The following result was known to
Chinese mathematics
in the first century AD. By way of notation, if
a,b
are integers and
n is a positive
integer we write
a cgt
b (mod
n) if
n divides a-b.
Theorem Let n1,....nm
be pairwise coprime positive integers, that is,
(ni,nj)=1
for each distinct i,j. Then if a1,...
am are integers the system of equations
a cgt a1 (mod n1), ...,
a cgt to am (mod nm)
has an integer solution that is uniquely determined
modulo n1...nm.
Here's an example:
n1=3, n2=4, a1=1, a2=2
Then we can take
a=10 (or
a=...,-2,22,34,...).
A generalisation of this result to ring theory nowadays gets
called the Chinese Remainder Theorem. Let's state and prove this
generalisation.
Theorem Let I1,...,Im
be ideals in a ring R and suppose that Ii+Ij=R
for each distinct i,j. Then if a1,...
am are elements of R there exists
a in R such that the images of a and ai
in the quotient ring R/Ii coincide. Further
a is uniquely determined up to its coset a+I, where
I is the intersection of all the Ii.
Proof that the first theorem follows from the second:
Take the ring
R=Z and consider the ideals
Ii=niZ. One has
that (ni,nj)=1 iff
Ii+Ij=Z. For, if ni and
nj are coprime then Euclid's algorithm for the
highest common factor supplies r,s such that
1 = nir + njs
On the other hand, if
Ii+Ij=
Z
then we have an equation like the displayed one. Clearly any common
divisor of
ni and
nj is a
unit.
Finally, note that the quotient ring
Z/n
Z is the ring of
integers modulo n and that
n1...nmZ
is the intersection of the
Ii.
Proof of the second theorem:
Consider the inductive statement
S(k) R=I1 + I2 n....n Ik
for
k=2,....,m. Here
n denotes intersection.
By assumption,
S(
2) is true. I will
show that
S(
k) is true for all these values of
k, by
induction. So suppose that
S(
k-1) holds. Thus
R=I1 + I2 n....n Ik-1
Since
R.R=R (it contains 1) and
I1+Ik=R
we deduce that
R=(I1+Ik)(I1 + I2 n....n Ik-1)
Multiplying out the brackets we see that the LHS is contained in
I1 + Ik(I2 n....n Ik-1)
But the second of thse terms is contained in
I2 n....n Ik, and so we are done.
Thus
S(
m) is true.
The same argument (with different
labelling of the ideals) shows that Ii + the intersection
of all Ij, with j different from i, is
equal to R. Thus, for each i, we can choose bi
in Ii and ci in the intersection of
all Ij, for j different from i, such that
ai=bi + ci.
Now let a=c1+...+cn. Think about
the image of a in R/Ii. In this ring each
cj, for j different from i, has
image zero. Also in this ring ci has the the same image
as ai. Thus a has the required property that
it has the same image as ai in R/Ii.
If a' also has the same property then a-a' lies in each
Ii, hence the claimed uniqueness.
Finally a corollary that restates the result in terms of direct products
of rings.
Corollary With the hypotheses and notation of the theorem
there is an isomorphism of rings between R/I and the direct product
R/I1 x....x R/Im.
Proof: Define a ring homomorphism
f:R-->R/I1 x....x R/Im
by f(a)=(a+I1,...,a+Im). The theorem shows
that this map is surjective and has kernel I, hence we are
done by the first isomorphism theorem.