numbers, let us first examine the properties which a domain of positivity
P of the field of rational numbers must have, in case such a domain
exists. The set of integers contained in P is obviously a domain of positivity
for the ring of integers, so that in particular N Q P. But from D43.4) and
a^ ~ (—ay it follows that a domain of positivity of a ring must contain
all squares :^Q. Thus for a, b e N it follows from a/b = ab{b~^Y that
a/b e P. But if P contained a rational number that is not of this form,
then by D4i) it would also contain natural numbers a, b with (~a)/b e P,
which contradicts D43), since a/b -\- {—a)/b = 0^ P and a/b e P. Thus P
consists precisely of the quotients of natural numbers. But these quotients
do in fact form a domain of positivity, since a rational number 7^0 which
is distinct from these quotients has the form {~a)/b {a, b e A^), which
means that D43) is valid, while D4i.3.4) obviously hold. Thus the field of
rational numbers can be ordered in exactly one way. Since
b/b' - a/a' = b/b' + {-a)/a' = (ba' + b'(-a))/a'b' = (ba' - b'a)/a'b',
we have for integers a, a', b, b'\ a/a' < b/b' for a', b' > 0, if and only if
ab' < a'b. Since c > 0 implies c~^ = c(c~^y > 0 for every rational
number c, this result can easily be extended by D5) to arbitrary rational
numbers a, a', b, b'.
The ordering of the rational numbers is Archimedean; that is, for every
a, b > 0 there exists a natural number n with na > b}"^ For the proof we
first restrict ourselves to integers a, b. Then from b + \ > b and a > 1
it follows by the monotonic law that {b ^ \)a > ba ^ b\ = b, so that
na > b with n = b -\- I. But then for the rational numbers a/a', b/b' > 0
(a, a', b, b' natural numbers) we have a/a' = ab'/a'b', b/b' = a'b/a'b', so
that if we choose a natural number n with nab' > a'b, it follows by
multiplication with the positive number (a'b')-^ that n(a/a') > b/b', as
desired. In an arbitrary module, which may not contain the natural
numbers, we can always define na as being equal to X?=i ^' so that the
definition of "Archimedean" is applicable to any ordered module. At the
end of §4.3 we give an example of an ordered module in which the ordering
is not Archimedean.
The ordering of a field^^ is Archimedean if and only if 0 ^ a < aj-\
for all natural numbers n implies a = Q. For if a > 0 in an Archimedean
*' This statement is often called the "axiom of Archimedes" since it occurs as an
axiom in geometry (cf. 112, §1.2).
** If the field does not contain the natural numbers, then in the following inequality
(and in the proof) the natural number n must be replaced by the nth multiple of the unit
element 1 of the field.
128 PART 6 ARITHMETIC AND ALGEBRA
ordering, then there exists a natural number n with na > I, so that
a > n~'^. The same argument obviously holds for an ordered ring that
contains n~^ for every natural number n. Conversely, if 0 < a < n~^
implies a = 0, and if a, ^ > 0, so that ajb > 0, then the inequality
rtfl < ^ cannot hold for every natural number n, since it would imply
a/b < rt-i.
The absolute value | a | of the rational number a is defined as follows:
E1) I a I = max (a, —a);
where by max(a, b) we mean the number b if a < b and the number a if
a ^ b. Thus \ a\ = a or = —a and \a\ > 0, and these properties
obviously characterize the number | a \. Then we can at once derive
E2) I a^ I = I a I I ^ I.
Since ±a < I a I for all a, we have ±(a -\- b) ^ \ a \ -\- \ b \ and therefore
E3) I a + ^ I < I a I + I ^ 1.
Replacing ^ by—^, we see, since I —b \ = \b |,that| a — b\ <IaI + I^I,
and if we replace a — bby a, and a by a + ^, we have
I a I < I a + ^ I + I ^ I, and therefore la|-|^| <Ia + ^|.
Since the right-hand side is not altered by the interchange of a and b,
it follows that
E4) ||al-l^|| <|a + />|.
As in E3), we may replace a + ^ by a — ^. Of course, the definition E1)
of absolute value, and with it the consequences E2), E3), E4), are valid
for any ordered ring.
3.5. Endomorphisms
In view of the distributive law, the mapping x-^cx for any rational
number c is an endomorphism of the module (with respect to addition)
of the rational numbers. As in §2.4, we can show that for two
endomorphisms/, ^ of this module, the equality/(I) = ^A) implies/(x) = g{x)
for all integers. From C8) it is easy to prove /(Il?=i ^i) = Sr=i/(^f) by
complete induction. Thus for a rational a/a' (a, a' integers, a' > 0) and
two endomorphisms/, g with/(I) = ^A) we have:
a'f(a/a') = £ /(a/a') = f(a'(ala')) = f{a) = g{a) =