Processing math: 100%

The Stacks project

42.55 Calculating some classes

To get further we need to compute the values of some of the classes we've constructed above.

Lemma 42.55.1. Let (S, \delta ) be as in Situation 42.7.1. Let X be a scheme locally of finite type over S. Let \mathcal{E} be a locally free \mathcal{O}_ X-module of rank r. Then

\prod \nolimits _{n = 0, \ldots , r} c(\wedge ^ n \mathcal{E})^{(-1)^ n} = 1 - (r - 1)! c_ r(\mathcal{E}) + \ldots

Proof. By the splitting principle we can turn this into a calculation in the polynomial ring on the Chern roots x_1, \ldots , x_ r of \mathcal{E}. See Section 42.43. Observe that

c(\wedge ^ n \mathcal{E}) = \prod \nolimits _{1 \leq i_1 < \ldots < i_ n \leq r} (1 + x_{i_1} + \ldots + x_{i_ n})

Thus the logarithm of the left hand side of the equation in the lemma is

- \sum \nolimits _{p \geq 1} \sum \nolimits _{n = 0}^ r \sum \nolimits _{1 \leq i_1 < \ldots < i_ n \leq r} \frac{(-1)^{p + n}}{p}(x_{i_1} + \ldots + x_{i_ n})^ p

Please notice the minus sign in front. However, we have

\sum \nolimits _{p \geq 0} \sum \nolimits _{n = 0}^ r \sum \nolimits _{1 \leq i_1 < \ldots < i_ n \leq r} \frac{(-1)^{p + n}}{p!}(x_{i_1} + \ldots + x_{i_ n})^ p = \prod (1 - e^{-x_ i})

Hence we see that the first nonzero term in our Chern class is in degree r and equal to the predicted value. \square

Lemma 42.55.2. Let (S, \delta ) be as in Situation 42.7.1. Let X be a scheme locally of finite type over S. Let \mathcal{C} be a locally free \mathcal{O}_ X-module of rank r. Consider the morphisms

X = \underline{\text{Proj}}_ X(\mathcal{O}_ X[T]) \xrightarrow {i} E = \underline{\text{Proj}}_ X(\text{Sym}^*(\mathcal{C})[T]) \xrightarrow {\pi } X

Then c_ t(i_*\mathcal{O}_ X) = 0 for t = 1, \ldots , r - 1 and in A^0(C \to E) we have

p^* \circ \pi _* \circ c_ r(i_*\mathcal{O}_ X) = (-1)^{r - 1}(r - 1)! j^*

where j : C \to E and p : C \to X are the inclusion and structure morphism of the vector bundle C = \underline{\mathop{\mathrm{Spec}}}(\text{Sym}^*(\mathcal{C})).

Proof. The canonical map \pi ^*\mathcal{C} \to \mathcal{O}_ E(1) vanishes exactly along i(X). Hence the Koszul complex on the map

\pi ^*\mathcal{C} \otimes \mathcal{O}_ E(-1) \to \mathcal{O}_ E

is a resolution of i_*\mathcal{O}_ X. In particular we see that i_*\mathcal{O}_ X is a perfect object of D(\mathcal{O}_ E) whose Chern classes are defined. The vanishing of c_ t(i_*\mathcal{O}_ X) for t = 1, \ldots , t - 1 follows from Lemma 42.55.1. This lemma also gives

c_ r(i_*\mathcal{O}_ X) = - (r - 1)! c_ r(\pi ^*\mathcal{C} \otimes \mathcal{O}_ E(-1))

On the other hand, by Lemma 42.43.3 we have

c_ r(\pi ^*\mathcal{C} \otimes \mathcal{O}_ E(-1)) = (-1)^ r c_ r(\pi ^*\mathcal{C}^\vee \otimes \mathcal{O}_ E(1))

and \pi ^*\mathcal{C}^\vee \otimes \mathcal{O}_ E(1) has a section s vanishing exactly along i(X).

After replacing X by a scheme locally of finite type over X, it suffices to prove that both sides of the equality have the same effect on an element \alpha \in \mathop{\mathrm{CH}}\nolimits _*(E). Since C \to X is a vector bundle, every cycle class on C is of the form p^*\beta for some \beta \in \mathop{\mathrm{CH}}\nolimits _*(X) (Lemma 42.36.3). Hence by Lemma 42.19.3 we can write \alpha = \pi ^*\beta + \gamma where \gamma is supported on E \setminus C. Using the equalities above it suffices to show that

p^*(\pi _*(c_ r(\pi ^*\mathcal{C}^\vee \otimes \mathcal{O}_ E(1)) \cap [W])) = j^*[W]

when W \subset E is an integral closed subscheme which is either (a) disjoint from C or (b) is of the form W = \pi ^{-1}Y for some integral closed subscheme Y \subset X. Using the section s and Lemma 42.44.1 we find in case (a) c_ r(\pi ^*\mathcal{C}^\vee \otimes \mathcal{O}_ E(1)) \cap [W] = 0 and in case (b) c_ r(\pi ^*\mathcal{C}^\vee \otimes \mathcal{O}_ E(1)) \cap [W] = [i(Y)]. The result follows easily from this; details omitted. \square

Lemma 42.55.3. Let (S, \delta ) be as in Situation 42.7.1. Let i : Z \to X be a regular closed immersion of codimension r between schemes locally of finite type over S. Let \mathcal{N} = \mathcal{C}_{Z/X}^\vee be the normal sheaf. If X is quasi-compact (or has quasi-compact irreducible components), then c_ t(Z \to X, i_*\mathcal{O}_ Z) = 0 for t = 1, \ldots , r - 1 and

c_ r(Z \to X, i_*\mathcal{O}_ Z) = (-1)^{r - 1} (r - 1)! c(Z \to X, \mathcal{N}) \quad \text{in}\quad A^ r(Z \to X)

where c_ t(Z \to X, i_*\mathcal{O}_ Z) is the localized Chern class of Definition 42.50.3.

