The Stacks project

113.22 Commutativity of intersecting divisors

The results of this section were originally used to provide an alternative proof of the lemmas of Chow Homology, Section 42.27 and a weak version of Chow Homology, Lemma 42.29.5.

Lemma 113.22.1. Let $(S, \delta )$ be as in Chow Homology, Situation 42.7.1. Let $X$ be locally of finite type over $S$. Let $\{ i_ j : D_ j \to X \} _{j \in J}$ be a locally finite collection of effective Cartier divisors on $X$. Let $n_ j > 0$, $j\in J$. Set $D = \sum _{j \in J} n_ j D_ j$, and denote $i : D \to X$ the inclusion morphism. Let $\alpha \in Z_{k + 1}(X)$. Then

\[ p : \coprod \nolimits _{j \in J} D_ j \longrightarrow D \]

is proper and

\[ i^*\alpha = p_*\left(\sum n_ j i_ j^*\alpha \right) \]

in $\mathop{\mathrm{CH}}\nolimits _ k(D)$.

Proof. The proof of this lemma is made a bit longer than expected by a subtlety concerning infinite sums of rational equivalences. In the quasi-compact case the family $D_ j$ is finite and the result is altogether easy and a straightforward consequence of Chow Homology, Lemma 42.23.2 and Divisors, Lemma 31.27.5 and the definitions.

The morphism $p$ is proper since the family $\{ D_ j\} _{j \in J}$ is locally finite. Write $\alpha = \sum _{a \in A} m_ a [W_ a]$ with $W_ a \subset X$ an integral closed subscheme of $\delta $-dimension $k + 1$. Denote $i_ a : W_ a \to X$ the closed immersion. We assume that $m_ a \not= 0$ for all $a \in A$ such that $\{ W_ a\} _{a \in A}$ is locally finite on $X$.

Observe that by Chow Homology, Definition 42.28.1 the class $i^*\alpha $ is the class of a cycle $\sum m_ a\beta _ a$ for certain $\beta _ a \in Z_ k(W_ a \cap D)$. Namely, if $W_ a \not\subset D$ then $\beta _ a = [D \cap W_ a]_ k$ and if $W_ a \subset D$, then $\beta _ a$ is a cycle representing $c_1(\mathcal{O}_ X(D)) \cap [W_ a]$.

For each $a \in A$ write $J = J_{a, 1} \amalg J_{a, 2} \amalg J_{a, 3}$ where

  1. $j \in J_{a, 1}$ if and only if $W_ a \cap D_ j = \emptyset $,

  2. $j \in J_{a, 2}$ if and only if $W_ a \not= W_ a \cap D_1 \not= \emptyset $, and

  3. $j \in J_{a, 3}$ if and only if $W_ a \subset D_ j$.

Since the family $\{ D_ j\} $ is locally finite we see that $J_{a, 3}$ is a finite set. For every $a \in A$ and $j \in J$ we choose a cycle $\beta _{a, j} \in Z_ k(W_ a \cap D_ j)$ as follows

  1. if $j \in J_{a, 1}$ we set $\beta _{a, j} = 0$,

  2. if $j \in J_{a, 2}$ we set $\beta _{a, j} = [D_ j \cap W_ a]_ k$, and

  3. if $j \in J_{a, 3}$ we choose $\beta _{a, j} \in Z_ k(W_ a)$ representing $c_1(i_ a^*\mathcal{O}_ X(D_ j)) \cap [W_ j]$.

We claim that

\[ \beta _ a \sim _{rat} \sum \nolimits _{j \in J} n_ j \beta _{a, j} \]

in $\mathop{\mathrm{CH}}\nolimits _ k(W_ a \cap D)$.

Case I: $W_ a \not\subset D$. In this case $J_{a, 3} = \emptyset $. Thus it suffices to show that $[D \cap W_ a]_ k = \sum n_ j [D_ j \cap W_ a]_ k$ as cycles. This is Lemma 113.21.10.

