20.45 Glueing complexes
We can glue complexes! More precisely, in certain circumstances we can glue locally given objects of the derived category to a global object. We first prove some easy cases and then we'll prove the very general [Theorem 3.2.4, BBD] in the setting of topological spaces and open coverings.
Lemma 20.45.1. Let (X, \mathcal{O}_ X) be a ringed space. Let X = U \cup V be the union of two open subspaces of X. Suppose given
an object A of D(\mathcal{O}_ U),
an object B of D(\mathcal{O}_ V), and
an isomorphism c : A|_{U \cap V} \to B|_{U \cap V}.
Then there exists an object F of D(\mathcal{O}_ X) and isomorphisms f : F|_ U \to A, g : F|_ V \to B such that c = g|_{U \cap V} \circ f^{-1}|_{U \cap V}. Moreover, given
an object E of D(\mathcal{O}_ X),
a morphism a : A \to E|_ U of D(\mathcal{O}_ U),
a morphism b : B \to E|_ V of D(\mathcal{O}_ V),
such that
a|_{U \cap V} = b|_{U \cap V} \circ c.
Then there exists a morphism F \to E in D(\mathcal{O}_ X) whose restriction to U is a \circ f and whose restriction to V is b \circ g.
Proof.
Denote j_ U, j_ V, j_{U \cap V} the corresponding open immersions. Choose a distinguished triangle
F \to Rj_{U, *}A \oplus Rj_{V, *}B \to Rj_{U \cap V, *}(B|_{U \cap V}) \to F[1]
where the map Rj_{V, *}B \to Rj_{U \cap V, *}(B|_{U \cap V}) is the obvious one and where Rj_{U, *}A \to Rj_{U \cap V, *}(B|_{U \cap V}) is the composition of Rj_{U, *}A \to Rj_{U \cap V, *}(A|_{U \cap V}) with Rj_{U \cap V, *}c. Restricting to U we obtain
F|_ U \to A \oplus (Rj_{V, *}B)|_ U \to (Rj_{U \cap V, *}(B|_{U \cap V}))|_ U \to F|_ U[1]
Denote j : U \cap V \to U. Compatibility of restriction to opens and cohomology shows that both (Rj_{V, *}B)|_ U and (Rj_{U \cap V, *}(B|_{U \cap V}))|_ U are canonically isomorphic to Rj_*(B|_{U \cap V}). Hence the second arrow of the last displayed diagram has a section, and we conclude that the morphism F|_ U \to A is an isomorphism. Similarly, the morphism F|_ V \to B is an isomorphism. The existence of the morphism F \to E follows from the Mayer-Vietoris sequence for \mathop{\mathrm{Hom}}\nolimits , see Lemma 20.33.3.
\square
Lemma 20.45.2. Let f : (X, \mathcal{O}_ X) \to (Y, \mathcal{O}_ Y) be a morphism of ringed spaces. Let \mathcal{B} be a basis for the topology on Y.
Assume K is in D(\mathcal{O}_ X) such that for V \in \mathcal{B} we have H^ i(f^{-1}(V), K) = 0 for i < 0. Then Rf_*K has vanishing cohomology sheaves in negative degrees, H^ i(f^{-1}(V), K) = 0 for i < 0 for all opens V \subset Y, and the rule V \mapsto H^0(f^{-1}V, K) is a sheaf on Y.
Assume K, L are in D(\mathcal{O}_ X) such that for V \in \mathcal{B} we have \mathop{\mathrm{Ext}}\nolimits ^ i(K|_{f^{-1}V}, L|_{f^{-1}V}) = 0 for i < 0. Then \mathop{\mathrm{Ext}}\nolimits ^ i(K|_{f^{-1}V}, L|_{f^{-1}V}) = 0 for i < 0 for all opens V \subset Y and the rule V \mapsto \mathop{\mathrm{Hom}}\nolimits (K|_{f^{-1}V}, L|_{f^{-1}V}) is a sheaf on Y.
