show
that
all
2.
Show
3.
Prove that n 4
4.
5.
primes except 2,
are represented
by the forms 6n
and
1.
that there
is
an
is
number
infinite
of
composite when n
>
1.
1.
We
Unique Factorization.
34.
when
m is a prime this
ing manner:
Theorem
36.
If
is
b,
then p does
Take
<
p,
>
ka
is
<
<
(k
l)a
omitted because p
<
<
(p
ka
<
is
a prime.
Therefore
and
But
if
p divides
ab,
ka)b
<
ab
that there
the
a
b
= mip
= mp
2
+
+
ri
r2
<
ri,
r2
< p
PROPERTIES OF INTEGERS
29
Thus
= Kp
ab
rir 2
if p
than
less
We know
p.
nr 2 where both
r\
t,
{.
m =
PiPi
'
Pn
'
m=
qiq 2
a second factorization of
is
r,
>
where n
P1P2
pn
= 1, 2,
r.
+ 1 and qi, and
primes qh where j
divisors of qi are
Pi
'
Then
r.
qiq2
'
qr
Hence,
q\.
P2PZ
Pn
qiqz
'
'
qr
finally
have
Pr+l
and each
of the integers
'
Pn
p n must be a
Therefore, the
primes is unique.
It is obvious, then, that if we gather the equal primes together into the
power form p s any composite m can be written in the form pi n ^p 2 n2
n
p r *, where the factorization is unique except for the use of an associate
in the place of any prime and the order of the factors.
p r+i,
unit.
m into positive
well
remember that an
all
algebraic factorization of an
integers represented
30
these integers.
one
way
besides itself
expression
may have
and
1,
several factorizations.
It is
the expression a 2
value a = 5, a 2
represented by a 2
may have
by the
2
1 of a 1.
For instance, when
only rational factors a 1 and a
2
=
=
1
24 and this integer has the factors 3 and 8, 2 and 12, as
a
5, a
well as 4 and 6.
Hence, we must be careful to avoid drawing the conclu
sion that an integer lacks factors just because the algebraic expression of
which
it is
a value
fails
EXERCISES
Show that if both x and y are odd, there is no z such that x 2 + y 2 = z 2
Show that if {a, b) = 1 and ab = c n then a = s n b = t n
3. Show that an integer can be represented as a difference of two squares if and only
Show also that the representation is unique when
it is of the form 2n + 1 or 4n.
1.
2.
if
is a prime.
Find the positive integers x that make x(x
Find the positive integers x that make x(x
the integer
4.
5.
Method
+ 42)
+ 84)
a perfect square.
a perfect square.
of Infinite Descent.
* L.
Numbers,"
Vol. 2, p. 228.
PROPERTIES OF INTEGERS
31
is
more
of particular interest.
facility to
he claimed, in the
He admitted that
is
are the sides of a right triangle, the area cannot be a perfect square.
a, b, c
divisor of
mon
two
divisor
is
integers a
and
b.
If
one integer
is zero,
same
as
it is
common
divisor
therefore, with a
the greatest
>
we
b,
common
shall set
divisor of
is
if
the
b positive,
for finding
= rriib + r\
= ra ri + r 2
= m r2 + r3
a
b
ri
Because
>
r\
>
r2
Then
r k is
rrtiTii
+ Ti
= m k r k i +
= m k+x r k
r k !
Ti2
r k 2
<n < b
<r < n
< r3 < r
>
we must
the greatest
>
<
<
rk
r*
>
Ti
rk
0, it is
<
<
Tii
r*_i
common
divisor of a
and
b,
is
a divisor
for first of
all
rk
* Ibid., p. 233.
32
terms of a and
in
b,
for
mj)
h so that r 2
ri
Moreover,
2 a.
r2
2r
ra 2 (a
raib)
+ mim
(1
2 )b
+Mb
+Nb
= Mia
n = N a
Tii
we
ui
mi+iTi
r l+1
Qid
Q 2b
then
common
divisor of a
By
en
vr^
=1
common
divisor
r
s
of the di is
3
Pj in
pj
] [
>,
where each
a;.
Example.
573
Sj is
=1
291
282,
291
common
common
282
divisor
9,
divisor of 573
282
(31) (9)
and 291.
3,
(3) (3)
is 3.
EXERCISES
Find the greatest common divisor of 5040 and 4704.
Express the greatest common divisor of 168 and 525 as a linear function of these
numbers.
ba
3. If d = (a, b), then d is the number of integers in the sequence a, 2a, 3a,
Prove it.
that are divisible by b.
4. Show that the sum of a finite number of rational fractions in their lowest terms
cannot be an integer if the denominators are prime each to each.
2"
5. Prove that two integers having the form 2
+ 1 are relatively prime. From
1.
2.
He
many
primes.
(Integers of
PROPERTIES OF INTEGERS
33
Prove that the number of divisions required to find the greatest common divisor
two positive integers written in the scale of 10 by means of the Euclidean algorithm
6.
of
five
integers a
and
and
that a
b so
a Q d, b
the least
b d,
common
multiple of a
b is a b d.
Any common
But
form ma.
because (a
multiple of a and b
it
is
also a multiple of b
is
follows that b
1, it
b d ma d, and
common multiple
Thus
b d.
Hence, any
m.
common
multiple
Corollary
multiple
Corollary
common
W
if
integer
and only
is
common
the least
310.
If
common
and
the least
common
,
the least
is
is
the least
that
is
it
is
a multiple of a 3
multiple of
divides
all
it is
is
an
cini,
is
their
common
the least
a multiple
and a 3 is a multiple of L.
oi, a 2 and a 3 for it is a com
of a h a 2
common
multiples.
of integers
the least
Corollary.
an
multiple
then
common
L 12
is
common
a2
Then L is a common
of L 12 and a 3
But any common multiple of a and a 2 is a
common multiple L i2 For the same reason, if
common multiple
Therefore,
multiple of a\ and a 2
multiple of a h a 2 and a 3
ai,
is
a%.
multiple of a 1} a 2 and a 3
mon
if it is
multiple of them.
the least
of L.
common
An
2.
and
Theorem
of L12
is
If
1.
is
integers a
2,
a n _i.
integers a h a 2
is
a factor.
It is
]} p/
3
ni
>,
where
1,
2,
n, the
=1
r
least
common
multiple of the a {
is
J J p/**,
where Gj
is
.
34
EXERCISES
1.
If
where a and
(a, b),
show that ab
b are positive,
n,
by 2 k
divisible
is
is
is
a multiple of 2
is
fc
,
6.
Show that
common
the greatest
is
mon
di 2
The Divisors
d<,i, d*,*+i,
we mean
and we designate
the
by r(ra).
it
di r
m_i,
least
com
an Integer.
of
where
di },
divisors of an integer m,
integer
multiple of da,
38.
use the symbol <r(m) to represent the sum of the positive divisors of m.
It is evident that the number of divisors of a prime p is 2 and the sum
is just p + 1 if p > 0.
Moreover, the divisors of p a are 1, p, p 2
2
a + 1, and their sum is 1 + P + P +
of the divisors
is
Theorem
of
312.
m is (a\ +
<m
If
1)(2
Pl
1)
i+i
Pi
'
ai
pi p2
(r
p2
P2
pr
ar
r+i
Pr
their
number
a

the
Pr
and so
1)
2 +i
az
number
of divisors
of the divisors of
m is
If
m =
J Pi
ai
,
it is
i = l
of p\ ai is a
term
in the expression
Pi
Pi
P2
P2
'
of
'
both
Pi
ai
m and pi
(1)
ai .
In like man
(2)
P2">
If we multiply these
and only the divisors common to m and p 2 a2
of
the
result
is
a divisor of pi ai p 2 a2
terms
the
each
of
together,
two sums
and furthermore these terms give all the common divisors of pi ai p 2 a and
This product of (1) and (2) is
m.
give
all
PROPERTIES OF INTEGERS
1
Pi
Pi
'
'
Pi"
Vi
P1P2
Pi
35
+ ' +
+
P2 +
2>2
Pi
ai
ai
'
'
Pi p2
a*
and
(1
give
Pi
all
'
and only
Pl 0(l
tt
P2
'
'
'
P2
'
(1
+ Vt +
1)
'
'
Pr')
(3)
There are
divisors of m.
r(ro)
terms in this
The sum
(3)
and
is,
(i
l)(a 2
1)
(,
result,
of the divisors of
m is the
the
is
sum
11 (*
ti
!)
number
of the
of divisors of m.
terms in the same product
therefore,
r
<r(m)
= \\
i
(1
+P +
t
Pf) = J! g
1
^~T
EXERCISES
r
1.
p.(i+D
n
i
P<
=l
2.
Since 6
=61 =23 =
the required
number
smaller than 2 5
(ai
l)(a 2
5,
l),
m =
JJ
i=l
p*"** is
divisors.
or the pair
1, 2.
Evidently 2 2
12
is
4.
ri
perfect square.
Find all primes that are one less than a perfect square. Is there a prime that
than a perfect cube? Can you find a prime that is one less than n 4 ? Prove
a general statement to cover these results.
9. Find by trial positive integers n such that the sum of the divisors of n is a
8.
is
one
less
multiple of n.
Prove that a positive integer is the sum of consecutive positive integers if and
not a power of 2.
11. Prove that the number of divisors of a positive integer is odd or even according
as the integer is or is not a square.
10.
only
if it is
36
12.
the
number
n 8/2 where
,
s is
Prove that if r is the number of distinct prime factors of n > 0, the number of
which n can be factored into two relatively prime factors is 2 r_1
13.
ways
is
of divisors of n.
in
Examine the
Numbers.
39. Perfect
sum is
divisors of
12.
'
divisors
'
'
is
(1
2^ 1 )(l
2P
1)
2^(2^
1)
2 k+i q
But 2 k+l
1 is
2 *+i
l)(g
+ s)
+
2 k+l divides q
Upon
so that
2 k+l n
first
(2* +1
s,
equation,
we obtain
l)n
find that s
n.
k+l
is
a divisor of
Now
let
q.
us suppose that
1
Since q = 2 k+l 1, the exponent k
even perfect number has the form 2 P_1 (2 P
is
a prime.
1), in
Therefore, every
which both 2 P
and
p are primes.
*
A. Brauer, Bull.
hard,
1943.
H. A. Bern
PROPERTIES OF INTEGERS
37
with p a prime
after the
numbers has 77
that 2 P
is
more
quently, five
prime for p
integers are
now known
and 2281.
to be perfect.
Conse
It is particu
divisors
is
is
The
a multiple of n.
discovered.
Two
if
their
sum
is
sum
the
of the
The
Theorem
form
m =
314.
+
<
Any
a\r n
~l
<
positive integer
+
r
and
+
<
a$
<
r for i
1, 2,
n.
38
When m
rn
is
air"
+
1
\
a n with the
is
scale of r
m =
qir
\
an
qi
qtf
a n _i
ft
ft fir
=
=
ftil
?n
<
where
>
<
a*
arrive at q n
r,
for
which
1, 2,
qn  2
m =
This representation
rn
air 71
<
bj
<
"1
and
r(a r n1
is
n s
tion
is
>
0,
n.
less
+a
+ ar +
a Qr
a xrn
a r2
an
a n i
a,
and
a n j
and then
aorn
Since r
=
=
0, 1,
s,
>
Ol
unique, for
rn
Therefore, r divides b s
we
a n i
Then
0.
g n _i
with
ft^r
and
positive
is
_ s _!
b r
1
b r8
+ a
vl'^
and
bs
b^ '
8
++?>
then
s,
s 1
>
if
until,
a2
an
if
n =
(ln
fe 8
_i)
bs
an
_s _ i
all
be
unique.
in just
scale of 2.
Corollary.
one
way
as a
scales, for
127
is
2.
PROPERTIES OF INTEGERS
39
doubling.
4
8
16
32
Then
7584
all
237
474
948
1896
3792
7584
948
1896
10,665.
41
But 328
or 539
164
13(41)
+ 41 =
+ 6.
533,
82
164
328
4(41)
1(41)
6,
The digital idea is the basis for many of our computing machines.
Indeed even the abacus, which was probably the earliest mathematical
machine, makes use of it. Each rod of the abacus corresponds to the
place a digit occupies
is
Corresponding to the
machines use the contrivance of a
some
of the calculating
of a rotation.
40
needed
The
in
and
1,
are
each position.
calculator at
type.*
EXERCISES
1. Prove that any positive integer can be expressed uniquely as a sum of distinct
powers of 3 with coefficient 1, 0, or +1. From this representation show that a set
of five weights is sufficient to weigh any load of at most 121 lb if a balance scale having
two pans is used.
Then write 42 and 352 in the
2. Set up multiplication tables for the scale of 5.
Check the answer by
scale of 5, and find their product when they are so written.
311.
positive integer,
definition
<
<
is
a.
Theorem
is
of a
Prime That
315.
U? =
^~?
and
2,
Let
a and
(3,
where n
a,
= 3.
b greater
[[]'
_
than
0,
hi
lab]
"
&
so thaib
T)_
n = aa
a = 3b
/
+
+
<
<
"i
r 2
7*1
r2
<
<
a
b
Therefore,
n =
Mathematical Machines,
Sci.
ar 2
If
<
n.
= aa
Thus
Is a Factor of n\.
is
This
with
PROPERTIES OF INTEGERS
41
and
[sl'+N*]
However, r 2 is at most b
at most a(jb 1) + a
is
Corollary
If
1.
1,
and
r is at
a&
2,
If
2.
>
Corollary 3.
If ra, n,
Corollary
If
Letting
t,
4.
>
and
n = wi
and thus ar 2
1,
a2
s+,\
1,
then
aa
*)
where the n^
t ,
>[?] +
[>]
i;
(ai
1,
positive,
n2
= a we have n; =
n =
>
and a are
a positive prime,
is
p<
Corollary
most a
Therefore,
1.
ri
< n <
r2
+ [si
+
[
n with
for
Therefore,
a.
'
'
and
ai
a2
OL t
Hence,
highest power
of the highest
EJnl)
power
of
K+
p that divides n\
[?]

is
\M
42
The
p,
=E
is
p,
p"
(4),
ri
thereby obtaining the factor pL^l.
integer of the
Lp
We
P
o
2
new
set
set.
coeffi
All other
Therefore,
1, 2,
[;])
can, as before,
that
Likewise,
<
we remove
p
s+1
Corollary.
Theorem
n = a Qp s
is
a multiple of p
is
,? J from
remove the factor pLp
2
new
showing that
set,
and the
(4)
j] p)
.o[;]+*.0>
take out one factor p from each of these multiples of p that are in
the set
Hence,
(v2p
multiples of p in this
E p (n\)
V\
by p
cient of p
Now
2p,
ri
the factors pL? 
If
n =
317.
If
{ +
riJ
pLp
4
so that
^
\
#P
>
0,
while
a&,
((a6)I)
is
[?])
.
^
> a# p
0.
until
we
find that
Therefore,
(6!).
a s then
,
>
di
PROPERTIES OF INTEGERS
= ap
1, 2,
Because n
a,i
<
for
n
Lpj
n
aip s
.
~l
43
<
a s with
<
<
p and
s,
a p s_1
aip s_2
a ps
~2
ai7?
s_3
as _ 2p
as_ 2
a s _i
fe]Therefore
a,_i
or
tf,(n!)
a ps
aip s
~1
'
'
(27)5
We may
scale
(138
of
a s _ip
(a
ai
a s _ip
as
3;
1.
[?]
Hence
we have
6)/4
Oi
 a s i
a 8 i
 as
a.)
of the highest
27; 27
5 (138!)
53

(5)5
27
5,
ao
Example.
+ 0;
Pp
va ps
138
0(5 2 )
2(5)
power
^J =
2;
E p (n\).
of 5 in 138!.
(1)5
33.
Therefore
3.
5; 5
J,(138!)
33.
t
Theorem
318.
The
expression n!/ai!a 2
t \,
where
a{
>
n,
i=i
is
an
integer.
at
is
is
at least
w*  [=] + [>] +
+ [] +
+
Lp
(5)
44
Since a\
therefore
+
^
2
+ <H = n,
for
if
1, 2,
sum
in the
it
*;'""; [?] +
m <
for
exceeds n,
may
1,
.fe]
.;;;
exceeds each
and
a*,
+ ...+[p]
fe]
are also
it
Hence,

*^[] + [p] +
where, of course, some
s+1
be
0, in
which case
all
Likewise,
0.
+
.7.'.
[r]
L?J  L^J
As a
result of
summing by columns
(5)
by
divisible
Yv
'
the expressions
+ E p (a
ai \)
is
The product
1.
'
'
the
(6) for
EP (ai\), we
that
E p (n\) > E p
Corollary
l_^j
of
2 \)
+ E p (a
\)
an integer.
any n consecutive positive integers
is
n\.
(k + n  l)/n\ = (k + n  1) !/
+ l)(fc + 2)
and therefore is an integer.
This corollary shows that the coefficients in the expansion of (a + b) n
with n a positive integer are themselves integers, a fact otherwise known
from the multiplication itself. As a matter of fact Theorem 318 proves
that the coefficients in the expansion of (6i + b 2 +
+ b r ) n for n > 0,
obtained by means of the multinomial theorem are integers, for any term
of the expansion takes the form
The
(k
expression k(k
1) \n\
ai!a 2
where a
a2
'
'
In particular
it is
Corollary
If
b r) p
2.
'
now
is
ar
ar
bi
ai
b 2a *
'
b ra 
n.
evident that:
a positive prime,
multiples of p.
all
the coefficients of
&* p ,
where
1, 2,
(6i
.
b2
r,
are
PROPERTIES OF INTEGERS
45
with d\ + a 2
ar p, is
an integer, and furthermore since each a, either is positive and less than
p or is 0, there is no factor p in the denominator. Hence, (p 1)!/
The
atlas!
expression p\/ai\a 2
ar
is
ar
\,
coefficient is
Mp.
EXERCISES
If aP
1.
by p 2
2.
3.
show
bp
where p
is
a positive prime,
is
divisible
by
p,
prove that
it is
divisible
Prove that
If
re
is
if
m,
n,
number and
a real
integers,
[x] is
is less
than or equal to
x>
that:
a.
[(1
b.
[(1
+
+
+
\/o>]
+
+1
is
by
divisible
c.
[(3
4.
5.
is [4 n
is
even or odd.
2".
(2
\/2)
n
]
divisible
by 112?
is
inn!?
Applying Theorem 317, let (n l)/2 = 200, for the sum of the coefficients of the
powers of 3 necessary to express n in the scale of 3 is at least 1. Then n is at least 401.
However, 402 has the factor 3,
But the highest power of 3 contained in 401! is 3 196
so that 402!, 403!, and 404! have the factor 3 197 but 405 has the factor 3 4 so that 405!
Hence, there is no n such that n\ has the factor required.
has the factor 3 201
19
is the highest power of 5 contained inn!.
6. Find a positive integer n such that 5
With how many zeros does 100! end?
7. Show that 95! ends with 22 zeros.
8. Find the highest power of 12 contained in 500!.
9. Prove that the exponent of the highest power of 3 contained in (3 r 2)! is
.
(3*
10.
11.
2r
l)/2.
m >
and a(m)
is
the
sum
....+ [2]
a)+w + +[f]+.[5]+8[;] +
...
+ .[2]
14. If
(m
+n 
1)!
m\n\
is
an
16.
integer.
integers,
(2m)!(2n)!
7n\n\(m
is
an
integer.
n)\
(a!) 6 .
least
46
n =
17. If
Oi
0,2
a>r
with
>
all en
and
(ai,
a2
ar)
<2,
then
I
d(n
ai!a 2
an
is
1)!
ar
integer.
m,
and a are
18.
If
19.
Prove that
n,
what conditions
will
[t>[;]+*R]
.
(x+ri
i)
(x
w hen
l)(x 2
is
1)
(z r
1)
is
a factor of (x n
l)(x n+1
l)
a positive integer.
primes.
x2
41
is
gives 15
extraordinary, for
it
',
f(x'
where the
mp) = a Q (x
mp) n
a x {x
mp) n ~
and
in
+ a
each expansion
* W. H. Mills, Bull. Am. Math. Soc, Vol. 53, No. 6, p. 604, 1947.
Am. Math. Monthly, Vol. 58, No. 9, pp. 616618, 1951.
t R. C. Buck, Am. Math. Monthly, Vol. 53, No. 5, p. 265, 1946.
E.
M. Wright,
PROPERTIES OF INTEGERS
every term except the
mp) = a
f(x'
first is
a multiple of
x' n
aix'
n
Therefore,
p.
'
'
47
+ Kp
an
or
+ Kp
and f(x' + mp) is not a prime unless f(x' + ^p)
Hence, p f{x'
n 1, f(x''+ mp) =
Suppose that, for m = 0, 1, 2,
is p or p.
yield
the
equation f(x) = p of
of
m
can
or
other
value
no
Then
p,
p.
+ mp)
+ mp),
f{x'
= p
mp >
and /(V
mp)
a value of
for
which
x'
^ p.
senting function.
it
EXERCISE
~
Prove that for integral values of x an integral polynomial a x n + a\X n
+
has an infinite number of distinct prime factors.
(Assuming the
l
+ a n of degree n >
number
is finite,
=i
The number
of positive
7r(n)
= n
+r
Lf
lPi\
Z/ IViVA
M=
l
(!)'
P1P2
'
Pr\
2
where p h p 2
p r are all the positive primes such that pi < n
where i = 1,2,
r.
Determine I so that / is the largest integer whose square is less than or
equal to n. Then find all positive primes p h p 2
p r that are less
,
48
than or equal to
n are
divisible
I.
Since exactly
by p h none
integers from
1,2,...,
of the integers
of these multiples of
In like manner,
prime.
Mi =
p h except p\
itself, is
LPzj
p2
Of these multiples
and
pi,
p 2 however,
of
= [il
IPii
the
first
Mi
pi
and hence
M
is
by both
are divisible
fe]
 \jl]
LP*}
LP1P2J
1
.
LVkj
+ [=1+
LpiPz]
LP1P2]
iPkiPk]
+(rLP1P2
LP1P2P3]
pk \
n
Pk+ij
Of these integers
Pk+i
Pi
examine the
operate on
pi, for
if
we
p k+h 2p*+i,
how many
by
by
p\.
\^\
by p h p 2
we operated on n above.
or
p r we must
,
Hence,
[1
lPk+i
Pk+i.
Pk
determine whether
IPk+i]
Likewise, if we wish to determine
just as
Pk+i
P1P2
"
1
Pk+
L
PkiPk
(D
.Pk+i.
[pip 2
"
Pk:
49
PROPERTIES OF INTEGERS
the
is
also
r
*
LP1P2
Pt
jlI
_ r_5_i
_p k +ii
LPiPk+i]
Pk
fjLl
PlP2
by p k+ i and
Using the fact that
are divisible
pp*+iJ
we have
+
f_=_i
IPkPk+ij
\^\ +
LPiPzPk+i}
(8)
p*p*+i.
1}
=+...+
fc+1
zlI
p*J
fe]
+ r jl] _ [jl]
LP1P2J
Lpfc+iJ
JL ]+
+(i)4
PkPk+ij
LP1P2
PkPk+ij
number
the
of the
w(n)
= n
+r
the
written
ir(n)
<f>(n } r)
+r
where
50
some, but even though his method has been improved upon,* no expeditious method for finding the exact number of positive primes less than
a large n has been discovered.
On the other hand, due to the work of Legendre and Gauss (17771855)
in applying analysis to the theory of numbers, we have formulas which
approximate the number of primes not exceeding x. Legendre stated the
empirical formula
F(x)
logx
1.08366
which agrees very well with t(x) so long as x is not greater than
Gauss discovered "the integral logarithm of x,"
but
we have no
their
observation.
[*
dt
J 2 log
work on that
in developing
0) =
of
1,000,000.
it.
Basing
By
a refinement of Gauss'
is
is
called
theorem,"
lim
x
*& =
^ z/log
as 1948 Paul Erdos and Atle Selberg developed new and more
elementary methods for showing this limit.
As a matter of fact although many theorems about primes have been
demonstrated, we can still state a large number of theories that mathematicians believe to be true but which remain unproved. We have seen
that Euclid established an interval within which there must be a prime.
Again it has been proved that if p 1} p 2
Pni are the first n 1
primes, when n > 4, the next prime p n is such that
As recently
Vn
<
PlP2
'
'
Pnl
PROPERTIES OF INTEGERS
51
>
2,
n\
3,
n\
Hence, as we move out in the sequence of positive integers, there must be consecutive primes whose difference is larger
than any assigned integer.
Euler mentioned that Goldbach (1742) had stated the empirical
theorem that every even integer greater than 2 can be represented as the
sum of two positive primes. Although the truth of this conjecture has
been verified in many cases, it has never been proved. Goldbach also
said that every odd integer greater than or equal to 9 is the sum of three
odd primes. In 1937 Vinogradov proved by analytical means that this
theorem is true for sufficiently large odd integers.
If we examine a table of primes, we notice that there is at least one
2
prime between any two consecutive squares n 2 and (n
l)
but whether
or not this statement is always true, we do not know.
Again, we do not
know whether or not there is an infinite number of primes of the form
is
a prime
when n
1.
(2n) 2
1.
