the structure + , X , and < ) of the integers as first constructed to those described by pairs and equality. In either construction, the integers
Z
can be described as the minimal way, unique up to isomorphism, of embedding N in a larger structure in which subtraction is always possible, preserving all the algebraic properties of + and X . Here as elsewhere, what matters is not an exact description of what an integer
is,
but a description of the structure of all the integers, up to isomorphism.6.
Rational Numbers
To keep accounts, one often needs to divide numbers evenly into parts
and this often cannot be accomplished with whole numbers. Fractions provide the answer. They are introduced individually, as
1 /2, 2/3, 1 /5, 4/5,
etc., and then manipulated in the evident way :m m ' mn '
+m 'n
- +
n
-n ' -
-nn ' ' m m '
n n ' mm '
nn ' . ( I )
This practical process suggests a corresponding formalization. The ini
tial observation is that division is not always possible within the set N + of all positive natural numbers. Hence one is led to introduce the set Q + of all ordered pairs (
m,n
) ofpositive
natural numbers, defining addition and multiplication by the evident translations(
m,n
) + (m ',n '
) = (mn '
+nm ', nn '
), (m,n
)(m ',n '
) = (mm ',nn '
)(2)
of the practical rules
( I )
and taking care to define the equality of (m,n
)and
(r,s
) byms
=nr.
Whenm
in N + is identified with the pair (m, l
) this again defines a minimal expansion of the set N + to a larger set in which division is always possible and in which all the rules of arithmetic still hold. As before, there is nothing unique about the formulation of this construction. Instead, one might have used only those pairs (m,n
) in whichm
andn
have no common factor (exceptI);
in that case, one must modify addition and multiplication in( I )
to reduce each answer to lowest terms.With this inconvenience, the "artificial" definition of equality of pairs is avoided. Again, what matters is only the resulting structure up to isomor
phism.
The system Q of
all
rationals may then be obtained from Q + , the positive rationals, by simply adjoining zero and negative rationals. Alterna
tively, one may construct Q directly from
Z
by using all pairs(a,b )
of integersa,b
inZ,
with the same addition and multiplication as in (2)-and the important proviso that the "denominator"b
is neverO.
As in previous cases, what matters is not the explicit definition of a rational, but the resulting structure.
5
2 I I . From Whole Num bers to Rational N u m bers7 . Congruence
A typical clock runs up to the figure of 1 2 hours and then repeats, but one can still do arithmetic on the limited list of hours: Seven hours after nine o'clock is four o'clock. Similarly, in the decimal system there are only ten digits 0, 1 ,2 , . . . , 9 ; the usual rules for addition and multiplication, ignor
ing the carryover to the tens' place, work perfectly well for the manipula
tion of these digits by themselves:
6
+ 7 = 3,8
+ 7 =5,
8 · 3 =4,
3 · 9 = 7 .These rules ignore all the multiples of 1 0 ; in a sexagesimal system there are similar rules which ignore all the multiples of sixty. "Casting out nines" is a rule for checking arithmetic calculations. This rule for checking a multiplication says: Add up the digits of each factor, multiply the result
ing sums, and check this against the digit sum of the original purported answer. Thus 32 times 27 calculates to
864.
To check, 32 becomes5,
27 becomes 9,5
times 9 is 45, with digit sum 9. This checks with the digit sum in the purported answer, which is8
+6
+4
= 18
with digit sum 9.What happens here is that 32 is replaced by
5,
casting out the difference which is 27, or three nines. The reason it works is that factors differing by a multiple of nine will have a product differing (at most) by a multiple of 9. In brief, arithmetic operations are valid "casting out 9's" or "modulo 9".These examples each involve the use of a modulus: 1 2, 1 0, 60, or 9, as the case may be. The general procedure is similar. For integers
a,
b, and any natural numberm
oF 0 as modulus, one writesa
== b (modm),
or says thata
iscongruent
to b for the modulusm,
when the differencea
- b is a multiple ofm.
Then one readily proves the arithmetic rules: Ifa
== b andc
== d, both modm,
thena
+c
== b + d (modm), ac
== bd (modm) .
( 1 ) This congruence modulom
behaves like equality; also i t is reflexive, symmetric and transitive. ("Transitive" is defined in § 1 .
5
; a relation such as ==is
symmetric
whena
== b implies b ==a
for alla
and b.)Two integers
a
and b are congruent modulom
if and only if they leave the same remainderr,
with 0< r
<m,
upon division bym.
As a result, calculations modulom
amount to calculations with a finite list of objects (to wit, with the remainders 0, 1 ,2 , . . ., m - l).
All the rules for addition and multiplication-commutative, associative, and distributive laws-still hold for these finite calculations. Thus the remainders modulom
form an (abelian) group under addition. Under multiplication, the non-zero remainders modulo a primep
also form a group ofp
-I
elements. This is not the case for a composite modulusm,
such as 2 · 3, because there 2 · 3 == 0 (mod6)
so that neither 2 nor 3 can have a multiplicative inverse7. Congruence
53
modulo
6.
To get a multiplicative group for such a composite modulusm,
on must use only those remainders
r
which have no factor in common with the modulusm.
The number of such remainders is denoted bycp (m),
while
cp
is called Euler'scp
-function. For a primep
or for integersm,n
with greatest common divisor l one readily calculates that
cp (p )
=p - I , cp (pk )
=(p _ l )pk - l, cp
(mn
) =cp ( m )cp ( n ) . (2)
These formulas provide for a computation of any
cp (m)
from the prime decomposition ofm.
We cite them to emphasize that the formulation of congruence arises both from practice (multiplying hours or digits) and from number theory.To say that calculations with congruences
are
calculations with the remainders is a bit artificial. Thus modulo5
one could replace the five remainders0, 1 , 2, 3, 4
by the remainders- 2, - 1 , 0, 1 , 2
or by- 4, - 3, - 2,
-I , 0.
Here, as always, mathematicians strive for an invariant formulation. Each remainder
r
stands for (and may be replaced by) the"congruence class"
emr
ofall
integersa
witha
==r
(modm).
To add the classemr
to the classems
one may then take any representativea
inemr,
any
b
inems,
adda
andb,
and take the class of this suma
+b
as the sumemr
+ems.
One must then prove that the resulting sum of classes doesn't depend on the representativesa
andb
chosen-but this fact is just a restatement of the rule( I )
for adding two congruences. With this fact established, we see that the collectionZm
of all these congruence classesem
forms a system with binary operations of addition and multiplication, and that the functionem
fromZ
toZm ,
as incarries the addition and multiplication of integers to that of congruence classes. (It is thus a first example of a homomorphism of + and X .) This gives an "invariant" formulation of the calculation with remainders.
Thus we have (at least) three descriptions of the algebra of integers modulo
m:
As the ordinary integers taken with a new equality, congruence modulom;
as the algebra of remainders modulom;
or as the algebra of congruence classes, modulom.
The last description is the more invariant-and the more sophisticated, since it involves a set whose elements are sets (a collection of congruence classes). However, all three con
structions yield isomorphic results, and the results are useful (and practi
cally indispensable) for the statement of simple number theoretic facts.
For example, for any integer x, one has always
x2 == ° or
I
(mod4),
x 2 ==0, I ,
or4
(mod8) .
Another problem is that of finding a common solution x for two (simul
taneous) congruences:
54
I I . From W hole Num bers to Rational Num bers x ==b
( modm ),
x ==c
( modn ) .
(3) In this situation, the "Chinese remainder theorem" states that ifm
andn
have no common factors (except 1 ), there always is a solution, unique modulo the product