Case II: $W_ a \subset D$. In this case $\beta _ a$ is a cycle representing $c_1(i_ a^*\mathcal{O}_ X(D)) \cap [W_ a]$. Write $D = D_{a, 1} + D_{a, 2} + D_{a, 3}$ with $D_{a, s} = \sum _{j \in J_{a, s}} n_ jD_ j$. By Divisors, Lemma 31.27.5 we have

\begin{eqnarray*} c_1(i_ a^*\mathcal{O}_ X(D)) \cap [W_ a] & = & c_1(i_ a^*\mathcal{O}_ X(D_{a, 1})) \cap [W_ a] + c_1(i_ a^*\mathcal{O}_ X(D_{a, 2})) \cap [W_ a] \\ & & + c_1(i_ a^*\mathcal{O}_ X(D_{a, 3})) \cap [W_ a]. \end{eqnarray*}

It is clear that the first term of the sum is zero. Since $J_{a, 3}$ is finite we see that the last term agrees with $\sum \nolimits _{j \in J_{a, 3}} n_ jc_1(i_ a^*\mathcal{L}_ j) \cap [W_ a]$, see Divisors, Lemma 31.27.5. This is represented by $\sum _{j \in J_{a, 3}} n_ j \beta _{a, j}$. Finally, by Case I we see that the middle term is represented by the cycle $\sum \nolimits _{j \in J_{a, 2}} n_ j[D_ j \cap W_ a]_ k = \sum _{j \in J_{a, 2}} n_ j\beta _{a, j}$. Whence the claim in this case.

At this point we are ready to finish the proof of the lemma. Namely, we have $i^*D \sim _{rat} \sum m_ a\beta _ a$ by our choice of $\beta _ a$. For each $a$ we have $\beta _ a \sim _{rat} \sum _ j \beta _{a, j}$ with the rational equivalence taking place on $D \cap W_ a$. Since the collection of closed subschemes $D \cap W_ a$ is locally finite on $D$, we see that also $\sum m_ a \beta _ a \sim _{rat} \sum _{a, j} m_ a\beta _{a, j}$ on $D$! (See Chow Homology, Remark 42.19.5.) Ok, and now it is clear that $\sum _ a m_ a\beta _{a, j}$ (viewed as a cycle on $D_ j$) represents $i_ j^*\alpha $ and hence $\sum _{a, j} m_ a\beta _{a, j}$ represents $p_* \sum _ j i_ j^*\alpha $ and we win. $\square$

Lemma 113.22.2. Let $(S, \delta )$ be as in Chow Homology, Situation 42.7.1. Let $X$ be locally of finite type over $S$. Assume $X$ integral and $\dim _\delta (X) = n$. Let $D$, $D'$ be effective Cartier divisors on $X$. Assume $\dim _\delta (D \cap D') = n - 2$. Let $i : D \to X$, resp. $i' : D' \to X$ be the corresponding closed immersions. Then

  1. there exists a cycle $\alpha \in Z_{n - 2}(D \cap D')$ whose pushforward to $D$ represents $i^*[D']_{n - 1} \in \mathop{\mathrm{CH}}\nolimits _{n - 2}(D)$ and whose pushforward to $D'$ represents $(i')^*[D]_{n - 1} \in \mathop{\mathrm{CH}}\nolimits _{n - 2}(D')$, and

  2. we have

    \[ D \cdot [D']_{n - 1} = D' \cdot [D]_{n - 1} \]

    in $\mathop{\mathrm{CH}}\nolimits _{n - 2}(X)$.