52
Table of Primes
2003
2011
2017
2027
2029
2039
2053
2371
2377
2381
2383
2389
2393
2399
2411
2417
2423
2749
2753
2767
2777
2789
2791
2797
2801
2803
2819
1663
1667
1669
1693
1697
1699
1709
1721
1723
1733
2063
2069
2081
2083
2087
2089
2099
2111
2113
2129
2437
2441
2447
2459
2467
2473
2477
2503
2521
2531
2833
2837
2843
2851
2857
2861
2879
2887
2897
2903
1381
1399
1409
1423
1427
1429
1433
1439
1447
1451
1741
1747
1753
1759
1777
1783
1787
1789
1801
1811
2131
2137
2141
2143
2153
2161
2179
2203
2207
2213
2539
2543
2549
2551
2557
2579
2591
2593
2609
2617
2909
2917
2927
2939
2953
2957
2963
2969
2971
2999
1087
1091
1093
1097
1103
1109
1117
1123
1129
1151
1453
1459
1471
1481
1483
1487
1489
1493
1499
1511
1823
1831
1847
1861
1867
1871
1873
1877
1879
1889
2221
2237
2239
2243
2251
2267
2269
2273
2281
2287
2621
2633
2647
2657
2659
2663
2671
2677
2683
2687
3001
3011
3019
3023
3037
3041
3049
3061
3067
3079
1153
1163
1171
1181
1187
1193
1201
1213
1217
1223
1523
1531
1543
1549
1553
1559
1567
1571
1579
1583
1901
1907
1913
1931
1933
1949
1951
1973
1979
1987
2293
2297
2309
2311
2333
2339
2341
2347
2351
2357
2689
2693
2699
2707
2711
2713
2719
2729
2731
2741
3083
3089
3109
3119
3121
3137
3163
3167
3169
3181
233
239
241
251
257
263
269
271
277
281
547
557
563
569
571
577
587
593
599
601
877
881
883
887
907
911
919
929
937
941
1229
1231
1237
1249
1259
1277
1279
1283
1289
1291
1597
1601
1607
1609
1613
1619
1621
1627
1637
1657
71
283
293
307
311
313
317
331
337
347
349
607
613
617
619
631
641
643
647
653
659
947
953
967
971
977
983
991
997
1009
1013
1297
1301
1303
1307
1319
1321
1327
1361
1367
1373
73
79
83
89
97
101
103
107
109
113
353
359
367
373
379
383
389
397
401
409
661
673
677
683
691
701
709
719
727
733
1019
1021
1031
1033
1039
1049
1051
1061
1063
1069
127
131
137
139
149
151
157
163
167
173
419
421
431
433
439
443
449
457
739
743
751
757
761
769
773
787
797
809
179
181
191
193
197
199
211
223
227
229
467
479
487
491
499
503
509
811
821
823
827
829
839
853
857
859
863
2
3
5
7
11
13
17
19
23
29
31
37
41
43
47
53
59
61
67
461
463
521
523
541
1993
1997
1999
CHAPTER
PROPERTIES OF CONGRUENCES
41.
We
Congruent Integers.
9^
= qm
<
<
\m\ classes
\m\
+r
q m + r
giw
and
<
when
above equations be
difference a
a
r2
S3 b (mod
<
\m\,
m), for
r2
r\
Theorem
41.
their difference
by m.
b is divisible
then a
Therefore,
identical,
is
b
.
if
= qim
m(qi
q 2)
r\
it
<
\m\
b (mod
Conversely,
and
+n
m) the remainders
b
r2
q2
and
if
m+
ri
r2
km, then
where
< n,
b
r 2 is divisible
by m.
Thus we have:
Two
divisible
by
m^
modulo
if
and only
if
0.
Thus, because 12
m.
53
=
m
54
These definitions imply that each integer belongs to exactly one residue
modulus and that each residue class modulo m contains
one and only one of the integers 0, 1, 2,
\m\ 1.
Hence, there
class for a given
modulo m.
Any set of \m\ integers selected so that no two of them belong to the
same residue class modulo m forms a complete residue system modulo m.
Thus for the modulus m the set of integers 0, 1, 2,
\m\ 1 or the
.
1, 2, 3,
integers of the
\m\ is often
km
form
Any
r.
km
Since, as
\m\
km
1 is, therefore,
one and
integer
or
is
is
1,
km
not prime to
by
2,
modulo m.
not
prime to m, the set of remainders from 1 through \m\ 1 which are prime
These
to m represents all and only the integers that are prime to m.
integers, prime to m, are thus separated into residue classes modulo m
that are in onetoone correspondence with the positive integers from 1
through
Any
of
\m\
the modulus
division
is
or
is
set of integers
them belongs
its
prime to
for
For the
modulus
are iden
tical
= qm
<
for either
<
is
\m\
and we
as moduli,
making a
From
d
specific
the equations a
km,
tm,
we have
+ Lm.
5.
If
b (mod
m) and
= d(mod
m), then ac
bd(mod m).
PROPERTIES OF CONGRUENCES
6.
If
7.
If
55
integral
/(6)(mod m).
The
first
three statements
for a factor of a
"
When
it
term
of a congruence.
two
just
equality"
is
of those
to division:
If
is
a divisor of
2.
If
b (mod mi)
m =m
d.
We
Examples.
1.
4(mod
17),
EXERCISES
1.
21 10
is
2.
by
is
divided by 51;
is
divided
when
3 10
is
divided
by
51;
51.
by
127.
Do
the
same
for 5 66
when
56
3.
Prove that 2 U
4.
If a
by
9.
= b (mod
ra),
9.
is
divisible
is
by
divisible
by 8
if
and only
if
the
number formed by
its
8.
10. Prove that an integer is divisible by 1 1 if and only if the sum of the digits in the
oddnumbered places diminished by the sum of the digits in the evennumbered places
is divisible by 11.
11. If an integer N is written in the scale of r and then its digits are rearranged in
is divisible by r 1.
any way to form the integer M, the difference N
16
(2
10fc)(3
6); that
is,
100
16.
4 (mod 6) as well as 2 2 =
the same least positive residue 2 ^ 0(mod 6) can be
4(mod
class of 4
any prime, we
modulo
6.
If
we examine
the least
when
PROPERTIES OF CONGRUENCES
6(mod
(a, p) = 1, C\ = c 2 (mod p).
residue class modulo p.
6(mod
p)
and ac 2
When we
Hence,
b (mod
and
m),
C\
and
c2
it is
ac 2 (mod p),
and since
must come from the same
Therefore, aci
p).
57
evident that
we
We have,
is
an
have
do not belong to the same residue class for that modulus.
It may happen, of course, that all answers are in but one residue class as
But it is also posis true in the case of the congruence 5x = 1 (mod 6)
sible that there be no answer whatever, for a solution of 2x = l(mod 6)
would demand that 2x = 1 + 6/c and that 2 divide 1. Consequently, we
must proceed with care, for division modulo m is not always possible and,
when it is, need not yield a unique result.
One of the important ideas of arithmetic is that if ab = and a ^ 0,
then b must be 0. Suppose that
it is
possible to
results that
ab
0(mod m)
we
for the
modulus m?
0(mod 6) and
modulo 6. But when the modulus
that neither 2 nor 3 is congruent to
is a prime p and ab = 0(mod p), both a and b cannot come from the set
of integers 1, 2, 3,
p 1, for the product of any two of these
integers is prime to p.
Consequently, we see that in this case a product
.
is
not congruent to
6,
is
in the class of
for the
When the modulus m is composite, however, by factorthat m = n n^ where 1 < n\ < m, it follows that ttin =
given modulus.
ing
m^Oso
0(mod m).
EXERCISES
that although 2(6) = 26 (mod 14), 26 cannot be factored into integers such
in the class of 2 and the other in the class of 6 modulo 14.
2. Find numbers in the class of 10 modulo 11 that can, and some that cannot, be
expressed as a product of two integers, one from the class of 2 and the other from the
1.
Show
that one
class of 5
is
modulo
11.
58
How many
3.
m
=
a
=
2x
2x
2x
Sx
4.
6.
Compare the
the powers of
6 (mod 10)
3 (mod 4)
3 (mod 5)
6(mod
15)
12.
3, 5,
and 6 modulo
Can you
15.
2, 4,
find a
Function.
Leonhard Euler (17071783) worked in
pure and applied mathematics. His voluminous publications were concerned with algebra, the calculus of finite differences, the
differential and integral calculus, the calculus of variations, astronomy,
and analytical mechanics besides the theory of numbers. In the latter
field he discovered the theorems which taken together make up the
44. Euler's
many
fields of
the
function.
an integer m ^
is the number of positive
and prime to m. Thus 0(m) is the
number of integers in a reduced residue system modulo m, and 4>(m) =
Because of this last fact, it will be sufficient to use only positive
0( ra).
The
indicator <f>(m) of
m in
integers
considering the
Examples.
0(1)
It is evident that
Theorem
4>(p
n
)
To
42.
= p n~Kv 
when p
p
If
function.
0(5)
1,
is
is
4,
0(6)
2.
a positive prime,
4>(p) is
1.
a positive integer,
is
i).
p,
2p,
either divisible
by p
p
or
r
,
is
prime to
But p n ~
p.
n~ 1
n~ l
= p (p 1) of the
Therefore, p p
p.
through p n are prime to p.
r, is arithmetic
Any function of the variables Xi, where i = 1, 2,
if it assumes only integral values for the sets of integral values of the
The function <f>(x) is
variables Xi for which the function is defined.
integers
from
plicative
We
shall
if,
for
any a and
is
function
is
multiplicative.
f(a)f(b).
PROPERTIES OF CONGRUENCES
Theorem
43.
If
59
0(a) 0(6).
We suppose that the integers a and b are positive and arrange the
integers from 1 through m = ab in the following array suggested by the
residue classes
modulo
a:
2a
+
+
L)o+
(6
2a
+
+
a
2a
3a
fc
ba
2a
+
+
+k
a
2a
3
3
(6
l)a
sa
k(mod
a)
Therefore,
with
and
a,
if
prime to
is
a,
prime to
is
a.
There
How many
a.
of
No two
2a
...
(b
congruent modulo
sa
(t
s)a
=
=
ta
\)a
b,
for
/c(mod
0(mod
b)
= 0(mod
b)
if
b)
and
t
since
2,
= 1. But
1 modulo
b)
(a,
.
integers of
and
any column
to the integers
1, 2,
6,
so that unless
s,
s(mod
b).
0,
The
1,
are, therefore, in
b.
too.
Since 0(6) integers in each of the 0(a) columns of integers prime to a
6, the number of integers from 1 through ab that are
Theorem
.
(P.
If
6 is 0(a) 0(6).
m=
ni
p1 p2
r,
1)
44.
'
'
(Pr
1).
U2
0(m)
Pr
Tlr
,
pi'
il
lr
1, 2,
""!)
60
Since
m has
0(m)
By
we
from which we
Corollary.
0(pi ni )0(P2 n2
'
Vr
nr
)
find
<t>(p^)<t>(p2
n2
HPr nr )
m>
If
2, <t>(m) is
even.
less
1)
96.
EXERCISES
3.
n and prime
to
is
(i)+Sa>a*> *
03
f(ia
(n
1)!
^~
[]
1)1
cp*p<
 1)!
"
pi<r l)(Pfcl)
[1
i
=i
i<j
Set
6.
method
If
6.
up a method
for finding
G(n)
>
4>(d),
than n and
by
<j>(x)
x such that
<f>(x)
Use your
n.
=16.
is
multiplicative.
d\n
>
^>(d) is
sum over
read "the
the divisors of n of
</>(d),"
and
dn
Theorem 45.
2,
through
'
(1
r,
Urn =
U2
pi p 2
p^ *
are distinct positive primes, the
(1/Pi)).
2,
pr
nr
where the p
number
p*ism(l
of integers
(l/pi))(l
if
from
1,
(I/P2))
PROPERTIES OF CONGRUENCES
61
When
from
m=
ni
pr
nr
m that
are prime to p h p 2
m
Some
of integers
(m/pi) = ra(l
Proceeding by induction, suppose that the
through
number
it is
through
n*
pi p 2
number
and p k
from
of integers
is
05)0)"(s)
We
of these integers,
wish, there
through
Pi, P2,
from
set
To
Pk
find this
m that
through
number,
how many
is
or
is
or
is
are
TYl p k +i
Vk+l
not prime to
not prime to
integers of the set
as its coefficient c
They
p k+h 2p k+h 3p k+ i,
Since any integer cp k+1
first
is
pi,
p2
p,
p2
.
and p k according
and p k we must ask
.
m
1, 2, 3,
'
p2
and p k
we consequently know
(1),
Pk+i
'
that
m
Vk+
pj\
vj
(2)
Pk)
number
the
m
\
Pi)\~pJ
"
p k)
m
Pk+ii
Pi) \
\
\
Pi) \
pi)
pi)
Ph)
V ~
Pk) \
"
Pk~^)
62
as the
number
p and
k
2,
prime factors of m.
Example.
through
includes
it
some
from
of integers
also p*+i.
m =
If
23
32
Theorem
n from
m is
n and
number
the
23
is
32
of integers
5 3 (1
m =
46.
through
divisor of
53
)(1
J)
of integers
the greatest
is
through
3600.
number
kd, the
from
common
<f>(k).
through
Theorem
m^
0,
2,
1,
47.
then
d h d 2j
If
0(d)
0(di)
k,
m the
common
greatest
integers
0(/c)
divisor
from
d.
0(d 2 )
4>(d r )
\m\.
d\m
Each
integer
di,
where
in the set 1, 2,
1, 2,
has with
\m\
as greatest
r,
common
{,
symbol
of the
>
<j>(d) is
and
2.
sum
Examples. 1.
such that (n, 90)
78,
read, "the
If
m =
6
is
90, the
0(15)
number
The
of integers
>
tf>(d)
n from
through 90
8.
84.
The
1, 2, 5, 7, 10, 14,
m>
1,
and
35,
and
70.
More
1; 0(5)
4; 0(7)
6; 0(10) = 4; 0(14) = 6;
1; 0(2)
=
=
integers is 70.
and
the
sum
of
these
24;
0(35)
24; 0(70)
45. Residue Systems Modulo m.
Take any integer a prime to
over, 0(1)
let
r h r 2)
rm
Form
No two
of these
arj (mod
the products
ar m
modulo m,
m)
for
if i
^j
and
PKOPERTIES OF CONGRUENCES
63
then
Ti
which
Moreover,
modulo
ra.
if
n, r 2
r^m)
1,
ra,
is
is
is
r,(mod ra)
(a, ra)
the set
ar^m)
integers, all of
is itself
a
for
any a
in the set,
is
and
rh
r2,
rm
if
+n=
ry(mod m)
then
Ti
Sometimes
ry(mod m)
a in the form
= km
<
< m
system we keep the values of r which do not exceed ra/2, that is,
through ra/2 or (ra l)/2 depending upon whether
ra is even or odd.
But those values of r which exceed ra/2 are replaced
by the negative integers of least numerical value to which they are congruent for the modulus ra. Since
for this
ra
this
= r(mod
19
and when
odd,
ra is
ra
when
ra is even, is
ra
it is
ra
U, 1, A,
ra,
ra)
1
>
m 1
ra
,
3
j
2,
64
1, 2,
is
7.
d,
2d,
(&
l)d
ri,
for the
modulus
b,
and y takes
modulo
(3)
ka
(4)
a.
all
kh k2
These
integers.
ar s
rs )
fa
kj
ari
kj
A'*(mod ab)
then
Ufa
and a
But
(fa
since
kj(mod ab)
so that
kj),
no two
(mod
a)
modulo
a,
kj.
Hence,
a(ri
rs )
0(mod
ab)
rs
0(mod
b)
and
Ti
so
that,
as
above,
n =
rs
Consequently,
as
the
ari
bkj
ar s
a(ri
rs )
b(k
bk
(raod.
ab)
then
But then a\
fj
rs
(fa
fa),
and therefore
Av)(mod ab)
fa
fa.
Also
(rt
modulo
ab.
r s ), so that
PROPERTIES OF CONGRUENCES
65
It is easy to show also that if x has the values in (3) that are prime to b
while y has the values in (4) that are prime to a, then when (a, b) 1, the
by form a reduced residue system modulo ab.
integers ax
EXERCISES
Use the form ax + by with (a, b) = 1 to show that <(a&) = <t>(a)<t>{b).
Prove Theorem 45 by setting up the integers 1 through m = p^pf*
Pknk s
n
n
in an array of s complete residue systems modulo pi ip2
Phnk
n~ x
3. Show that a
+ y generates a complete residue system modulo a n if x has the
values in a complete residue system modulo a while y has the values in a complete
~
residue system modulo a n l
4. If f(x) is an integral polynomial and if there are \p(m) integers prime to m in the
,/(m), prove that when (a, b) = 1, xp(ab) = Ha)Hb).
set/(l),/(2),
5. Find the number of integers prime to m in the set:
1.
2.
a.
,
6.
2,
3,
1223
~T' 2~'
For m >
'
'
'
set
up
m(m +
m(m +
1).
1)
2
all
the permutations
fc
whose greatest common divisor is prime to m is 4>k{m). Find a formula for <}>k(p n ), and
show that this function is multiplicative.
7. Without using an enumeration according to size, show that if a, b, and c are
positive integers and a = be, there are in a complete residue system modulo a exactly
(Let ci, c 2
c c be a complete residue system
c integers that are divisible by b.
modulo c. Then consider the set be i, 6C2,
bc e .)
8. Can you find an integer the powers of which set up a complete residue system
modulo 13? Can all integers prime to 13 be used to form such a set?
9. By expanding (1 + 1 +
+ l) p prove that if p is a prime, a p = a (mod p)
and hence that when (a, p) = 1, a^ {p) = l(mod p).
.
CHAPTER
eration of the
a,2,
.
xn
congruence.
ai,
(all
fi(x 1} x 2 ,.
where
/i
and f2 are
If
two
ah x2
(x h x 2 ,.
,^)(modm)
fi(a h a 2 ,.
then x\
=f
,x n )
a2
,a n )
xn
/2 (ai,a 2 ,.
integral polynomials /i
an
is
and f 2
,fl)(mod
and
if
m)
xn
are such that the coefficients of like terms are congruent to each other for
,x n )
we sometimes
f2 (x h x 2 ,.
write
,x n
)(modm)
Corre
,x n ) with
spondingly, an integral rational algebraic function f(xi,x 2 ,.
if and
integral coefficients is identically congruent to zero for the modulus
.
if all its
66
67
gruence will be satisfied regardless of the integral values that are assigned
to the variables.
On
the other hand, we shall call a congruence of the above form, but in
left and righthand members /i and f 2 are not identically con
which the
know from
consequently a solution
r(mod
is itself
ra).
8(mod
10).
Let us recall that in algebra* two polynomials fi(x) and f2 (x) are said
to be identically equal if and only if they are equal for all values of x, and
hence if and only if corresponding terms have the same coefficients. In
particular, a polynomial vanishes identically if and only if it vanishes for all
In
values of x, which means if and only if all its coefficients are zero.
contrast, consider the congruence x 3 x = 0(mod 3), and observe that
although for the modulus 3 the polynomial x 3 x is congruent to zero for
x, yet not all its coefficients are congruent to zero modulo 3.
members of the congruence x 3 2 = x + l(mod 3) have the
same values modulo 3 for all values of x, but the members are not identically congruent modulo 3.
In other words a congruence may be
satisfied by all integers and still not be an identical congruence according
Examples of such conditional congruences are
to the above definition.
all
values of
Again, both
xb
x
Sx 2
x
2x
= 0(mod
= 0(mod
2x
4(mod
6)
x2
2(mod
5)
5)
6)
Likewise,
and
first one having two incongruent soluwhile the second has no solution whatever.
On
modulo
6,
x2
*
(x
M. Bocher, "Introduction
pany,
New
York, 1931.
2)(x
2)
0(mod
4)
1,
68
and
Qx 2
15
3(mod
6)
EXERCISES
Determine whether the following congruences are identical or conditional, and
find the solutions
2x 2
xz
x2
x4
6.
4
1
= 0(mod 5).
= (x  l)(x 
52. Equivalent
if
if
operation listed
42),
trial
+ Sx = 5(mod 7).
+ x = x  z (mod
1.
3.
5.
by
2).
2){x
3)(x
x4
4.
2x 3
4)(mod
6).
5).
Congruences.
among
+ x m 0(mod 10).
+ 3x + x = 0(mod
2.
The
first
operation
is
said to be
1. Adding to or subtracting from each member of a congruence congruent integers or other expressions that are identically congruent for the
given modulus.
2.
is
if
F(x)
Multiplying
member
of a
fi(x h x 2 ,.
,x n )
f2(xi,x 2 ,.
,x n
)(modm)
,x n )
can be reduced to an equivalent congruence of the form f(xi,x 2 ,.
= 0(mod m). For example, the congruence x 2 + 10 = 7a:(mod 6) is
4 = 0(mod 6), 5x 2
equivalent to each of the congruences x 2 x
.
5x
= 0(mod
6),
and 3x 2
Sx
12
0(mod
18).
,x n ) =
a congruence has been written in the form f(xi,x 2 ,.
0(mod m), the degree of the congruence is defined as the degree of the term
When
,x n )
whose
coefficient or coeffi
6x 3
is
of degree three,
of
7x
0(mod
12)
Sx 2
5x
0(mod
12)
Thus
although
12z 3
is
10a;
69
3x 2
5x
0(mod
12)
has the same solutions, for the two are equivalent. Again, consider the
congruence 2x = 4 (mod 6). Every integer that satisfies this congruence
also satisfies the equivalent
lies in
all
congruence x
2 (mod 3).
The
distinction
modulus
3,
The first conthe classes of the residues 2 and 5 for the modulus 6.
gruence, therefore, has two incongruent solutions modulo 6, while the
second one has but one solution modulo 3.
It is not true, however, that two congruences that have the same soluBoth 2x =
tions for a given modulus are necessarily equivalent.
4 = 0(mod 6) have the solutions x = 2 (mod 6)
4 (mod 6) and x 2 x
and x
where
(m, m/)
1,
it is
Notice
70
satisfies
is
satisfied
by
all
On
the
where x
Example.
7
0(mod
To determine whether
15)
and 9x
Sx
= 0(mod
11
we
f
try
y will satisfy Sy
We reject
3 since
9(mod
it is
15)?
The
not prime to
solutions are 3,
15.
8,
Multiplying each
member
of the
first
equivalent.
EXERCISES
1.
2.
3.
4.
5.
6.
2
7.
8.
2>x
= 0(mod
10)
ss
ax 2 (mod m)
Xi
.x 2
and
(mod m)
71
Theorem
52.
When
is
axib
6 (mod
m)
cd,
On
b.
if
d divides
b,
let
a x
c(mod
know
for
modulo m.
class
We
and only
solutions
modulo m; that
is,
we should
same residue
see that
Xi
if
more
which values
kmo
Xi
0(mod m)
sm (mod m)
if
(k
s)
and only if k
Consequently, when k ranges
s = 0(mod d).
through d 1, the integers xi + kmo represent exactly d solutions
that are incongruent for the modulus m and all solutions of the given
congruence lie in one of these d classes modulo m.
Example. Solve: 15x = 12(mod 36).
Since (15, 36) = 3 and 3 12, we reduce the congruence to hx =
4(mod 12) of which there is one solution x = 8(mod 12). Hence, the
solutions of the original congruence are x = 8, 20, 32 (mod 36).
that
is, if
from
EXERCISES
State with reasons the
number
3.
5.
3x m 5(mod 9).
6z = 3(mod 18).
12x s 36 (mod 56).
2.
4.
5x = 3(mod 27).
49z = 23(mod 125).
and
72
If fi(x)
+m
integral polyno
where g(x) is
an integral polynomial, by f 2 (x), obtaining integral polynomials q(x) and
mials and
if it is
g(x),
r(x) so that
fi(x)
} 2 (x)q(x)
(mod m)
r(a;)
is
of
we
As
modulo
2a;
l(mod
x2
there
On
6)
no value of a that
is
divide
modulus
6,
but
it is
2x
is
we
by
a;
3 using
is,
of
b that
x by 2x
2x
4a;
modulo 6
is
5\
\x 2
\x 2
lOx
4x
4x
l(mod
10
11
If
divide x 2
lOz
Hence, 4z 2
a remainder 12 which
3.
possible,
2a;
1 by Sx 2 modulo 6, for if
would make (Sx 2) (ax f b) +
an identical congruence, then 3a = l(mod 6). But
is
we cannot
stands,
it
6 of
we
if
for the
or
(x)
f 2 (x).
is
the division
if
is
(2x
try to divide
4a;
5) (2x
+x+
2)
6)
5(mod
6)
5 (mod 6).
by 2x
modulo
6,
we
find
we cannot
carry out the division far enough to obtain a remainder of the required
form, for
2x
5[
4x 2
4x 2
10a;
11a;
But 2y
5 (mod 6) has
identical congruence
However,
instead of
no
2
solution,
2x
5a;
l(mod
6)
if
2a;
4a;
(2x
of the
5)(2x)
5x
is
to write the
l(mod
6).
4x
5 4a;
+
+
2
2
4a;
\x +
4x
4a;
0(mod
6 ==
Hence,
4a;
+ +
a;
We could
5a;
73
6)
= (4x  5)(x +
first
l)(mod 6).
adding Qx 2 to 4a; 2
+x+
for
we
find that
2a
10z 2
10x 2
x .+
5s
25a;
26a;
2a;
l(mod
2a;
We,
therefore,
have
4a;
by
3a;
1,
(2x
5) (5a;
6)
= 0(mod
is 6,
6)
l)(mod
we can
6).
divide
3a;
for
3a;
3a;
+
+
=
=
(3a;
(3a;
2
2
l)(x)
+ x + l(mod 6)
+ 3a; + l(mod 6)
l)(3a;)
and
3a;
When
the modulus
is
(3a;
l)(5x)
5x
l(mod
6)
it
easy to show that the division modulo p of fi(x) by fz(x) can always be
accomplished, for any term present in either one of these polynomials has
is
a coefficient that
6(mod
fi(x) =J2{x)qi{x)
ri(aO(mod p)
r 2 (x)
and
fi(x)
= f2(x)q
(x)
(mod
p)
then
f2(x)[qi(x)
If q\{x)
q 2 (x) f
0(mod
q 2 (x)]
r 2 (x)
ri(x)(mod p)
be
0(mod
p),
and
let
ri(x).
identically
This
is
impossible,
7*1(0;)
and we
r 2 (o;)(mod p).
qiiv)
(mod
of
p)
74
EXERCISES
1.
why
Explain
division
modulo
x2
+3
cannot be divided by
4a;
modulo
8.
