Recall
Lemma 12.5.
Suppose
is an algebraically closed field.
If
V(I) is finite, then so is
B(I).
Before we can prove this lemma, we need the following result.
Hilbert's Nullstellensatz (strong form).
Suppose
is algebraically closed and
I is an ideal in
.
For any
, if f(P)=0 for every
, then some
power of f must lie in
I.
Note that if
is NOT algebraically closed, then Hilbert's Nullstellensatz
might not hold.
For example, let
,
.
Then
.
For f =1, we have f(P)=0 for every
,
but
.
Proof (of Lemma 12.5).
Since
V(I) is finite, for each
,
the ith coordinate
of points in
V(I) has only finitely many values. Let
be
a nonzero polynomial that vanishes at these values. Viewing fi as
a polynomial in
we have
fi(P)=0 for all
.
By Hilbert's Nullstellensatz,
for some mi.
The leading
term of
fimi is of the form
xini for some ni, thus
for
.
By Lemma 11.4,
B(I) is finite.
Next we want to see when equality occurs. That is, when does
?
Proof. From Lemma 12.3,
we know that
.
Let
where
t=|V(I)|.
There are polynomials
such that
Suppose first that
I is radical.
For any
,
define a new function
Conversely, suppose that
t=|B(I)|.
Let
be such that
for some m.
Then
(g(P)m = gm(P)=0 for every
.
Hence
g(P) = 0 for every
.
Since
,
the linear independence of
implies that
they form a basis for
over
.
We can express g as
The functions
constructed in the proof are very important.
They are called orthogonal idempotents. That is, they satisfy
Proof.
Let
,
and let fi,
,
be
defined as above.
We know from the proof of previous theorem that
form a basis for
over
,
that is,
This map is a Fourier Transform!!!
Remark.
When
I is not a radical ideal, the ring structure of
can also be determined quite easily.
Suppose one is familiar with the concept of multiplicity
of a point as a common zero of polynomials in an ideal.
Also, for an integer m, denote by
the ring
.
Theorem. Suppose that
is algebraically closed and
an ideal. Suppose further that
(so is finite) and Pi has multiplicity mi for
.
Then we have the ring isomorphism