You are on page 1of 13

168

4. Chow Groups

On the other hand, since the cycle associated to X0 is insensitive to the


embedded points, we have
[X] [X0 ] = [D],
as claimed.
There are many other interesting examples of flows t for which we
could make similar constructions, such as the one-parameter subgroups
Gm Aut(P n ). Since every representation of the multiplicative group Gm
can be diagonalized over an algebraically closed field, one can write such
an action, in suitable coordinates, in the form
Gm 3 t : (x0 , . . . , xn ) 7 (tw0 x0 , . . . , twn xn )
for some integral weights wn . An example is given in Exercise 4.42.

4.2 Rational equivalence through divisors


Another important way to construct rational equivalences is by using rational functions on subvarieties of X (in fact, Corollary 4.18 shows that
this method can be used to construct all rational equivalences, and thus
gives an alternate presentation of A(X).) ****insert picture**** To set
the stage, we review the definition of the divisor associated to a nonzero
rational function on an arbitrary variety.
Consider first the affine case, and the divisor associated to a nonzero
principal ideal in a domain. Let U be an affine variety with coordinate ring
A, and let f A be a nonzero function. Krulls Principle Ideal Theorem 0.1
asserts that each irreducible component Y of the subscheme V (f ) defined
by f is of codimension 1 in X. Thus the local ring R = OU,Y of the generic
point of Y is 1-dimensional, so, as above, the factor ring R/(f ) is of finite
length. We define the order of f at Y to be this length, for which we will
write ordR (f ) or ordY /U (f ), so that
ordR (f ) = ordY /U (f ) = length OU,Y /f OU,Y .
We define the divisor of f to be the divisor of the subscheme defined by f ,
X
Div(f ) :=
ordY (f ) Y.
{Y |Y is a
component of V (f )}
Note that, by Krulls Principal Ideal Theorem all the components of V (f )
have codimension exactly 1 in U .
What makes the case of a principal ideal special is that the association
f 7 Div(f ) turns multiplication into addition: that is,

4.2 Rational equivalence through divisors

169

Proposition 4.9. If f, g are nonzerdivisors in a local Noetherian ring R


of dimension 1, then
DivX (f g) = Div(f ) + Div(g).
Proof. Since g is a nonzerodivisor, multiplication by g defines an isomorphism R/(f ) (g)/(f g). The resulting exact sequence
0

- R/(f )

- R/(f g)

- R/(g)

- 0,

gives ordR (f g) = ordR (f ) + ordR (g).


This additivity has a crucially important consequence: we can define
the divisor not just of a regular function, but also of a rational function.
Suppose that K(X) is a rational function on a (not necessarily affine)
variety X, and Y is any codimension 1 subvariety in X, then we can choose
nonzero f, g R := OX,Y such that = f /g. The number ordR (f )
ordR (g) does not depend on the choice of f and g, since if f /g = f 0 /g 0 we
will have f g 0 = f 0 g, whence by additivity ordR (f ) + ordR (g 0 ) = ordR (f 0 ) +
ordR (g), so that ordR (f ) ordR (g) = ordR (f 0 ) ordR (g 0 ). Thus we can
unambiguously define ordY () = ordR (f ) ordR (g).
We claim that, for any nonzero rational function , there are only finitely
many codimension 1 subvarieties Y X such that ordY () 6= 0. To see
this, take a finite covering by affine open subsets Ui . For each i, write
|Ui = fi /gi with fi , gi OX (Ui ). If Y appears with a nonzero coefficient
in Div(), and Ui is one of the open sets such that Y Ui is dense, then
Y Ui must be among the finitely many irreducible components of V (fi )
or V (gi ). This leaves only finitely many possibilities for Y .
Proposition 4.10. Let X be a variety of dimension n. There is a homomorphism of groups
Div : K(X) Ratn1 (X) Zn1 (X)
taking K(X) to
Div() =

ordY ()Y.

is a
codimension 1
subvariety of X.}
{Y |Y

Proof. The argument above shows that the map Div : K(X) Zn1 (X)
is well-defined. To see that its image is in Ratn1 (X), let be a rational
function on X. The element defines a rational map : X P 1 and the
graph of is a subset of X P 1 birational to X. What we have called
Div() is, under this isomorphism, precisely 1 (0) 1 ().

