Loading [MathJax]/extensions/TeX/mathchoice.js

The Stacks project

42.52 Properties of localized Chern classes

The main results in this section are additivity and multiplicativity for localized Chern classes.

Lemma 42.52.1. In Situation 42.50.1 assume E|_{X \setminus Z} is zero. Then

\begin{align*} P_1(Z \to X, E) & = c_1(Z \to X, E), \\ P_2(Z \to X, E) & = c_1(Z \to X, E)^2 - 2c_2(Z \to X, E), \\ P_3(Z \to X, E) & = c_1(Z \to X, E)^3 - 3c_1(Z \to X, E)c_2(Z \to X, E) + 3c_3(Z \to X, E), \end{align*}

and so on where the products are taken in the algebra A^{(1)}(Z \to X) of Remark 42.34.7.

Proof. The statement makes sense because the zero sheaf has rank < 1 and hence the classes c_ p(Z \to X, E) are defined for all p \geq 1. The result itself follows immediately from the more general Lemma 42.49.6 as the localized Chern classes where defined using the procedure of Lemma 42.49.1 in Section 42.50. \square

Lemma 42.52.2. In Situation 42.50.1 let Y \to X be locally of finite type and c \in A^*(Y \to X). Then

P_ p(Z \to X, E) \circ c = c \circ P_ p(Z \to X, E),

respectively

c_ p(Z \to X, E) \circ c = c \circ c_ p(Z \to X, E)

in A^*(Y \times _ X Z \to X).

Proof. This follows from Lemma 42.49.5. More precisely, let

b : W \to \mathbf{P}^1_ X \quad \text{and}\quad Q \quad \text{and}\quad T' \subset T \subset W_\infty

be as in the proof of Lemma 42.50.2. By definition c_ p(Z \to X, E) = c'_ p(Q) as bivariant operations where the right hand side is the bivariant class constructed in Lemma 42.49.1 using W, b, Q, T'. By Lemma 42.49.5 we have P'_ p(Q) \circ c = c \circ P'_ p(Q), resp. c'_ p(Q) \circ c = c \circ c'_ p(Q) in A^*(Y \times _ X Z \to X) and we conclude. \square

Remark 42.52.3. In Situation 42.50.1 it is convenient to define

c^{(p)}(Z \to X, E) = 1 + c_1(E) + \ldots + c_{p - 1}(E) + c_ p(Z \to X, E) + c_{p + 1}(Z \to X, E) + \ldots

as an element of the algebra A^{(p)}(Z \to X) considered in Remark 42.34.7.

Lemma 42.52.4. Let (S, \delta ) be as in Situation 42.7.1. Let X be locally of finite type over S. Let Z \to X be a closed immersion. Let

E_1 \to E_2 \to E_3 \to E_1[1]

be a distinguished triangle of perfect objects in D(\mathcal{O}_ X). Assume

  1. the restrictions E_1|_{X \setminus Z} and E_3|_{X \setminus Z} are isomorphic to finite locally free \mathcal{O}_{X \setminus Z}-modules of rank < p_1 and < p_3 placed in degree 0, and

  2. at least one of the following is true: (a) X is quasi-compact, (b) X has quasi-compact irreducible components, (c) E_3 \to E_1[1] can be represented by a map of locally bounded complexes of finite locally free \mathcal{O}_ X-modules, or (d) there exists an envelope f : Y \to X such that Lf^*E_3 \to Lf^*E_1[1] can be represented by a map of locally bounded complexes of finite locally free \mathcal{O}_ Y-modules.

With notation as in Remark 42.52.3 we have

c^{(p_1 + p_3)}(Z \to X, E_2) = c^{(p_1)}(Z \to X, E_1)c^{(p_3)}(Z \to X, E_3)

in A^{(p_1 + p_3)}(Z \to X).

Proof. Observe that the assumptions imply that E_2|_{X \setminus Z} is zero, resp. isomorphic to a finite locally free \mathcal{O}_{X \setminus Z}-module of rank < p_1 + p_3. Thus the statement makes sense.

Let f : Y \to X be an envelope. Expanding the left and right hand sides of the formula in the statement of the lemma we see that we have to prove some equalities of classes in A^*(X) and in A^*(Z \to X). By the uniqueness in Lemma 42.35.6 it suffices to prove the corresponding relations in A^*(Y) and A^*(Z \to Y). Since moreover the construction of the classes involved is compatible with base change (Lemma 42.50.4) we may replace X by Y and the distinguished triangle by its pullback.

