Given a finitely generated -graded module
over a graded ring
(finitely generated over
, which is an Artinian local ring), the Hilbert function of
is the map
such that, for all
,
(1)
|
where
denotes the length. If
is the dimension of
,
then there exists a polynomial
of degree
with rational coefficients (called the Hilbert
polynomial of
)
such that
for all sufficiently large
.
The power series
(2)
|
is called the Hilbert series of . It is a rational function
that can be written in a unique way in the form
(3)
|
where
is a finite linear combination with integer coefficients of powers of
and
. If
is positively graded, i.e.,
for all
, then
is an ordinary polynomial with integer coefficients in
the variable
.
If moreover
,
then
,
i.e., the Hilbert series is a polynomial.
If
has a finite graded free resolution
(4)
|
then
(5)
|
Moreover, if
is a regular sequence over
of homogeneous elements of degree 1, then the Hilbert function
of the
-dimensional
quotient module
is
(6)
|
and in particular,
(7)
|
These properties suggest effective methods for computing the Hilbert series of a finitely generated graded module over the polynomial ring , where
is a field.
The Hilbert series of ,
which has dimension
,
can be obtained by considering the maximal regular sequence
of
, and the Hilbert function of the 0-dimensional quotient
ring
,
which is the same as
.
Now
,
and
for all
.
Hence
.
It follows that
is the constant polynomial 1, so that
(8)
|
This approach can be applied to all Cohen-Macaulay quotient rings , where
is an ideal generated by homogeneous polynomials. The first
step is to find a maximal regular sequence
of
composed of homogeneous polynomials of degree 1; here, by
virtue of the Cohen-Macaulay property,
. This will produce a 0-dimensional ring
(a so-called Artinian reduction
of
)
whose Hilbert series is the polynomial
. By (5) and (6) the result is
(9)
|
If, for example, ,
which is a 1-dimensional Cohen-Macaulay ring,
an Artinian reduction is
. Its Hilbert
series can be easily determined from the definition:
for all
, whereas, for all
,
, since the length of a vector space over
is the same as its dimension. Since
in
all multiples of
and
are zero, we have
(10)
| |||
(11)
| |||
(12)
|
Hence,
This is
.
By (8) it follows that
(13)
|
The same result can be obtained by first constructing a graded free resolution of over
,
(14)
|
which yields ,
whereas the remaining
are zero. Hence, by (4) and (7),
(15)
| |||
(16)
| |||
(17)
|
as above. We rewrite it in the form of a power series,
(18)
| |||
(19)
|
From this, according to (2), we can retrieve the values of the Hilbert function ,
(20)
|
It follows that the Hilbert polynomial of is the constant polynomial
.
More generally, the graded free resolution of , where
is the ideal
of
, and
is a polynomial of degree
, is
(21)
|
and the Hilbert series of is
(22)
|
For more complicated ideals , the computation requires the use of Gröbner
bases, with the techniques explained by Eisenbud (1995), Fröberg (1997),
or Kreuzer and Robbiano (2000).
Historically, the Hilbert function arises in algebraic geometry for the study of finite sets of points in the projective
plane as follows (Cayley 1843, Eisenbud et al. 1996). Let be a collection of
distinct points. Then the number of conditions imposed by
on forms of degree
is called the Hilbert function
of
. If curves
and
of degrees
and
meet in a collection
of
points, then for any
, the number
of conditions imposed by
on forms of degree
is independent of
and
and is given by
(23)
|
where the binomial coefficient is taken as 0 if
(Cayley 1843).