170

4. Chow Groups

Example 4.11. Let P P 3 be a plane. Any two curves D, D0 P, both


of degree d, are rationally equivalent in P and thus in P 3 : for if g = 0 and
g 0 = 0 are the equations of D and D0 respectively, then the ratio = g/g 0
is a rational function on P because g and g 0 are homogeneous of the same
degree. Further, Div() = [D] [D0 ].

Here are some observations that will help us with further examples:
Proposition 4.12. Let X be a scheme.
(a) A(X) = A(Xred ).
(b) If X is a variety of dimension d, then Ad (X)
= Z, with generator [X],
called the fundamental class of X. More generally, if X is reduced, then
Ad (X) is the free abelian group on the classes [Xi ] of the irreducible
components of X that have dimension d.
(c) If Y X is a closed subscheme and U = X \ Y is the complementary
open subscheme, then there is a right-exact sequence
A(Y )

- A(X)

- A(U )

- 0.

(d) If X1 , X2 are closed subschemes of X, then there is a right exact sequence


A(X1 X2 )

- A(X1 ) A(X2 )

- A(X1 X2 )

- 0.

Proof of Proposition 4.12. (a) By definition the generators of Z(X) are


varieties in X, and since they are reduced they are contained in Xred . The
same argument shows that Z(X P 1 ) = Z(Xred P 1 ), so the cokernel of
is unchanged if we replace X by Xred .
(b) Since X P 1 is the only subvariety of X P 1 of dimension d + 1, the
map X : Zd+1 (X P 1 ) Zd (X) is zero, so Ad (X) = Zd (X) = Z [X].
For the second statement, let X1 , . . . , Xm be the distinct irreducible
components of X of dimension d. By definition, Ad (X) is generated by
[X1 ], . . . , [Xm ]. Any relation among the [Xi ] comes from a variety W
X P 1 dominating P 1 and containing Xi {0}.
S
Since W = i (W (Xi P 1 )) and W is irreducible, we must have
W Xi P 1 . Since W dominates P 1 , it is strictly larger than Xi 0, so
dim W 1+dim Xi = dim(Xi P 1 ). Thus W = Xi P 1 . Thus the relation
on Ad (X) provided by W is of the form [Xi ] = [Xi ]the trivial relation.

4.2 Rational equivalence through divisors

171

(c) There is a commutative diagram


- Z(Y P 1 ) - Z(X P 1 ) - Z(U P 1 )

Y
0

- 0

?
- Z(Y )

?
- Z(X)

?
- Z(U )

?
A(Y )

?
- A(X)

?
- A(U )

?
0

?
0

?
0

- 0

where the map Z(Y ) Z(X) takes the class [A] Z(Y ), where A is a
subvariety of Y , to [A] itself, considered as a class in X, and similarly for
Z(Y P 1 ) Z(X P 1 ). The map Z(X) Z(U ) takes each free generator
[A] to the generator [A U ], and similarly for Z(X P 1 ) Z(U P 1 ).
The two middle rows and all three columns are evidently exact. A diagram
chase shows that the map A(X) A(U ) is surjective, and the bottom row
of the diagram above is right exact, yielding part (c).
(d) Let Y = X1 X2 . We may assume X = X1 X2 . We may argue
exactly as before from the diagram
0 - Z(Y P 1 ) - Z(X1 P 1 ) Z(X2 P 1 ) - Z(X P 1 ) - 0

?
- Z(Y )

?
- Z(X1 ) Z(X2 )

?
- Z(X)