Proof. For any x \in Z we can choose an affine open neighbourhood \mathop{\mathrm{Spec}}(A) \subset X such that Z \cap \mathop{\mathrm{Spec}}(A) = V(f_1, \ldots , f_ r) where f_1, \ldots , f_ r \in A is a regular sequence. See Divisors, Definition 31.21.1 and Lemma 31.20.8. Then we see that the Koszul complex on f_1, \ldots , f_ r is a resolution of A/(f_1, \ldots , f_ r) for example by More on Algebra, Lemma 15.30.2. Hence A/(f_1, \ldots , f_ r) is perfect as an A-module. It follows that F = i_*\mathcal{O}_ Z is a perfect object of D(\mathcal{O}_ X) whose restriction to X \setminus Z is zero. The assumption that X is quasi-compact (or has quasi-compact irreducible components) means that the localized Chern classes c_ t(Z \to X, i_*\mathcal{O}_ Z) are defined, see Situation 42.50.1 and Definition 42.50.3. All in all we conclude that the statement makes sense.

Denote b : W \to \mathbf{P}^1_ X the blowing up in \infty (Z) as in Section 42.53. By (8) we have a closed immersion

i' : \mathbf{P}^1_ Z \longrightarrow W

We claim that Q = i'_*\mathcal{O}_{\mathbf{P}^1_ Z} is a perfect object of D(\mathcal{O}_ W) and that F and Q satisfy the assumptions of Lemma 42.51.2.

Assume the claim. The output of Lemma 42.51.2 is that we have

c_ p(Z \to X, F) = c'_ p(Q) = (E \to Z)_* \circ c'_ p(Q|_ E) \circ C

for all p \geq 1. Observe that Q|_ E is equal to the pushforward of the structure sheaf of Z via the morphism Z \to E which is the base change of i' by \infty . Thus the vanishing of c_ t(Z \to X, F) for 1 \leq t \leq r - 1 by Lemma 42.55.2 applied to E \to Z. Because \mathcal{C}_{Z/X} = \mathcal{N}^\vee is locally free the bivariant class c(Z \to X, \mathcal{N}) is characterized by the relation

j^* \circ C = p^* \circ c(Z \to X, \mathcal{N})

where j : C_ ZX \to W_\infty and p : C_ ZX \to Z are the given maps. (Recall C \in A^0(W_\infty \to X) is the class of Lemma 42.48.1.) Thus the displayed equation in the statement of the lemma follows from the corresponding equation in Lemma 42.55.2.

Proof of the claim. Let A and f_1, \ldots , f_ r be as above. Consider the affine open \mathop{\mathrm{Spec}}(A[s]) \subset \mathbf{P}^1_ X as in Section 42.53. Recall that s = 0 defines (\mathbf{P}^1_ X)_\infty over this open. Hence over \mathop{\mathrm{Spec}}(A[s]) we are blowing up in the ideal generated by the regular sequence s, f_1, \ldots , f_ r. By More on Algebra, Lemma 15.31.2 the r + 1 affine charts are global complete intersections over A[s]. The chart corresponding to the affine blowup algebra

A[s][f_1/s, \ldots , f_ r/s] = A[s, y_1, \ldots , y_ r]/(sy_ i - f_ i)

contains i'(Z \cap \mathop{\mathrm{Spec}}(A)) as the closed subscheme cut out by y_1, \ldots , y_ r. Since y_1, \ldots , y_ r, sy_1 - f_1, \ldots , sy_ r - f_ r is a regular sequence in the polynomial ring A[s, y_1, \ldots , y_ r] we find that i' is a regular immersion. Some details omitted. As above we conclude that Q = i'_*\mathcal{O}_{\mathbf{P}^1_ Z} is a perfect object of D(\mathcal{O}_ W). All the other assumptions on F and Q in Lemma 42.51.2 (and Lemma 42.49.1) are immediately verified. \square

Lemma 42.55.4. In the situation of Lemma 42.55.3 say \dim _\delta (X) = n. Then we have

  1. c_ t(Z \to X, i_*\mathcal{O}_ Z) \cap [X]_ n = 0 for t = 1, \ldots , r - 1,

  2. c_ r(Z \to X, i_*\mathcal{O}_ Z) \cap [X]_ n = (-1)^{r - 1}(r - 1)![Z]_{n - r},

  3. ch_ t(Z \to X, i_*\mathcal{O}_ Z) \cap [X]_ n = 0 for t = 0, \ldots , r - 1, and

  4. ch_ r(Z \to X, i_*\mathcal{O}_ Z) \cap [X]_ n = [Z]_{n - r}.

Proof. Parts (1) and (2) follow immediately from Lemma 42.55.3 combined with Lemma 42.54.5. Then we deduce parts (3) and (4) using the relationship between ch_ p = (1/p!)P_ p and c_ p given in Lemma 42.52.1. (Namely, (-1)^{r - 1}(r - 1)!ch_ r = c_ r provided c_1 = c_2 = \ldots = c_{r - 1} = 0.) \square


Comments (0)


Your email address will not be published. Required fields are marked.

In your comment you can use Markdown and LaTeX style mathematics (enclose it like $\pi$). A preview option is available if you wish to see how it works out (just click on the eye in the toolbar).

Unfortunately JavaScript is disabled in your browser, so the comment preview function will not work.

All contributions are licensed under the GNU Free Documentation License.