Proof.
Lemma 20.32.6 tells us H^ i(Rf_*K) is the sheaf associated to the presheaf V \mapsto H^ i(f^{-1}(V), K) = H^ i(V, Rf_*K). The assumptions in (1) imply that Rf_*K has vanishing cohomology sheaves in degrees < 0. We conclude that for any open V \subset Y the cohomology group H^ i(V, Rf_*K) is zero for i < 0 and is equal to H^0(V, H^0(Rf_*K)) for i = 0. This proves (1).
To prove (2) apply (1) to the complex R\mathop{\mathcal{H}\! \mathit{om}}\nolimits (K, L) using Lemma 20.42.1 to do the translation.
\square
Situation 20.45.3. Let (X, \mathcal{O}_ X) be a ringed space. We are given
a collection of opens \mathcal{B} of X,
for U \in \mathcal{B} an object K_ U in D(\mathcal{O}_ U),
for V \subset U with V, U \in \mathcal{B} an isomorphism \rho ^ U_ V : K_ U|_ V \to K_ V in D(\mathcal{O}_ V),
such that whenever we have W \subset V \subset U with U, V, W in \mathcal{B}, then \rho ^ U_ W = \rho ^ V_ W \circ \rho ^ U_ V|_ W.
We won't be able to prove anything about this without making more assumptions. An interesting case is where \mathcal{B} is a basis for the topology on X. Another is the case where we have a morphism f : X \to Y of topological spaces and the elements of \mathcal{B} are the inverse images of the elements of a basis for the topology of Y.
In Situation 20.45.3 a solution will be a pair (K, \rho _ U) where K is an object of D(\mathcal{O}_ X) and \rho _ U : K|_ U \to K_ U, U \in \mathcal{B} are isomorphisms such that we have \rho ^ U_ V \circ \rho _ U|_ V = \rho _ V for all V \subset U, U, V \in \mathcal{B}. In certain cases solutions are unique.
Lemma 20.45.4. In Situation 20.45.3 assume
X = \bigcup _{U \in \mathcal{B}} U and for U, V \in \mathcal{B} we have U \cap V = \bigcup _{W \in \mathcal{B}, W \subset U \cap V} W,
for any U \in \mathcal{B} we have \mathop{\mathrm{Ext}}\nolimits ^ i(K_ U, K_ U) = 0 for i < 0.
If a solution (K, \rho _ U) exists, then it is unique up to unique isomorphism and moreover \mathop{\mathrm{Ext}}\nolimits ^ i(K, K) = 0 for i < 0.
Proof.
Let (K, \rho _ U) and (K', \rho '_ U) be a pair of solutions. Let f : X \to Y be the continuous map constructed in Topology, Lemma 5.5.6. Set \mathcal{O}_ Y = f_*\mathcal{O}_ X. Then K, K' and \mathcal{B} are as in Lemma 20.45.2 part (2). Hence we obtain the vanishing of negative exts for K and we see that the rule
V \longmapsto \mathop{\mathrm{Hom}}\nolimits (K|_{f^{-1}V}, K'|_{f^{-1}V})
is a sheaf on Y. As both (K, \rho _ U) and (K', \rho '_ U) are solutions the maps
(\rho '_ U)^{-1} \circ \rho _ U : K|_ U \longrightarrow K'|_ U
over U = f^{-1}(f(U)) agree on overlaps. Hence we get a unique global section of the sheaf above which defines the desired isomorphism K \to K' compatible with all structure available.
\square
Lemma 20.45.6. In Situation 20.45.3 assume
X = U_1 \cup \ldots \cup U_ n with U_ i \in \mathcal{B},
for U, V \in \mathcal{B} we have U \cap V = \bigcup _{W \in \mathcal{B}, W \subset U \cap V} W,
for any U \in \mathcal{B} we have \mathop{\mathrm{Ext}}\nolimits ^ i(K_ U, K_ U) = 0 for i < 0.