5.
Divide x 2  2x + 5 by 2x  3 modulo 7.
Divide 3x 2  2x + 4 by 2x  1 modulo 15.
2
4. Divide x 3  2x + 5x  1 by 2x  3 modulo 11.
5. Divide 2x 2 2x + 5 by 2x 3 modulo 6, and then change the divisor to the
form 4x 3 and carry out the division modulo 6.
2x
1 by 2>x  1 modulo 11.
6. Divide Ax z  3z 2
n
and g(x) = b Q x + friz'" 1 +
7. If fix) = a x
ai^ 1
+b
are integral polynomials with t < n, and if (b m) = 1, do polynomials g(x) and r(x)
exist so that /(a:) h= g(x)q(x) +r(a;)(mod m) with r{x) lower in degree than g{x)l
If so, are these polynomials unique modulo m?
2.
3.
++
conversely.
remainder
f(r) is the
when f(x)
(x
r)q(x)
 f(r) =
(x
r)q(x)
f(x)
+ f(r)
or
f(x)
identically,
But
f(r) is
n1
/(r)
b 1x n
~2
= 0(mod
&_i.
f(r)
are
f( x )
f( r)
and so
f(x)
identically,
=
=
(x
r)q(x)
(x
r)q(x)(mod m)
(mod m)
f(x)
(x
f(r)
= 0(mod
showing that x
Conversely,
if
r)g(a:)(mod m)
then
and x
r(mod m)
is
m)
a solution oi fix)
0(mod m).
modulus m.
quotient
55.
is
r is
75
a factor modulo
of f(x), the
unique.
of differential
He gave a
x2
by 2
in integers.
If
is
+.+ o in which a ^
integral polynomial a x n
ai^ n_1
0(mod
p) has at
is
an
0(mod
p),
modulo p.
We have observed that the congruence f(x)
solution,
if
there
is
a solution x
f(x)
(x
ri)
ni
gi(x)(mod p)
= 0(mod p)
(r 2
and
rx
ri)
ni
gi(r 2 )(mod p)
r 2 (mod p), it is
as above,
qi(x)
(x
r 2 ) m q2(x)(mod p)
76
so that
f(x)
(x
ri)
ni
(x
r 2 ) n2 g 2 (V)(mod p)
identically.
= O(mod
modulus
p,
congruence
f(x)
+n +
a (x
rtl
ri)
(x
r 2 ) n2
(x
r k ) nk (mod p)
(1)
solution of multiplicity
/()
But s
0(mod
Ti
p),
rii,
 r^is 
a Q (s
r 2 ) n2
(s
r k ) n *(mod p)
(2)
and x
f(s)
There
fore, there
It is easy to
f(x)
(x
rxYq^x)
(x
r 1 ) v q 2 (x)(mod p)
But
is
if
ri) [(x
ri)
u v
qi(x)
q 2 (x)]
= 0(mod
p)
u~v
p,
ri)
rh
qi(x)
and
if
qi(x)
of p.
(x
Substituting
r\
for x,
we
(ri
ri)
w y
giO)
q 2 (x)(mod p)
gi(ri)
# 2 (ri)(mod p)
find
ri)
w y
= 0(mod
result q^ri)
which
a factor
r\ is
when n >
It is
modulo p
of #2(2),
is
0(mod
and x
p),
77
p) can
Again,
us observe that
let
we
if
p,
distinct solutions
modulo
p.
of f(x)
l)st solution
2,
4,
k,
for
is
Hence, we
p.
an
of multi
n, is
m being counted m times), each aiy where i = 0, 1,
congruent to zero modulo p and the congruence is an identical congruence.
Example. Find by trial the solutions of x*
x2 x
2 = 0(mod 5).
4
2
If f(x) = x
Hence, x
1 is a factor of
x x
2, /( 1) =5.
3
By using synthetic division we find f(x) = (x
f(x) modulo 5.
1) (x
plicity
+
+
Then/i(3) =
45, and therefore f(x) = (x f l)(x + S)(x 2  4x + 4) (mod 5). It is
now evident that f(x) = (x + l)(x + 3) (a; 2) 2 (mod 5) and that
besides the solutions x = l(mod 5), x = 3 (mod 5), there is a double
solution re = 2 (mod 5).
Theorem 57. If p is a prime, the congruence f(x) = 0(mod p) of
degree n < p has a solution # = a (mod p) of multiplicity r < n if and
only if /(a) = 0(mod p), /'(a) = 0(mod p),
/^(a) = 0(mod p),
and/^(a) ^ 0(mod p).
We recall that x = a (mod p) is a solution of multiplicity r of /(z) =
0(mod p) if and only if (x a) r is a factor modulo p of the polynomial
r+1 is not
Furthermore, we
a factor modulo p of /(x).
f(x), but (# a)
x2
2x
3)
(mod
5).
Let/iO) =
x*
x2
2x
3.
= a xn
~
n~ 2
a n as f(x) = atfix n l
l)x
a n _i,
i(^
the application of Taylor's theorem to the polynomial f(x) sets up the
notice that
ai# n1
we
if
/(a)
If /(a)
identity
f(x)
+ (x
a)f(a)
0(mod p),/'(a)
0(mod p) for
+
=
r
(x
a)
O(modp),
<
n, then, the
+..+.(*.
and/^a) =
{i)
(a)/i\
a)
(a;
 aY J^ + +(*
/(x)
Or
and
a) r Q(x)(mod p)
a)
^
O(modp),
being integers,
have
/(*)
^p
(mod
p)
we
78
If (x
a) r+1
(x
and the
a) r+1 q(x)
then
of f(x),
(x
a) r Q(x)(mod p)
identical congruence
(x
a) r [(x
a)q(x)
Q(x)]
0(mod
p)
implies that
(x
a)q(x)
Q(x)(mod
p)
n r
f(x)
a Qx n
ciiX"
+ +
a*
(b Q x
xs
b r )(c
'
+C
'
8)
then
f'(x)
a n^ n_1
kXcowr
ai(n
1
+"+ &ni
+ c _i) + (b rx ~ +
l)x n
"
~2
(b
xr
& r_i)(co^
+
On
method
of induction
shows that
it is
are polynomials in
d k (uv)
dk u
~ ~
~dx~
dx~k+
Consequently,
if
we may,
x,
f(x)
dv d k
~l
c.)
If
u and
therefore, write
u
+
dx dx ^
1
valid to apply
k(k
1)
d 2 v dk
~2
2!
+
dx~ d^F
2
'
'
dkv
"
dx~k
(x
modulo p
^ [(X
X)]
~dx^
that/ (r) (a)
To make
of f(x),
^ ^+
ir)
'
'
'
0(mod
it
r ( r!)(
^ ~
WW +
( r!)
^W
p).
Theorem
solutions
'
79
a\X n
~x
Since r
is
r)Q(^)(mod
= a (mod m).
(ic
modulo m of a n
*Theorem 59.
Hence, a n
n~ 1
= r6 n _i(mod
aix n
++ =
m) and
6 n _i(mod
is
a factor
If r satisfies
Substi
6(mod
12) for r
1, 2,
The values
,11.
4 and 8 are immediately ruled out, for the corresponding linear con
Then we
values r
(r
\)y
rowed to
= 8, and we
= 8 (mod
The
12).
2, 3, 5, 6, 9,
and
modulo
11
12,
nar
of these
is
x such that
x
=
=
ai(mod mi)
a 2 (mod mi)
a n (mod
mn
Let
\\
t=i
rrti
and Mi =
M/rrii,
where
1, 2,
n.
Then
set
up
the n congruences
Mix
In each case (M{, mi) =
of each congruence (4).
1,
l(mod
and there
Now
mi)
(4)
is
.x\(mod
X = M iXidi + M2X2CL2 +
+ M nx n a (mod M)
and substitute X for x in x = ai(mod mi). Since M and each M
Mi
every integer of
are congruent to
0,
z)
consider
n
for the
except
modulus mi,
a! (mod mi).
In like
80
manner the
modulo
M satisfy
all
the
given congruences.
But
congruences, for
and Xi
if
is
X=
and since the
shows that
Xi(mod
prime in
ra z are relatively

X=
pairs, the
m,)
pairs,
Xi(mod M)
prime in
congruences x
2,
1,
a; (mod
n, are relatively
m)
M=
55a:
S5x
=
=
=
6,
when
Hence,
X=
385).
M* = 55
x = 77
3 = 35
2x = l(mod 5) has the solution x = 3 (mod 5)
7) or Qx = l(mod 7) has the solution x = 6 (mod 7)
11) or 2x = l(mod 11) has the solution x = 6 (mod 11)
l(mod
l(mod
l(mod
27(mod
57.
2,
5, 7,
M
77x
m^
Example.
divided by
\\
385
5) or
when
(mod
385),
We shall
or
Xm
demonstrate
it exists,
divisor of
a solution
m^
mj,
corresponding a
of the set,t all
least
common
a,j.
common
When
the integer
n,
satisfies
multiple of the
m,
and
is
any
any pair
with
^ j,
of moduli,
divides the
each congruence
where L is the
X + Lt,
integer.
=
=
ai(mod mi)
a 2 (mod m 2 )
No.
Vol. 59,
with d\2
(mi,
2 ),
let
81
Then
since
=
=
x
Xq
we
ai(mod
a 2 (mod
m
m
x)
2)
infer that
=
=
ai(mod
di 2 )
a 2 (mod
c? i2 )
ai
a 2 (mod d i2 )
Xq
and that
Conversely,
first
congruence
cti
miy
a 2 (mod
miy
a2
2)
or
But
since
dn
(a 2
i),
there
is
ai(mod
2)
a simultaneous solution a\
x1
=
=
x (mod mi)
x (mod
2)
Xi
^ (mod L)
X\
where
is
the least
integer of the
common
multiple of mi and
+ Lt is a
form x
common
Furthermore, any
2.
if
Xo
is,
Then
for
if
there
is
common
the least
If
tainly
dij
i t* j.
is
dji
2,
X
X +L
solution
k it,
m^i,
and
and only
where L k i
all
the parameter.
is
the
first
for
t,
1,
2,
it is
.
But
X=
ai(mod m*)
X=
and
fc
(mod m&)
1, 2,
true that
1,
and
82
Hence,
X=
X=
a k (mod d ik )
a (mod d ik )
(mod dik )
a*
and therefore
di
Consequently,
Conversely,
dij
divides a*
if
dy divides
solutions of the
where
fc
first
for
aj
di
i,
for
aj
1, 2,
i,
where
k,
2,
1,
k,
^ j.
common
X +
moreover, possible to
determine a value for the parameter t so that this expression will produce
To prove this statement, consider the
a solution of x = a (mod m k ).
congruence
Lkit,
is
a simultaneous solution.
It
is,
fc
X + Lkit =
in the
Lkit
We
a k (mod
k)
form
X (mod m k
ak
is
common
divisor of
L k i and
=
rrik
(mod
k)
divides ak
by prov
ra 4, let
ar
= 0(mod m
X 
ar
= 0(mod
r)
and hence
m/
)
Since a k
is
ar
divisible
X 
ar
= 0(mod
ar
X 
ak
 0(mod
m ')
and
ak
0(mod p
vls ')
Therefore,
ak
is
is
m>)
true for
divisible
83
a k is divisible
powers of the primes that are contained in D that is,
by D. Thus there is a value of t that forces a solution of the first k 1
congruences to satisfy the kih one, x = a (mod m&).
Again, any two solutions of the set of k congruences x = a* (mod mi)
;
fc
of mi,
2,
X =
X =
Xi(mod L i)
Xi(mod m k )
X =
Xi(mod Lk)
fc
and hence
Lk
where
2,
is
,
the least
common
Moreover,
k.
Let d!
evident
(18,
that
dx
15)
\
(7
3,
L k i and m k and
form X + L k
common solution.
multiple of
integers of the
all
if
x
x
x
=
=
=
7(mod
10(mod
l(mod
d2
(18,
d2
10),
is
(7
14)
hence of m h
satisfy each
18)
15)
14)
1),
dz
2,
d3
(10
14)
(15,
1.
It is
Therefore,
substitute 7
f
1).
solution exists.
Substituting 7
second congruence,
18 in the
18*
3(mod
15)
l(mod
15)
we
find
so that
Since 90
18
is
the least
common
we
= 24(mod
90s
14)
and therefore
Hence, the
common
solution
is
3 (mod 14)
X=
EXERCISES
1.
2.
3.
a.
6.
x2
2z 3
+3 =
+1 =
0(mod
0(mod
5)
3)
84
Find the
4.
common
m 2(mod
= 4(mod
= 9 (mod
x
x
x
= 11 (mod 21)
= 2 (mod 12)
s 4(mod 10)
a
=
=
x
x
a.
b.
c.
Find a multiple
6.
respectively,
58.
If
by
of 7 that
2, 3, 4, 5,
The Number
and
14)
12 (mod 46)
l(mod 31)
16 (mod 28)
1, 2, 3, 4,
and 5 when
it is
divided,
6.
of Solutions of the
Congruence
f(x)
11)
15)
0(mod m).
m =
ni
pi p 2
n*
J{x)
where /(#)
is
0(mod w)
an integral polynomial,
if
and only
(6)
if
there
is
a simultaneous
Six)
=
=
0(mod pi ni )
0(mod p 2 )
/(x)
0(mod p r nr )
S{x)
But
and the
= Xi(mod
m).
The
integer
X\
satisfies
Xi
x z (mod pi ni )
and therefore
SiX,)
=f(xi)(mo&pf)
But
= 0(mod
pi**)
x)
0(mod
p^i)
SiX,)
0(mod m)
Sixi)
so that
SiX
and
both the
85
the set of congruences (8) we replace the solution of just one of the
congruences of (7), say the first one, by a solution x[ distinct from X\
If in
x
x
=X
(mod m)
= xi(mod pi ni
= ^(mod pfi)
be distinct modulo
were the same,
will
2, 3,
(mod m)
=X
(mod
Xi =
Xi
of (8), for
if
the solutions
Xi
pi
ni
1, 2,
Then
xi
xJ(mod pi ni )
fix)
m of f(x) =
0(mod m).
there
no solution
is
Theorem
512.
if
If
one congruence of
(7) fails to
have a
solution,
0(mod m).
of f(x)
1,
2,
r,
have
fc
>
]
i
k{
=l
of f(x)
= 0(mod
m), where
m=
J[ pi
=i
ni
and the
problem
of solving a
congruence
is
of a prime.
= 0(mod
p* 1 )
The
Suppose that f(x) = 0(mod p 81 ) has a solution x = a;' (mod p*~ ).
Under what conditions will one of these integers, x' + kp 51 be a solution
of f(x) = 0(mod p s )?
We must determine which values of k, if any, will
l
86
kp 8 '
fix'
n~ l
= 0(mod
When/(x) = a x + a\X
+
+ a n is a rational integral function
we have defined the derivative of fix) with respect to x as a Q nx n ~ 1 +
n~ 2
+
+ a n \. The expansion of f(x + h) according to
diin l)x
Taylor's theorem is finite and is of necessity valid for integral values of x.
n
of x,
Consequently,
fix'
kp 8 ^)
f(x')
k V 8 y{x')
fc
2 * 2
0^
+
_+_
frn
f
n S n J
(n)
(x')
\_>_
nl
mQ( J p s\
an identical congruence, the expressions (r) (re') /r! having been shown
~
Moreover, if s > 2, then p rs r > p 8 f or r > 2 and hence
to be integers.
all except the first two terms of the expansion are congruent to
for the
modulus p s so that
is
But
+ kp
+
81
that fix')
if
x'
kp 8 1 )
fix'
= fix') + kp  f'ix')imod p
satisfy fix) = 0(mod p
this congruence
)
),
by p s and
know
to the congruence
When
tp
t
1
,
to
is
kf'ix')
common
0(mod
kp s
l
last
therefore k
f'ix
shows
must be
0(mod p 8 ).
We
p).
the greatest
= (mod
divisor of fix')
p)
0(mod p 8 ).
When
divide
t,
the greatest
common
the solution x
divisor of fix')
= a/ (mod p ~
= 0(mod p
8
and p
of fix)
But
is p,
0(mod p ~
s
does not
).
if
71
To
1
Since /(4)
=
19,
if
solve 2x 2
0(mod
t
19)
Sx
and
13ft
0(mod
= O(mod
19 2 ),
first
Also f'(x)
1.
that x
87
0(mod
308 (mod
4x
19 2 )
3,
19).
4,
),
f(x)
take f(x)
7(mod
19).
a solution
is
of
the
original
congruence.
Let the student show that x = 235 (mod 19 2 ) is the solution derived
from x = 7 (mod 19).
After showing that x = 3 (mod 4) is the solution of f(x) = 0(mod 4),
apply the Chinese remainder theorem to find the solutions of f(x) =
0(mod 2 2
19 2 ).
EXERCISES
Solve the following congruences.
1. 3a: 2
3.
5.
7.
9.
2x
17a:
+3 =

0(mod
72).
= 0(mod 500).
 x + 20a: + 4 = 0(mod 1089).
2a:
 2a;  2 = 0(mod 2000).
x x 2x 8x  Q = 0(mod 357).
6a:
a:
20
2.
4a: 3
4.
a:
6.
8.
10.
a;
a:
CHAPTER
proof.
is
a prime, but
it
1,
p.
92.
D. H.
ar 2
rn(mod m)
r 2 (mod m)
ar^m)
n>(m)(mod m)
ari 53
89
Therefore,
(t>(
m)
r 1r 2
= nr
r^ (m)
r^ (ro)
(mod m)
and
,<Kro)
Corollary
1.
If
Corollary 2.
If
56
Examples.
p
p
l(mod m)
is
a prime and a
is
a prime and a
48
7); 2
l(mod
prime to
is
any
l(mod
p,
av
~l
integer, a p
is
= l(mod p).
= a (mod p).
105).
EXERCISES
1.
2.
7?
3.
4.
5.
6.
for
any
positive n.
7.
Prove that a n
8.
Can you
b ls is divisible
by 133
whose
if
a and
b are
prime to 133.
by 7?
By
other
primes?
Fermafs Theorem
The solution of ax = b (mod m), where
6a^ (w)1 (mod m).
62. Applications of
Theorem
is
62.
(a,
m)
1,
i(
a^ m) b
6(mod m)
mo d m)
Hence,
and therefore x
Theorem
63.
greater than
2,
m)
is
If f(x)g(x)
~
xp l
x pi
the solution of ax
l(mod
b (mod m).
where p
is a prime
each of the congruences f(x) = 0(mod p) and g(x) =
0(mod p) has the maximum number of incongruent solutions modulo p
permitted by its degree.
Fermat's theorem shows that the congruence
has exactly p
distinct solutions
0(mod
modulo
p),
(1)
p)
p.
When
>
2, if
we
factor
90
x pi
modulo p into
tively, the
and
f(x)
g(x) of degrees r
and p
r,
respec
congruence
f(x)
= 0(mod
p)
(2)
p)
(3)
has at most p
f(x)g(x)
0(mod
^ 0(mod
r incongruent
p) has exactly p
1 ==
l)(a;P2
p.
But
since
modulo
solutions
1
maximum number
its
3PI
l ==
0(mod
of
1)
has exactly p
.
_j_
^p3
{
2 incongruent solutions
p)
2, 3,
P i.
If p is a prime greater than 2, each of the congruences
0(mod p) andx^ /2 + 1 = 0(modp) has exactly (p  l)/2
that are incongruent modulo p.
Corollary
x (pd/2
>
2,
solutions
For p
1.
x pi
modulo
p.
0(mod
(a^D/ 2
of the congruences x
p) has exactly (p
l)(x^' 2
(p ~ 1)/2
1)
1
0(mod
p),
and
Furthermore, since
(p
a)^ 1)/2
= (^^(mod
p)
if
EXERCISES
1.
modulo 35.
8
2. Find the solutions of x +
x* + 1 = 0(mod 19) and x  1
x8
it
= 0(mod
17),
and
also of
How many solutions has the congruence x 3 1 = 0(mod 13)? Find them.
Prove that the congruence x 2 + 1 = 0(mod p) in which p is a prime of the form
4n + 1 has two distinct solutions modulo p.
3.
4.
91
*Theorem
64.
n 1
xp
has
all
0(mod
by x p
p)
by the modulus.
Moreover,
if
we
than
the
remainder
Q(x)
and
R(x)
are
both
the
quotient
that
we
observe
p,
integral polynomials and that
divide f(x)
x until
f(x)
is
(x p
x)Q(x)
R(x)(mod
is less
p)
distinct solution of
}{x)
R(x)
and conversely.
Example. 2x 4
0(mod
p)
modulo p.
*Theorem
a n = 0(mod
gruence b Q x n
The
65.
p) with a
n~ l
+ bix
congruence
0(mod
a xn
and p a prime
p)
f(x)
bn
0(mod
is
p) in
a\x n
~x
equivalent to a con
which
l(mod
p).
and multiply
x Q a Qx n
+ Xodtf" +
1
Obviously, x f(x)
(x Q , p)
by
f(x)
1.
s=
l(mod
p)
Xo so that
0(mod
x an
p)
is
xn
bix n
equivalent to fix)
6 n (mod p)
0(mod
p),
for
92
71
'
"
exactly n
congruence
xp
f(x)Q(x)
R(x)(mod
p)
xp
is
identical
f(x)Q(x) (mod p)
= 0(mod
p) has
Theorem
Wilson's
Theorem
67.
If
is
a positive prime, (p
2,
1,
+1=
1)!
x pi
is
i)( x
an identical congruence.
It
2)
(z
p).
l(mod
by x
p),
Hence,
1.
l)(mod
therefore, satisfied
is,
0(mod
p)
= 0(mod
p)
1 =
If
it is
odd,
it
Corollary.
sum
1, 2,
Because
If
is
l)!(mod p)
1
= 0(mod
1,
p),
2,
the set
(_i)Pi(p
1)
~
xp l
modulo
1 is
1 is
powers
except for
1+2+
of integers selected
l)(x
powers of x
when p >
sum
rata
if
p =
2,
2, the
time from
a multiple of p.
identically congruent to (#
and
p.
2)
in these
(#
two poly
of x occurring in the
are congruent to
modulo
p.
But
~
the coefficient of x p 2 in this product is the
~
the coefficient of x p z is the sum of products
its sign,
two
ways without
repetition
1,
X Pri
2,
for r
2,
1,
1.
.
is
2,
selected r at
93
sum
the
of products of integers
All these
set.
sums
are, therefore,
multiples of p.
Wilson's theorem.
Theorem
68.
If (n
Suppose that n
>
1 is
are factors of (n
Hence, n
<
0(mod
n),
then n
is
a prime.
(n
1)!
1)!
<
n2
rii,
n,
0(mod
But neither
1)!.
n)
can divide
of these integers
a prime.
Theoretically Wilson's theorem and
and n 2
ri\
1.
is
when n
prime, but
is
large, the
amount
its
entirely impracticable.
is
EXERCISES
Use Wilson's theorem to show that 23 is a prime.
Show that, for p > 5, (p 1) + 1 has a prime factor different from the prime p.
3. If p is a prime of the form 4.n + 1, prove that (2n)! is a solution of the congruence x 2 = l(mod p).
4. If p is a prime of the form 4n 1, show that (2n 1)! is a solution of the congruence x 2 = l(mod p).
p_1 is
5. If p is a prime, prove that each of the coefficients of the expansion of (1 z)
one greater than some multiple of p.
k p are both complete residue systems for
6. If r\, r 2
r p and h\, k 2
r p k p form a complete residue system
the modulus p, a prime, can rjci, r 2 k 2
modulo p?
r p _i is any reduced residue system modulo p, a prime,
7. Show that if ri, r 2
1.
2.
p1
1 rt
= l(mod
p).
i=l
r p _i is a reduced residue system modulo p, an odd prime, then p
any integral rational symmetric function of the n, where i = 1, 2,
p 1, whose degree is less than p 1.
9. Develop another proof of Wilson's theorem by making use of the solutions of
x 2 = l(mod p) and ax = l(mod p), where (a, p) = 1 and p is an odd prime.
Notice
that of the integers 1, 2, 3,
p 1 only 1 and p 1 satisfy the first congruence
and that when a is selected from the set 2, 3,
p 2, there is a solution of the
second congruence that is in this very set and is distinct from a.
8.
If
n, r 2
divides
If the p
where i = 1, 2,
r, are
nr
and if m = pi ni p 2 n2
we saj that
pr
if any exponent n > 1
ju(m) =
that m(^) = ( l) r if each w< = 1 and
that p(m) = 1 if each n = 0, which means that /*( + 1) = 1.
When it is
so defined, /x(m) is called the Mobius function.
64.
The Mobius
Function.
i}
94
*Theorem
Take (a,
69.
b = qi si q 2 *
powers
of
distinct
into
positive primes.
factorizations
are
the
where
q
=
=
where
where
or
S/,
.
.
r,
.
If any n iy
i
1, 2,
1, 2,
j
,
, t, is
But then /*(a&) = 0.
greater than 1, jn(a)/*(b) = 0.
r
If all rii and s, are 1, then /*(a) = ( l) and /*(&) = (l 1)*, so that
r+f
r+t
=
(~l)
=
(il)
But
M (a)/*(6)
If either a = 1, or a = 1 and & = 1, then ju(a&) = p(a)n(b).
b)
St
mW
*Theorem
610.
or
> ii{d) is
according as
greater than or
\m\ is
d\m
equal to
If
1.
m
1,
J mM
m(1)
1.
rfi
m=
+pi ni p 2 W2
n
Pr % since p(d)
'
'
,Ci(
Co M(l)
1)
2>
,c 2 (I) 2
But r CoKCi(l)
hence p(di)
n{d 2 )
'
viViVi)
rC r (l) r
C 2 (l)
"(Pi)
H{pip 2
++ n(d
'
'
Pr)
+rCr(l) r = (1
'
s)
where
0,
di,
d2
1)'
.