In the proof of Lemma 42.46.7 we have seen that conditions (2)(a), (2)(b), and (2)(c) imply condition (2)(d). Combined with the discussion in the previous paragraph we reduce to the case discussed in the next paragraph.

Let \varphi ^\bullet : \mathcal{E}_3^\bullet [-1] \to \mathcal{E}_1^\bullet be a map of locally bounded complexes of finite locally free \mathcal{O}_ X-modules representing the map E_3[-1] \to E_1 in the derived category. Consider the scheme X' = \mathbf{A}^1 \times X with projection g : X' \to X. Let Z' = g^{-1}(Z) = \mathbf{A}^1 \times Z. Denote t the coordinate on \mathbf{A}^1. Consider the cone \mathcal{C}^\bullet of the map of complexes

t g^*\varphi ^\bullet : g^*\mathcal{E}_3^\bullet [-1] \longrightarrow g^*\mathcal{E}_1^\bullet

over X'. We obtain a distinguished triangle

g^*\mathcal{E}_1^\bullet \to \mathcal{C}^\bullet \to g^*\mathcal{E}_3^\bullet \to g^*\mathcal{E}_1^\bullet [1]

where the first three terms form a termwise split short exact sequence of complexes. Clearly \mathcal{C}^\bullet is a bounded complex of finite locally free \mathcal{O}_{X'}-modules whose restriction to X' \setminus Z' is isomorphic to a finite locally free \mathcal{O}_{X' \setminus Z'}-module of rank < p_1 + p_3 placed in degree 0. Thus we have the localized Chern classes

c_ p(Z' \to X', \mathcal{C}^\bullet ) \in A^ p(Z' \to X')

for p \geq p_1 + p_3. For any \alpha \in \mathop{\mathrm{CH}}\nolimits _ k(X) consider

c_ p(Z' \to X', \mathcal{C}^\bullet ) \cap g^*\alpha \in \mathop{\mathrm{CH}}\nolimits _{k + 1 - p}(\mathbf{A}^1 \times X)

If we restrict to t = 0, then the map t g^*\varphi ^\bullet restricts to zero and \mathcal{C}^\bullet |_{t = 0} is the direct sum of \mathcal{E}_1^\bullet and \mathcal{E}_3^\bullet . By compatibility of localized Chern classes with base change (Lemma 42.50.4) we conclude that

i_0^* \circ c^{(p_1 + p_3)}(Z' \to X', \mathcal{C}^\bullet ) \circ g^* = c^{(p_1 + p_2)}(Z \to X, E_1 \oplus E_3)

in A^{(p_1 + p_3)}(Z \to X). On the other hand, if we restrict to t = 1, then the map t g^*\varphi ^\bullet restricts to \varphi and \mathcal{C}^\bullet |_{t = 1} is a bounded complex of finite locally free modules representing E_2. We conclude that

i_1^* \circ c^{(p_1 + p_3)}(Z' \to X', \mathcal{C}^\bullet ) \circ g^* = c^{(p_1 + p_2)}(Z \to X, E_2)

in A^{(p_1 + p_3)}(Z \to X). Since i_0^* = i_1^* by definition of rational equivalence (more precisely this follows from the formulae in Lemma 42.32.4) we conclude that

c^{(p_1 + p_2)}(Z \to X, E_2) = c^{(p_1 + p_2)}(Z \to X, E_1 \oplus E_3)

This reduces us to the case discussed in the next paragraph.

Assume E_2 = E_1 \oplus E_3 and the triples (X, Z, E_ i) are as in Situation 42.50.1. For i = 1, 3 let

b_ i : W_ i \to \mathbf{P}^1_ X \quad \text{and}\quad Q_ i \quad \text{and}\quad T'_ i \subset T_ i \subset W_{i, \infty }

be as in the proof of Lemma 42.50.2. By definition

c_ p(Z \to X, E_ i) = c'_ p(Q_ i)

where the right hand side is the bivariant class constructed in Lemma 42.49.1 using W_ i, b_ i, Q_ i, T'_ i. Set W = W_1 \times _{b_1, \mathbf{P}^1_ X, b_2} W_2 and consider the cartesian diagram

