Lemma 76.53.1. Let S be a scheme. Let f : X \to Y be a flat, proper morphism of finite presentation of algebraic spaces over S. Let \mathcal{E} be a finite locally free \mathcal{O}_ X-module. For a morphism g : Y' \to Y consider the base change diagram
\xymatrix{ X' \ar[d]_{f'} \ar[r]_{g'} & X \ar[d]^ f \\ Y' \ar[r]^ g & Y }
Assume \mathcal{O}_{Y'} \to f'_*\mathcal{O}_{X'} is an isomorphism for all g : Y' \to Y. Then there exists an immersion j : Z \to Y of finite presentation such that a morphism g : Y' \to Y factors through Z if and only if there exists a finite locally free \mathcal{O}_{Y'}-module \mathcal{N} with (f')^*\mathcal{N} \cong (g')^*\mathcal{L}.
Proof.
Let y : \mathop{\mathrm{Spec}}(k) \to Y be a field valued point. Then the fibre X_ y of f at y is connected by our assumption that H^0(X_ y, \mathcal{O}_{X_ y}) = k. Thus the rank of \mathcal{E} is constant on the fibres. Since f is open (Morphisms of Spaces, Lemma 67.30.6) and closed we conclude that there is a decomposition Y = \coprod Y_ r of Y into open and closed subspaces such that \mathcal{E} has constant rank r on the inverse image of Y_ r. Thus we may assume \mathcal{E} has constant rank r. We will denote \mathcal{E}^\vee = \mathop{\mathcal{H}\! \mathit{om}}\nolimits (\mathcal{E}, \mathcal{O}_ X) the dual rank r module.
By cohomology and base change (more precisely by Derived Categories of Spaces, Lemma 75.25.4) we see that E = Rf_*\mathcal{E} is a perfect object of the derived category of Y and that its formation commutes with arbitrary change of base. Similarly for E' = Rf_*\mathcal{E}^\vee . Since there is never any cohomology in degrees < 0, we see that E and E' have (locally) tor-amplitude in [0, b] for some b. Observe that for any g : Y' \to Y we have f'_*((g')^*\mathcal{E}) = H^0(Lg^*E) and f'_*((g')^*\mathcal{E}^\vee ) = H^0(Lg^*E'). Let j : Z \to Y and j' : Z' \to Y be the locally closed immersions constructed in Derived Categories of Spaces, Lemma 75.26.6 for E and E' with a = 0 and r = r; these are characterized by the property that H^0(Lj^*E) and H^0((j')^*E') are locally free modules of rank r compatible with pullback.
Let g : Y' \to Y be a morphism. If there exists an \mathcal{N} as in the lemma, then, using the projection formula Cohomology on Sites, Lemma 21.50.1, we see that the modules
f'_*((g')^*\mathcal{E}) \cong f'_*((f')^*\mathcal{N}) \cong \mathcal{N} \otimes _{\mathcal{O}_{Y'}} f'_*\mathcal{O}_{X'} \cong \mathcal{N}\quad \text{and similarly }\quad f'_*((g')^*\mathcal{E}^\vee ) \cong \mathcal{N}^\vee
are locally free of rank r and remain locally free of rank r after any further base change Y'' \to Y'. Hence in this case g : Y' \to Y factors through j and through j'. Thus we may replace Y by Z \times _ Y Z' and assume that f_*\mathcal{E} and f_*\mathcal{E}^\vee are locally free \mathcal{O}_ Y-modules of rank r whose formation commutes with arbitrary change of base.
In this situation if g : Y' \to Y is a morphism and there exists an \mathcal{N} as in the lemma, then the map (cup product in degree 0)
f'_*((g')^*\mathcal{E}) \otimes _{\mathcal{O}_{Y'}} f'_*((g')^*\mathcal{E}^\vee ) \longrightarrow \mathcal{O}_{Y'}
is a perfect pairing. Conversely, if this cup product map is a perfect pairing, then we see that locally on Y' we have a basis of sections \sigma _1, \ldots , \sigma _ r in f'_*((g')^*\mathcal{L}) and \tau _1, \ldots , \tau _ r in f'_*((g')^*\mathcal{E}^\vee ) whose products satisfy \sigma _ i \tau _ j = \delta _{ij}. Thinking of \sigma _ i as a section of (g')^*\mathcal{L} on X' and \tau _ j as a section of (g')^*\mathcal{L}^\vee on X', we conclude that
\sigma _1, \ldots , \sigma _ r : \mathcal{O}_{X'}^{\oplus r} \longrightarrow (g')^*\mathcal{E}
is an isomorphism with inverse given by
\tau _1, \ldots , \tau _ r : (g')^*\mathcal{E} \longrightarrow \mathcal{O}_{X'}^{\oplus r}
In other words, we see that (f')^*f'_*(g')^*\mathcal{E} \cong (g')^*\mathcal{E}. But the condition that the cupproduct is nondegenerate picks out a retrocompact open subscheme (namely, the locus where a suitable determinant is nonzero) and the proof is complete.
\square
Comments (0)