0,
and
d s are the
positive divisors of m.
*Theorem
611.
If
is
a positive integer,
fi(n)
1.
n=l
mWi) + /ift) +
dil
But
will
ju(l) will
d 2 2
+ J
dmm
"W) =
occur
1,
2,
m, so that
of the
times.
of the integers
r
r
from
95
fi(d) will
occur
Hence,
n=
d\n
and
m
=l
>
0,
is
so
defined that
g(m)
f(d)
d\m
then
f{m)
^ "to (f)
(4)
a\m/d
and
Kd)g
(j)
mW
/(a)
am/d
Hence,
2^
d\m
)?
fe)
M(rf)
d\m
XX
d\m a\m/d
/(a)
a\m/d
M(rf)/(a)
96
d\m
a\m/d d\m
But to say that a ranges through the positive divisors of m/d while d takes
on the values of the positive divisors of m is the same as saying that a
ranges through the positive divisors of m while d is a positive divisor of
m/a. Consequently,
d\m
a\m
=
But
y(d)
>
unless a
d\m/a
/(a)
m(0
d\m/a
m, so that
d\m
An
65.
shown that
if
m>
m =
0,
>
We
Application of the
4>{d).
have
Consequently,
d\m
</>(ra)
5>3
d\m
d\m
Thus
if
0(m)
ra
7?i
/?l
=
I
(
1,
then
<j>(m)
1,
but
Pi
if
m=
h
H
Pr
ni
pi p 2
"
"
n2
pr
nr
,
then
+
Prl??r
PlZ>2
P1P2P3
+(1)'
PlP2
'
'
)
Pr)
and
^..(il^ii)..
.(11)
EXERCISES
1.
functions?
/(ra)
If F(ri)
is 0,
then/(n)
Prove
this state
d\n
din
d
[] F(n/d)^ K
97
member
it
by taking the
m
3.
Prove that
for
any positive
integer m,
m
of
V
n=
(  Tl)
l
together with
v(n)/n
>
Theorem 611.1
<
1.
CHAPTER
ON BELONGING TO AN EXPONENT
The
71.
X Function.
We
for
to
a 4>(m)
l(
mo d m)
we have
exhibited
some
cases in which
a particular integer a
<f>(m), of
is
sufficient
modulo m even when a ^ 1. Take, for example, the congruence x 2 ss l(mod 5), which has the solution 4, as well as 1, modulo 5,
and the congruence x z = l(mod 14), with solutions 9, 11, and 1 modulo
to produce
It
14.
is,
whether there
is
a positive integer
2k
r
1.
<f>(m)
Hence,
(2k
l)
4fc(fc
1)
or
a2
= l(mod2
3
)
But
(
a 2)2
(l
a 22
a 22
l(mod
2 4)
2t
l(mod
2k)
2 3 s) 2
or
2 4s
+ 2V
so that
In like manner,
if
a
it
follows that
or
a2
"' 1
m l(mod2 A:+1
ON BELONGING TO AN EXPONENT
Hence, we conclude that
>
if
2n ' 2
99
2,
l(mod 2 n )
>
Consequently, for n
power
this integral
modulus 2 n
of
is
we proceed
Accordingly,
= l(mod2 w
name
to give a
to this
number
n
0(2 )/2 as well as to other numbers closely related to <f>(m).
It was R. Carmichael who used the symbol X(ra) to designate the arith
2.
3.
4.
m
If m
li m
If m
If
distinct
X(pi0,
X(Pr
'
4,
2,
r,
being
multiple of X(2 n ),
n ').
n
or 2p n for
0, 1, or 2, the form p
and p an odd prime, the X function has the same value as the
function, but when m has the factor 2 n with n > 2, or 2 2 and an odd
prime factor, or two factors that are powers of distinct odd primes, the
X function is at most half of the
function.
n
n2
=
When the p^ where i
r, are odd primes, if m = 2 pi ni p 2
1, 2,
nr
=
know,
therefore,
that
1, we
p r and (a, m)
Thus, when
>
<f>
a X(2)
ni
au P
i
But X(m)
X(ra)/X(2
=
=
i( mo d2)
l(mod pi ni )
is
n
)
a xC2))x<)/x(2.)
i(
mo d2
a X(p^))X(m)/X( Pi ")
l(
mo d p
TC
and
(
Finally, since 2
n
,
pi ni
pr
nr
a x(m)
..)
mo d m)
for n = 0,
i(
cf>
Theorem
1,
a 60
71.
(mod 2800).
100
of the
that
ak
(a d ) 3
ar
l(mod m)
a r (mod m)
is
But d
is
l(mod
ra),
and there
Consequently, d k.
Corollary.
The exponent d to which an integer a, prime to m, belongs
is a divisor of 4>(m) and of X(m).
modulo
This corollary shows that we need try only divisors of X(m) to find the
exponent to which an integer belongs modulo m. For instance, to find
the exponent to which 7 belongs modulo 55, we try only the exponents
fore r
0.
2,
4,
5,
36(mod
10,
and
55), 7
20,
for X(55)
32(mod
55), 7
20.
10
r2
r3
r4
...
...
m
r h 2ri, 3fi,
(m
r m i
of
l)ri
ON BELONGING TO AN EXPONENT
101
under t.
Thus,
m=
if
modulo
7,
7 and
we form the
we wish
12
14
Then
3 (mod 7),
table
and so we move to 3
We
find 2
under 3, and hence 3 2 = 2 (mod 7). Moving to 2 in the first row, we find
Continuing in this manner, we find
6 under it, and have 3 3 = 6 (mod 7).
35
modulo
7.
34
6
4,
powers
5,
and
36
l(mod
The
7).
integer
of 3 are in order 3, 2, 6, 4, 5,
1,
and
3,
first
it is
therefore, belongs to
six positive integral
is
is
integers.
EXERCISES
Find the exponents to which 5 and 7 belong modulo 11.
Find the remainder when 7 182 is divided by 675.
3. Prove that if a and b are prime to 1729, then a 36 6 36 is divisible by 1729.
4. If p is a prime, (a, p) = 1, and a d = 6 d (mod p), where d is the least positive
exponent for which the congruence is true, then d is a divisor of p 1.
5. If a belongs to d modulo p an odd prime, and if d is even, then a d/2 = 1 (mod p).
6. If p is an odd prime, and if a k = l(mod p), where k is the least positive
integer for which the congruence is true, then 2k is the exponent to which a belongs
1.
2.
modulo
p.
102
a, where
< a < p 2, belongs to the exponent 3
modulo p.
belongs
to
6
p,
8. If the integer a belongs to d modulo p, a prime, show that the product of all the
distinct residues of the powers of a is congruent to 1 or 1 according as d is odd or even.
K(m) ~ (mod m) is a solution of ax = b (mod m) if
9. Show that x = ba
(a, m) = 1.
7.
Prove that
modulo
the integer
if
a prime, then a
Compare the fact that the powers of 2 will generate all the solutions of x 4
= l(mod 5) with the corresponding property of the root i of the equation x 4 1 = 0.
What do you notice about the other solutions?
10.
Theorem
74.
If
prime,
modulo
p.
When p =
p,
<f>(d)
xd
l(mod
integer
belongs to
p)
prime to p and
satisfies
a2 az
,
ad
(a 8 )*
(a d ) s
l(mod
p)
According to Theorem 73, furthermore, no two of these integers are conConsequently, these powers of a give all the solutions
p.
of x d = 1 (mod p) for the modulus is a prime and there are, therefore, no
more than d solutions. Hence, any integer that belongs to d is congruent
modulo p to an integer of the set a, a 2 ... a d and none of these integers
belongs to an exponent greater than d.
However, if (s, d) = 1, a s belongs to d modulo p, for if we assume that
a 8 belongs to k < d,
gruent modulo
(a 8 ) k ss
l(mod
p)
== l( mo(l
p)
and
a sk
But
since a belongs to d
than
On
modulo
p,
if (s,
= n ^
d)
d modulo
p.
so that s
1,
(a 8 ) do ss a Sod 3=
<j>(d)
Hence, d
sk.
k,
and k
is
not
less
d.
l(mod
d
p, for
<f>{d)
,
1, 2,
is less
ns
= nd
then
p)
powers a 8 where
d,
than
d.
and prime to
d,
exactly
ON BELONGING TO AN EXPONENT
103
If
1.
If
2.
d/n modulo p.
To prove Corollary 3, notice that if p = 2, d can be only
In the case of an odd prime, if d = d n, we saw that
Now
(a s ) d ==
l(mod
p)
l(mod
p)
<
with
<
Then d
do.
follows that do
Theorem
75.
This
t.
and a s belongs to
do
if s
impossible
modulo
When
But
st.
is
(s,
then
s n,
when
<
(s
is
odd.
(ab)
if
1, if
(a') '(&')'
modulo m,
a6 belongs to k
st
and k
<
ks, so
kt,
so that s
it
do,
a belongs to
and
s,
belongs to
so that
l(mod m)
l(mod m)
k.
kt
Since
b ks
l(mod m)
In like manner,
that t\k.
(ab)
Moreover,
st.
(ab) ks
and therefore
and
1,
so that
(ab) k
then k
do)
p.
t)
st
Therefore,
do.
and
for a
suppose that
(a s Y
But
1,
(s,
kt
l(mod m)
1, it
follows that
st
and
<
fc.
Consequently, k = st.
Theorem 76 (H. G. Erlerus, 1841). When pi and p 2 are odd primes,
if m = ai(mod pi) and m = a 2 (mod p 2 ), and if in addition ai belongs to
di
modulo
common
Since
pi
and a 2 belongs
to d 2
modulo p 2 then
,
m =
dl
l(mod
pi)
m<* 2
l(mod p 2),
(m dl) L/di
l(mod
(m d *) L/d2
l(mod p 2)
and
pi)
and
if
is
the least
common
104
Therefore,
mL =
But
if
m belongs to
modulo
l(mod
P1P2)
L.
Again, from
m =
l(mod
P1P2)
m =
l(mod
pi)
we
infer that
k
1,
k, d 2
k, and L
k.
Hence, k = L.
Examples. Because 3 belongs to 5 modulo 11 and 10 belongs to 2 for
the same modulus, and since the exponents 5 and 2 are relatively prime,
it is evident from Theorem 75 that 30, and hence 8, belongs to the exponent 10 modulo 11.
Again, 7 belongs to 10 for the modulus 11, and 5 belongs to 4 modulo 13.
Thus, according to Theorem 76, the integer 18 belongs to 20 modulo 143.
Therefore, di
EXERCISES
1. Find an integer that belongs to 2 modulo 19 and one that belongs to 3 modulo 19.
Using these results, find an integer that belongs to 6 modulo 19.
2. Set up the least positive residues of the powers of 2, 3, and 6 modulo 17.
3. Find all the integers that belong to 16 modulo 17.
10
= l(mod 31).
4. Solve the congruences x 3 = l(mod 7) and x
5. Show that 2 belongs to 12 modulo 13, and thus find the exponent to which 8
belongs modulo 13. Do any other integers belong to this exponent modulo 13?
Then determine
6. Find the integer to which 7 belongs modulo 5 and modulo 11.
the integer to which 7 belongs modulo 55.
n
7. When p is a prime, if a and b are prime to p, and if a = b (mod p ), show that
pr
pr
n+r
n+1
p
=
=
induction
and
hence
by
that
a
6
(mod
(mod
6
a?
).
),
p
p
8. When the modulus m is composite, prove theorems analogous to Corollaries 2 and
3 of
Theorem
If pi
9.
74.
and p 2
is
Generalize the
theorem.
73.
the integer
am
is it
true that
6
6, and so a
6 l(mod
2
21
is
m is a prime?
~l
l(mod m)
This question
is
l(mod 21) for all integers a that are prime to 21. Thus
2
Hence, 8 20 = l(mod21), and yet
21), and 8 = l(mod 21).
not a prime.
Is it
m is a
prime
a mi
if,
i(
prime to m,
mo d m)
in
ON BELONGING TO AN EXPONENT
but where m is a composite.
3(11) (17) satisfies the congruence
Any
is fulfilled
a 80
because X(561)
80.
105
(mod 561)
Consequently,
a 56o
i(
mo d
561)
An
Assuming that
But 0(m)
m is
< m
m>
1
1 is
composite, since
because
(a,
m)
1,
and m.
Therefore,
a prime.
E. Lucas,
Am.
J. Math., Vol.
1,
p. 301, 1878.
106
Theorem
78.
modulo p belong
integers from
to
Therefore,
d.
through p
if
we
represent the
let \p{d)
number
of
m)
<
0(d)
It is
1.
0(<
where the
with
di,
But we have
1.
4>(di)
0(d 2 )
1, 2,
also
+
.
r,
Hdr)
= V 
shown that
<f>(d 2 )
= p 
<f>(d r )
Because no yp{d ) can exceed the corresponding 0(d/), if any \p(di) were less
than the corresponding 0((&), these statements could not both be true.
z
Therefore, for
all
i,
f(di)
In particular
cf)(p
1)
0(cfc)
when
4>(cf>(p))
If
integers from
of 3
reduce to
3,
9(mod
33
37
39
3 11
3 13
3 15
modulus
belongs to the exponent 8 modulo 17,
Notice that 3 2
and
3,
17)
6,
and
17.
for
ON BELONGING TO AN EXPONENT
107
EXERCISES
Find all primitive roots of p = 5, 7, 11, and 13.
Prove that the product of all the integers that belong to a particular exponent
d > 2 modulo p, an odd prime, is congruent to 1 modulo p.
341
2 is divisible by 341 even though 341 is not a prime.
3. Show that 2
4. Find by trial composites m other than 561 such that a m_1 = l(mod m) for all
integers prime to m.
1.
2.
75. Gauss'
Method
Root Modulo
p.
To
find
None
modulo
p.
satisfies
the congruence
select
z
,
a\ dx
x dl
Now
ai
l(mod
less
p)
hd %
l(
mo d
p)
and
a:
congruence.
If d 2 is a multiple of d h but not p 1, we have found an integer that
belongs to an exponent modulo p that is greater than d\.
Then according
Theorem
to Corollary 3 of
for
if
p that
(b 2
1.
Moreover, di/bi
1.
But
(6i,
i)
= bd
= 1 also,
2
i,
is
is
d 02 )
74, ai bl
modulo
d
and
d\
2
d\d 2 /b,
p,
and
this
exponent
is
the least
common
multiple of
108
41.
Because the integer 3 does not occur among the power residues of 2, we
determine the power residues of 3 modulo 41. They are 3, 9, 27, 40, 3,
9, 27, 40. Consequently, 3 belongs to 8 modulo 41.
We know
Since (20, 8) = 4, we use the factors 4 and 1 as &i and b 2
Consequently,
that 2 4 belongs to 5 and 3 belongs to 8 modulo 41.
.
16
is,
The power
residues of 7
7,
common
modulo 41 are
35,
40
5,
modulo p n
n
<f>(<j>(p
))
incon
Suppose that the integer a is a primitive root modulo p. This statement means, of course, that every integer in the class with a belongs to
We shall show that we can choose
the exponent p 1 for the modulus p.
Specifically,
primitive
root of p n
is
a
class
that
residue
this
in
an integer
.
if
a pi
we
shall
prove that a
is
shall
mo d
2
)
aP 1
we
= i(
l(mod
p'2
if
kp
satisfies
the condition
(a
kpY~
l(mod p 2 )
and that
First, let
l(modp
).
us suppose that
Then
when a belongs
n
.
to
modulo
p,
av
ON BELONGING TO AN EXPONENT
(a
kp) p
~l
=
=
and consequently
is divisible by p.
a? 1
ap
(a
+~(p 2
kp)
there
l(mod p
r*
= p n~
l(mod
divides
(fcp^^mod p 2 )
modulo p 2
if
and only
if
/cp will
ftp)*'
l(mod p 2 )
kp of p such that
p_1
).
an exponent
4>(p
congruent to
is
to
is
modulo p but
(a
is,
kp
If
that
~2
(mod p
/cp
p~
l)a p
109
But
congruence
if s
r*
n)
But does
1,
More than
p
1).
that, since
it is
so,
is
belong
a divisor of
that p 1
has the form p s (p 1) with s = 0, 1, 2,
or
did not exceed n 2, we could raise each member of the
l(mod p n ) to the p n ~ 2 ~ s power and obtain
and
Hence,
if.
1.
r<^ = l(mod p n ).
smaller than 4>(p n )?
If
p)
modulo p
(p
p)
(r,
belongs to
modulo
p, it is clear
pn2(pl)
l( m()( J pn)
We shall show, however, that the last congruence cannot be true and
hence that r cannot belong to an exponent less than </>(p n ) modulo p n
To do so, use the fact that
.
r vi == i(
in the
(r p
i
form
rp
(1
n *
~l
cp,
where
mo d
p)
(c,
<p)
1.
Then
cp )2
+
+
The (m
l)st
term
p
n2(
of this
n2
expansion
is
j)
n2
 m+
1/
m!
obviously a multiple of p n
is
it
n2
1)
(m
1,
that
is,
(yn2
^T=
_m+
j)
if
m>
is
an
integer.
;,n_2
(mod p n )
1)
and
(cp)
and
p
if
(m, p)
But n
m =
fcp'
1,
for
with
1):
p ^' >
>
>
kp
n.
But
fcp*
t + 2.
Each
term
p
after the second of the expansion is, therefore, a multiple of p n
Hence, f or n > 3
(k,
n
p)
{
r p2(pl)
_j_
pnl c
mo(J p n)
110
and because
(c,
p)
1,
rpHP i) = i(
n =
If
2,
we know
mo d
n
)
that
rp
l(mod p 2 )
i(
whereas
r p( P
Consequently,
Now
i)
r is a primitive root
2
)
>
modulo p n where n
,
1.
Each
mo d
l(mod p n )
==
of the integers
y
yP n_1 (p 1)
y2
satisfies this
prime to
is
They
p.
are,
assuming u
s
>
r v(
mo d
pn)
w, u
1, 2,
that r w_ "
we conclude
n~ 1
(p
1)
l(mod p n ).
This result is
than p
Hence, these </>(p n )
(p 1).
integers are distinct solutions modulo p n of the conditional congruence.
But only integers prime to p can be solutions of this congruence, and since
there are exactly <t>{p n ) integers from 1 through p n that are prime to p, we
have found all the solutions.
Furthermore, the integers that belong to 4>{p n ) modulo p n must be
v,
impossible because u
n~ 1
v is less
in
only
if
s is
prime to p
n~ 1
(p
1).
There
are,
then, exactly
n
<f>(4>(p ))
3 16
l(mod
prime to 0(17 2 )
primitive root of 17 3
17 4
....
EXERCISES
2.
ON BELONGING TO AN EXPONENT
4. Prove that
d modulo p, and
n ~h
if
if
is
is
111
modulo p n ?
Theorem
roots
710.
modulo 2p n
Any
cf>(<j>(2p ))
l(mod 2p n )
.<f>(2pn)
of
is
tive root of
n
,
if
2p n must be odd.
n
it
1) modulo 2p
,
pni( P i)
i(
mo d
a pi(pD
i(
mo d p n
implies that
But
is
if
an
a primi
for
and
incongruent primitive
a belonged to d
l(mod 2p n ).
The
n~ 1
n
then since a is odd,
(p 1) modulo p
congruence
is,
of
course,
impossible, for a is
last
for
the primitive roots
We must, therefore, look
<
a
a primitive root of 2p n
of 2p n among the primitive roots of p n
.
2p n )
But
if r is
is
because
it is
in the
same residue
class as r
(f>(<l>(2p ))
It is also
112
49(mod 289)
is
EXERCISES
Find all the primitive roots of 50 and 250.
Find the primitive roots of 98. Determine the integers that belong to the
exponent 3 modulo 98. Find a primitive root of 686.
3. Prove that the congruence x^ 2 ?") = l(mod 2p n ) has exactly 4>{2p n ) solutions
modulo 2p n if p is an odd prime.
4. If r is a primitive root of 2p n p being an odd prime, show that r 8 belongs to
1.
2.
if
and only
if (s,<j>(2p ))
1.
rs
When m =
2 n2
modulo
2n
"primitive X roots of
>
where n
for
to be congruent to
if
m.*
2,
modulo 2 n
That
2 "" 3
would have
(22
I)*
so that, for
=
n
>
3,
2n
~3
and hence
it is
in this case 3
2* 4 (2" 3
2n 3
2n " 3
is
pk
*,
2n  2
(mod
2")
3,
X(2 3 )
2 and
we have now
is
mi =
where the p
2 i(mod 2 n )
with p a prime.
there is always a primitive X root of m.
tive X root of
1)2 4
is
l(mod 2 n ). When n
2 modulo 2 3
this relation
22
i}
ni
pi p2
with
n2
Pki
2,
1,
He showed
nk  1
.
and
that
if
is
a primi
r 2 is a primitive X root of
k,
si r 2 (mod
ri(mod mi)
and
is
a primitive X root of
suppose that
m =
belongs to
mip k nk
s for
R. Carmichael, Bull
")
To show
the modulus m.
r s ss
*
pk
is
true,
Then
l(mod mi)
ON BELONGING TO AN EXPONENT
113
and
rs
But
can be
is
the least
exactly A(m).
is
l(mod p k Uk )
modulo mi and
and \(pk nk ) divide
r belongs to A (mi)
m) =
modulus p& n *.
common
It is
The
If
Therefore, r
1.
solution
is
we have a set of
17 5 = 5(mod 21).
0(6)
=2
of 17 that
is
They
A (21)
6,
are 17 and
There are but two incongruent primitive roots of 7, and they are 3 and
When we use the integer 5 with the only primitive root of 3 to form
the set of congruences
x = 5 (mod 7)
x = 2(mod 3)
5.
is x = 5 (mod 21).
But the set of powers, 5 and
two primitive A roots of 21 already found. This situation,
however, does not mean that there are no other primitive A roots of 21,
for the number 2 belongs to 6 modulo 21.
Moreover, 2 and 2 5 =
II (mod 21) form a new set of two primitive A roots of 21.
we
5 5 repeat the
,
H. P. Lawther,
Jr.,
114
EXERCISES
Prove that
if r is
is
*Theorem
711.
If
>
and
not
d,
divides 2
2,
2n
Modulo
n~ 2
modulo
2 n but
,
if
2,
modulo 2\
We have shown that the integer 3 is a primitive X root of 2 n if n > 3.
On this basis we shall show that for the modulus 2 n any integer having
the form 8/c + 3 cannot belong to an exponent smaller than 2 n_2
If
there are just three incongruent integers that belong to d
>
3,
(8/c
3)
2 " 3
(3
8/c)
2 " 3
2n ~ 3
2 "  1
3
2* 3
(2 0
(3)
~
W 3
2 1
3
3
3(2
/c)
(2
/c)
 (mod
3
2")
Hence,
(8k
But
since 3
2n " 3
l(mod 2 n ),
2 " 3
3)
8/c
32
"~ 3
(mod
2")
3 belongs to 2 M
"2
modulo 2 n
if
>
3.
yielding
all
less
than 2\
X root of 2 3
modulo 2 3
Again, if n >
having the form
3,
8/c
1,
in
which k
2" 3 (2 3 /c)
is
for the
prime to
modulus 2 n
2,
integers
all
2"~ 3 for
,
(8/c
On
(8/c
l)
2 "" 3
I)
2"
4
=
=
if
l(mod 2
n >
(2
2 " 3
/c)
2")
4,
2" 4 (2 3 /c)
2" 5 (2 4
3
2
1)(2 A0
+
=
(mod
(2
2n_4
A)
(mod
2 n)
2"" /c(mod 2 n
)
1
Hence, if (/c, 2) = 1 and n > 4, the integers 8k 1 belong to the expoMoreover, in this case values of k that are prime
nent 2 n_3 modulo 2 n
~
2 n_3 in 4>(2 n s ) ways,
can
from
the integers 1,2,3
2
be
chosen
to
.
ON BELONGING TO AN EXPONENT
and 7 belong to
~4
(2V
also 2 4
l)
2n " 4
if
2r with
l(mod
1 through
not only the integers 9
15 belongs to 2 modulo 2 4
4,
2)
(r,
1,
2n)
is
we
Similarly,
But
if
(8Jfe
divisor of 2 n
and that 2 4 r
modulo 2
2 modulo 2 4 but
We
2 n_3
115
l)
>
n~a
= 1 + 2 (2
= l(mod2 n
>
s
s
r)
3,
8
(2 r)
2n
~'(mod 2 n )
If,
in addition,
s
(2 r
l)
2 "" s_1
2)
(r,
>
and n
2 s  1
s
(2 r)
1,
n ~ s 2
{2
n s~ l
l)(2 s r) 2
(2Y)
2n ~ s1
(niod 2 n )
2 w V(mod 2 n )
belongs to 2 n
~s
modulo 2 n
n~ s
<f>(2
n~ 2
of
=
=
~
to 2 n 2 modulo 2 n
~
2 n 3 belong to 2"~ 3
~
~
2 n 4 belong to 2 W 4
2n
them belong
n 3
2</>(2
)
n~ 4
)
2<K2
20(2 n
2,
~s
)
~s
+ 1=3
20(2)
1 belongs to
But
22
23
accounted for
all
Therefore, the
number
the 2
n~ 1
ft
When
to
>
ft
~s
belong to 2
modulo 2 n
2n
~2
2 n ~\
of integers given in
when
belong to 2 n
3.
3,
modulo 2 3
116
The congruence x 2k
Corollary.
2 k+1 solutions
'2 n
l(mod
<
where
),
<n
2,
n > 2.
Theorem 711 shows that the congruence x 2 = l(mod 2 n ), for n > 2,
has 2 2 solutions modulo 2 n since the solutions must be prime to 2 and
2k
belong to the divisors of 2. For a like reason x = l(mod 2 n ), with
2
3
has 1 + 3 + 2 + 2 +
1 < k < n
2,
+ 2 k = 2 k+l solutions
n
modulo 2 if n > 2.