Proof. Part (2) is a trivial consequence of part (1). Let us write $[D]_{n - 1} = \sum n_ a[Z_ a]$ and $[D']_{n - 1} = \sum m_ b[Z_ b]$ with $Z_ a$ the irreducible components of $D$ and $[Z_ b]$ the irreducible components of $D'$. According to Chow Homology, Definition 42.28.1, we have $i^*D' = \sum m_ b i^*[Z_ b]$ and $(i')^*D = \sum n_ a(i')^*[Z_ a]$. By assumption, none of the irreducible components $Z_ b$ is contained in $D$, and hence $i^*[Z_ b] = [Z_ b\cap D]_{n - 2}$ by definition. Similarly $(i')^*[Z_ a] = [Z_ a \cap D']_{n - 2}$. Hence we are trying to prove the equality of cycles

\[ \sum n_ a[Z_ a \cap D']_{n - 2} = \sum m_ b[Z_ b \cap D]_{n - 2} \]

which are indeed supported on $D \cap D'$. Let $W \subset X$ be an integral closed subscheme with $\dim _\delta (W) = n - 2$. Let $\xi \in W$ be its generic point. Set $R = \mathcal{O}_{X, \xi }$. It is a Noetherian local domain. Note that $\dim (R) = 2$. Let $f \in R$, resp. $f' \in R$ be an element defining the ideal of $D$, resp. $D'$. By assumption $\dim (R/(f, f')) = 0$. Let $\mathfrak q'_1, \ldots , \mathfrak q'_ t \subset R$ be the minimal primes over $(f')$, let $\mathfrak q_1, \ldots , \mathfrak q_ s \subset R$ be the minimal primes over $(f)$. The equality above comes down to the equality

\[ \sum _{i = 1, \ldots , s} \text{length}_{R_{\mathfrak q_ i}}(R_{\mathfrak q_ i}/(f)) \text{ord}_{R/\mathfrak q_ i}(f') = \sum _{j = 1, \ldots , t} \text{length}_{R_{\mathfrak q'_ j}}(R_{\mathfrak q'_ j}/(f')) \text{ord}_{R/\mathfrak q'_ j}(f). \]

By Chow Homology, Lemma 42.3.1 applied with $M = R/(f)$ the left hand side of this equation is equal to

\[ \text{length}_ R(R/(f, f')) - \text{length}_ R(\mathop{\mathrm{Ker}}(f' : R/(f) \to R/(f))) \]

OK, and now we note that $\mathop{\mathrm{Ker}}(f' : R/(f) \to R/(f))$ is canonically isomorphic to $((f) \cap (f'))/(ff')$ via the map $x \bmod (f) \mapsto f'x \bmod (ff')$. Hence the left hand side is

\[ \text{length}_ R(R/(f, f')) - \text{length}_ R((f) \cap (f')/(ff')) \]

Since this is symmetric in $f$ and $f'$ we win. $\square$

Lemma 113.22.3. Let $(S, \delta )$ be as in Chow Homology, Situation 42.7.1. Let $X$ be locally of finite type over $S$. Assume $X$ integral and $\dim _\delta (X) = n$. Let $\{ D_ j\} _{j \in J}$ be a locally finite collection of effective Cartier divisors on $X$. Let $n_ j, m_ j \geq 0$ be collections of nonnegative integers. Set $D = \sum n_ j D_ j$ and $D' = \sum m_ j D_ j$. Assume that $\dim _\delta (D_ j \cap D_{j'}) = n - 2$ for every $j \not= j'$. Then $D \cdot [D']_{n - 1} = D' \cdot [D]_{n - 1}$ in $\mathop{\mathrm{CH}}\nolimits _{n - 2}(X)$.

Proof. This lemma is a trivial consequence of Lemmas 113.21.10 and 113.22.2 in case the sums are finite, e.g., if $X$ is quasi-compact. Hence we suggest the reader skip the proof.