- 0

?
A(Y )

?
- A(X1 ) A(X2 )

?
- A(X)

- 0

- Z(X1 ) Z(X2 ) takes a generwhere, for example, the map Z(Y )


ator [A] Z(Y ) to ([A], [A]) Z(X1 ) Z(X2 ) and the map Z(X1 )
Z(X2 ) - Z(X) is addition.
Proposition 4.12 allows us to answer Keynote Question (b) in the negative. Indeed, we can generalize as follows.
Corollary 4.13. If X P n is a variety of dimension m and degree d,
then Am (P n \ X)
= Z/(d), while if m < m0 n then Am0 (P n \ X) = Z. In
particular, m and d are determined by the isomorphism class of P n \X.
Proof. Part (c) of Proposition 4.12 shows that there are exact sequences
Ai (X) Ai (P n Ai (P n \X 0. Furthermore Am (X) = Z by Part ?? of
Proposition 4.12, while Am0 (X) = 0 for m < m0 n. By Theorem 1.20 we

172

4. Chow Groups

have Ai (P n ) = Z for 0 i n, and the image of the generator of Am (X)


in Am (P n ) is d times the generator of Ai (P n ) . The results of the Corollary
follow.

4.3 Proper push-forward


To get much further we need stronger tools. One of them is very compactly
expressed below by saying that A(X) behaves like a covariant functor with
respect to proper maps, and this preserves the grading by dimension. This
can be used to prove, for example, that the class of a given cycle is nonzerowe just exhibit a proper map that takes it to a cycle whose class we
already know to be nonzero. We will see examples of this below.
If f : X X 0 is a proper map of schemes, then the image of a subvariety
Y X is a subvariety f (Y ) X 0 . This would make it possible to define
a push-forward map f on cycles by taking any subvariety to its image.
However, this map gives no information about the Chow groups because it
does not preserve rational equivalence (see Exercise ??). But if we introduce
certain multiplicities, rational equivalence will be preserved, and we get a
map on Chow groups. Here is how this is done:
Definition 4.14. Let f : X X 0 be a proper map of schemes, and let
Y X be a subvariety.
(a) If f (Y ) has lower dimension than Y , then we define f (hY i) = 0.
(b) If dim f (Y ) = dim Y , then the map f |Y : Y f (Y ) is generically
finite. If n := [K(Y ) : K(f (Y ))] is the degree of the extension of fields
of rational functions, we say that f is generically a cover of degree n,
and we define f hY i = nhf (Y )i.
P
(c) If z =
ni hYi i Z(X) is a P
formal Z-linear combination of subvarieties, then we define f (z) = ni f hYi i.
Theorem 4.15. If f : X X 0 is a proper map of schemes, then the map
f : Z(X) Z(X 0 ) defined above preserves rational equivalence, and thus
induces a map f : A(X) A(X 0 ).
We will prove Theorem 4.15 and the closely related Theorem 4.17 below
after developing an important algebraic tool in Lemma 4.19.
Before starting the proof of Theorem 4.15 we discuss some geometric
consequences. A first hint of the importance of the apparently bland statement of the theorem is that it makes possible the existence of a degree
function A0 (X) Z when X is projective over a field. In particular, it
allows us to prove that A0 (P nK ) = Z [p] for any closed K-rational point
p P n , something we could not prove in the last section.

4.3 Proper push-forward

173

a + b + c d + e + f 2g + h
c
X

b
a

X0
a0 b0

c0

d0

g0

h0

a0 + b0 + c0 3d0 2g 0 + h0
FIGURE 4.4. Pushforwards of equivalent cycles are equivalent