*Theorem 712. If d is a divisor of X(2 n ) and n = 0, 1, or 2, there is
exactly one positive integer less than 2 n that belongs to d modulo 2 n
has
if
(f>(p
Modulo p n
*Theorem
713.
If
(a
rap^
a*
Hence,
if
(m, p)
(a
(mp) (mod p 2 )
fc
1,
mp)*
l(mod p 2 )
1.
cp with
3= 1
+ ka  mp +
+ ka ~ mp(mo&
(c,
(r
&),p
(1
cp)
wpn
and
(r*)^""
~\mod p n )
l
wcp n (mo& p n )
l(mod p n )
=z
But
(^p!
Therefore, r belongs to kp
Furthermore,
if
== i(
n~ l
mo d
n
)
modulo p when n
ft modulo p and
>
1.
a belongs to
fc
ss
l(mod p s_1 )
l(mod p s )
but
a*
there
is
an
a(mod
p*),
where
r k ss
for since a*
(a
+ mp
2p
s
s_1
with
k
ss 1
(,
1, 2,
1,
such that
l(mod p s ),
p)
\
=
=
1,
ft_1
fca
s_1
mp*~ 1 (mod p s )
+ ka k ~ mp ~ (mod p
l
ON BELONGING TO AN EXPONENT
But an integer
117
~l
mp
~l
tp s
~x
{mod p s )
ka k
Hence,
+ mp ~
s
{a
Let r
i
1,2,
rap s_1
Then
~l
m=
k
(mod p)
l(mod p s )
ss
/c
for each
p*,
where
s.
gp
gp
l(mod p s+1 )
as
v is in
r for
(v
=
=
^
=
a) WP i
but
(^)p
n_s
~s
+ cp wpn ~\mod
+ wcp^^mod p w
s
n
)
l(mod p n )
l(mod p n )
~s
Therefore, v belongs to kp n
implies that there
(1
is
modulo p n
modulo p n
*Theorem 714.
.
Moreover, a
if
(s,
d)
modulo p n
where
1,2,
Therefore,
1.
<j>(d)
Consequently,
l(mod p
n
)
d,
if \j/(d) is
the
number
of incongruent integers
n
,
and so
f(di)
where the
di,
with
+
i
ffa)
1, 2,
Hdm) = v n~\v 
l)
4>{p
over,
<f>(d 1 )
\f/(di)
4>(d 2 )
<f>(di).
<t>(d
m)
= p\p 
1)
).
More
118
*Theorem
2p
belongs to
If r
<t>(2p
modulo
integers
modulo
are
modulo 2p n
of (f>(2p n )
r,
r2,
r*
(2pn)
.
These are
l(mod 2p n )
the integers from
all
through 2p n
if
4>{2p
(s,
))
1,
rs
belongs to
n
<f>(2p
)/d modulo 2p n
modulo 2p n
d
n
<j>(2p ))
(s,
But
if
ranges over
belongs to
d,
all
<f>(2p
and so does
n
<f>(2p
the integer
<f>(2p ),
is,
n
<j>(2p )/d.
(5<f)M2 P )/<i
and
divisors of
There
divisor of
n
...
2,
1,
any
&
*(2p)
l(
m0 d
2p
(2pn)
n
)
the congruence
b satisfies
gruent to a power of
divisor
<j>(2p
author's students, Bernard Sussman, and the author have developed the
following proof of a
of
method
but also
It
all
if
m =
n
]
J pj
>,
a primitive X root
a,
can
common
solution of the
set of
= ^(mod p/0
,r
j = 0, 1,
root of m.
We have seen, moreover, that
x
is
a primitive X
solution
generates
by means
of its
powers
v
,
(1)
with
this
common
prime to X(w),
<f>(\(m))
none
of
which
is in
a pre
ON BELONGING TO AN EXPONENT
ceding
set.
119
m is
it is
<t>(k(pj
nj
))
^(X(2"))(mod2 w
<>)
Y[
yi
cf>(\(pj
0)
congruences
(1),
Y is
s ^(mod
This integer
pfi)
0, 1, 2,
... ,r
YHm)
i(
mo d
YHm)
l(
mo d m
and that
py
(2)
evident that
it is
w ')
to da
modulo
pj
nj
,
=1
r
the d^.
\l/(di
) (mod
2 n )
<f>(da)
=1
determined by one choice of the
incon
0, 1,
r,
having least
obviously a multiple of
common
dy,
where
number
<f>(\(m)).
r
Each
set of
\l/(d i0 )
(mod 2 n )
=1
120
of
of
</>(X(m))
i3
gi
where
Si
and
s2
Sl
gi
(mod m)
s<l
S2t
g2
(mod m)
1*1
g2
**i
g 2 (mod
m)
is false.
There
will,
there
fore,
*(<fco)(mod2) II *(*/)
0(X(m))
more, assuming that d^h is distinct from di2h and that /3 belongs to d^h
modulo ph nh and 7 belongs to d^n modulo ph nh let us suppose that (3 V =
7 (mod m) with v and w prime to X(m). Then /3 U = 7 w (mod p^**). But
,
i>
and
(pv)*i x *
d
((3
hh) v
l(mod
p^
However,
nA
)
Hence,
7 >)<V
= l(mod prt
wd^.
and thus
d,,*
d it h divides
If
divides
d^h,
and
finally d^h
dij,
d;^,
which
In like manner,
is
m determined by
m and which is, thereprime to each of the p3 ni belongs to exactly one exponent that is a
Hence, from all possible choices of sets of
divisor of \(p3 nj ) modulo p3 nj
r,
r + 1 positive integers d i3 where i = 1, 2,
q3 j = 0, 1, 2,
fore,
ON BELONGING TO AN EXPONENT
121
which the least common multiple of the da in a set is X(m), we find all
the incongruent primitive X roots modulo m, for if a belongs to X(ra)
for
modulo m, then
a\(pfi)
i(
mo d
py
dij
0, 1,
\(pj
ni
).
If, however, the least common multiple of these dq were u < X(m), a
would belong to u modulo m. Thus when \f/(X(m)) (mod m) is the number
of incongruent integers that belong to X(m) for the modulus m, we have:
*Theorem
If
0,
1,
2,
r,
are distinct
^(X(m))(modm) = V
i(dU)(mod
2") [] <K<fe)
y=i
9
r
m=
1, 2,
qj,
of the X(py nj )
whose
least
common
summation
multiple
is
X(m)
and where each of the q sets contains just one divisor of each of the
ni
These primitive X roots of m can be separated into subsets, of
\(pj )>
0(X(m)) integers generated by any member of the subset.
It is, furthermore, obvious that the method with an analogous proof
holds for finding the integers that belong to any divisor of X(m) for the
modulus m.
Examples. At the end of Sec. 77 we showed that the primitive roots
of 7 and 3 yield exactly two distinct primitive X roots of 21.
But it is
evident that X(21)
common
=6
multiple of 0(7)
and <(3) = 2, but also as the least common multiple
of the factors 6 of 0(7) and 1 of 0(3), and again as the least common multiple of the factors 3 of 0(7) and 2 of 0(3), and in no other way from
factors of 0(7)
and
0(3).
= 3(mod
= l(mod
7)
3)
x
x
=
=
10 5
The
19(mod
5(mod
l(mod
of 21.
belongs to
7)
3)
21).
integers 2
7,
and 2 belongs
x
x
=
s
2(mod
2(mod
7)
3)
=
=
4(mod
2(mod
7)
3)
to 2
modulo
3.
122
observe that 2 5
11
primitive X roots of
least
common
factor of X(7).
= 2, and X(7) = 6, we
2,
which X(168) = 6 can be set up as the
multiple of one factor of X(2 3 ), one factor of X(3), and one
These sets of factors are listed in the following table.
must determine
23
all
Since X(2 3 )
7.
possible
ways
X(3)
in
Number
Case
Factors of A(3)
Factors of X(2 3 )
Factors of A (7)
of incongruent
primitive X roots of
168 determined
(1)
(2)
(3)
(4)
(5)
(6)
2
2
2
2
6
2
6
(7)
Case
1.
The
sets of congruences to
x
x
x
x
x
x
x
x
x
The
=
=
=
=
=
=
=
=
=
3 (mod 8)
be satisfied are
=
=
=
=
=
=
=
=
=
x
x
x
2 (mod 3)
3 (mod 7)
5 (mod 8)
x
x
x
2 (mod 3)
3 (mod 7)
7 (mod 8)
2 (mod 3)
x
x
x
3 (mod 7)
3 (mod 8)
2 (mod 3)
5 (mod 7)
5 (mod 8)
2 (mod 3)
5 (mod 7)
7 (mod
2(mod
5 (mod
8)
3)
7)
5,
143,
and 47, respectively, and they occur in sets of 0(6) = 2, which are
5 and 5 5 = 101 (mod 168); 47 and 47 5 = 143 (mod 168); and 59 and
59 5 = 131(mod 168).
Case
The
2.
The congruences
solutions are x
are
x
x
x
= l(mod
= l(mod
= 3(mod
73 and x
x
x
x
8)
3)
7)
73 5
=
=
=
l(mod
l(mod
5(mod
145(mod
8)
3)
7)
168).
ON BELONGING TO AN EXPONENT
Case
3.
The congruences
x
a;
The
solutions are x
Case
4.
ss
l(mod
2 (mod 3)
7)
17 and x
a;
ss
=
=
x
x
x
a:
3=
a;
s=
3=
x
x
x
8)
3(mod
x
a;
are
s
=
The congruences
17 5
3(mod
l(mod
3(mod
5(mod
l(mod
3(mod
7(mod
l(mod
3(mod
The congruences
x
x
x
The
solutions are x
Case
6.
x
x
=
=
7)
a;
s=
8)
a:
3)
x
x
7)
3)
x
x
7)
a;
8)
x
x
x
a;
3e
3=
a;
3s
=
=
3(mod
l(mod
5(mod
3=
5(mod
l(mod
5(mod
=
=
=
7(mod
l(mod
5(mod
8)
3)
7)
8)
3)
7)
8)
3)
7)
and
103, respectively,
168).
=
=
=
x
x
x
8)
65 and x
168).
65
l(mod
8)
2 (mod 3)
4(mod
137(mod
7)
168).
are
3(mod
l(mod
4(mod
x
x
x
=
=
=
=
=
=
8)
a;
3e
3)
a;
7)
a;
7(mod
l(mod
4(mod
3(mod
l(mod
2(mod
8)
5(mod
l(mod
2(mod
8)
7(mod
l(mod
2(mod
3)
7)
3)
7)
x
x
x
=
=
5(mod
l(mod
4(mod
8)
3)
7)
8)
3)
7)
8)
3)
7)
modulo 168 are 163, 67, 37, 109, 79, and 151.
= 109(mod 168); 67 and 67 5 = 163(mod
151(mod 168).
solutions
79 6
l(mod
7)
3=
7)
are
2(mod
=
=
=
=
8)
5(mod
89(mod
3)
2 (mod 3)
The congruences
x
x
x
The
=
=
=
l(mod
2 (mod 3)
8)
3=
=
=
are
The
Case
123
They form
168); 79
and
124
Case
The congruences
7.
x
x
x
x
x
x
x
x
x
The
are
= 3(mod
= 2 (mod
= 2(mod
= 5(mod
= 2(mod
= 2(mod
= 7 (mod
= 2 (mod
= 2 (mod
8)
hh
3 (mod 8)
3)
7)
=
=
2 (mod 3)
4 (mod 7)
8)
3)
s=
7)
5 (mod
2(mod
4(mod
=
= 7 (mod
= 2(mod
= 4 (mod
8)
3)
7)
8)
3)
7)
8)
3)
7)
modulo 168 are 107, 11, 149, 53, 23, and 95, and they form
= 107 (mod 168); 23 and 23 5 = 95(mod 168); 53 and
149(mod 168).
solutions
53 5
EXERCISES
1.
2.
3.
4.
5.
Find
Find
Find
Find
Find
CHAPTER
INDICES
When p
Prime Modulus.
primitive root of p,
r,
is
2
.
a prime and r is a
~l
are incongruent
rp
modulo
p.
integers
1, 2,
n
Gauss
To
called the
exponent
we
s of r
(mod
p)
write s
when
there is no danger of confusion as is true in the case of the congruence r indn = ft (mod p).
As thus defined, the index of n modulo p is unique for the modulus
and
1,
if
for
>
if
t,
*
l(mod
2(mod p
(mod
p)
then
p)
and
It
is,
1)
s
Obviously the index of n
r (mod p) as the index of n modulo p.
n
modulo 2 is useless.
Of course, the index of n determined by the primitive root r may be
For
different from that determined by another primitive root of p.
instance, for the modulus 7, ind 3 2 = 2(mod 6) but inds 2 = 4(mod 6).
Theorem
81.
If
m=
n(mod
a particular
md mn =
r
ind r
m+
ind r
n(mod p
1).
125
ELEMENTARY THEORY OF
126
Since
r indn (mod p)
and
m
it
53 r indm
(mod
p)
follows that
mn =
indm+indn (mod
mn =
indmn (mod
mn =
ind r
p)
But
p)
Therefore,
ind r
Theorem 83.
n(mod p
k ind r
If
m+
ind r
a prime and n
is
w(mod p
\)
prime to
is
p,
then ind r n k
1).
If
r indn (mod
nk
r fcindn (mod p)
nk
r indn^ m()CJ p)
ind r n k
k ind r
p)
then
Also
Hence,
n(mod p
1)
It is evident from these laws that the index of an integer plays a role
which is analogous to that played by the logarithm of a number. This
analogy is further emphasized by the following formula for changing the
base of a system of indices from one primitive root of p to another
Theorem
ind r2 n ind ri
Let
ri
84.
If
r 2 (mod
and
r2
a prime and n
is
is
prime to
p,
then ind ri n
1).
Then
p.
rh
ri
ind r2 n ind n
1)
ind ri n
1)
ind ri n ind ri
r 2 (mod
or
Theorem
ind n
r2
ind r2
85.
ri
ind ri
r2
ind ri r2
ind r2
r x ss
roots of p, a prime,
then
r 2 (mod p)
ind r2
r2
(mod p
1)
and
ind r r 2 ind r2
,
If
we
r\
l(mod p
1)
r of p,
we
127
INDICES
+ n ind x =
ind 6(mod p
1)
Therefore,
n ind x
and unless d
(n,
ind 6
divides ind 6
1)
a(mod p
ind
ind
1)
no value of ind x
is no solution of
a,
But
if
(ind b
ind
a),
ra
2ra
as
(d
l)ra
Therefore,
j.
r i+A;mo(
mo ^
p)
0, 1,
By
referring to a table of
determined.
Of course,
this
method
of solving a
indices:
ind
10
10
202.
1,
p. 185.
128
Then
ind 5
ind x
ind x
7(mod
ind
a;
3 (mod 10)
= 8(mod
or
10)
and
Therefore,
2.
Solve: 7x*
= 3(mod
ind 7
11)
11).
3 ind x
=
=
ind x
7 (mod 10)
7(mod
3 ind x
Hence,
and
3.
Solve:
3a:
2(mod
ind 3
4 ind x
4 ind x
But
(4,
10)
and 2
2,
11)
11).
3,
Jf
=
=
so that there
is
no solution
of the given
congruence.
4.
Solve: bx 2
3(mod
11).
ind 5
=
=
2 ind x
2 ind x
(2,
10)
2 and 2
4.
ind x
2(mod
10)
4(mod
11)
ind x
7(mod
10)
7(mod
11)
and
5.
Solve: 7x
Any
4(mod
121).
ind 7
ind x
ind x
Hence, x
the form x
10(mod
10
11),
and
11/c.
7(10
all
must
=
=
satisfy 7x
4 (mod 11).
But
Therefore,
11/b)
4(mod
77k
55(mod
7k
5(mod
121)
or
121)
Hence,
11)
129
INDICES
and
ind 7
ind k
ind k
= 7(mod
10)
7(mod
11)
87(mod
so that
and
Therefore,
82. Euler's
moduli
Solvability of x n
for the
Criterion
121)
c(mod m).
For
test
= c(mod m) with
= b (mod m) with a
form x n
m) =
1.
Because any binomial congruence ax
prime to m can be reduced to this form by multiplying each member
of the congruence by the solution of ax = l(mod m), the problem is also
(c,
and
fc
Conversely,
<j>(m)/d
s4>(m)/d
<}>(m)\s/d
(mod
Til)
if
c <Km)/d
i(mod m)
r s<t>(m)/d
i(
then
Therefore, scf)(m)/d
result,
is
values modulo
cf>(m)
s(mod
of k.
and s/d
is
4>{m)) is satisfied
by
a multiple of
the congruence nk
mo d m)
<t>(m),
an
integer.
As a
just d incongruent
modulo
c(mod m).
If p is a prime and d = (n,p 1), there are (p l)/d
incongruent values modulo p of c, prime to p, such that x n = c(mod p)
Theorem
87.
has a solution.
According to Theorem 86, the congruence x n = c(mod p) has a solution
~ 1)/d =
if and only if c (p
l(mod p) where d = (n, p 1). But the con~ 1)/d
ip
= l(mod p) has a solution. There are, thus, exactly
gruence x
130
integer for a
same
c is
modulus
if
modulo p
one
first
an integer
c is
and these
of 'this congruence,
solvable.
is
and
if
(n,<(m)), the
^m)/d == i( mo d
if
the congruence x d
fc
fc
we turn
EXERCISES
2
Prove that the least positive residues of l 2 2 2
(p l) modulo p, where
p is an odd prime, repeat themselves exactly twice.
2. If p is a prime and n is prime to p 1, prove that the integers l n 2 n 3 n
n
Thus show that if p is a prime of
(p l) form a reduced residue system modulo p.
3
the form 3n + 2, the integers l 3 2 3 3 3
(p l) form a reduced residue system
1.
modulo
p.
is
a.
x3
b.
x*
6.
if
(a,
7.
a.
b.
= 5(mod
= 7(mod
13)
13)
Show
that
5x
5x 2
m 4 (mod
= 6 (mod
13)
13)
INDICES
8.
9.
10.
131
the congruences:
a.
b.
c.
d.
Sx
3x
7x
x3
=
es
2 (mod 193)
191 (mod 193)
157 (mod 193)
=
= 64(mod
193)
132
192
34
84
4
68
118
104
8
102
168
10
35
11
12
17
152
14
138
16
183
13
141
15
ind
85
136
31
18
10
145
20
69
21
188
22
25
23
162
24
25
186
26
175
27
60
28
172
29
123
30
119
31
82
32
170
33
75
34
65
35
105
36
44
37
ind
38
179
39
33
40
103
41
151
42
30
43
24
44
59
45
169
46
4
47
29
48
28
49
16
50
36
51
115
52
53
77
54
94
55
184
56
17
14
57
37
58
157
59
148
60
153
61
63
80
12
65
142
66
109
67
18
68
99
69
47
62
116
64
ind
54
70
139
71
177
72
78
73
91
74
39
75
86
76
21
77
95
78
67
79
167
80
137
81
144
82
185
83
122
84
64
85
32
86
58
87
88
93
89
147
90
15
91
53
92
38
93
166
94
63
95
146
96
62
97
158
98
50
99
159
100
70
ind
n
ind
n
ind
n
ind
n
ind
n
ind
19
11
101
103
107
104
51
105
189
111
107
154
108
128
109
160
110
134
102
149
106
ind
111
89
112
48
113
41
114
71
115
163
116
191
117
117
118
182
119
135
120
187
121
174
122
81
123
43
124
150
125
ind
126
114
127
13
128
46
129
108
130
176
131
132
143
133
57
134
52
135
61
136
133
137
110
138
88
139
190
140
173
143
132
144
112
145
124
146
125
147
100
148
73
149
155
150
120
ind
ind
20
26
141
113
142
ind
151
155
83
156
101
157
140
159
161
160
154
129
158
126
152
55
153
ind
161
19
171
ind
74
162
178
163
23
164
27
165
76
166
156
167
79
168
98
169
90
170
66
n
ind
171
121
172
92
173
165
174
49
175
106
176
127
177
40
178
181
179
42
180
45
181
182
87
183
131
184
72
185
6
186
8
187
22
188
97
189
164
190
180
ind
56
191
ind
130
192
96
INDICES
133
2
25
125
4
46
14
56
{Continued)
37
185
153
8
186
158
15
16
87
49
17
52
18
67
142
20
131
10
18
ind
11
12
90
64
13
127
ind
21
76
22
187
23
163
24
43
25
22
26
110
27
164
28
48
29
47
30
42
ind
31
17
32
85
33
39
34
35
10
36
50
37
57
38
92
39
74
40
177
41
113
42
179
43
123
44
36
45
180
46
128
47
61
48
112
49
174
50
98
51
104
52
134
53
91
54
69
55
152
56
181
57
133
58
86
59
44
60
27
61
135
62
96
63
94
64
84
34
66
170
67
78
68
4
69
20
70
100
71
72
184
73
148
161
75
33
76
165
77
53
78
72
79
167
80
63
83
155
84
85
15
86
75
87
182
88
138
89
111
90
169
n
ind
n
ind
n
ind
n
ind
74
65
19
114
ind
81
122
82
91
73
92
172
93
88
94
54
95
77
96
192
97
188
98
168
99
68
100
147
101
156
102
8
103
104
40
105
35
106
175
107
103
108
129
109
66
110
137
111
106
112
144
113
141
114
126
115
51
116
62
117
117
118
6
119
30
120
150
121
171
122
83
123
29
124
145
125
146
126
151
127
176
128
108
129
154
130
191
131
183
132
143
133
136
134
101
135
119
136
16
137
80
138
14
139
70
140
157
141
13
142
65
143
132
144
81
145
19
146
95
147
89
148
59
149
102
150
124
151
41
152
12
153
60
154
107
155
149
156
166
157
58
158
97
159
99
160
109
161
159
162
23
163
115
164
189
165
173
166
93
167
79
168
9
169
45
170
32
ind
171
160
172
28
173
140
174
121
175
26
176
130
177
71
178
162
179
38
180
190
ind
181
178
182
118
183
184
55
185
82
186
187
120
188
21
189
105
190
139
191
116
192
n
ind
n
ind
n
ind
n
ind
n
ind
n
ind
n
ind
n
ind
n
ind
31
11
24
CHAPTER
QUADRATIC RESIDUES
91. The General Quadratic Congruence
Theorem 91. A solvable quadratic congruence a y 2 + a y + a 2
0(mod ri), where a ^ 0(mod n), can be reduced to the form x 2
a(mod m) in which (a, m) 1.
By multiplying the modulus and each member of the congruence
x
by 4a
we
?/
dry
a2
0(mod
=
=
n)
4a
y + 4a
aii/
4a a 2
0(mod 4a
n)
2a y
ai
4a n
z(mod m)
4a a 2
6(mod m)
(2a y
Now let
ai)
4a a 2 (mod 4a n)
and
i
The
form
b(mod m)
ekw(mod m)
k 2w 2
b (mod
kw 2
m)
or
If (k,
m =
)
s,
member
is
of the last
s
6 there is no solution.
no solution unless s = 1.
congruence by k, and let
unless
sequently, there
(mod
= kw (mod m
134
But
(6
If s
ra
1,
1.
Con
multiply each
QUADRATIC RESIDUES
135
Then
x2
and
if
we
s= b
k(mod
set
b k ss
a(mod
and
linear congruences.
Therefore,
12,
42(mod
w =
1,
4 (mod
5),
60).
EXERCISES
Solve the congruences:
a.
x2
b.
x2
=
=
28 (mod 84)
64(mod420)
When
(a,
m)
is
example, 2
is
satisfies
x2
2 (mod
7),
is no solution of
but 2 is a quadratic nonresidue modulo
x 2 = 2 (mod 5). Can you find a number whose square gives the remainder 4 when it is divided by 15?
If an integer a is prime to m > 0, its quality of being a quadratic
residue or nonresidue modulo m is called its quadratic character with
respect to m.
Obviously all integers in the class with a modulo m have the
same quadratic character with respect to m.
The problem of determining the quadratic character of a is, therefore,
equivalent to that of testing the solvability of the congruence x 2 =
a (mod m). We have already shown that by factoring m into powers of
primes we can reduce the discussion to the question of solving the congruence x 2 ss a (mod p n ) with p a prime, and finally to the case x 2 =
5 because there
a(mod p). Theorem 513 shows that when p is an odd prime, these last
two congruences either are both insolvable or have the same number of
solutions,
and therefore
in this case
it will
136
were all very much interested in the theory of quadratic residues, but
Gauss (17771855) was the one who contributed most to this subject.
The fact is that Gauss was one of the greatest mathematicians of all
It is really no wonder that his name is connected with so much
time.
that has been produced in the theory of numbers, for it was his favorite
He considered it the " queen of mathematics." His "Disstudy.
quisitiones arithmeticae," published in 1801, is the classic of the theory of
numbers and exhibits very well the elegance of form and rigor of presentaSome of the topics to which he made great
tion for which he is noted.
contributions are quadratic forms, biquadratic residues, and the theory of
congruences.
Theorem 92
is
diX
a(mod
p)
1, 2,
a (mod p)
a k a,j
a (mod p)
p.
and
i f^
then
di
= a^mod
p)
whereas these integers are distinct modulo p. The integers a are thereby
separated into (p l)/2 pairs, and the product of these pairs implies that
z
i2
'
cipi
a (p_1)/2 (mod p)
(p
dpi
1)!
= l(mod
p)
QUADRATIC RESIDUES
137
Therefore,
= l(modp)
If a is a quadratic residue modulo p and x = r(mod p) is one solution of
ss a(mod p), then x = p r(mod p) is the other solution, for (p r)
(pi)/2
x2
s= r
(mod
p).
has x
S3
modulo
p.