Then a solution exists and is unique up to unique isomorphism.
Proof.
Uniqueness was seen in Lemma 20.45.4. We may prove the lemma by induction on n. The case n = 1 is immediate.
The case n = 2. Consider the isomorphism \rho _{U_1, U_2} : K_{U_1}|_{U_1 \cap U_2} \to K_{U_2}|_{U_1 \cap U_2} constructed in Remark 20.45.5. By Lemma 20.45.1 we obtain an object K in D(\mathcal{O}_ X) and isomorphisms \rho _{U_1} : K|_{U_1} \to K_{U_1} and \rho _{U_2} : K|_{U_2} \to K_{U_2} compatible with \rho _{U_1, U_2}. Take U \in \mathcal{B}. We will construct an isomorphism \rho _ U : K|_ U \to K_ U and we will leave it to the reader to verify that (K, \rho _ U) is a solution. Consider the set \mathcal{B}' of elements of \mathcal{B} contained in either U \cap U_1 or contained in U \cap U_2. Then (K_ U, \rho ^ U_{U'}) is a solution for the system (\{ K_{U'}\} _{U' \in \mathcal{B}'}, \{ \rho _{V'}^{U'}\} _{V' \subset U'\text{ with }U', V' \in \mathcal{B}'}) on the ringed space U. We claim that (K|_ U, \tau _{U'}) is another solution where \tau _{U'} for U' \in \mathcal{B}' is chosen as follows: if U' \subset U_1 then we take the composition
K|_{U'} \xrightarrow {\rho _{U_1}|_{U'}} K_{U_1}|_{U'} \xrightarrow {\rho ^{U_1}_{U'}} K_{U'}
and if U' \subset U_2 then we take the composition
K|_{U'} \xrightarrow {\rho _{U_2}|_{U'}} K_{U_2}|_{U'} \xrightarrow {\rho ^{U_2}_{U'}} K_{U'}.
To verify this is a solution use the property of the map \rho _{U_1, U_2} described in Remark 20.45.5 and the compatibility of \rho _{U_1} and \rho _{U_2} with \rho _{U_1, U_2}. Having said this we apply Lemma 20.45.4 to see that we obtain a unique isomorphism K|_{U'} \to K_{U'} compatible with the maps \tau _{U'} and \rho ^ U_{U'} for U' \in \mathcal{B}'.
The case n > 2. Consider the open subspace X' = U_1 \cup \ldots \cup U_{n - 1} and let \mathcal{B}' be the set of elements of \mathcal{B} contained in X'. Then we find a system (\{ K_ U\} _{U \in \mathcal{B}'}, \{ \rho _ V^ U\} _{U, V \in \mathcal{B}'}) on the ringed space X' to which we may apply our induction hypothesis. We find a solution (K_{X'}, \rho ^{X'}_ U). Then we can consider the collection \mathcal{B}^* = \mathcal{B} \cup \{ X'\} of opens of X and we see that we obtain a system (\{ K_ U\} _{U \in \mathcal{B}^*}, \{ \rho _ V^ U\} _{V \subset U\text{ with }U, V \in \mathcal{B}^*}). Note that this new system also satisfies condition (3) by Lemma 20.45.4 applied to the solution K_{X'}. For this system we have X = X' \cup U_ n. This reduces us to the case n = 2 we worked out above.
\square
Lemma 20.45.7. Let X be a ringed space. Let E be a well ordered set and let
X = \bigcup \nolimits _{\alpha \in E} W_\alpha
be an open covering with W_\alpha \subset W_{\alpha + 1} and W_\alpha = \bigcup _{\beta < \alpha } W_\beta if \alpha is not a successor. Let K_\alpha be an object of D(\mathcal{O}_{W_\alpha }) with \mathop{\mathrm{Ext}}\nolimits ^ i(K_\alpha , K_\alpha ) = 0 for i < 0. Assume given isomorphisms \rho _\beta ^\alpha : K_\alpha |_{W_\beta } \to K_\beta in D(\mathcal{O}_{W_\beta }) for all \beta < \alpha with \rho _\gamma ^\alpha = \rho _\gamma ^\beta \circ \rho ^\alpha _\beta |_{W_\gamma } for \gamma < \beta < \alpha . Then there exists an object K in D(\mathcal{O}_ X) and isomorphisms K|_{W_\alpha } \to K_\alpha for \alpha \in E compatible with the isomorphisms \rho _\beta ^\alpha .
