The Stacks project

62.16 Relative cycles in the non-Noetherian case

We urge the reader to skip this section.

Let $f : X \to S$ be a morphism of schemes of finite presentation. Let $r \geq 0$. Denote $Hilb(X/S, r)$ the set of closed subschemes $Z \subset X$ such that $Z \to S$ is flat, of finite presentation, and of relative dimension $\leq r$. We consider the group homomorphism

62.16.0.1
\begin{equation} \label{relative-cycles-equation-cycle-classes-general} \begin{matrix} \text{free abelian group} \\ \text{on }Hilb(X/S, r) \end{matrix} \longrightarrow \begin{matrix} \text{families of }r\text{-cycles} \\ \text{on fibres of }X/S \end{matrix} \end{equation}

sending $\sum n_ i[Z_ i]$ to $\sum n_ i[Z_ i/X/S]_ r$.

Lemma 62.16.1. Let $S$ be a quasi-compact and quasi-separated scheme. Let $f : X \to S$ be a morphism of finite presentation. Let $r \geq 0$ and let $\alpha $ be a family of $r$-cycles on fibres of $X/S$. The following are equivalent

  1. there exists a cartesian diagram

    \[ \xymatrix{ X \ar[r] \ar[d] & X_0 \ar[d] \\ S \ar[r] & S_0 } \]

    where $X_0 \to S_0$ is a finite type morphism of Noetherian schemes and $\alpha _0 \in z(X_0/S_0, r)$ such that $\alpha $ is the base change of $\alpha _0$ by $S \to S_0$

  2. there exists a completely decomposed proper morphism $g : S' \to S$ of finite presentation such that $g^*\alpha $ is in the image of (62.16.0.1).

Proof. Let a diagram and $\alpha _0 \in z(X_0/S_0, r)$ as in (1) be given. By Lemma 62.6.9 there exists a proper surjective morphism $g_0 : S'_0 \to S_0$ such that $g_0^*\alpha _0$ is in the image of (62.16.0.1). Namely, since $S'_0$ is Noetherian, every closed subscheme of $S'_0 \times _{S_0} X_0$ is of finite presentation over $S'_0$. Setting $S' = S \times _{S_0} S'_0$ and using base change by $S' \to S'_0$ we see that (2) holds.

Conversely, assume that (2) holds. Choose a surjective proper morphism $g : S' \to S$ of finite presentation such that $g^*\alpha $ is in the image of (62.16.0.1). Set $X' = S' \times _ S X$. Write $g^*\alpha = \sum n_ a [Z_ a/X'/S']_ r$ for some $Z_ a \subset X'$ closed subscheme flat, of finite presentation, and of relative dimension $\leq r$ over $S'$.

Write $S = \mathop{\mathrm{lim}}\nolimits S_ i$ as a directed limit with affine transition morphisms with $S_ i$ of finite type over $\mathbf{Z}$, see Limits, Proposition 32.5.4. We can find an $i$ large enough such that there exist

  1. a completely decomposed proper morphism $g_ i : S'_ i \to S_ i$ whose base change to $S$ is $g : S' \to S$,

  2. setting $X'_ i = S'_ i \times _{S_ i} X_ i$ closed subschemes $Z_{ai} \subset X'_ i$ flat and of relative dimension $\leq r$ over $S'_ i$ whose base change to $S'$ is $Z_ a$.

To do this one uses Limits, Lemmas 32.10.1, 32.8.5, 32.8.7, 32.8.15, 32.13.1, and 32.18.1 and and More on Morphisms, Lemma 37.78.5. Consider $\alpha '_ i = \sum n_ a [Z_{ai}/X'_ i/S_ i]_ r \in z(X'_ i/S'_ i, r)$. The base change of $\alpha '_ i$ to a family of $r$-cycles on fibres of $X'/S'$ agrees with the base change $g^*\alpha $ by construction.

Set $S''_ i = S'_ i \times _{S_ i} S'_ i$ and $X''_ i = S''_ i \times _{S_ i} X_ i$ and set $S'' = S' \times _ S S'$ and $X'' = S'' \times _ S X$. We denote $\text{pr}_1, \text{pr}_2 : S'' \to S'$ and $\text{pr}_1, \text{pr}_2 : S''_ i \to S'_ i$ the projections. The relative $r$-cycles $\text{pr}_1^*\alpha '_ i$ and $\text{pr}_1^*\alpha '_ i$ on $X''_ i/S''_ i$ base change to the same family of $r$-cycles on fibres of $X''/S''$ because $\text{pr}_1^*g^*\alpha = \text{pr}_1^*g^*\alpha $. Hence the morphism $S'' \to S''_ i$ maps into $E = \{ s \in S''_ i : (\text{pr}_1^*\alpha '_ i)_ s = (\text{pr}_1^*\alpha '_ i)_ s\} $. By Lemma 62.6.12 this is a closed subset. Since $S'' = \mathop{\mathrm{lim}}\nolimits _{i' \geq i} S''_{i'}$ we see from Limits, Lemma 32.4.10 that for some $i' \geq i$ the morphism $S''_{i'} \to S''_ i$ maps into $E$. Therefore, after replacing $i$ by $i'$, we may assume that $\text{pr}_1^*\alpha '_ i = \text{pr}_1^*\alpha '_ i$. By Lemma 62.5.9 we obtain a unique family $\alpha _ i$ of $r$-cycles on fibres of $X_ i/S_ i$ with $g_ i^*\alpha _ i = \alpha '_ i$ (this uses that $S'_ i \to S_ i$ is completely decomposed). By Lemma 62.6.3 we see that $\alpha _ i \in z(X_ i/S_ i, r)$. The uniqueness in Lemma 62.5.9 implies that the base change of $\alpha _ i$ is $\alpha $ and we see (1) holds. $\square$

Discussion. If $f : X \to S$, $r$, and $\alpha $ are as in Lemma 62.16.1, then it makes sense to say that $\alpha $ is a relative $r$-cycle on $X/S$ if the equivalent conditions (1) and (2) of Lemma 62.16.1 hold. This definition has many good properties; for example it doesn't conflict with the earlier definition in case $S$ is Noetherian and most of the results of Section 62.6 generalize to this setting.

We may still generalize further as follows. Assume $S$ is arbitrary and $f : X \to S$ is locally of finite presentation. Let $r \geq 0$ and let $\alpha $ be a family of $r$-cycles $\alpha $ on fibres of $X/S$. Then $\alpha $ is an relative $r$-cycle on $X/S$ if for $U \subset X$ and $V \subset S$ affine open with $f(U) \subset V$ the restriction $\alpha |_ U$ is a relative $r$-cycle on $U/V$ as defined in the previous paragraph. Again many of the earlier results generalize to this setting.

If we ever need these generalizations we will carefully state and prove them here.


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 0H6U. Beware of the difference between the letter 'O' and the digit '0'.