\xymatrix{ W \ar[d]_{g_1} \ar[rd]^ b \ar[r]_{g_3} & W_3 \ar[d]^{b_3} \\ W_1 \ar[r]^{b_1} & \mathbf{P}^1_ X }

Of course b^{-1}(\mathbf{A}^1) maps isomorphically to \mathbf{A}^1_ X. Observe that T' = g_1^{-1}(T'_1) \cap g_2^{-1}(T'_2) still contains all the points of W_\infty lying over X \setminus Z. By Lemma 42.49.3 we may use W, b, g_ i^*\mathcal{Q}_ i, and T' to construct c_ p(Z \to X, E_ i) for i = 1, 3. Also, by the stronger independence given in Lemma 42.51.2 we may use W, b, g_1^*Q_1 \oplus g_3^*Q_3, and T' to compute the classes c_ p(Z \to X, E_2). Thus the desired equality follows from Lemma 42.49.7. \square

Lemma 42.52.5. Let (S, \delta ) be as in Situation 42.7.1. Let X be locally of finite type over S. Let Z \to X be a closed immersion. Let

E_1 \to E_2 \to E_3 \to E_1[1]

be a distinguished triangle of perfect objects in D(\mathcal{O}_ X). Assume

  1. the restrictions E_1|_{X \setminus Z} and E_3|_{X \setminus Z} are zero, and

  2. at least one of the following is true: (a) X is quasi-compact, (b) X has quasi-compact irreducible components, (c) E_3 \to E_1[1] can be represented by a map of locally bounded complexes of finite locally free \mathcal{O}_ X-modules, or (d) there exists an envelope f : Y \to X such that Lf^*E_3 \to Lf^*E_1[1] can be represented by a map of locally bounded complexes of finite locally free \mathcal{O}_ Y-modules.

Then we have

P_ p(Z \to X, E_2) = P_ p(Z \to X, E_1) + P_ p(Z \to X, E_3)

for all p \in \mathbf{Z} and consequently ch(Z \to X, E_2) = ch(Z \to X, E_1) + ch(Z \to X, E_3).

Proof. The proof is exactly the same as the proof of Lemma 42.52.4 except it uses Lemma 42.49.8 at the very end. For p > 0 we can deduce this lemma from Lemma 42.52.4 with p_1 = p_3 = 1 and the relationship between P_ p(Z \to X, E) and c_ p(Z \to X, E) given in Lemma 42.52.1. The case p = 0 can be shown directly (it is only interesting if X has a connected component entirely contained in Z). \square

Lemma 42.52.6. In Situation 42.7.1 let X be locally of finite type over S. Let Z_ i \subset X, i = 1, 2 be closed subschemes. Let F_ i, i = 1, 2 be perfect objects of D(\mathcal{O}_ X). Assume for i = 1, 2 that F_ i|_{X \setminus Z_ i} is zero1 and that F_ i on X satisfies assumption (3) of Situation 42.50.1. Denote r_ i = P_0(Z_ i \to X, F_ i) \in A^0(Z_ i \to X). Then we have

c_1(Z_1 \cap Z_2 \to X, F_1 \otimes _{\mathcal{O}_ X}^\mathbf {L} F_2) = r_1 c_1(Z_2 \to X, F_2) + r_2 c_1(Z_1 \to X, F_1)

in A^1(Z_1 \cap Z_2 \to X) and

\begin{align*} c_2(Z_1 \cap Z_2 \to X, F_1 \otimes _{\mathcal{O}_ X}^\mathbf {L} F_2) & = r_1 c_2(Z_2 \to X, F_2) + r_2 c_2(Z_1 \to X, F_1) + \\ & {r_1 \choose 2} c_1(Z_2 \to X, F_2)^2 + \\ & (r_1r_2 - 1) c_1(Z_2 \to X, F_2)c_1(Z_1 \to X, F_1) + \\ & {r_2 \choose 2} c_1(Z_1 \to X, F_1)^2 \end{align*}

in A^2(Z_1 \cap Z_2 \to X) and so on for higher Chern classes. Similarly, we have

ch(Z_1 \cap Z_2 \to X, F_1 \otimes _{\mathcal{O}_ X}^\mathbf {L} F_2) = ch(Z_1 \to X, F_1) ch(Z_2 \to X, F_2)

in \prod _{p \geq 0} A^ p(Z_1 \cap Z_2 \to X) \otimes \mathbf{Q}. More precisely, we have

P_ p(Z_1 \cap Z_2 \to X, F_1 \otimes _{\mathcal{O}_ X}^\mathbf {L} F_2) = \sum \nolimits _{p_1 + p_2 = p} {p \choose p_1} P_{p_1}(Z_1 \to X, F_1) P_{p_2}(Z_2 \to X, F_2)

in A^ p(Z_1 \cap Z_2 \to X).