Corollary 4.16. Let K be a field, and let X be a scheme that is proper


over Spec K. There is a map A0 (X) Z that sends the class [Y ] of a
zero-dimensional subscheme Y X to deg Y := dimK OY . In particular
A0 (X) 6= 0.
Proof. First, the definition of deg : Z0 (X) Z makes sense because any
zero-dimensional scheme, proper over Spec K, is finite over Spec K. Let
f : X Spec K be the constant map. If p is a closed point of X then,
by definition, f (hpi) = dimK (p) hSpec Ki, so deg(hpi) = deg f (hpi) =
dimk (p). By Theorem 4.15, the map deg f is well-defined on A0 (X).
The idea of this proof also shows that we cannot drop the assumption of
properness in Theorem 4.15: for example, if we could define f on A0 (A nK )
when f : A nK Spec K, we would deduce, as above, that A0 (A nk ) 6= 0 and
we have already seen that A0 (A nK ) = 0. See Exercises 4.29 and 4.30 for
further examples.
The statement of Theorem 4.15 is obvious when f is a closed immersion since in this case any rational equivalence in Z(X) is also a rational
equivalence in Z(X 0 ). But already in this case, we can use the result to
answer Keynote Question (a) in the negative: If X P 3 is a surface with
a map : X P 1 then the fibers of are rationally equivalent in X. By
Theorem 4.15 they are also rationally equivalent in P 3 . By Theorem 4.4
this implies that they have the same degree as projective curves.
We can sharpen Theorem 4.15, in the generically finite case, by giving an
explicit formula for the push-forward of the divisor of a rational function.

174

4. Chow Groups
a

a0

ha0 i = hai
hai + hai = N hai
FIGURE 4.5. The pushforward of the divisor of a function is the divisor of its
norm

Theorem 4.17. Let f : X Y be a proper, dominant, generically finite morphism of varieties, and let be a rational function on X. If
N : K(X) K(Y ) denotes the norm function,



N () := det K(Y ) K(X) - K(X)


then f (Div()) = Div(N ()) Z(Y ). In particular, the push-forward of
the divisor of a rational function under a generically finite morphism is
again the divisor of a rational function.
We saw (just before Proposition 4.12) that if W is a subvariety of X
and is a rational function on W , then Div(), regarded as a cycle on X,
is rationally equivalent to zero. As a consequence of Theorem 4.17 we can
show that the divisors of rational functions suffice to generate the relation
of rational equivalence:
Corollary 4.18. For any scheme X, the kernel of the map Z(X) A(X)
is generated by the divisors of the form Div() where ranges over rational
functions on subvarieties of X.
Proof of Corollary 4.18. We already know that such Div() are in the kernel. On the other hand, the kernel is generated by elements of the form
hV0 i hV i where the Vi are subschemes of X such that there exists a
variety V X P 1 generically finite over its image in X, and dominating
P 1 , with V X {0} = V0 and V X {} = V . As a cycle on V we
have hV0 {0}i hV {}i = Div(), where is the rational function
corresponding to the projection map V P 1 .

4.3 Proper push-forward

175

Let V 0 be the image of V in X, and note that V 0 is a subvariety. As


cycles on X we have V0 = f (hV0 {0}i), and similarly for V . Thus
hV0 i hV i = f (Div()) = Div(N ()), the divisor of a rational function
on V 0 , completing the argument.

4.3.1

The determinant of a homomorphism

The proof of Theorems 4.15 and 4.17 rely on a fundamental result in commutative algebra relating a homomorphism to its determinant.
Suppose that R is a domain with quotient field K. If M is a finitelygenerated torsion-free R-module and : M M is an endomorphism,
then R K is an endomorphism of the finite dimensional vector space
M R K. We define the determinant det() of to be the usual determinant
of R K, an element of K.
For example, if r R and is multiplication by r, then det = rrank M ;
this is because the endomorphism R K of M R K = K rank M is also
multiplication by r, represented by a diagonal matrix

r 0 ... 0
0 r . . . 0

..