Proof.
In this proof \alpha , \beta , \gamma , \ldots represent elements of E. Choose a K-injective complex I_\alpha ^\bullet on W_\alpha representing K_\alpha . For \beta < \alpha denote j_{\beta , \alpha } : W_\beta \to W_\alpha the inclusion morphism. Using transfinite recursion we will construct for all \beta < \alpha a map of complexes
\tau _{\beta , \alpha } : (j_{\beta , \alpha })_!I_\beta ^\bullet \longrightarrow I_\alpha ^\bullet
representing the adjoint to the inverse of the isomorphism \rho ^\alpha _\beta : K_\alpha |_{W_\beta } \to K_\beta . Moreover, we will do this in such that for \gamma < \beta < \alpha we have
\tau _{\gamma , \alpha } = \tau _{\beta , \alpha } \circ (j_{\beta , \alpha })_!\tau _{\gamma , \beta }
as maps of complexes. Namely, suppose already given \tau _{\gamma , \beta } composing correctly for all \gamma < \beta < \alpha . If \alpha = \alpha ' + 1 is a successor, then we choose any map of complexes
(j_{\alpha ', \alpha })_!I_{\alpha '}^\bullet \to I_\alpha ^\bullet
which is adjoint to the inverse of the isomorphism \rho ^\alpha _{\alpha '} : K_\alpha |_{W_{\alpha '}} \to K_{\alpha '} (possible because I_\alpha ^\bullet is K-injective) and for any \beta < \alpha ' we set
\tau _{\beta , \alpha } = \tau _{\alpha ', \alpha } \circ (j_{\alpha ', \alpha })_!\tau _{\beta , \alpha '}
If \alpha is not a successor, then we can consider the complex on W_\alpha given by
C^\bullet = \mathop{\mathrm{colim}}\nolimits _{\beta < \alpha } (j_{\beta , \alpha })_!I_\beta ^\bullet
(termwise colimit) where the transition maps of the sequence are given by the maps \tau _{\beta ', \beta } for \beta ' < \beta < \alpha . We claim that C^\bullet represents K_\alpha . Namely, for \beta < \alpha the restriction of the coprojection (j_{\beta , \alpha })_!I_\beta ^\bullet \to C^\bullet gives a map
\sigma _\beta : I_\beta ^\bullet \longrightarrow C^\bullet |_{W_\beta }
which is a quasi-isomorphism: if x \in W_\beta then looking at stalks we get
(C^\bullet )_ x = \mathop{\mathrm{colim}}\nolimits _{\beta ' < \alpha } \left((j_{\beta ', \alpha })_!I_{\beta '}^\bullet \right)_ x = \mathop{\mathrm{colim}}\nolimits _{\beta \leq \beta ' < \alpha } (I_{\beta '}^\bullet )_ x \longleftarrow (I_\beta ^\bullet )_ x
which is a quasi-isomorphism. Here we used that taking stalks commutes with colimits, that filtered colimits are exact, and that the maps (I_\beta ^\bullet )_ x \to (I_{\beta '}^\bullet )_ x are quasi-isomorphisms for \beta \leq \beta ' < \alpha . Hence (C^\bullet , \sigma _\beta ^{-1}) is a solution to the system (\{ K_\beta \} _{\beta < \alpha }, \{ \rho ^\beta _{\beta '}\} _{\beta ' < \beta < \alpha }). Since (K_\alpha , \rho ^\alpha _\beta ) is another solution we obtain a unique isomorphism \sigma : K_\alpha \to C^\bullet in D(\mathcal{O}_{W_\alpha }) compatible with all our maps, see Lemma 20.45.6 (this is where we use the vanishing of negative ext groups). Choose a morphism \tau : C^\bullet \to I_\alpha ^\bullet of complexes representing \sigma . Then we set
\tau _{\beta , \alpha } = \tau |_{W_\beta } \circ \sigma _\beta
to get the desired maps. Finally, we take K to be the object of the derived category represented by the complex
K^\bullet = \mathop{\mathrm{colim}}\nolimits _{\alpha \in E} (W_\alpha \to X)_!I_\alpha ^\bullet
where the transition maps are given by our carefully constructed maps \tau _{\beta , \alpha } for \beta < \alpha . Arguing exactly as above we see that for all \alpha the restriction of the coprojection determines an isomorphism
K|_{W_\alpha } \longrightarrow K_\alpha
compatible with the given maps \rho ^\alpha _\beta .