Consequently,
0,10,2
,'
r(p
^zia^^modp)
'
r)
But
r(p
r) s=
r = a(mod
2
p)
Therefore,
1)!
33=
a (pD/2
(p
tf(p/J(modp)
and
i(
m odp)
(r
2)*(m)/2
= a ^^ )/2 (mod
m)
But
r *()
i(
mo d m)
a <^(m)/2
i(
mo d m)
Hence,
It is
obvious that
if
m>
2, <(ra)
may
proof.
This
result,
l(mod
48), still
7(mod
(7,
48)
1,
\(48)/2
2,
and 7 2
138
Theorem
The quadratic
94.
residues of an
r,
r2,
r 2k (mod p)
r 2k+1 (mod p)
(a,
p)
rp
~l
1.
Then
if r is
form a reduced
or
In the
evident that a
first case, it is
a (mod p).
(r
is
2fc+i)(pi)/2
i(
(r
2
)
mc.dp),
l)/2
But then (2k
the exponent of r must be a multiple of p 1.
would have to be an integer, and that is impossible. Hence, in the second
case a
is
a quadratic nonresidue of
Thus the
p.
of
consists of the
nonresidue of
When
p.
a table of indices
residues
Theorem
95.
The
integers
2
,
22
M
I=
these integers
no two
of
is
them
modulo
p.
2
,
Each
22
of
modulo
di
p, for
if
a 2 2 (mod p)
then
(ai
a 2 )(i
a 2)
= 0(mod
p)
QUADRATIC RESIDUES
139
residues of p.
Examples.
Because 2
of 13 are 2 2
4,
24
3,
is
26
8
12, 2
ee 3, 5 2
12, 6
10(mod
95,
9,
2 10
they are
10,
l
and 2 12
1,
22
l(mod
4,
13).
32
9,
13).
EXERCISES
1.
is
modulo
4.
p.
If
is
is
+ 7x
i
36
= 0(mod
p) has a
solution.
5.
x2
If
1 is divisible
by
1,
is
p.
x2
If b is
tive root
modulo
p.
p,
is
divisible
10.
93.
tic
by
p.
ellip
geometry which was so well received that at the time it rivaled Euclid's
" Elements" in popularity.
In 1830 he published two volumes on the
theory of numbers that organized his own researches and those of his
predecessors in this subject. In this work he partly proved the remarkable law of quadratic reciprocity.
If
is
according as a
is
(a,
p)
1,
by
letting
(j = flor
1.
It is evident that
if
is
(a,
p)
1,
then
=
J
140
2.
If d\
p,
then
1,
then
(?)  (?)
3.
J
4.
a (p 1)/2 (mod
1,
where
is
(a,
p)
p).
Corollary 3 of
(ai} p)
if
Theorem
1, 2,
n,
then
if
is
^\
\^\
(?)
5.
If ai
p,
) (
Furthermore,
(?) indicate
as well as
1,
= 1 and
=
J
respect to p.
94. The Prime Moduli of Which an Integer Is a Quadratic Residue.
Having solved the problem of determining the quadratic residues of a
prime,
we now ask
if
we can
find the
a quadratic residue.
the integer
integer
is
odd primes.
Suppose that a
If
for
Then because
= k
)
b,
^=
)>
>
0.
these primes be q h q 2
qn
~ \ v
Then
)\v)\v)
\v)
where q
is
a quad
J,
and
an odd prime.
V,
prime
p.
evident that
= +1
QUADRATIC RESIDUES
Theorem
the form 4n
An
The
96.
1
integer
141
a quadratic residue of
is
of all
all primes of
primes of the form
3.
it is
= 2n
l)/2
12(mod29).
Theorem 97 (The Lemma
prime to
q,
2q,
Find the
p.
[(p
The
of Gauss).
l)/2]q.
1.
If
is
() = (!)"
integers
q, 2J, 3q,
^q
(1)
p.
are prime to p
modulo p
(2)
Let
ai,
(Li,
au
represent the least positive residues greater than p/2 of the integers in
while
(1),
bi,
b2
bv
The
(p
less
Then
than p/2.
l)/2.
ai,
a2
au
a/(mod p)
a,j
1, 2,
v;
modulo
1, 2,
Moreover,
p.
then
bi
However, both
therefore sq
set
1, 2,
+
.
h and
p)
= 0(mod
(p l)/2.
tq
= 0(mod
(1), and
where s and 2 are distinct integers of the
Hence, s + t = 0(mod p). But since both
p),
142
and
by p.
6t
a,
61, b 2 ,
form a
and
are positive
Thus
set of (p
positive
and
integers
1, 2,
less
.
bv ,
(p
b v (p
616.
ah p
a2
au
As a
l)/2.
di)(p
a 2)
result,
(p
au )
au
^^
(mod
p)
(mod
p)
and
(l) u bA
But the
b v aia 2
bi
6162
b v a\a 2
au
'
of the
2q
products in
(1).
Therefore,
g(mod
p)
P h ^^(mod
niv
p)
!
and
(1) U ^7T^^ (P
By
multiplying by
1) M
~ 1)/2
= ^o^!(modp)
[(p
= l) M (mod
p)
and dividing by
f(pl)/2
l)/2]!,
we have
But
g( p
 1)/2
(mod
p)
Hence,
(j)(l)(modp)
but since both
and
and
1,
it
follows
that(j) = (l).
Theorem
98.
[?H
Let (p
q, 2a,
...
If
l)/2
s(?
s.
is
+
Also
q
)
r2
then
p,
and
if
(1)*.
p\
let
be in order n,
is
rs
Then
QUADRATIC RESIDUES
= v
2q
= p
M+"
_P
sq
= p
sq
_P_
rs
But these
i
1, 2,
Let
a*
u,
bj
in
1, 2,
v,
of
where
Theorem
97.
pM + A + B
(3)
fey
= pu A
v2
1
(q
(3),
l)
we
2).
+B
(4)
find
p (M
u)
2A
(5)
Consequently,
u(mod
(p
Then
B.
p
By
+r
r2
that
fact
yi
l__l 9 =
But
fry,
= A and
=i
are in
ri
where k = 1, 2,
where j
and the
integers r k
.
pM +
143
M=
p,
by
we have
= (!) (1)"
On
the
hand,
other
if
2,
then
2q
0,
0,
V
(P
l)g/2 l
P
J
since p = 2& +
p
LP
[
0.
Hence,
0,
and Eq.
1,
p<
= (2k + l)M(mod
u = w(mod 2)
2)
(5)
shows that
144
Therefore,
by Theorem
97,
(_1)
The
Corollary.
form Sn
when p = Sn
3, (p
then (p 2
1,
l)/8
l)/8
of all
= Sn
2
= Sn Qn
2
(_ 1 )(22)(24)/8 = +lj
()
a quadratic residue of
is
When p = Sn
Examples
integer 2
(1)(p D/8
and
2n,
and
1,
but
= +1,

(I) 
but
< 1) (28)(30)/8
1.
tinct
odd primes, (
^J
C?) = (l)^
If
p and q are
dis
'^.
v^f
^<;
boundary, there
The equation
are, therefore,
of the line
Iff
2
OE is py =
integer,
is
the
number
lattice points.
qx,
k,
and
)
it
intersects
Therefore,
of lattice points
on x
if
any
line
a positive
is
1,
2,
...
145
QUADRATIC RESIDUES
(p
lattice points
number
on
1,
kq/p
~ l)/2, we
OEC is
(<7
triangle
is
[?H?] +
OAE
is
number
"
t}
for
1, 2,
[ViH
We
= (1) M and
that
(^J
OAEC
(1)*.
is
There
fore,
d) (,)
and consequently when p and
_<_,,
g are distinct
odd primes,
He
plished this feat in 1796, when he was but eighteen years of age.
appraised the theorem so highly as to call it the "gem of higher arithmetic " and developed six different proofs of it. Among the leading
mathematicians who have also proved the theorem are Cauchy, EisenIndeed, the
stein, Jacobi, Kronecker, Kummer, Liouville, and Zeller.
interest that it has continued to arouse is evidenced by the fact that it was
proved in about fifty ways* during the nineteenth century. More than
that, the number of proofs keeps growing, but, of course, not all of them
are essentially different.
Because
of the
meaning
of
proofs:
* P.
146
M [
N
we wish
&
(p
p.
course,
if
+r+
most s + 1.
=
sp
M + 2V =
no
+p
2p
because p q
p
<
2p.
Assuming that
for
sp
at
is
ft
be
<
M whose value
q,
We
last
<
with
and
s.
at
is
term
of
<
M which
[**'
most
if
less
Of
1.
]
p,
then
^+
Therefore,
1.
observe that
integer, for
most
written
(p
q.
can
Iff
an
 =0, and
LPJ
[P
so that kq
ff
/bg/p is
q]
P.
2p
pq
Moreover,
s,
p and
L q J
l)/2,
But
g.
fe]
to prove that
1, 2,
2q
<
l)g
1)g
(P
is
is
at
Dffl
2p
2p
(k
(2
D/2,
is
Since
1,
Hence,
and therefore
k
is
number
the
where
term
all
<
of the last
<
(q
[*>]
term
l)/2.
nonnegative
<
(q
l)/2, the
1 is
number
of
number
s,
of the last
Consequently, for
terms of
QUADRATIC RESIDUES
>
p^i  [f
M
2g
B]
(q
is
(p
l)/2
Therefore,
I^f]
Hi
P.
Oyg
2
([?]
LQ
L 2 J
1
J
["
[^!]+^^
[][f]
N+p 1 q
+ JV =
2p
+
+
ikf
3p
[?])
Hence,
l)/2
147
^^
^^
>
follows.
Corollary
If at least
1.
of the
is
(J)*Ms)primes
both
and
Corollary
the
If
2.
q are of the
form 4n
3,
ry
) (
form 4^
1,
then
6) (!)
=
Examples.
Since 15
Also
=)
2.
1.
= 2(mod
^) =
(1) 8
= +1, and
17),
=
= +1, and
(~J
there
is
(l)tt>U>/8
+L
77
)*
Hence,
Since
67
S 22(mod
Furthermore,
for
Test z 2
89
is
of the
oq)
89),
= +1,
form 8n
we
for 89
find
is
of the
and (oq)(tt)
form 4n
+1
+ l;(^j=+l
according to Corol
148
lary
L^J =
Thus
above.
I J = +
1.
Factoring immediately
= +1, and
x2
= 67 (mod 89).
is a solution of x
33 a quadratic residue of 89?
Is
we have
(g)
() Qjj
[~
y) =
(3)
= +1.
Consequently, (oq)
above,
Therefore, there
3.
However,
= !
Q0 ()
Therefore,
But
= 1, and
the congruence
s=
problem
means
Theorem
of Corollary 3 of
If q is of
+ 1,
the form 4n
5,
then
1,
Hence,
if
q.
and
to be
is
if I
For example,
is
if
l(mod
1,
2)
of
() = +1.
quadratic residue of
) is
and one
Thus p
q.
in order that
an odd prime
and one
if (
is
to be 1,7?
q
We
94.
to be
1,
all
4(mod
Hence, p
5)
1,
9(mod
10A;
10); that
1.
On
is,
is
l(mod
2,
2)
of the congruences
the form
3,
7 (mod 10),
10/c
If the given
and
and 10k +
prime
is
is
3 (mod 5)
a quadratic nonresidue of
all
primes of
7.
of the
form 4n
3,
then l^J
yH
(i)l*/i
149
QUADRATIC RESIDUES
and so (^) = (l)^
1 )' 2
( Y
In this case
if
(\
is
to be
l(mod4)
and
(^) = +1
= 3(mod4)
and
1^1 = 1
1,
then either
+1,
either
or
If
is
to be
= 3(mod4)
and
(^
and
{) =
= +1
or
Accordingly,
if
l(mod4)
7 and
to be
and one
is
l(mod
of the congruences
and one
satisfy
4)
xsl,2, 4(mod
giving
1
s L
9,
25 (mod 28), or p
the congruence
satisfies
3 (mod 4)
3, 5,
7)
of
6(mod
7)
15, 28fc
17,
and 28k
+L
Then
()
a quadratic residue of p,
23.
Now
4n
=
it
(?) (j)
of the
is
(_l)(f/
(j).
= l(mod
8)
and
= 3(mod8)
and
(]= ~
?J
= +1
Ox'
If
2g
is
to be
150
If either
m l(mod8)
and
m 3(mod
and
(?
or
then 2g
is
When
is
even
or
a quadratic nonresidue of
evident that
f^J =
so that
(^J
is
of the
1,
'
fl
p.
But \S\
( j.
P\
The exponent
3 (mod 8).
Hence,
==
(p
will
be
and
() =
and
() =
1
= 4n
(_i)(pd/2
+
+
4p
1
3, it
5)/8
when
is
either
when
p
The
conditions for
5,
7(mod
XP/
to be
8)
are
now
obvious.
primes
(!)
= (1)^^
(1)(p +4 P 5)/8
and only if p
if
8)
and
qi
q2
with q 1
By combining
= 4n
the
first
and
q2
two cases
qi and
by the following statements:
They satisfy x = l(mod 2).
is
4s
it is
3.
Then
i^\ =
characterized
1.
2.
They
They
q\.
satisfy x
of q 2
or they
2.
3.
They
They
l(mod
3 (mod 4)
q\.
of q 2 or they
,
satisfy x
4)
QUADRATIC RESIDUES
151
a
=
a;
ss 5, 7 (mod 8)
sa 3, 5,
a?
1,
1,
3(mod 8)
2, 4(mod
7)
and
From
the
first set
we
find that
6(mod
7)
51(mod
and from
56),
to be
+1.
and one
giving p
Hence, p
satisfies
be
+1
or both
for
= +l(mod
10)
of
1, 3, 9, 19,
25,
27(mod
28)
1,9, 19, 29, 31, 59, 81, 109, 111, 121, 131,
satisfies
and one
of
3,
7(mod
13,
139(mod
140), or p
10)
23(mod
28)
EXERCISES
1.
Evaluate:
2.
Is there
3.
4.
5.
6.
a solution of x 2
q for
21 (mod 41)?
which
(A) = +
1.
1, 3, 9,
39 (mod 40).
7.
Prove that
is
a quadratic residue of
all
and a
10.
152
95.
if
1,2,
m is
,r,
P =
Let
pip 2
p r where the
with
pi,
Then
following manner:
\p) ~ xpJ
771
p
is
defined in the
'
where the symbols to the right of the equality sign are Legendre symbols.
When P = pip 2
p r and Q = qiq 2
q s with the pi and qj,
s, positive odd primes, the properties of the Jacobi
where j = 1, 2,
symbol are expressed by the following theorems:
Theorem 910. If m is prime to both the positive odd integers P and
'
'
("pj^l )( )
(?)  fe)
0) h(?)(s)'=
(!) (f) " d "*"*" (?) (l)  fe)
Theorem
911.
/mn\ _ (m\

If
'
'
'
i)
fe)
n\
yv)
\p) \pf
(mn\
(mn\ (mrb\
 (?)
if
m=
.
r,
912.
n(mod
Because
.
H (")
fe>
Theorem
(mn\
If
,.
 fe) fe)
(s)
(mn\
fe) fe)
P), then
m=
_,
(pJ
n(mod P)
p)*
m=
implies that
n(mod p ), where
%
1,
Jacobi symbol.
Theorem
913.
If
is
?
=
According
to
the
definition
=r
1) (PD/2
(
V[( Pi l)/2]
(
1)
2,
*
,
where
1,
2,
r.
But
P =
p xp 2
"
pr
QUADRATIC RESIDUES
(pi
(P2
'
1)}
{1
(Pr
J= (l)+ J=
i,k
P =
and thus
1)},
1)}{1
153
(pi1)(p1)+
i<k
 O(mod
so that
2),
P=
>
i
(p.
l)(mod
P
Hence,
4).
=l
V ^^ (mod
i
Therefore,
2).
(^j\ = (l)^' 2
Theorem
914.
If
is
pi j = ( l)<
1 >' 8
.
1)J{1
(?>2
1)}
{1
(p,*
1)}
+ 2
"
1)
^
(Pi
=l
r
2
(Pi
i)fe
1)
+"
+ n
0(mod
follows that
8), it
P =
2
!)
l)(mod
But
since
>
Pi2
(mod
Theorem
8).
915.
If
Hence,
P and Q
Y"\
fe
z=i
i,&=i
fe
and
64),
=1
(^\ =
(1)Cp*u/8.
(0^(s)te)(^"factoring
p)
in like
pairs
where
r we find
) (
1, 2,
and
1,3
1, 2,
For a
fixed
*.
j,
(p*
l)(
a 
1)
i=l
fe
1)
(p<
1).
But we
i=l
saw that \
i
(p
l
1)
= P
l(mod
4),
and because
qj
1 is
even,
it is
154
evident that
fe
(lk
1)
(P
i)(*
i)
J {(
1)
1)(
l)(mod
8)
=i
Therefore,
(^
X(
1)
(p<
p " !)fe
i)}
l)(mod8)
 (PBecause
P 
1 is
even, (P
1)
fe
l)(mod8)
Jfo
1)
1)
(P
1)(Q
l)(mod
8).
Hence,
Therefore,
<
w
)
= +1
or
1.
are
(^
(?)
)
is
In the
+1 when
first
all
case the
congruences
x2
have a solution
ra(mod
for each
p,)
1, 2,
= m(mod
But
(7)
has no solution.
However,
if
=5
is
(6)
a solution of
(7)
have a solution
Hence,
a quadratic residue of P.
P)
tain
p)
and therefore
if
for cer
= 1,
it is
m be
155
QUADRATIC RESIDUES
Jacobi symbols,
'
zr=~ )
find
(773
'
 (I) 
(w)
Therefore,
we
= +*
is
a prime,
Since 173
Legendre symbols.
we
+1
and
"' (iS) = +1
(735
is
is
a prime, this
(g)
\j35j
 +1
* (I)
'
a Legendre symbol
is
Thus
(jjjlj )
QL
But
j (jfg) (jfg
(AX?) 
and (3
r)
= 1.
= 1.
Hence, (770)
+1.
(A)(t)But
(?)
so
'
 1
(A)
Now
In this case
prime.
/2l\ =
+1?
)
Hence, \kfo
Qj {^fj
but this
is
Furthermore,
Hence

solution.
tion of
() (i)
x = 21 (mod
2
and
11).
since
(it) =
1
(tt) 
Hence, there
is
'
is
not a
Qj =
= +1, and
=
(^fj
+1.
we can reach no
However,
(S)
+1,
(m)
'
there
no solution
is
of
no solu 
the given
congruence.
EXERCISES
1.
to determine
gruence x 2
2.
3.
q=
>
r^ry
>
^rz
156
The
96.
the congruence x
(a, 2)
ss
o(
= a(mod
2)
the congruence x 2
1,
l(mod
Theorem
a ss l(mod
a
Solution of x 2
mod
2 n ).
1,
if
if
4), in
8).
Suppose x
satisfies
Then x
the congruence.
a(mod
2 n ),
and there
a(mod 8)
But x is odd, and its square is, therefore, congruent
Hence, a = l(mod 8).
to 1 modulo 8.
Theorem 917. If a = l(mod 8), there are exactly four distinct solutions modulo 2 n where n > 3, of the congruence x 2 = a(mod 2 n ).
We shall first show by induction that the congruence x 2 = a(mod 2 n )
with a s= 1 (mod 8) has a solution. We know that under the given condiAssuming that, for a given k, x 2 =
tion x 2 s= a(mod 8) has a solution.
We, therefore,
a (mod 2*) is satisfied by x we infer that x 2 a = 2 k h.
wish to determine t so that #o + 2 k ~H satisfies x 2 = a(mod 2 k+1 ), that is,
fore x
ss
so that
(x
2*x
+
+
2
2
1
2A;
2 (h
and
A;
~2
s=
a (mod2 k+l )
=
=
0(mod
0(mod
a;
2 k+l )
2 k+1 )
finally
/i
Because
which x
this
= 0(mod
Xo t
2)
2 fc1 satisfies x 2
= a(mod
2 k+l ).
is
a value of
for
a:i(mod 2 n ),
it
Xi
a; 2
0(mod
(mod
2 n)
and
(xi
x\
[(#i
x 2 )(xi
x 2)
2n)
However,
x 2 )/2]
x2
Xl
rr 2
[(^i
a; 2
se
)/2]
0(mod 2 W ~ 2 )
iCi,
and
therefore
one
of
(iCi + x 2 )/2
~
Hence, one of the congruences (xi x 2 )/2 = 0(mod 2 n 2 ) holds.
~
But when x\
Therefore, x 2 = ^i(mod 2 n l ), or x 2 = Xi(mod 2 n_1 ).
2
n
=
Consequently,
all
four integers
satisfies x
a(mod 2 ),
X\ does also.
2
n
n~ l
=
a(mod
incongruent
and
they
are
2
satisfy x
X\, xi + 2
),
(iCi
2 n_2
x 2 )/2 and
modulo
2n
is
odd.
As a
by
QUADRATIC RESIDUES
157
congruent to
values of
a.
Example. In the case of x 2 = a(mod 16), a can have only the values
1 and 9 modulo 16.
The solutions of x 2 = l(mod 16) are 1, 7, 9, and 15
modulo 16, and those of x 2 = 9 (mod 16) are 3, 5, 11, and 13 modulo 16.
EXERCISE
First find the values that a can
64),
and then
have
be a solution of x 2
= a (mod
CHAPTER
10
is
sum
than 9 cubes, not more than 19 fourth powers. He gave no proof of his
assertion and may have had only calculations to support it, but the
problem implied by his statement has ever since challenged the best
mathematicians and has been only recently solved. It is generally
agreed that Waring meant to imply that for every positive integer k there
exists a smallest positive integer g(k) such that any positive integer n can
be expressed as a sum of at most g(k) positive kth. powers.*
Certain results connected with this problem of representing a positive
integer as a sum of like powers of integers had long been conjectured even
though not much progress in proving them was made until the eighteenth
Fermat, for instance, was much interested in the theorem that
century.
every prime of the form 4n + 1 can be expressed as a sum of two squares
and, barring the use of negative integers, in but one way.
Thus 5 = l 2 +
2
2
2
It remained for Euler, however, to demonstrate
2 and 13 = 2 + 3
Moreover, that any integer n is a sum of two
the theorem satisfactorily.
squares if and only if it has the form n = 2 a 2 P, with t >
and P a
product of different primes of the form 4s + 1, had been determined in
the seventeenth century. Again, mathematicians had asserted that any
integer not of the form 4 r (8s + 7) is expressible as a sum of three squares
and that integers of this form fail to be expressible as such a sum. Furthermore, by 1770 Lagrange had proved that every integer is a sum of at
.
From the time Waring enunciated his theorem, it took 139 years to
prove that every integer is a sum of at most nine cubes. Although
Liouville proved in 1859 that there exists a smallest integer #(4) such that
every n is a sum of at most #(4) fourth powers, still it has not yet been
proved that in this case #(4) actually is 19. In 1909 Hilbert proved the
general theorem that for each k there exists a positive integer g(k), independent of n, such that every integer n is a sum of at most g(k) kth. powers,
but his proof merely shows the existence of algebraic identities for deter*
G. H. Hardy,
of the
Theory
of
Numbers."
159
mining g(k) and sheds no light on the actual value of g(k). Since then
Hardy and Littlewood have developed by analytical means a formula
that determines an upper bound for g(k) for every k.
From these few remarks we can obtain some idea of the magnitude of
this problem, and certainly a perusal of a few of the original proofs will
give an appreciation of the ingenious adaptation of the tools of the theory
sum
expressed as a
Let us recognize
first
(a 2
of at
b 2 )(c 2
For instance,
(2
ever, that (2 2
)(3
1)(2
1)
52
(ac
two sums
d2)
bd) 2
of squares as a
32
(ad
be) 2
sum
7
2
.
two squares.
We notice, howof
formula gives but one representation of the product as a sum of two positive squares.
Does the formula ever fail to give at least one solution
when (a, b) = 1 and (c, d) 1? In this case ac = bd, and ad = be.
But if ac = bd, then a = d and b = c, and so the two given sums are
2
2
Hence, in the single
identical.
If ad = be as well, then a = b = 1.
2
case (1 + 1)(1 + 1) = 2 f0, the formula fails to give a sum of two
positive squares.
It is also
(a, b)
and
d)
(c,
1,
the squares in
the expression for the result of the product need not be relatively prime,
for example,
(ac
or
bd,
if (c,
d)
Euler's
ad
(8
be)
1 )(9
1,
2 2)
then
k, it is clear
first
(a, b)
70 2
+
1
25 2
and
On
(c,
d)
if
k,
1,
for
if (a, b)
bd and ad
represented as a
sum
instructive to study
it
be.
1
can be
of
The second
proof
enormous improvement in the directness of the presentation that a mathematician often attains when the initial proof is reviewed.
Lemma 10la. If a prime p = c 2 + d 2 and if there is a q > 1 such
that pq = a 2 + b 2 with (a, b) = 1, then q is a sum of two relatively prime
exemplifies the
squares.
*
t Ibid.,
Vol. 2, p. 281.
Theory
of
Numbers,"
160
If
the prime p
c (a
d 2 then
)
d2)
b 2c 2
2
fr
d)
(c,
(c
and
1,
if
pg
pq
a 2d 2
(be
a2
2
a p
52
we have
= mp
But
2
c {a
2
)
a 2 (c 2
ad) (be
ad)
+r
= td\s
tc
and
Then
cr
be
ds
ad
 ds
be
tc
and
id 2
so that
cr
ad
d2)
en
k(c 2
d 2)
t(c
Hence,
= ds
= dn
cr
r
But
if
be
ad
/cp,
and
the equations
6
kc
kd
cr
be
kc 2
ds
ad
kd 2
ds
be
ad
cr
= ds
and
give
and
Hence,
cr
and
In this case
r
In the
first
pq
and
= dn
and
= en
case
a2
=
=
=
td\
en) + (tc +
+ n )(c + d
p(t + n
(
(r2
in the second
pq
=
=
=
en) +
+ n )(c +
p(fc + n
2
(fed
(k 2
(fcc
d 2)
dn) 2
dn) 2
prime
is
sum
of
if (a, 6)
1,
for
if
161
n)
(t,
u,
rela
the equations
is
tc
n)
(A;,
that
'
'
'
'
squares.