Here is the proof in the general case. Let $i_ j : D_ j \to X$ be the closed immersions Let $p : \coprod D_ j \to X$ denote coproduct of the morphisms $i_ j$. Let $\{ Z_ a\} _{a \in A}$ be the collection of irreducible components of $\bigcup D_ j$. For each $j$ we write

\[ [D_ j]_{n - 1} = \sum d_{j, a}[Z_ a]. \]

By Lemma 113.21.10 we have

\[ [D]_{n - 1} = \sum n_ j d_{j, a} [Z_ a], \quad [D']_{n - 1} = \sum m_ j d_{j, a} [Z_ a]. \]

By Lemma 113.22.1 we have

\[ D \cdot [D']_{n - 1} = p_*\left(\sum n_ j i_ j^*[D']_{n - 1} \right), \quad D' \cdot [D]_{n - 1} = p_*\left(\sum m_{j'} i_{j'}^*[D]_{n - 1} \right). \]

As in the definition of the Gysin homomorphisms (see Chow Homology, Definition 42.28.1) we choose cycles $\beta _{a, j}$ on $D_ j \cap Z_ a$ representing $i_ j^*[Z_ a]$. (Note that in fact $\beta _{a, j} = [D_ j \cap Z_ a]_{n - 2}$ if $Z_ a$ is not contained in $D_ j$, i.e., there is no choice in that case.) Now since $p$ is a closed immersion when restricted to each of the $D_ j$ we can (and we will) view $\beta _{a, j}$ as a cycle on $X$. Plugging in the formulas for $[D]_{n - 1}$ and $[D']_{n - 1}$ obtained above we see that

\[ D \cdot [D']_{n - 1} = \sum \nolimits _{j, j', a} n_ j m_{j'} d_{j', a} \beta _{a, j}, \quad D' \cdot [D]_{n - 1} = \sum \nolimits _{j, j', a} m_{j'} n_ j d_{j, a} \beta _{a, j'}. \]

Moreover, with the same conventions we also have

\[ D_ j \cdot [D_{j'}]_{n - 1} = \sum d_{j', a} \beta _{a, j}. \]

In these terms Lemma 113.22.2 (see also its proof) says that for $j \not= j'$ the cycles $\sum d_{j', a} \beta _{a, j}$ and $\sum d_{j, a} \beta _{a, j'}$ are equal as cycles! Hence we see that

\begin{eqnarray*} D \cdot [D']_{n - 1} & = & \sum \nolimits _{j, j', a} n_ j m_{j'} d_{j', a} \beta _{a, j} \\ & = & \sum \nolimits _{j \not= j'} n_ j m_{j'} \left(\sum \nolimits _ a d_{j', a} \beta _{a, j}\right) + \sum \nolimits _{j, a} n_ j m_ j d_{j, a} \beta _{a, j} \\ & = & \sum \nolimits _{j \not= j'} n_ j m_{j'} \left(\sum \nolimits _ a d_{j, a} \beta _{a, j'}\right) + \sum \nolimits _{j, a} n_ j m_ j d_{j, a} \beta _{a, j} \\ & = & \sum \nolimits _{j, j', a} m_{j'} n_ j d_{j, a} \beta _{a, j'} \\ & = & D' \cdot [D]_{n - 1} \end{eqnarray*}

and we win. $\square$

Lemma 113.22.4. Let $(S, \delta )$ be as in Chow Homology, Situation 42.7.1. Let $X$ be locally of finite type over $S$. Assume $X$ integral and $\dim _\delta (X) = n$. Let $D$, $D'$ be effective Cartier divisors on $X$. Then

\[ D \cdot [D']_{n - 1} = D' \cdot [D]_{n - 1} \]

in $\mathop{\mathrm{CH}}\nolimits _{n - 2}(X)$.

First proof of Lemma 113.22.4. First, let us prove this in case $X$ is quasi-compact. In this case, apply Lemma 113.21.11 to $X$ and the two element set $\{ D, D'\} $ of effective Cartier divisors. Thus we get a proper morphism $b : X' \to X$, a finite collection of effective Cartier divisors $D'_ j \subset X'$ intersecting pairwise in codimension $\geq 2$, with $b^{-1}(D) = \sum n_ j D'_ j$, and $b^{-1}(D') = \sum m_ j D'_ j$. Note that $b_*[b^{-1}(D)]_{n - 1} = [D]_{n - 1}$ in $Z_{n - 1}(X)$ and similarly for $D'$, see Lemma 113.21.5. Hence, by Chow Homology, Lemma 42.25.4 we have

\[ D \cdot [D']_{n - 1} = b_*\left(b^{-1}(D) \cdot [b^{-1}(D')]_{n - 1}\right) \]

in $\mathop{\mathrm{CH}}\nolimits _{n - 2}(X)$ and similarly for the other term. Hence the lemma follows from the equality $b^{-1}(D) \cdot [b^{-1}(D')]_{n - 1} = b^{-1}(D') \cdot [b^{-1}(D)]_{n - 1}$ in $\mathop{\mathrm{CH}}\nolimits _{n - 2}(X')$ of Lemma 113.22.3.

