Processing math: 100%

The Stacks project

Lemma 99.16.1. Let S be a scheme. Let f : X \to Y be a morphism of algebraic spaces over S. Assume f is proper, flat, and of finite presentation. Let K, E \in D(\mathcal{O}_ X). Assume K is pseudo-coherent and E is Y-perfect (More on Morphisms of Spaces, Definition 76.52.1). For a field k and a morphism y : \mathop{\mathrm{Spec}}(k) \to Y denote K_ y, E_ y the pullback to the fibre X_ y.

  1. There is an open W \subset Y characterized by the property

    y \in |W| \Leftrightarrow \mathop{\mathrm{Ext}}\nolimits ^ i_{\mathcal{O}_{X_ y}}(K_ y, E_ y) = 0 \text{ for }i < 0.
  2. For any morphism V \to Y factoring through W we have

    \mathop{\mathrm{Ext}}\nolimits ^ i_{\mathcal{O}_{X_ V}}(K_ V, E_ V) = 0 \quad \text{for}\quad i < 0

    where X_ V is the base change of X and K_ V and E_ V are the derived pullbacks of K and E to X_ V.

  3. The functor V \mapsto \mathop{\mathrm{Hom}}\nolimits _{\mathcal{O}_{X_ V}}(K_ V, E_ V) is a sheaf on (\textit{Spaces}/W)_{fppf} representable by an algebraic space affine and of finite presentation over W.

Proof. For any morphism V \to Y the complex K_ V is pseudo-coherent (Cohomology on Sites, Lemma 21.45.3) and E_ V is V-perfect (More on Morphisms of Spaces, Lemma 76.52.6). Another observation is that given y : \mathop{\mathrm{Spec}}(k) \to Y and a field extension k'/k with y' : \mathop{\mathrm{Spec}}(k') \to Y the induced morphism, we have

\mathop{\mathrm{Ext}}\nolimits ^ i_{\mathcal{O}_{X_{y'}}}(K_{y'}, E_{y'}) = \mathop{\mathrm{Ext}}\nolimits ^ i_{\mathcal{O}_{X_ y}}(K_ y, E_ y) \otimes _ k k'

by Derived Categories of Schemes, Lemma 36.22.6. Thus the vanishing in (1) is really a property of the induced point y \in |Y|. We will use these two observations without further mention in the proof.

Assume first Y is an affine scheme. Then we may apply More on Morphisms of Spaces, Lemma 76.52.11 and find a pseudo-coherent L \in D(\mathcal{O}_ Y) which “universally computes” Rf_*R\mathop{\mathcal{H}\! \mathit{om}}\nolimits (K, E) in the sense described in that lemma. Unwinding the definitions, we obtain for a point y \in Y the equality

\mathop{\mathrm{Ext}}\nolimits ^ i_{\kappa (y)}(L \otimes _{\mathcal{O}_ Y}^\mathbf {L} \kappa (y), \kappa (y)) = \mathop{\mathrm{Ext}}\nolimits ^ i_{\mathcal{O}_{X_ y}}(K_ y, E_ y)

We conclude that

H^ i(L \otimes _{\mathcal{O}_ Y}^\mathbf {L} \kappa (y)) = 0 \text{ for } i > 0 \Leftrightarrow \mathop{\mathrm{Ext}}\nolimits ^ i_{\mathcal{O}_{X_ y}}(K_ y, E_ y) = 0 \text{ for }i < 0.

By Derived Categories of Schemes, Lemma 36.31.1 the set W of y \in Y where this happens defines an open of Y. This open W then satisfies the requirement in (1) for all morphisms from spectra of fields, by the “universality” of L.

Let's go back to Y a general algebraic space. Choose an étale covering \{ V_ i \to Y\} by affine schemes V_ i. Then we see that the subset W \subset |Y| pulls back to the corresponding subset W_ i \subset |V_ i| for X_{V_ i}, K_{V_ i}, E_{V_ i}. By the previous paragraph we find that W_ i is open, hence W is open. This proves (1) in general. Moreover, parts (2) and (3) are entirely formulated in terms of the category \textit{Spaces}/W and the restrictions X_ W, K_ W, E_ W. This reduces us to the case W = Y.

