The Stacks project

Lemma 86.9.4. Let $S$ be a scheme. Let $X \to Y$ be a proper, flat morphism of algebraic spaces which is of finite presentation. Let $(\omega ^\bullet , \tau )$ be a pair consisting of a $Y$-perfect object of $D(\mathcal{O}_ X)$ and a map $\tau : Rf_*\omega ^\bullet \to \mathcal{O}_ Y$. Assume we have cartesian diagrams

\[ \xymatrix{ X_ i \ar[r]_{g_ i'} \ar[d]_{f_ i} & X \ar[d]^ f \\ Y_ i \ar[r]^{g_ i} & Y } \]

with $Y_ i$ affine such that $\{ g_ i : Y_ i \to Y\} $ is an étale covering and isomorphisms of pairs $(\omega ^\bullet |_{X_ i}, \tau |_{Y_ i}) \to (a_ i(\mathcal{O}_{Y_ i}), \text{Tr}_{f_ i, \mathcal{O}_{Y_ i}})$ as in Definition 86.9.1. Then $(\omega ^\bullet , \tau )$ is a relative dualizing complex for $X$ over $Y$.

Proof. Let $g : Y' \to Y$ and $X', f', g', a'$ be as in Definition 86.9.1. Set $((\omega ')^\bullet , \tau ') = (L(g')^*\omega ^\bullet , Lg^*\tau )$. We can find a finite étale covering $\{ Y'_ j \to Y'\} $ by affines which refines $\{ Y_ i \times _ Y Y' \to Y'\} $ (Topologies, Lemma 34.4.4). Thus for each $j$ there is an $i_ j$ and a morphism $k_ j : Y'_ j \to Y_{i_ j}$ over $Y$. Consider the fibre products

\[ \xymatrix{ X'_ j \ar[r]_{h_ j'} \ar[d]_{f'_ j} & X' \ar[d]^{f'} \\ Y'_ j \ar[r]^{h_ j} & Y' } \]

Denote $k'_ j : X'_ j \to X_{i_ j}$ the induced morphism (base change of $k_ j$ by $f_{i_ j}$). Restricting the given isomorphisms to $Y'_ j$ via the morphism $k'_ j$ we get isomorphisms of pairs $((\omega ')^\bullet |_{X'_ j}, \tau '|_{Y'_ j}) \to (a_ j(\mathcal{O}_{Y'_ j}), \text{Tr}_{f'_ j, \mathcal{O}_{Y'_ j}})$. After replacing $f : X \to Y$ by $f' : X' \to Y'$ we reduce to the problem solved in the next paragraph.

Assume $Y$ is affine. Problem: show $(\omega ^\bullet , \tau )$ is isomorphic to $(\omega _{X/Y}^\bullet , \text{Tr}) = (a(\mathcal{O}_ Y), \text{Tr}_{f, \mathcal{O}_ Y})$. We may assume our covering $\{ Y_ i \to Y\} $ is given by a single surjective étale morphism $\{ g : Y' \to Y\} $ of affines. Namely, we can first replace $\{ g_ i: Y_ i \to Y\} $ by a finite subcovering, and then we can set $g = \coprod g_ i : Y' = \coprod Y_ i \to Y$; some details omitted. Set $X' = Y' \times _ Y X$ with maps $f', g'$ as in Definition 86.9.1. Then all we're given is that we have an isomorphism

\[ (\omega ^\bullet |_{X'}, \tau |_{Y'}) \to (a'(\mathcal{O}_{Y'}), \text{Tr}_{f', \mathcal{O}_{Y'}}) \]

Since $(\omega _{X/Y}^\bullet , \text{Tr})$ is a relative dualizing complex (see discussion following Definition 86.9.1) there is a unique isomorphism

\[ (\omega _{X/Y}^\bullet |_{X'}, \text{Tr}|_{Y'}) \to (a'(\mathcal{O}_{Y'}), \text{Tr}_{f', \mathcal{O}_{Y'}}) \]

Uniqueness by Lemma 86.9.3 for example. Combining the displayed isomorphisms we find an isomorphism

\[ \alpha : (\omega ^\bullet |_{X'}, \tau |_{Y'}) \to (\omega _{X/Y}^\bullet |_{X'}, \text{Tr}|_{Y'}) \]

Set $Y'' = Y' \times _ Y Y'$ and $X'' = Y'' \times _ Y X$ the two pullbacks of $\alpha $ to $X''$ have to be the same by uniqueness again. Since we have vanishing negative self exts for $\omega _{X'/Y'}^\bullet $ over $X'$ (Lemma 86.9.2) and since this remains true after pulling back by any projection $Y' \times _ Y \ldots \times _ Y Y' \to Y'$ (small detail omitted – compare with the proof of Lemma 86.9.3), we find that $\alpha $ descends to an isomorphism $\omega ^\bullet \to \omega _{X/Y}^\bullet $ over $X$ by Simplicial Spaces, Lemma 85.35.1. $\square$


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 0E60. Beware of the difference between the letter 'O' and the digit '0'.