.
0 0 ... r
Lemma 4.19 (Determinant Lemma). Suppose that R is a 1-dimensional
local Noetherian domain, and let : M M be an endomorphism of
a finitely generated torsion-free R-module M . If det is nonzero, then
coker is a module of finite length, and its length is equal to ordR (det ).
In particular, if r R and M has rank n, then
length(M/rM ) = n length(R/rR).
The Lemma implies, in particular, that ordR det 0, even when det
is not in R. This comes about because det is always in the integral closure
of R (use, for example, Eisenbud [1995], Corollary 4.6).
When M is free and the endomorphism is defined by a diagonal matrix,
the result says that
Y
X
length R/( xi ) =
length R/(xi ),
i

an easy special case of Proposition 4.9. The idea of the proof we will give
is to reduce to this situation by moving to the integral closure of R and
then localizing so that R becomes a discrete valuation ring. We can then
use the structure of matrices over such rings.

176

4. Chow Groups

Proof of Lemma 4.19. The second statement follows from the first because,
taking to be multiplication by r on M , the remark just before the lemma
shows that det = rn .
In the remainder of the proof we will assume that the domain R is a
localization of an algebra finitely generated over a field, so that the integral
closure R0 of R is a finite R-module. This suffices for the applications we
will make. See the remarks after the proof for various ways of removing
this restriction and generalizing the lemma further.
Let : M M be an endomorphism with nonzero determinant D.
Since R has dimension 1, the cokernel of is a module of dimension zero,
and thus finite length. Let R0 be the integral closure of R in its quotient
field K, and let M 0 = M R R0 /(torsion), so that M M 0 K R M (in
fact, M 0 is the image of M R R0 in M R K).
We will use three properties of this construction. First, the endomorphism extends to 1 : M R R0 M R R0 , and preserves the torsion
submodule, so it induces a map 0 : M 0 M 0 extending the map on M .
Second, since M 0 M R K, the quotient M 0 /M is torsion, and thus of
finite length. Of course 0 induces an endomorphism of M 0 /M , which we
will call 0 . Finally, since M R K = M 0 R K = M R0 K, the maps
and 0 have the same determinant D = det = det 0 . We will reduce to
the case where R = R0 , and then use the structure theorem for modules
over a principal ideal domain.
With these points in mind, consider the diagram
0

Ker 0

?
- M

?
- M0

?
- M 0 /M

- 0

?
- M

0
?
- M0

0
?
- M 0 /M

- 0

?
?
?
coker - coker 0 - coker 0

- 0.

A diagram chase (the Snake Lemma) yields an exact sequence


0 Ker 0 coker coker 0 coker 0 0,
from which we deduce that
length coker 0 length coker
= length coker 0 length Ker 0 .

4.3 Proper push-forward

177

But from the exact sequence


0 Ker 0 M 0 /M

- M 0 /M coker 0 0,

we similarly deduce that


0 = length M/M 0 length M/M 0
= length coker 0 length Ker 0 ,
proving that length coker = length coker 0 .
As a special case, we may take M = R, and we see that length R/D =
length R0 /D. Thus it suffices to prove the lemma in the case R = R0 ;
that is, we may assume that the 1-dimensional ring R is integrally closed.
By the Chinese Remainder Theorem, any R-module of finite length is the
direct sum of its localizations at various maximal ideals, so we may further
assume that R is a discrete valuation ring, with parameter , say.
Any finitely-generated torsion-free module over such a ring R is free, so
the endomorphism is defined by a square matrix of elements of R. After
row and column operations, we can reduce this matrix to a diagonal matrix
with powers of on the diagonal,

a
0
1

.. .
..
= ...
.
.
0

am

Since length(R/ a ) = a, we see that


length coker = length

m
M

(R/ ai ) =

m
X

ai

i=1

i=1

while
length R/(det ) = length R/(

m
Y
i=1

ai ) =

m
X

ai

i=1

as required.
As we remarked in the beginning of the proof, the restriction to algebras
essentially of finite type over a field, made in order to guarantee the
finiteness of the integral closure, is unnecessary. One way to avoid it is
to pass to the completion, and use the result that the integral closure of
any complete local Noetherian ring is finite. This is the path taken by
Grothendieck [1967] 21.10.17.2. There is an attractive direct proof along
very different lines in Fulton [1984], Appendix A1A3. But the result given
here and in these references is actually a special case of a far more general
story involving multiplicity: the essential point is that two modules with
the same multiplicity will have the same length after factoring out a regular
sequence modulo which they have finite length. This idea can be used to