Assume W = Y. We claim that for any algebraic space V over Y we have Rf_{V, *}R\mathop{\mathcal{H}\! \mathit{om}}\nolimits (K_ V, E_ V) has vanishing cohomology sheaves in degrees < 0. This will prove (2) because

\mathop{\mathrm{Ext}}\nolimits ^ i_{\mathcal{O}_{X_ V}}(K_ V, E_ V) = H^ i(X_ V, R\mathop{\mathcal{H}\! \mathit{om}}\nolimits (K_ V, E_ V)) = H^ i(V, Rf_{V, *}R\mathop{\mathcal{H}\! \mathit{om}}\nolimits (K_ V, E_ V))

by Cohomology on Sites, Lemmas 21.35.1 and 21.20.5 and the vanishing of the cohomology sheaves implies the cohomology group H^ i is zero for i < 0 by Derived Categories, Lemma 13.16.1.

To prove the claim, we may work étale locally on V. In particular, we may assume Y is affine and W = Y. Let L \in D(\mathcal{O}_ Y) be as in the second paragraph of the proof. For an algebraic space V over Y denote L_ V the derived pullback of L to V. (An important feature we will use is that L “works” for all algebraic spaces V over Y and not just affine V.) As W = Y we have H^ i(L) = 0 for i > 0 (use More on Algebra, Lemma 15.75.6 to go from fibres to stalks). Hence H^ i(L_ V) = 0 for i > 0. The property defining L is that

Rf_{V, *}R\mathop{\mathcal{H}\! \mathit{om}}\nolimits (K_ V, E_ V) = R\mathop{\mathcal{H}\! \mathit{om}}\nolimits (L_ V, \mathcal{O}_ V)

Since L_ V sits in degrees \leq 0, we conclude that R\mathop{\mathcal{H}\! \mathit{om}}\nolimits (L_ V, \mathcal{O}_ V) sits in degrees \geq 0 thereby proving the claim. This finishes the proof of (2).

Assume W = Y but make no assumptions on the algebraic space Y. Since we have (2), we see from Simplicial Spaces, Lemma 85.35.1 that the functor F given by F(V) = \mathop{\mathrm{Hom}}\nolimits _{\mathcal{O}_{X_ V}}(K_ V, E_ V) is a sheaf1 on (\textit{Spaces}/Y)_{fppf}. Thus to prove that F is an algebraic space and that F \to Y is affine and of finite presentation, we may work étale locally on Y; see Bootstrap, Lemma 80.11.2 and Morphisms of Spaces, Lemmas 67.20.3 and 67.28.4. We conclude that it suffices to prove F is an affine algebraic space of finite presentation over Y when Y is an affine scheme. In this case we go back to our pseudo-coherent complex L \in D(\mathcal{O}_ Y). Since H^ i(L) = 0 for i > 0, we can represent L by a complex of the form

\ldots \to \mathcal{O}_ Y^{\oplus m_1} \to \mathcal{O}_ Y^{\oplus m_0} \to 0 \to \ldots

with the last term in degree 0, see More on Algebra, Lemma 15.64.5. Combining the two displayed formulas earlier in the proof we find that

F(V) = \mathop{\mathrm{Ker}}( \mathop{\mathrm{Hom}}\nolimits _ V(\mathcal{O}_ V^{\oplus m_0}, \mathcal{O}_ V) \to \mathop{\mathrm{Hom}}\nolimits _ V(\mathcal{O}_ V^{\oplus m_1}, \mathcal{O}_ V) )

In other words, there is a fibre product diagram

\xymatrix{ F \ar[d] \ar[r] & Y \ar[d]^0 \\ \mathbf{A}_ Y^{m_0} \ar[r] & \mathbf{A}_ Y^{m_1} }

which proves what we want. \square

[1] To check the sheaf property for a covering \{ V_ i \to V\} _{i \in I} first consider the Čech fppf hypercovering a : V_\bullet \to V with V_ n = \coprod _{i_0 \ldots i_ n} V_{i_0} \times _ V \ldots \times _ V V_{i_ n} and then set U_\bullet = V_\bullet \times _{a, V} X_ V. Then U_\bullet \to X_ V is an fppf hypercovering to which we may apply Simplicial Spaces, Lemma 85.35.1.

Comments (0)

There are also:

  • 1 comment(s) on Section 99.16: Moduli of complexes on a proper morphism

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.