\square
Using transfinite induction we can prove the result in the general case.
Theorem 20.45.8 (BBD gluing lemma).reference In Situation 20.45.3 assume
X = \bigcup _{U \in \mathcal{B}} U,
for U, V \in \mathcal{B} we have U \cap V = \bigcup _{W \in \mathcal{B}, W \subset U \cap V} W,
for any U \in \mathcal{B} we have \mathop{\mathrm{Ext}}\nolimits ^ i(K_ U, K_ U) = 0 for i < 0.
Then there exists an object K of D(\mathcal{O}_ X) and isomorphisms \rho _ U : K|_ U \to K_ U in D(\mathcal{O}_ U) for U \in \mathcal{B} such that \rho ^ U_ V \circ \rho _ U|_ V = \rho _ V for all V \subset U with U, V \in \mathcal{B}. The pair (K, \rho _ U) is unique up to unique isomorphism.
Proof.
A pair (K, \rho _ U) is called a solution in the text above. The uniqueness follows from Lemma 20.45.4. If X has a finite covering by elements of \mathcal{B} (for example if X is quasi-compact), then the theorem is a consequence of Lemma 20.45.6. In the general case we argue in exactly the same manner, using transfinite induction and Lemma 20.45.7.
First we use transfinite recursion to choose opens W_\alpha \subset X for any ordinal \alpha . Namely, we set W_0 = \emptyset . If \alpha = \beta + 1 is a successor, then either W_\beta = X and we set W_\alpha = X or W_\beta \not= X and we set W_\alpha = W_\beta \cup U_\alpha where U_\alpha \in \mathcal{B} is not contained in W_\beta . If \alpha is a limit ordinal we set W_\alpha = \bigcup _{\beta < \alpha } W_\beta . Then for large enough \alpha we have W_\alpha = X. Observe that for every \alpha the open W_\alpha is a union of elements of \mathcal{B}. Hence if \mathcal{B}_\alpha = \{ U \in \mathcal{B}, U \subset W_\alpha \} , then
S_\alpha = (\{ K_ U\} _{U \in \mathcal{B}_\alpha }, \{ \rho _ V^ U\} _{V \subset U\text{ with }U, V \in \mathcal{B}_\alpha })
is a system as in Lemma 20.45.4 on the ringed space W_\alpha .
We will show by transfinite induction that for every \alpha the system S_\alpha has a solution. This will prove the theorem as this system is the system given in the theorem for large \alpha .