Lemma
is
sum
103a.
of
If
a prime p divides a 2
2
,
where
(a, b)
then p
1,
two squares.
b 2 with
Let the prime p divide a 2
not a sum of two squares. Set
(a, b)
1,
= mp ri
< n <
= np
r2
<
is
and
r2
<
Then
2
7"i
?*2
a2
= Qp <
/cp
~2
Any common divisor of ri and r must divide Q, and thus the last equation
can be reduced to ai 2 + bi = Pp, with (a h hi) = 1. According to
Lemma 102a, it is now evident that P has a prime factor pi that is not a
sum of two squares, and furthermore, pi < p/2. Using the fact that pi
2
bi
a
l
sum
of
two squares.
2.)
sum of two
Theorem
(3
22
13,
32
22
10,
2
5,
must be a
squares.
101.
Every prime
of the
form An
is
sum
of
two
162
is
expressible as a
Now
sum
two squares.
of
suppose that p
a2
and
b, c
b2
d,
Then
odd.
d2
h2
c)(a
+c) =
\d
b)(d
so
that a
n(d
a2
and
(a
Let
(a
(m, n)
c,
= 1.
b)
r,
if
b)
= rm and
with
rn,
Therefore,
m(a
Hence,
we
let (a
c,
b)
c)
s,
b)
we have
\
= ms
ns
and
If
a and
But
if
case both
(r
c are
m
s
)(m 2
n 2) =
6)
(a
c)
)(m 2
a2
+
2
62
cP
= p
and s are even, p has been factored into the integers (r 2 + s 2 )/4
and m + n 2 both of which are greater than 1. But if r and s are odd,
r and s cannot both be 1, nor can both m and n be 1, for in either case
a = d and b = c. Consequently, when r and s are odd, p is equal to the
product of the integers (r 2 + s 2 )/2 and (m 2 + ^ 2 )/2, neither of which is 1.
Since such a factorization is impossible, we must conclude that the prime
p = \n + 1 has a unique representation as a sum of two squares.
Let us start the second proof of Theorem 101 with the statement that
since 1 is a quadratic residue of every prime of the form 4*1 + 1, there
are integers a and m that satisfy the equation a 2 + 1 = mp.
The fact
that a is a quadratic residue of p means that a can be chosen positive and
not greater than {p l)/2.
Consequently, a 2 + 1 < (p 2 /4) + 1, But
Thus,
if r
2
(p /4)
<
2
.
< m <
Set a
m/2.
= qim
As a
a2
< m
and
ri
m>
If
p.
mi
positive integer
we
1,
such that
b
q2
= mp,
<
(p
l)/2,
shall
rri\p is
m+
true that
it is
interval
163
r2
result
= n +
b2
+m
r22
( qi
g2
2
)
2m(r l9l
r 2q 2)
(1)
and
mp =
ri
+ mK
r22
so that
fi
r22
= mim
r22
<
But
ri
for not
both
ri
and
\r 2
feY
(a, b)
1.
Hence,
mim <
and
<m
mi
Lemma
r 2 )fei
g2
?2
(r iqi
r 2q 2 )
10la,
(nq 2
we have
r 2qi )
or
mmiqi 2
(2)
where
s
ngi
+ r<#
r x qi
and
Recalling Eq.
(1),
we observe
that
mp = mim
+m
(qi
r 2 qi
<?2 )
2ms
and
p
= mi
= mi
m(qi 2
q2
2s
Therefore,
m
and, according to
mim(qi 2
g2
(2),
mp =
x
mi 2
or
mip = (mi
2
f s)
2mis
2miS
164
Hence,
clear that
it is
We
we have
of
con
two squares.
x y
x y x y
+ (xiy
+ (xiy* x y + x y x y
If i has the usual meaning, \J 1, this identity can be proved by find2
2/4 )
4)
f
4)
2)
x2
Xi
Xz
xz2
+ ix
+ ixi
x 4 2 and
+ v*
y*
x3
+ 1x4
Vi
Xi
y*  iy*
ix 2

z)
W2
y^
2/3
iy*
y\
iyi
A  %B C iD
C  iD A +iB
A + B + C + D'

where
Lemma
1
< <
t
Let
lie
Xi,
p,
1026.
If
A =
B =
xiy 2
C =
x xy z
D =
xiy 4
is
such that tp
where
x{
xzyi
x 4 yi
0, 1, 2,
<
in the interval
+ xzyz + x y
+ x y xyz
+xy xy
xzyi
+xy
x 2y 2
x 2 yi
xiyi
<
Xi
(p
x 22
,
Xz
(p
exists
x42
an integer
(xi
Xi
than
where
l)/2.
t,
l)/2 values
p, for if
it
x?
follows that
which
p.
let yi
1 y
1 yi = 1
numbers
x' as
2
?/
(mod
p)
165
and
x2
so that x
>
and
tp,
<
Moreover, x 2
2p > p2 + 2 and
which shows that p
,
Lemma
1036.
such that tp
xi
= 0(mod
where
tp,
a positive integer.
is
p)
2
2
2
p /4. Therefore, (p /4) + (p /4) +
2
However, since p > 2, it is evident that
1 > tp.
2
2
2
2
p > (p /2) + 1. Hence, p > (p /2) + 1 > tp,
>
If
<
and y 2
/^,
\
(p /2)
t.
is
+ x% +
x 22
x 2 then
,
is
odd.
( xi +
2
\
a? 2
S2Y
/ a?i
""
/ ffs
"*"
gj
(x3__Xi\
"*"
_ xi
x 22
'
#3
Thus there
Lemma
less
Consequently,
1046.
than
If
p, satisfying
is
2/3*
Choose
Xi
than
s less
x 22
:r 3
+ X4
st
is
not
yi
1,
2/4
?/,,
Then y^
the condition tp
is
four squares.
t,
2
Xi
where
2
a;*
(mod
=
t),
1, 2, 3, 4,
so that y t
Xi(mod
t),
and let
\yt
<
t/2.
and
2
Vi
X ^ m d
2
Hence,
4
V
Therefore, y x 2
2/2
y%
2
?/,
?A
tp
0(mod
The
s.
integer
s is
not
or each
In this case t 2
2/ would be 0, and then each Xi would be divisible by t.
would divide tp, and t would divide p. Hence, t would be 1. Moreover,
4
since
\y {
<
t/2,
it
follows
that
yi
<
/4 and
V
i=l
<
2
,
and
<
t.
Consequently,
<
<
t.
y 1
<
2
.
Hence,
166
Theorem
102.
is
sum
of at
squares.
p =
p >
If
If
3,
observe that 3
3,
+
+
+
+
x22
yi
+x =
+ yi =
x, 2
yz
tp
<t <
st
<
Remember
We
<
that
find
is
upon applying
Euler's
identity that
t
sp
(xtfji
{xiyz
But
since yi
Xiyi
+x
x 2y 2
Xzyi
+ x y + x y^) + (x y  x yi + x y x yz)
+ xy x y^) + {x y x^yi + x yz x y
2
= ^(mod
(3)
2)
t),
+ x y% + xy = Xi + x +
xz2
+ Xzyt
y2
x 2
tp
= 0(mod
t)
Also,
xiy 2
x 2 yi
xy z
Xix 2
righthand
But
member
of (3)
2
t
Xix 2
x d Xi
sp,
with
<
theorem
sum
t),
and that
is
x&i = 0(mod
t,
is
sum
of four squares.
of four squares.
Hence,
least posi1,
and the
proved.
is
Theorem
103.
Every integer
is
sum
of at
squares.
Upon
EXERCISES
1.
as a
sum
7) with r
and n
>
cannot be expressed
of three squares.
2. Write
prove?
(xi
x22
x 32
rc 4 )
as a
sum
of three squares.
What
The Equation x 2
102.
From
2
.
167
integers x
1,
of
2
2
Moreover, because (y, z) = 1, one of the integers v, w is
y = v w
even, while the other is odd.
If (v, w) = 1 and one is even, while the other is odd, and k is an arbi.
= 2kvw
k(v 2
2
)
k(v 2
+w
2
)
by these
values.
and
are x
EXERCISES
2.
3.
Is there
1.
an
whose three
+y =
2
625,
and x 2
+y
2704.
4. Find expressions for the sides of all integral right triangles whose hypotenuse is
one greater than a side.
5. Show that if x 2 + y 2 = 2z 2
then x = k(v 2 + 2vw w 2 ), y = (r 2  2vw
w 2 ), z = k(v 2 + w 2 ), where k is an arbitrary integer, (v, w) = 1 and one of v and iv
,
is
is
odd.
168
103.
Fermat's Last Theorem. About 1637 Fermat stated that there
no solution in positive integers of the equation x n + y n = z n if n > 2.
This theorem is known as Fermat's last theorem, and about it he wrote,
"I have discovered a truly remarkable proof but this margin is too small
To this day mathematicians have been baffled by the
to contain it."
statement, for they have been able neither to prove nor to disprove the
The equations (x m ) 4 + (y m ) 4 = (z m ) 4 and (x m ) p +
general theorem.
m p = (z m p show that the proof can be broken up into the cases in
(y )
)
which n = 4 and n is an odd prime. In 1747 Euler published a proof,
which we shall reproduce below, that there is no solution in the first case.
No proof for every odd prime has been discovered although by 1857
is
It
this
reflection
on
his
own
error
made
if
n <
100.
in attempting to prove
mathematicians were able to set up various conditions for the insolvability of the equation.
H. S. Vandiver* has given a complete account of
the present status of the problem, including his own recent contributions
"toward its solution
Theorem
tion x 4
105.
of the
equa
2
.
The method we
shall use in
Fermat's method of
infinite descent.
is a neat example of
assume that there are
show that under this condition
We
shall
and shall
having a smaller z. It is evident that if
there is a solution of the equation in which the integers x, y, z are not relatively prime in pairs, then there is another in which the restriction holds.
We shall assume, therefore, that x, y, and z are prime each to each. Then
2 2
2
2
2
(x 2 ) 2 + (y ) = z so that according to Theorem 104 we have x = a
2
2
2
2
b y = 2ab, and z = a + b with (a, b) = 1, and where we may choose
a odd and b even. Because y 2 = a(2b), it follows that a = u 2 and
Therefore, a = r 2 + s 2 and b = 2rs with
2b = v 2
But a 2 = x 2 + b 2
Hence, 26 = 4rs = v 2 and r = Xi 2 s = yi 2
Thus xi A +
(r, s) = 1.
A
2
and, furthermore, 1 < u < a < z.
Thus there is a set of
yi = u
It is imposintegers with a smaller z that satisfies the given equation.
sible that this always be so, and consequently there is no solution of
integers that satisfy the equation
there
set
xA
y = z
Corollary.
x4
2
.
There
is
no solution
4
.
of
If
Vol. 53, No. 10, pp. 555578, 1946; ibid., Vol. 60, No. 3,
169
= m
= m n2 =
n2
u2
and
As a
result
2m = u
2
and
2n 2
Thus
(u, v)
and the
1,
even integers.
= u
last
Hence,
(u
and 2n =
2
even,
is
 2
y
^ =
2s 2
8q
and u
(u
v)(u
are
+ v).
1_ p
<
even.
is
>
either
+ v)
v){u
and
Accordingly,
t*
^1
2s<
2s 2
let
or
^1 =
and
Then
2n 2
SsH'
n2
4sH'
and
Furthermore, in the
and
is
2s
2s 2
case
and
and
2s 2
m =
2
n2 =
t*
4sH 2
4s 4
4s 2 2
in the second
But
first
Dickson, op.
cit.,
Vol. 2, p. 615.
Thus m
and 2s are the arms. But
2
(t )
(2s 2 ) 2
170
a2
b2
new
original.
2
and a
v )/S,
(g.
of the
follows that z
it
triangle
new
>
>
m, so
is
triangle
the integer
is
2 2
A = a H a* But
6)
p .)
(**)
>
^^
(^)
(V)
for
>
On
statement
Theorem
x2
\
If
x2
both p 2
whose
107.
,
it is
true:
is
sides are
= w
2p q
and p 2 q 2 = n 2
4
A
p q and p + q*.
p q (p
if
= p
is
Consequently, there
is
no solution
in integers of the
moduli.
If
m=
4, the
n
,
let
p =
m l(mod
and hence
p),
(J(pl)/2)p.i
and
t
171
(_]_
 _X
PHpl)/2
k p )p*
__
f
Mpn
Therefore,
f*(p)/2 ==
l(mod p n )
and
P = l(mod
m=
2p n
n
)
let s
let
satisfy
is
P=
But
is
(p
~ 1)/2
= l(mod
odd, and
m =
(j>(2p
n
)
n >/ 2
we
find that
(mod 2p n )
= l(mod
P = l(mod
p),
n
Therefore,
<t>{p ).
n
p j. However,
2p n ).
2U
where u
P = (iy
of these pairs,
u ~2
= +l(mod
2 U ).
When m =
x
x
=
=
s(mod pi)
l(mod 2p 2 p 3
'
p r)
'
'
Then
.
is
4>(m)
the
di is
such that
P =
But
since
i(piD/2
ni
(f>{p i
= l(mod
is
pi),
*<
and
m>' 2 (mod m)
$*&>i*i>/* ==
<t>(pi'
n
)<f>(p2
l(mod
*)
= +i(modpi0
'
pi ni )"
<f>{p r
),
However,
172
m)/2 =
(1 + 2p 2 ps
p r k, so that t^
2u ~
n2
n3
nr
k)4>(m)/2
anc
j t<i,(m)/2 = f l (mod p 2
Furthermore,
t
r
)
P2
Vr
p
2
U
M
2
= +l(mod 2 ), and thus t+' = +l(mod 2 ). Therefore, t*' =
+ l(mod m), and P = +l(mod m).
106. The Pellian Equation.
The equation x 2 by 2 = 1, in which b
Moreover,
2p 2 p 3
"
'
'
'
Dirichlet
Lemma
10lc.
If
is
any
and
< s < m.
Let x have the values
<
such that
m>
number and
real, irrational
\r
sa\
is
an
< 1/m
and
0, 1, 2,
There
m+
are, then,
<
values of y
excluding
{y x
(x 2
Xi)a\
x2
1(2/2
x 2 a)
Xia)\
<
ni
or
(?/2
But the
s
x2
Xi,
we conclude
<
Lemma
an
infinite
inequality
\r
2/i)
sa\
<
If b is
of pairs of integers r
s b\
<
and
number
\r
Hence,
if
y%
y\
and
that
102c.
<
< m
<
2 \/b.
and
is
with
< m
>
is
Lemma
exist
when
and
S\
such that
<
In
Vb\ <
si
'
Now
173
<
< m
si
< fllV^I
mi
Then
r2
and
s2
<
\r 2
s2
<
'mi
<
s/b\
< mi
s2
Hence,
\r 2
By
Vb\ <
s2
continuing to choose
1, 2, 3,
<
ri+ i
\n
we obtain
\/b\
Si
This means
we can
<
\/b\
si+1
\A <
Si
sufficiently large,
m,:
<
mi
for
\n
<
s i+1
<
rrii
Tfli
Thus
\ri+1
and
in this
way we can
satisfying the
But
for
y/b\
s i+1
set up an
above condition.
any pair
<
\fi
y/b\
Si
number
infinite
of pairs of integers
r, Si
ri} s i}
<
\n
Si
Vb\ <
< 
rriii
Si
Vb
Hence,
<
\n
8i
Vb\ < 
2si
Si
and
<
\n
exists
an
r
an integer k
infinite
number
many
s b\ lies
If 6 is
Si b\
/6
between
and
of pairs of integers x
and
and
with
1+2 y/b.
y.
is
by 2
is
satisfied
by
174
Since b
most
is
2[1
not a square,
pairs of integers
of pairs of integers
by 2
\r
number
r,
s,
0, is
k.
Theorem
109.
the equation x 2
manner
If b is
by 2
like
of y, into the
tive.
Accordingly,
if
=
=
X\
2/i
Now
Xi
2/2(mod k)
2/i
2
x2
^
^
x2
Xi,
x2
2/2
2/i,
2/2
=
=
 by, 2
 by 2 2
>
>
k
k
x2
From
2 (mod k)
Xi
y/b
2/1
x x x 2 
2/12/26
O12/2
x*2
Vb
2/2
\/b
X2IJ1)
y* 2h
^1^2
2/i2/2?>
Xi
x 2y x
Xiiji
byi
ss
0(mod
k)
and
xiy 2
xiiji
= 0(mod
k)
Therefore, let
XiX 2
2/12/26
_ u
the
#i,
among
k,
ano
Xiy 2
x 2 yi
and thus
xi
2/1
Vb =
(u
s/b)(x 2
2/2
y/b)
2/1
V6 =
{u
\A)(>2
2/2
\/b)
Likewise,
by i 2
(u 2
k(u 2
bv 2
bv 2 )(x 2 2
175
by 2 2 )
or
v b)
so that
u2
and
(xi
2/1
VS)(^2
2/2
V^) =
2/3
V&
Oi
2/1
V&)(#2
2/2
y/b)
^3
2/3
V&
also
Consequently, x 3
mined
(xi
fo/3
1,
is
2/1
of integers z 3
by 2
2/3
so deter
by induction, that
It follows,
1
integral n.
Moreover,
/
Oi
and
X4
x2
if xi, 2/1, is
2/4
fo/
Again,
y/b
1,
2/1
v7rT
fr)
<
Xl
~ ^
)n
2/1
y/b) n for n
2/1
1,
2,
x*, y
is
determined by
also a solution of
1
n =
if
2/
0,
we
find
(xi
\A)
= Oi
2/1
*\A)
and
0.
Furthermore, when
y,
are solutions.
integers x,
of
x2
by 2
2/
71
1.
But the
176
is,
(*i
By
member
multiplying each
if
we
yi b
we
1,
2/1
Vb) n+1
(xi
y1
Vb
y\
Vb) n and
see that
Vb) n
Vl
(xi
statement by
of this
Y Vb <
< (X + Y Vb)(x 
But
X+
Vh) n <
2/i
<x +
1
let
(X
+ Y Vb)(xi ~
Vi
Vb) n =
x'
+ y'Vb
 Y Vb)(x +
Vl
Vb) n =
x'
y'Vb
necessarily giving
(X
then x n
by' 2
and
<
But
x'
y'
Vb =
(x
x'
y'
Vb)~ and
l
<
Adding
<
(4)
and
shows that
(5)
Vb <
y'
+ y\ Vb
X + Y Vb
Xi
Vb
Vb <
x'
>
0,
and subtracting
y'
>
y'
(4)
therefore
y'
was
Vl
x'
however,
x1
(5)
(5)
from
(4) gives
0.
Vb
be
less
than X\
y\
Vb,
for
Therefore,
the formula x
(xi
Vb) n
2/1
+ y Vb
Since
= (#i
y\
to
Vb) n
EXERCISES
1.
2.
X2
_
3.
= 1
Show
V3 =
from x
1,
1?
(2
+ V3)
n
.
all
solutions of x 2
3y 2
is
CHAPTER
11
POLYNOMIALS
111. Integral
Domains and
Fields.
Let us
of
rational integers has certain salient properties with respect to the oper
The sum
1.
the
of
two elements
in a certain order is a
unique element of
set.
Addition is commutative.
Addition is associative.
4. Each element has an inverse with respect to addition.
= a for any a.
such that a
It follows then that there is an element
There is a second law of combination for the set of rational integers
which we called multiplication, and it has the following properties.
1. Multiplication is unique, and the product is in the set.
2. Multiplication is commutative.
2.
3.
3.
Multiplication
is
associative.
4.
Multiplication
is
5.
There
6.
0,
Any
then
is
c.
integral
When
be an
domain.
the set has at least two elements and in addition to the above
properties each element of the set except zero has an inverse with respect
is called a field.
Thus, in a field, ax = 1 with
can be solved for x, and the value is unique. Hence, division by
= 1, then a(bx ) = b and a divides b. The
a t^
is possible, for if ax
rational integers fail to form a field, for division of an integer b by a ^
is not always possible, but the set of the rational numbers (ratios a/b of
two rational integers with b t^ 0) is a field. The real numbers as well as
the complex numbers also form fields.
Now consider the set of all polynomials J(x) a x n + aixtt1 j
n, are selected from
+ a n where the coefficients a with i = 0, 1,
This set includes the field of the coefficients itself, and consea field.
a^O
t,
177
178
it
also holds.
if
is
Then
f(x)q(x) identically.
quotient of g{x)
most that
A
A
there
by f(x).
f(x)
Furthermore,
is
if
such that
g(x)
0,
is
the
is
at
of g(x).
is
Theorem
111.
If g(x)
f(x)q(x),
field F,
where
f(x)
unique.
If
g(x)
f(x)qi(x)
f(x)q 2 (x)
then
f(x)[ qi (x)
q 2 (x)}
Hence,
qi(x)
Theorem
q 2 (x)
and
q x (x)
q 2 (x)
112.
Zero is the null polynomial of the set of polynomials
with
coefficients
in F.
f(x)
The equation
shows that the quotient of zero by zero exists
f(x) =
but is indeterminate. It also shows that zero divides only itself. Fur
POLYNOMIALS
179
thermore, every polynomial f(x) divides zero, so that there can be but one
null polynomial.
Theorem 113. All the elements except zero of a field F are unit poly
common
greatest
If
the greatest
cients in a field
Theorem
divisor
is 1,
114.
coefficients in
If f(x)
kb
and
f(x)q(x)
r(x)
with
Do
is satisfied.
likewise
if
g(x)
is
of lower degree
=
=
0.
and
r(x)
than f(x)
g{x)
and the
~
not of lower degree than f(x), take qi(x) = kx n m where
~
m
n
Then ri(x) = g(x) kx f(x) is lower in degree than g(x)
If g(x)
common
~
a xn
an
aix n 1
~l
m
++&, take q(x)
bix
theorem
a
If g(x)
b xm
divisor of
either r(x)
is
common
is
and
g(x)
f(x)q 1 {x)
n(x)
If ri(x)
f(x)q 2 (x)
g(x)
f(x)[q!(x)
Thus we obtain
r 2 (x)
and
Again,
if
r 2 (x)
or
if its
degree
is
q*(x)]
r 2 (x)
180
and
#2(2)
number
r 2 (x),
r s _i(z)
f(x)q 8 (x)
g(x)
f(x)[qi(x)
if
is
repeated
r s (x)
Hence,
and
but
we obtain
of steps
++
q 2 (x)
q.(x)]
r a (x)
either rs (x)
Now
Then
stated conditions.
g(x)
f(x)q(x)
r(x)
f(x)Q(x)
R(x)
and
f(x)[q(x)
Q(x)]
r(x)
R(x)
s=
Theorem
The
115.
greatest
common
divisor D(x) of
two polyno
mials f(x) and g(x), not both zero, with coefficients in a field F exists and
can be expressed in the form D(x) = F(x)f{x)
G(x)g(x), where the
<m<
Then
n.
g(x)
f(x)
Ri{x)
Rti(x)
= f(x)Qi(x) + Ri(x)
= Ri(x)Q (x) + R (x)
= R (x)Q (x) + R${x)
2
= R
Ri(x)
is
of degree
(x) is of
Rz{x)
(x)Qt+i(x)
is
< m
m < mi
m <m
m\
degree
of degree
+ R t+1
We must finally arrive at a step in which R t+ i is a constant, for the polynomials Ri(x), where i = 1, 2,
Then
t
1, decrease in degree.
by making use of an argument that parallels that in the proof of the
.
Euclidean algorithm, we see that f(x) and g(x) have a greatest common
It is also
divisor R (x) different from a constant if and only if R t+ i = 0.
evident that a necessary and sufficient condition that f(x) and g(x) be
relatively prime is that R +i be a constant different from zero.
If Ri(x) f^ 0, solve for each Ri(x) that is not zero, and substitute the
expression in the succeeding equation of the algorithm.
Thus when
t
+ Qi(x)Q
Rz(x)
=
=
= F 1 (x)f(x)
R.{x)
(x)
[1
[1
2 (x)]f(x)
Q2(x)Q 3 (x)]g(x)
G!(x)g(x)
 Q {x)g{x)
 [Q (x)Q 2 (x)Q
2
z (x)
+ Q
1 (x)
Q*(x)]f(x)
POLYNOMIALS
Hence
if
and
f(x)
D(x)
However,
f(x)
if
and
+G
(x)f(x)
(x) is
not a monic
we have
G,(x)g(x)
if
leading coefficient,
its
F,(x)f(x)
R t+1 = F
and dividing by
prime and
181
we
find
(x)g(x)
we obtain
Rt+i,
F(x)f(x)
+ G(x)g(x)
In the last case the term of highest degree of F(x) comes from the
product Qi(x)Q 2 (x)
Q +i(x). The degree of Qi(x) is n m, of
Q 2 (x) is m mi, and the degree of each Qj(x), where j = 3, 4,
Therefore, when m > 1, the degree of their
t
1, is m/_ 2 W/_i.
(m mi) +
product is less than n, the degree of #(#), for n m
In like manner, when
(m _i m ) = n m
(mi m 2 )
m > 1, the degree of G(x) is determined by the degree of Q 2 (x)Q z (x)
Q t+ i(x) and is less than m, the degree of f(x).
In the special case where R\(x) = 0, fix) is a greatest common divisor,
and then f(x) = f(x)
g(x) and the theorem holds.
Notice that although the method is applicable when the polynomial
f(x) is a constant not zero, the statement about the degrees of F(x) and
G(x) does not hold. For example, if g(x) = 4x
3 and f(x) = 2, we
write 1 = 2 J
If f(x) = 0, the result is obvious.