Proof. Choose proper morphisms b_ i : W_ i \to \mathbf{P}^1_ X and Q_ i \in D(\mathcal{O}_{W_ i}) as well as closed subschemes T_ i \subset W_{i, \infty } as in the construction of the localized Chern classes for F_ i or more generally as in Lemma 42.51.2. Choose a commutative diagram

\xymatrix{ W \ar[d]^{g_1} \ar[rd]^ b \ar[r]_{g_2} & W_2 \ar[d]^{b_2} \\ W_1 \ar[r]^{b_1} & \mathbf{P}^1_ X }

where all morphisms are proper and isomorphisms over \mathbf{A}^1_ X. For example, we can take W to be the closure of the graph of the isomorphism between b_1^{-1}(\mathbf{A}^1_ X) and b_2^{-1}(\mathbf{A}^1_ X). By Lemma 42.51.2 we may work with W, b = b_ i \circ g_ i, Lg_ i^*Q_ i, and g_ i^{-1}(T_ i) to construct the localized Chern classes c_ p(Z_ i \to X, F_ i). Thus we reduce to the situation described in the next paragraph.

Assume we have

  1. a proper morphism b : W \to \mathbf{P}^1_ X which is an isomorphism over \mathbf{A}^1_ X,

  2. E_ i \subset W_\infty is the inverse image of Z_ i,

  3. perfect objects Q_ i \in D(\mathcal{O}_ W) whose Chern classes are defined, such that

    1. the restriction of Q_ i to b^{-1}(\mathbf{A}^1_ X) is the pullback of F_ i, and

    2. there exists a closed subscheme T_ i \subset W_\infty containing all points of W_\infty lying over X \setminus Z_ i such that Q_ i|_{T_ i} is zero.

By Lemma 42.51.2 we have

c_ p(Z_ i \to X, F_ i) = c'_ p(Q_ i) = (E_ i \to Z_ i)_* \circ c'_ p(Q_ i|_{E_ i}) \circ C

and

P_ p(Z_ i \to X, F_ i) = P'_ p(Q_ i) = (E_ i \to Z_ i)_* \circ P'_ p(Q_ i|_{E_ i}) \circ C

for i = 1, 2. Next, we observe that Q = Q_1 \otimes _{\mathcal{O}_ W}^\mathbf {L} Q_2 satisfies (3)(a) and (3)(b) for F_1 \otimes _{\mathcal{O}_ X}^\mathbf {L} F_2 and T_1 \cup T_2. Hence we see that

c_ p(Z_1 \cap Z_2 \to X, F_1 \otimes _{\mathcal{O}_ X}^\mathbf {L} F_2) = (E_1 \cap E_2 \to Z_1 \cap Z_2)_* \circ c'_ p(Q|_{E_1 \cap E_2}) \circ C

and

P_ p(Z_1 \cap Z_2 \to X, F_1 \otimes _{\mathcal{O}_ X}^\mathbf {L} F_2) = (E_1 \cap E_2 \to Z_1 \cap Z_2)_* \circ P'_ p(Q|_{E_1 \cap E_2}) \circ C

by the same lemma. By Lemma 42.47.11 the classes c'_ p(Q|_{E_1 \cap E_2}) and P'_ p(Q|_{E_1 \cap E_2}) can be expanded in the correct manner in terms of the classes c'_ p(Q_ i|_{E_ i}) and P'_ p(Q_ i|_{E_ i}). Then finally Lemma 42.51.1 tells us that polynomials in c'_ p(Q_ i|_{E_ i}) and P'_ p(Q_ i|_{E_ i}) agree with the corresponding polynomials in c'_ p(Q_ i) and P'_ p(Q_ i) as desired. \square

[1] Presumably there is a variant of this lemma where we only assume F_ i|_{X \setminus Z_ i} is isomorphic to a finite locally free \mathcal{O}_{X \setminus Z_ i}-module of rank < p_ i.

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.