The case where \alpha = \beta + 1 is a successor ordinal. (This case was already treated in the proof of the lemma above but for clarity we repeat the argument.) Recall that W_\alpha = W_\beta \cup U_\alpha for some U_\alpha \in \mathcal{B} in this case. By induction hypothesis we have a solution (K_{W_\beta }, \{ \rho ^{W_\beta }_ U\} _{U \in \mathcal{B}_\beta }) for the system S_\beta . Then we can consider the collection \mathcal{B}_\alpha ^* = \mathcal{B}_\alpha \cup \{ W_\beta \} of opens of W_\alpha and we see that we obtain a system (\{ K_ U\} _{U \in \mathcal{B}_\alpha ^*}, \{ \rho _ V^ U\} _{V \subset U\text{ with }U, V \in \mathcal{B}_\alpha ^*}). Note that this new system also satisfies condition (3) by Lemma 20.45.4 applied to the solution K_{W_\beta }. For this system we have W_\alpha = W_\beta \cup U_\alpha . This reduces us to the case handled in Lemma 20.45.6.
The case where \alpha is a limit ordinal. Recall that W_\alpha = \bigcup _{\beta < \alpha } W_\beta in this case. For \beta < \alpha let (K_{W_\beta }, \{ \rho ^{W_\beta }_ U\} _{U \in \mathcal{B}_\beta }) be the solution for S_\beta . For \gamma < \beta < \alpha the restriction K_{W_\beta }|_{W_\gamma } endowed with the maps \rho ^{W_\beta }_ U, U \in \mathcal{B}_\gamma is a solution for S_\gamma . By uniqueness we get unique isomorphisms \rho _{W_\gamma }^{W_\beta } : K_{W_\beta }|_{W_\gamma } \to K_{W_\gamma } compatible with the maps \rho ^{W_\beta }_ U and \rho ^{W_\gamma }_ U for U \in \mathcal{B}_\gamma . These maps compose in the correct manner, i.e., \rho _{W_\delta }^{W_\gamma } \circ \rho _{W_\gamma }^{W_\beta }|_{W_\delta } = \rho ^{W_\delta }_{W_\beta } for \delta < \gamma < \beta < \alpha . Thus we may apply Lemma 20.45.7 (note that the vanishing of negative exts is true for K_{W_\beta } by Lemma 20.45.4 applied to the solution K_{W_\beta }) to obtain K_{W_\alpha } and isomorphisms
\rho _{W_\beta }^{W_\alpha } : K_{W_\alpha }|_{W_\beta } \longrightarrow K_{W_\beta }
compatible with the maps \rho _{W_\gamma }^{W_\beta } for \gamma < \beta < \alpha .
To show that K_{W_\alpha } is a solution we still need to construct the isomorphisms \rho _ U^{W_\alpha } : K_{W_\alpha }|_ U \to K_ U for U \in \mathcal{B}_\alpha satisfying certain compatibilities. We choose \rho _ U^{W_\alpha } to be the unique map such that for any \beta < \alpha and any V \in \mathcal{B}_\beta with V \subset U the diagram
\xymatrix{ K_{W_\alpha }|_ V \ar[r]_{\rho _ U^{W_\alpha }|_ V} \ar[d]_{\rho _{W_\beta }^{W_\alpha }|_ V} & K_ U|_ V \ar[d]^{\rho _ U^ V} \\ K_{W_\beta } \ar[r]^{\rho _ V^{W_\beta }} & K_ V }
commutes. This makes sense because
(\{ K_ V\} _{V \subset U, V \in \mathcal{B}_\beta \text{ for some }\beta < \alpha }, \{ \rho _ V^{V'}\} _{V \subset V'\text{ with }V, V' \subset U \text{ and }V, V' \in \mathcal{B}_\beta \text{ for some }\beta < \alpha })
is a system as in Lemma 20.45.4 on the ringed space U and because (K_ U, \rho ^ U_ V) and (K_{W_\alpha }|_ U, \rho _ V^{W_\beta }\circ \rho _{W_\beta }^{W_\alpha }|_ V) are both solutions for this system. This gives existence and uniqueness. We omit the proof that these maps satisfy the desired compatibilities (it is just bookkeeping).
\square
Comments (4)
Comment #2796 by Tanya Kaushal Srivastava on
Comment #2899 by Johan on
Comment #5514 by Zili Zhang on
Comment #5709 by Johan on