(4#
3).
'
'
'
t.
Theorem
and
If
Di(x) and
\
1,
= cD
2 (x)
D (x)
D (x).
2
(x).
common
Also
118.
is
+ G(x)g(x)
2 (x)
1.
unique.
and g(x),
than or equal to the
Hence, di = d 2
d\.
divisors of f(x)
d\ of Di(:r) is less
Di(x), so that d 2
<
(x).
If fi(x)
with coefficients in a
/i(z) divides
and Di(x) =
Theorem
are
if
greatest
result Di(x)
Di(x)
and only
then Di(x)
degree d 2 of
As a
The
117.
if
field
F and if /i(x)
relatively
(x).
1,
fi(x)F 1 (x)
+f
prime polynomials
(x)F 2 (x)
182
Then
MiWW +
Theorem
119.
/,(*)
it is
and
If f(x)
at least the
is less
than that
of g(x)
G(x)g{x)
of G(x) is less
nomials, then
F(x)f(x)
G(x)g(x)
= F 1 {x)f{x)
{x)g{x)
Hence,
f(x)[F(x)
But
and
since f(x)
Unless F(x)
F{x)
Fi(x)
F,(x)]
g{x)[G 1 {x)
G(x)]
F\(x)
is less
this division
than that
would be impossible,
of g(x).
It is
Fi(x).
infinitely
many
Theorem
solutions of ax
and
by
But
1.
and
condition
is
it is
to be noticed that
many
G(x).
If
+ G(x)g(x) =
F(x) + k(x)g(x)
and G(x)
form
Are there any others?
nomials
On
is
When
1,
and
the second
for
if
and
Fix)
and G{x)
now be obvious.
common divisor of
when
the greatest
is
Theorem
not
1,
119, for
(x 2
4x
3)(
+ l) +
(2.r
6)
(
lj
= x 
and
(x 2
When
the
field
4x
3)(l)
(2x
6)
is
of algebra,
we know
POLYNOMIALS
183
that, except for the order of the factors, f(x) can be factored into linear
factors,
fix)
a xn
where the
aix n
with
Vi,
factorization of f(x)
'
'
1 7 2,
a>n
The reader
+
+
'
cto(x
r l )(x
r 2)
(x
r)
unique
however, that such a factorizaIn the field of the real numbers
realizes,
all fields.
reducible in F.
rational
The
field of
and
so
is
we
shall consider a
field
that con
primitive polynomial
is
coefficients are
relatively prime.
Theorem
positive.
is
1110.
If fi(x)
++
71
and
f 2 (x)
xm
b x x 1
++&.
with n > m, have integral coefficients that are relatively prime. Their
product necessarily has integral coefficients, but suppose that a prime p
divides each of these coefficients.
Then there is a first a;, say a r and a
,
first bj,
say
a r + s bo
6S
that
is
not divisible by
a r +ib s i
a rb s
p.
Now
a rib s+ i
+ a r+s m b m
184
all
is
if
fi(x)
pg(x),
where g(x)
p[g(x)f 2 (x)]
an integral polynomial
and the product is not
is
primitive.
Theorem
1111.
If f(x) is
leading coefficient unity, fix) is factorable into the product of two monic
polynomials in the field of the rational numbers if and only if it is factor
domain
able in the
Let f(x)
fi(x)f2 (x),
rational coefficients,
factors are integers.
of
both g(x)
1.
If f{x) is factorable in
able rationally
it is,
of course, factor
Theorem
1112.
of the rational
If
numbers
and
h(x).
in the field
the
first
c is
of the rational
cg(x),
where g(x)
is
POLYNOMIALS
f(x)
finite
185
We now
f{x)
Cpi(x)p 2 (x)
'
way and
after a
p r (x)
'
cpi(x)p 2 (x)
Clearly, c
k, for
q s (x)
where
1, 2,
Moreover,
Hence,
s.
Thus
qi(x)
the identical
tpi(x),
we have
q r +i(x)
is 1,
q s (x)
and the
original factorization
is
unique.
Theorem 1114. If fi(x) and f2 (x) are integral polynomials, not both
we can choose a greatest common divisor of them so that it is an
zero,
integral polynomial.
Since the coefficients of fi(x) and f 2 (x) are rational, their greatest
divisor D(x) exists and can be written
common
D(x)
Let
and h be the
d, g,
common
least
g(x)/g,
and
2 (x)
h(x)/h.
fl{x)
(x)F 2 (x)
Fxix)
+f
fi{x)Fi(x)
and
h.
d(x)/d,
common
multiple of
Thus
k,d(x)
where kid(x)
m+ Mx) m
Let D(x)
Then
is
fi(x)[k 2 g(x)]
an associate
of the
+f (x)[kzh(x)]
2
coefficients.
Very often we use the primary associate of kid(x) in place of the greatest
divisor of fi(x) and f 2 (x) even though we may not be able to
write it in the above form with k 2 g(x) and k z h(x) integral polynomials.
Example. The greatest common divisor x of 3a; 3 2x 2 3x + 2
and 3a; 2 8a; + 4 can be expressed in the form
common
#=
i(3a;
2a;
3a;
2)
(a
2)(3a; 2
Sx
+ 4)
186
Its associate
9x
6
Instead of 9x
9x
3a;
6 can be written
(3x z
2x 2
Sx
however, we
6,
+ 2) 
may
(3x 2
Sx
4)0
2)
2.
EXERCISES
Can
1.
2x
5x + 3 be expressed in
polynomials?
2
common
in
divisor of 2x 2
and
uses integral
+ 6,
where r(x)
Hence,
is
kf2 (x)g(x)
r(x)
= f2(x)[kg(x)] +
fi(x)
and
r(x)(mod p)
kg{x)
of fi(x)
is
by f 2 (x).
ax = b (mod p) has a solution and we can infer that a divides any integral
polynomial modulo p. There can be no other units modulo p, for 1 is not
divisible modulo p by a multiple of p or by any polynomial of the first
degree or higher. Since r is identically congruent to r + kp modulo p,
1, 2,
p 1 represent all the units modulo p. Moreany integer congruent to
modulo p represents the null element.
In like manner, the rational integers congruent to 1 modulo p denote the
unity element of the set of integral polynomials modulo p. The definition
of an identical congruence stated in Chap. 5 is the basis for the statement
the integers
over,
POLYNOMIALS
that
if
f(x)g(x)
0(mod
187
+1
+2
+3
+4
x2
x2
x2
x2
z
x2
x2
x2
x2
2
+l
+2
+3
+4
+x
+x
+x
+x
+z
x2
x2
z2
x2
z2
+1
+2
+3
+4
+ 2x
+ 2x
+ 2z
+ 2x
+ 2z
x2
+
+
+
+
z2
x2
2
a;
+ Sx
+ Sx + 1
+ 3x + 2
+ 3x + 3
+ 3z + 4
The
+3
+4
+2
x
x
z2
zfl
2
z2
x2
x
x2
z2
+
+ +
f
+ 2^+3
+ 2^ + 4
+ 3z + 4
modulo
ic
z2
z2
z2
+ 4z
+ 4z
+ 4x
+ 4z
+ 4z
+1
+2
+3
+4
5:
2
+ 4x +
x2
5 of the
degree.
common
divisor modulo p, a prime, of a set of integral polycongruent to zero modulo p, is a common divisor of the
set that is divisible modulo p by every common divisor of the set.
We
refer to the primary associate of a greatest common divisor modulo p as
the greatest common divisor modulo p.
greatest
nomials, not
Theorem
all
If fi(x)
any
difference in degree,
=f
(x)q!(x)
riO)(mod
p)
188
either r\(x)
this process
/2O)
n(x)
r k  2 (x)
r k i(x)
=
=
r 2 (x)q 3 (x)
+r
+r
+
r k i(x)q k (x)
2
3
(x)(mod p)
(x)(mod p)
r k (x)(mod p)
divisor
modulo p
of fi(x)
and/2 (V)
divides r k (x),
r k (x) is
r k (x)
s= ri(x)q 2 (x)
nix)
= fi(x)hi(x) + f
and f 2 (x), we
and
Therefore,
By solving
find
(x)h 2 (x)(mod p)
D(x)
^f^g^x) + f
(x)g 2 (x)(modp)
less
oif2 (x).
It will now be easy for the reader to prove the following theorems:
Theorem 1116. If the integral polynomials fi(x) and f 2 (x) are relatively prime modulo p, a prime, and if fi(x) divides f 2 (x)fz(x) modulo p,
then fi(x) divides f^(x) modulo p.
Theorem 1117. If p is a prime, the integral polynomials fi(x) and
f2 (x) are relatively prime modulo p if and only if there exist integral polynomials gi(x) and g 2 (x) such that fi(x)gi(x) + f2 (x)g 2 (x) = l(mod p).
Theorem
is
A Method
p.
is
a prime,
all
for Solving a
distinct solutions
POLYNOMIALS
f(x)
0(mod
p)
may have
189
EXERCISES
Write the primary prime polynomials modulo 3 of degrees 0, 1, and 2. Then
all the prime polynomials modulo 3 of degree 2 that are incongruent modulo 3.
2. Factor 2x 5 + x 3 + 2z 2 + 2x + 2 into prime factors modulo 3.
2
4
3. Find the solutions of x  2x  3 = 0(mod 5).
4. Find the solutions of x 5 4x 3
Sx = 0(mod 7).
5. Find the solutions of x 4 x
1 m 0(mod 7).
4
6. Find the solutions of 2x  10z  27 = 0(mod 35).
7. If f(x) is an integral polynomial and p is a prime, develop a method for solving
the congruence /(x) = 0(mod p) by using the derivatives oif(x) with respect to x and
1.
write
certain greatest
common
divisors
modulo
p.
CHAPTER
12
PARTITIONS
121.
Numbers.
of
mathematicians who paid particular attention to developing the theory concerned with the separation of an integer into all possible
summands selected from a given set, for example, the representation of 4
by 4, 3 + 1, 2 + 2, 2 + 1 + 1, 1 + 1 + 1 + 1, where selections are
made from 1, 2, 3, and 4. To Euler, however, is due a large part of the
This additive theory of numbers is a difficult subject. We
basic theory.
The student can refer to the
shall develop only the fundamental ideas.
work of G. H. Hardy, one of the modern experts in this field, for an
extensive treatment of this topic.
If from any set of positive integers a, where i = 1, 2, 3, . . . , finite
among
the
or infinite,
first
we
select
numbers
in
it
so that
this chapter
we
is
n = ax
of
n as a sum
made from
a2
1,
2,
q or
is
In
addends from the
a m the
a partition of the
of primes is a partition
contains repetitions,
from
all
considering both the case where repetitions are permitted and where they
are prohibited,
summands
in each situation
being
irrelevant.
When
are prohibited,
we
shall
parts.
If
number
of parts,
we
U,
190
PARTITIONS
If
1,
2,
Q(n, U,
and
191
always to include
is
q).
remaining n
Thus when n
To
m parts have
m units to
> m and the
selections are
to
parts.
we
integers,
have
= P(n  m,
P(n, m)
1)
+ P(n 
m, 2)
+
+
P(n
m, m)
But then
P(n
1,
m
= P(n 
1)
m,
1)
P(n
m,
2)
+
+
P(n
m,
1)
Hence,
= P(n
P(n, m)
1,
1)
P(n
m, m)
Theorem
equal to
When
121.
integers, the
P(n
number
of partitions of
P(n
the number
1,
To
Example.
find P(7, 3)
find
P(6, 2)
recursion formula of
1)
P(4, 3)
P(3, 2)
=
=
=
=
P(5,
1)
1)
P(3, 2)
P(2, 1)
into
121,
P(3,
is
m, m).
P(4, 3).
Theorem
P(6, 2)
P(4, 2)
we
we
of the
obtain
+ P(4,
+ P(2,
+ P(l,
+ P(l,
2)
2)
3)
2)
=
=
=
=
1
1
+
+
P(4, 2)
1
P(3, 2)
1
Hence,
P(7, 3)
= P(n 1, m 1).
and P(n m, m) = 0.
Corollary 2.
P(2n, 2) = n, and P(2n + 1, 2) = n.
The partitions of an even integer 2n into two parts are
Corollary
If
m>
If
1.
n/
2,
m > n/2,
m < m
P(n, m)
(2w
2)
(2n
n)
an odd integer 2n
+
+
f
The
are
partitions of
(2n
1)
greater than
into
two parts
192
+
+
2n
(n
By making
1,
(2n
1)
1)
1,
P(n, n
P(n,
1)
P(n  3, 2)
+ P(n  3,
3)
1)
1,
as well as
P(n, 2)
P(w, 3)
= P(n = P(w 
=
P(n, k)
1, 1)
1,
2)
+ P(n 
+ P(n  2, 2) =
+ P(n  3, 3) =
3, 2)
= P(n  1, k = 1 + P(n  *,
P(n
+ P(w + P(w 
1)
2)
+ P( 
P(n
2, 2)
2, 1)
3, 3)
ft)
fc,
*, 3)
P(n
k, k)
m parts selected
Values of n
m, the
number
of parts
10
12
11
15
10
13
11
2
3
5
6
7
8
9
10
11
12
The
row
first
and P(n, n
1)
42
32=1.
123.
exactly
Diagrams
of Partitions.
If
1, 2,
we
.
set
.
up
q,
193
PARTITIONS
least
one
q,
from each
of the
P(n, m, q) =
m 1 parts selected from 1,2,
q; that
_ q m i <g). If repetitions are prohibited, Q(n, m, q) =
q, m 1, <q 1).
We can diagram a partition of n into m parts that include q as a greatest
n
P( n
Q(n
q into
is,
way and
made above:
1
1
...
... 1
(q units)
(m rows)
Moreover,
we read
if
we have a
partition of
Theorem
from
1, 2,
122.
.
q
titions being permitted in each case.
Theorem
of
123.
n into at most
parts
repetitions permitted of
P(n,
<m) =
P(n,
<m) =
The number
n
is
of partitions
into parts
P(n, U, <m).
Furthermore, by subtracting P(n,
repetitions
<m
1)
P(n, U,
<m
1)
is,
from
Corollary.
of
that:
than or equal to m.
124. Generating Functions for the Number of Partitions.
Consider
the addends that are summed to determine the exponents of x in the
product (1
z 2 )(l + x*) = 1
x 1+2 + x 3 + x 1+ *
x)(l
x + x2
i+2+3
j j s evident that these exponents are the results arising
a2+3 _[_ C
from all possible selections of one, two, and three distinct summands from
the set 1, 2, 3.
Consequently, the coefficient 2 of x 3 gives the number of
ways 3 can be produced by adding together distinct integers selected from
Similarly the coefficient of x n in the expansion of (1 +
1, 2, and 3.
x)(l f x 2 )
(1 + x q ) is the number of partitions of n into dis
<
194
from the
tinct integers
set
Q(n, U, <q).
Now take the product
2,
1,
+ zx){\ +
(1
and
zx 2 )
is,
(1
zx q ).
The
pres
The
1, 2, 3,
effect of multiplying
~
z qiq 1)/2 is
(1
It,
q.
zx){\
zx 2 )
(1
zx q )
by 1 +
number
zx n
<m,
value of Q(n,
<q).
The fact that the series 1
xm
x 2m
developed from the
m
quotient 1/(1
x ), is absolutely convergent for
< x < 1 enabled
Kronecker* to prove that the coefficients of the expansion of the generating function 1/(1 x)(l x 2 )(l x z )
give the number of partitions of n with repetitions permitted.
The discovery of this theorem is
due to Euler. We shall merely indicate the truth of the statement by the
),
first
,
following argument:
of the factors (1
+x +x
n
(1
2n
f
+x+x +
2
determined by
),
(1
x2
the generating
The
function.
first
first
1 factors
the exponents,
a, 26, 3c,
...
nk,
is n,
number
we have
a partition of n.
of partitions of
n with
Thus
repetitions
permitted.
Let us illustrate the use of the generating function by finding the partiof 5.
We need but the factors (1 x + x 2 x 3 + x 4 x b ),
2
4
x
x
x 3 ), (1
x 4 ), and (1
z 5 ). The product is to be
+
(1
), (1
tions
2, p. 104.
PARTITIONS
195
x l+l
x l+l+1
x 1+1+1+1
a;i+i+i+i+i)(l
+
+
+
+
1+1+1
1+2
+
+
+
+
x 1+1+1+1
z 1+1 + 2
partitions of
1, 2, 3, 4,
__
^ 1+1+ l+i+l
^1+1+1+2
x 2+ 2
_J_
x l+2+2
1+3
__
^1+1+3
+
The
Z 5)
is
1+1
(1
+ x + x 2+2
+ x )(l + z )(l +
+
+
+
x*
as well as
5,
X 2+ 3
Z 1+4
5
+
+
+
+
+
+
+
and the
enumerates
Similarly we can see that 1/(1 x)(l x 3 )(l x h )
the partitions of n into odd integers with repetitions permitted and that
1/(1
x 2 )(l
x A )(l
x 6)
when the
EXERCISES
1.
2.
3.
4.
Show
Show
that P(n,
<m,
q)
P(n,
<
q,
in).
m+
BIBLIOGRAPHY
P., "Die Lehre von der Kreistheilung," B. G. Teubner, Leipzig, 1921.
"Grundlehren der neueren Zahlentheorie," Walter De Gruyter & Com
Bachmann,
,
York, 1914.
Chrystal, G., "Algebra," A.
&
I,
1932.
Chicago, 1929.
,
Chicago, 1930.
,
of
Numbers," University
of
Chicago
Hancock, H., "Foundations of the Theory of Algebraic Numbers," The Macmillan Company, New York, Vol. I, 1931, Vol. II, 1932.
Hardy, G. H., "Some Famous Problems of the Theory of Numbers," Oxford
University Press, New York, 1920.
and E. M. Wright, "The Theory of Numbers," Oxford University Press,
New York, 1938.
Hecke, E., "Theorie der algebraischen Zahlen," Akademische Verlagsgesellschaft m.b.H., Leipzig, 1923.
&
I,
Sons, Inc.,
New
York, 1951.
Ore, Oystein,
Its History,"
New
York, 1948.
Reid, L. W., "The Elements of the Theory of Algebraic Numbers," The Macmillan Company, New York, 1910.
Smith, D. E., "History of Mathematics," Ginn & Company, Boston, Vol. I,
Inc.,
198
Stewart, N. M.,
in
"Theory
of
New
Inc.,
York,
1952.
J. V., and M. A. Heaslet, "Elementary Number Theory," McGrawBook Company, Inc., New York, 1939.
Uspensky,
Hill
New
York, 1939.
&
Sons,
INDEX
Chrystal, G., 197
Abacus, 39
Absolute value, 7
Addition, 3, 54, 177
Additive theory of numbers, 190
Ahmes papyrus, 1
Algebra, 1, 16, 67
Algorithm, 31, 180
Amicable integers, 37
Archibald, R. C, 37n.
Archimedes, 9
principle of, 9, 12, 28
an integral right
Arithmetic function, 58
Area
of
Closure,
3,
Commutative
Composite,
179, 187
Conditional congruence, 67
Congruence, 53
conditional, 67
triangle,
168
degree
of,
68
identical, 66
properties
solution
of,
5455
67
of, 66,
Congruences,
integer, 8
of a polynomial, 179
177
25
9,
Composite polynomial,
"Arithmetica," 169
an
3,
Complementary factors, 7
Complex numbers, 47, 182
Arithmetica, 2
Associates, of
54jf.
equivalent, 6870, 91
3,
linear, 7071,
177
89
simultaneous, 7983
operations on, 68
quadratic, 134135
Bachet,
C, 169
Bachmann,
solution
P., 145n.,
of, 66,
67
Congruent
197
integers, 53ff.
Consecutive integers, 3, 6
Consecutive squares, 51
Crelle, A. L., 100, 106
Criterion for solvability, of ax
&(mod m), 71
of x n = c(mod m), 129
Cunningham, A., 127
by induction, 3, 4
Denary system of notation, 39
Definition
Cahen, R, 197
Descartes, R.,
Calculating machines, 39
Cancellation law, 3, 177
Carmichael, R.
Cauchy,
C,
A., 145,
197
168
Chinese, 88
2,
2, 3940
Diophantine equations, 16
linear, 17, 1923
simultaneous, 2324
Digits,
199
x, 77,
86
88
127/i.,
200
of,
16,
21
method of descent,
Fermat numbers, 32
Diophantus, 1, 16, 88
"Arithmetica," 169
Dirichlet, L., 27, 172
Field, 177
Discrete set, 3
" Disquisitiones arithmeticae," 136
Form, 14
Finite induction,
degree
30,
168
1011
14
linear, 14r16
quadratic, 46
Franqui, B., 37n.
178
7,
of,
4,
89, 105
by 3, 55
by 8, 9, and 11, 56
modulo ra, 55, 57, 72
Fundamental theorem
of arithmetic, 29
Divisor, 7
common, 7
an
Divisors, of
number
integer,
34
3435
sum of, 3435
of zero, 57, 178
Egyptians,
Goldbach, C., 51
Goldberg, B., 127
Greater than, 3
Greatest common divisor,
of,
1,
1415, 18,
188
used for solving congruences, 188
Greeks, 1, 16
Gupta, H., 37n.
39
9,
7,
"Elements,"
139
Equals, 4
c(mod m),
Ideals, 168
129, 136
Identical congruence, 66
identity, 164
<t>
Even
integer, 10
178
common, 7
greatest, 7
Factor,
7,
n!,
Incongruent integers, 53
Index of n, 125
Indicator, 58
Indices, 125131
used in solving congruences, 127129
Advanced Study, 40
Institute for
164166
terms of a base, 37
INDEX
Mathematical induction,
Mathews, G., 197
unique factorization
of,
201
2829
4,
1011
Meissel, E., 49
amicable, 37
Mersenne, M., 37
3, 6,
p,
51
W.
by
indices, 127
H., 46n.
even, 10
Mills,
negative, 5
Mobius function, 93
Mobius inversion formula, 9596
odd, 10
Modul, 16
Modulo m, 53
Modulus, 53
Multinomial theorem, 44
Integral logarithm of
Multiple,
positive, 3
rational,
square, 51
x,
50
7,
common, 8
Inverse, 5
least, 8,
33
modulo m, 72
Multiplication, 3, 54, 177
Jacobi,
C,
127, 145
0(mod
p), 76,
f(x)
77
Kronecker,
Kummer,
E., 145,
Lagrange,
J.,
194
Nonresidue,^quadratic, 135
Null element, 8, 186
168
Lambda
75, 136,
158
Number,
to d
to d
Less than, 3
Littlewood,
J.,
159
an
integer, 34
d modulo
p,
102
of divisors of
modulo p n 117
modulo 2p n 118
,
Numbers,
202
of,
Polynomials, domain
190
178
of,
185, 187
37
1,
13
3,
Power residues
multiply, 37
of a
modulo m, 101
Prime, 8
Mersenne, 37
common
greatest
183
signed, 2
Prime
Odd
integer, 10
Onetoone reciprocal correspondence,
Ordering relation, 3
Ore, O., SOn., 197
factor, of a composite,
notation
number
for,
for,
of,
of, 50,
n\,
4143
51
Partitions, 190
diagrams
25
40
of n!,
1
formula
of form, 2 n
192
190191
for,
generating functions
191192
193195
for,
2 2"
An
An
1,
27, 141
1,
number
table
not exceeding
of,
of,
n,
4750
52
n
,
108110
of 2p
111
and quadratic
integral, 14
159164
many, 2627
form An  1, 27
Peano's postulates, 13
Pellian equation, 172176
Perfect numbers, 3637
37
infinitely
multiply, 37
Phi function, 58, 88, 96, 100
Polynomial, 14
with coefficients in a field, 178183
28,
46, 151
Pascal, B., 88
of
1,
1,
residues, 138
Pythagorean
triangle, 167,
168
Pythagoreans, 10
primary associate
of,
183, 187
Quadratic forms, 46
reducible, 183
with
with
183185
prime modulus,
rational coefficients,
respect
to
186189
unique factorization
unit, 178
of,
184185
7,
10
203
INDEX
Sussman, B., 118
Symbol, definition
Radix, 38
Rational integers, 2, 3, 18, 183
Rational numbers, 177, 183
Real numbers, 183
Reciprocity law, 144147
si,
66
a\b,
Ep {m),
g(k),
(a, 6)
41
158
d,
40
139
in pairs, 9
Remainder in division, of a by
of g{x) by f(x), 72, 178
b,
10
(see
A(m), 99
93
58
tt(x), 47
<r(m), 34
r(m), 34
fi(m),
53
of,
<t>(m),
Symmetric property,
Primitive roots)
4,
55
Sarrus, F., 88
American, 40n.
Selberg, A., 27, 50
Sequence, 4
Shapiro, H. N., 27
Sieve of Eratosthenes, 25
Simultaneous congruences, 7983
Singlevalued function, 58
Smith, D. E., 197
Smith, H. J. S., 21n.
Transitivity, 4, 55
Scientific
Solution, of f(x)
= 0(mod
of
= b (mod
of x n
+ by = n,
+ yn = zn
ax
of ax
p),
8586
Uhler, H.
Unique
S.,
37
8, 178, 186
Unit polynomial, 178, 179
Unity, 177
Unity element, 177
Uspensky, J., 198
Unit,
m), 89
1920
168
J., 172
Waring, E., 158
Wertheim, G., 127
Wilson's theorem, 9293, 136
generalization of, 170172
Wright, E. M., 46n., 197
Wright, H. N., 198
Wallis,
21
= 0(mod
of f(x)
primes between, 51
Stewart, N., 198
Substitution, 4, 55
Subtraction, 3, 5, 54
of the divisors of an integer, 3435
over the divisors of an integer, 60, 62
of 2 squares, 159164
of 4 squares, 158, 164166
Sum,
Zassenhaus, H., 27
Zeller, C., 145
Zero,
1,
45, 178
no degree, 178
as null element, 8