178

4. Chow Groups

show, for example that if : Rp Rq is a map over a Cohen-Macaulay


ring R of dimension p q + 1 and the cokernel of has finite length, then
its length is the same as that of R/I, where I is the ideal generated by all
the q q minors of ; our Lemma 4.19 contains the case p = q. This result
is attributed in Bruns and Vetter [1986] to Angeniol and Giusti, though
the paper by Angeniol and Giusti seems never to have appeared. In any
case the result is generalized in the cited paper. The proof by multiplicities
and specialization was given by Hochster and Huneke (also unpublished).
Example 4.20. The second statement of Lemma 4.19 is already interesting
when the rank n is 1; a good example is the case where M = K[t], R =
K[t2 , t5 ] and r = t5 . In this case the length is simply the dimension as a
K-vector space. It is easy to see that the length of
M/rM = K[t]/ < t5 , t6 , t7 , >
is 5. The reader may check that the quotient
R/rR = K[t2 , t4 , t5 , t6 , . . . ]/ < t5 , t7 , t9 , t10 , t11 , > .
is generated as a vector space by 1, t2 , t4 , t6 , and t8 , so the length is 5 in
this case too, though the two vector spaces are not related in a very obvious
way.
Proof of Theorem 4.15. It suffices to show that the diagram
Z(X P 1 )

(f 1)
Z(X 0 P 1 )

X
?
Z(X)

X 0
f

?
- Z(X 0 )

commutes. To this end, suppose W X P 1 is a subvariety dominating


P 1 , and let W 0 = (f 1)(W ). Since f is proper, W 0 is a subvariety of
X 0 P 1 . Of course the projection W P 1 factors through W 0 , so W 0
dominates P 1 if and only if W does. Write W0 and W00 for the preimages of
0 P 1 in the varieties W and W 0 . We may think of W0 as a subvariety of
X = (X P 1 )0 and similarly for W00 and X 0 , and from this point of view,
W0 = f 1 (W00 ) as schemes.
****insert picture****
Suppose first that W 0 has strictly smaller dimension than W , so that
(f 1) hW i = 0. By the Principal Ideal Theorem, every component of
W0 or of W has dimension one less than that of W , and similarly the
0
components of W00 and W
have dimension one less than that of W 0 , so
f (hW0 i hW i) = 0 as well.
Thus we may assume that dim W 0 = dim W , so that W W 0 is generically finite. Write n for its degree, so that (f 1) hW i = nhW 0 i.

4.3 Proper push-forward

179

Let V 0 be an irreducible component of W00 , and suppose that hV 0 i appears in the cycle associated to W00 with multiplicity p. We will show that
hV 0 i appears in the cycle f (hW0 i) with multiplicity np. Summing over the
contributions of all the components of W00 we see that f (hW0 i) = nhW00 i.
The same argument will of course apply to W , proving the Theorem.
Since f induces a surjective map W W 0 , and V 0 W 0 has codimension
1, the map W W 0 must be finite over the generic point of V 0 (otherwise
the preimage of V 0 would have dimension at least that of W , so W would
be reducible). Because the map f is proper, the induced injection
f : R0 := OW 0 ,V 0 f OW OW 0 OW 0 ,V 0 = OW,f 1 V 0 =: R0
is finite.
Both R0 and R have dimension 1. Saying that V 0 occurs with multiplicity
p in W00 means that p = lengthR0 (R0 /xR0 ), where x is a coordinate on P 1
that vanishes at 0. Since f 1 : W W 0 has degree n, the ring R has
rank n as an R0 -module. The Determinant Lemma tells us that
lengthR0 (R/xR) = n lengthR0 (R0 /xR0 ) = np.
The maximal ideals m1 , . . . , mk of R correspond precisely to the subvarieties V1 , . . . , Vs W0 . If the multiplicity of x at mP
i is ei , then the Chinese
Remainder Theorem shows that lengthR0 R/xR = i ei ni , where ni is the
degree of the field extension K(Vi ) : K(V 0 ). From the definition of f we
see that f (hVi i) = ni hV 0 i. Putting this together we compute that the
multiplicity of hV 0 i in f (hW0 i) is
X
X
f (
ei hVi i) =
ei ni hV 0 i = nphV 0 i
i