Note that in the proof above, each referenced lemma works also in the general case (when $X$ is not assumed quasi-compact). The only minor change in the general case is that the morphism $b : U' \to U$ we get from applying Lemma 113.21.11 has as its target an open $U \subset X$ whose complement has codimension $\geq 3$. Hence by Chow Homology, Lemma 42.19.2 we see that $\mathop{\mathrm{CH}}\nolimits _{n - 2}(U) = \mathop{\mathrm{CH}}\nolimits _{n - 2}(X)$ and after replacing $X$ by $U$ the rest of the proof goes through unchanged. $\square$

Second proof of Lemma 113.22.4. Let $\mathcal{I} = \mathcal{O}_ X(-D)$ and $\mathcal{I}' = \mathcal{O}_ X(-D')$ be the invertible ideal sheaves of $D$ and $D'$. We denote $\mathcal{I}_{D'} = \mathcal{I} \otimes _{\mathcal{O}_ X} \mathcal{O}_{D'}$ and $\mathcal{I}'_ D = \mathcal{I}' \otimes _{\mathcal{O}_ X} \mathcal{O}_ D$. We can restrict the inclusion map $\mathcal{I} \to \mathcal{O}_ X$ to $D'$ to get a map

\[ \varphi : \mathcal{I}_{D'} \longrightarrow \mathcal{O}_{D'} \]

and similarly

\[ \psi : \mathcal{I}'_ D \longrightarrow \mathcal{O}_ D \]

It is clear that

\[ \mathop{\mathrm{Coker}}(\varphi ) \cong \mathcal{O}_{D \cap D'} \cong \mathop{\mathrm{Coker}}(\psi ) \]


\[ \mathop{\mathrm{Ker}}(\varphi ) \cong \frac{\mathcal{I} \cap \mathcal{I}'}{\mathcal{I}\mathcal{I}'} \cong \mathop{\mathrm{Ker}}(\psi ). \]

Hence we see that

\[ \gamma = [\mathcal{I}_{D'}] - [\mathcal{O}_{D'}] = [\mathcal{I}'_ D] - [\mathcal{O}_ D] \]

in $K_0(\textit{Coh}_{\leq n - 1}(X))$. On the other hand it is clear that

\[ [\mathcal{I}'_ D]_{n - 1} = [D]_{n - 1}, \quad [\mathcal{I}_{D'}]_{n - 1} = [D']_{n - 1}. \]

and that

\[ \mathcal{O}_ X(D') \otimes \mathcal{I}'_ D = \mathcal{O}_ D, \quad \mathcal{O}_ X(D) \otimes \mathcal{I}_{D'} = \mathcal{O}_{D'}. \]

By Chow Homology, Lemma 42.68.7 (applied two times) this means that the element $\gamma $ is an element of $B_{n - 2}(X)$, and maps to both $c_1(\mathcal{O}_ X(D')) \cap [D]_{n - 1}$ and to $c_1(\mathcal{O}_ X(D)) \cap [D']_{n - 1}$ and we win (since the map $B_{n - 2}(X) \to \mathop{\mathrm{CH}}\nolimits _{n - 2}(X)$ is well defined – which is the key to this proof). $\square$

Comments (0)

Post a comment

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.

In order to prevent bots from posting comments, we would like you to prove that you are human. You can do this by filling in the name of the current tag in the following input field. As a reminder, this is tag 0AYE. Beware of the difference between the letter 'O' and the digit '0'.