as required.
P
Proof of Theorem 4.17. Write Div() = ai hVi i, where the Vi are subvarieties. The cycle f (Div()) is the sum of the terms ai f (hVi i). By definition, such a term is zero unless dim f (Vi ) = dim(Vi ), in which case the
restriction f |Vi is generically finite. Since the dimension of X is the same
as that of Y , the subset of X over which f has finite fibers must contain
an open set of any codimension 1 component of Y . Thus we may restrict
to this set and assume, without loss of generality, that f is finite.
The components of f (Div()) and Div(N ()) are all of codimension 1,
so it suffices to prove the equality of the theorem at the generic point of
a codimension 1 subvariety V 0 Y . That is, we may suppose that Y is
the spectrum of a one-dimensional local ring R = OY,V 0 with quotient field
K(Y ). Since X is finite over Spec(R) it is also affine, say X = Spec(S),
where S = OX,f 1 (V 0 ) is a one-dimensional ring with quotient field K(X),
finite over R.

180

4. Chow Groups

We may write = r/s for some elements r, s S. Since the norm is multiplicative, N () = N (r)/N (s). Further, since Div takes multiplication to
addition, it suffices to show that f (Div(x)) = Div(N (x)) for any element
x S.
The components Vi of Div(x) that map onto V 0 correspond to the maximal ideals Pi of S that contain x, and f hVi i = [K(Vi ) : K(V 0 )]hV 0 i. By the
Chinese Remainder Theorem, S/xS = i SPi /xSPi , so lengthR (S/xS) =
P
ordVi (x)0 [K(Vi ) : K(V 0 )]. Thus f Div(x) = lengthR (S/xS)hV 0 i.
We can now apply Lemma 4.19 to the endomorphism of S given by
multiplication by x. The determinant of this map is N (x), and we see
that lengthR (S/xS) = lengthR (R/N (x)), which is the coefficient of V 0 in
hDiv(N (x))i, as required.

4.4 Flat pullback


In this section we will study and apply a different sort of functoriality of
the Chow groups:
Theorem 4.21 (Flat Pullback). If f : Y X be a flat map of equidimensional schemes, then there is a unique group homomorphism f : A(X)
A(Y ) satisfying
f ([A]) = [f 1 (A)]
for any variety A X. This homomorphism preserves codimension.
In the next chapter we will define an intersection product on the Chow
ring of a smooth quasiprojective variety, and it will follow from the methods
there that flat pullback between such varieties preserves products (Corollary 5.5).
We will use Theorem 4.21 in the next subsections to analyze affine and
projective bundles. It will also be useful as a step in the proof of existence
of pullbacks along arbitrary proper morphisms of smooth quasiprojective
varieties, a topic of Chapter 5.
Proof of Theorem 4.21. We first prove that the flat pullback preserves codimension. Suppose that W X is a subvariety, and that V is an irreducible
component of f 1 (W ).
Suppose that y V is a closed point not belonging to any other component of f 1 (W ), and x = f (y), which is again a closed point by the
Nullstellensatz. Because f |V : V W is flat in a neighborhood of y, the