The Stacks project

56.14 Deducing fully faithfulness

It will be useful for us to know when a functor is fully faithful we offer the following variant of [Lemma 2.15, Orlov-K3].

reference

Lemma 56.14.1. Let $F : \mathcal{D} \to \mathcal{D}'$ be an exact functor of triangulated categories. Let $S \subset \mathop{\mathrm{Ob}}\nolimits (\mathcal{D})$ be a set of objects. Assume

  1. $F$ has both right and left adjoints,

  2. for $K \in \mathcal{D}$ if $\mathop{\mathrm{Hom}}\nolimits (E, K[i]) = 0$ for all $E \in S$ and $i \in \mathbf{Z}$ then $K = 0$,

  3. for $K \in \mathcal{D}$ if $\mathop{\mathrm{Hom}}\nolimits (K, E[i]) = 0$ for all $E \in S$ and $i \in \mathbf{Z}$ then $K = 0$,

  4. the map $\mathop{\mathrm{Hom}}\nolimits (E, E'[i]) \to \mathop{\mathrm{Hom}}\nolimits (F(E), F(E')[i])$ induced by $F$ is bijective for all $E, E' \in S$ and $i \in \mathbf{Z}$.

Then $F$ is fully faithful.

Proof. Denote $F_ r$ and $F_ l$ the right and left adjoints of $F$. For $E \in S$ choose a distinguished triangle

\[ E \to F_ r(F(E)) \to C \to E[1] \]

where the first arrow is the unit of the adjunction. For $E' \in S$ we have

\[ \mathop{\mathrm{Hom}}\nolimits (E', F_ r(F(E))[i]) = \mathop{\mathrm{Hom}}\nolimits (F(E'), F(E)[i]) = \mathop{\mathrm{Hom}}\nolimits (E', E[i]) \]

The last equality holds by assumption (4). Hence applying the homological functor $\mathop{\mathrm{Hom}}\nolimits (E', -)$ (Derived Categories, Lemma 13.4.2) to the distinguished triangle above we conclude that $\mathop{\mathrm{Hom}}\nolimits (E', C[i]) = 0$ for all $i \in \mathbf{Z}$ and $E' \in S$. By assumption (2) we conclude that $C = 0$ and $E = F_ r(F(E))$.

For $K \in \mathop{\mathrm{Ob}}\nolimits (\mathcal{D})$ choose a distinguished triangle

\[ F_ l(F(K)) \to K \to C \to F_ l(F(K))[1] \]

where the first arrow is the counit of the adjunction. For $E \in S$ we have

\[ \mathop{\mathrm{Hom}}\nolimits (F_ l(F(K)), E[i]) = \mathop{\mathrm{Hom}}\nolimits (F(K), F(E)[i]) = \mathop{\mathrm{Hom}}\nolimits (K, F_ r(F(E))[i]) = \mathop{\mathrm{Hom}}\nolimits (K, E[i]) \]

where the last equality holds by the result of the first paragraph. Thus we conclude as before that $\mathop{\mathrm{Hom}}\nolimits (C, E[i]) = 0$ for all $E \in S$ and $i \in \mathbf{Z}$. Hence $C = 0$ by assumption (3). Thus $F$ is fully faithful by Categories, Lemma 4.24.4. $\square$

Lemma 56.14.2. Let $k$ be a field. Let $X$ be a scheme of finite type over $k$ which is regular. Let $x \in X$ be a closed point. For a coherent $\mathcal{O}_ X$-module $\mathcal{F}$ supported at $x$ choose a coherent $\mathcal{O}_ X$-module $\mathcal{F}'$ supported at $x$ such that $\mathcal{F}_ x$ and $\mathcal{F}'_ x$ are Matlis dual. Then there is an isomorphism

\[ \mathop{\mathrm{Hom}}\nolimits _ X(\mathcal{F}, M) = H^0(X, M \otimes _{\mathcal{O}_ X}^\mathbf {L} \mathcal{F}'[-d_ x]) \]

where $d_ x = \dim (\mathcal{O}_{X, x})$ functorial in $M$ in $D_{perf}(\mathcal{O}_ X)$.

Proof. Since $\mathcal{F}$ is supported at $x$ we have

\[ \mathop{\mathrm{Hom}}\nolimits _ X(\mathcal{F}, M) = \mathop{\mathrm{Hom}}\nolimits _{\mathcal{O}_{X, x}}(\mathcal{F}_ x, M_ x) \]

and similarly we have

\[ H^0(X, M \otimes _{\mathcal{O}_ X}^\mathbf {L} \mathcal{F}'[-d_ x]) = \text{Tor}^{\mathcal{O}_{X, x}}_{d_ x}(M_ x, \mathcal{F}'_ x) \]

Thus it suffices to show that given a Noetherian regular local ring $A$ of dimension $d$ and a finite length $A$-module $N$, if $N'$ is the Matlis dual to $N$, then there exists a functorial isomorphism

\[ \mathop{\mathrm{Hom}}\nolimits _ A(N, K) = \text{Tor}^ A_ d(K, N') \]

for $K$ in $D_{perf}(A)$. We can write the left hand side as $H^0(R\mathop{\mathrm{Hom}}\nolimits _ A(N, A) \otimes _ A^\mathbf {L} K)$ by More on Algebra, Lemma 15.73.14 and the fact that $N$ determines a perfect object of $D(A)$. Hence the formula holds because

\[ R\mathop{\mathrm{Hom}}\nolimits _ A(N, A) = R\mathop{\mathrm{Hom}}\nolimits _ A(N, A[d])[-d] = N'[-d] \]

by Dualizing Complexes, Lemma 47.16.4 and the fact that $A[d]$ is a normalized dualizing complex over $A$ ($A$ is Gorenstein by Dualizing Complexes, Lemma 47.21.3). $\square$

Lemma 56.14.3. Let $k$ be a field. Let $X$ be a scheme of finite type over $k$ which is regular. Let $x \in X$ be a closed point and denote $\mathcal{O}_ x$ the skyscraper sheaf at $x$ with value $\kappa (x)$. Let $K$ in $D_{perf}(\mathcal{O}_ X)$.

  1. If $\mathop{\mathrm{Ext}}\nolimits ^ i_ X(\mathcal{O}_ x, K) = 0$ then there exists an open neighbourhood $U$ of $x$ such that $H^{i - d_ x}(K)|_ U = 0$ where $d_ x = \dim (\mathcal{O}_{X, x})$.

  2. If $\mathop{\mathrm{Hom}}\nolimits _ X(\mathcal{O}_ x, K[i]) = 0$ for all $i \in \mathbf{Z}$, then $K$ is zero in an open neighbourhood of $x$.

  3. If $\mathop{\mathrm{Ext}}\nolimits ^ i_ X(K, \mathcal{O}_ x) = 0$ then there exists an open neighbourhood $U$ of $x$ such that $H^ i(K^\vee )|_ U = 0$.

  4. If $\mathop{\mathrm{Hom}}\nolimits _ X(K, \mathcal{O}_ x[i]) = 0$ for all $i \in \mathbf{Z}$, then $K$ is zero in an open neighbourhood of $x$.

  5. If $H^ i(X, K \otimes _{\mathcal{O}_ X}^\mathbf {L} \mathcal{O}_ x) = 0$ then there exists an open neighbourhood $U$ of $x$ such that $H^ i(K)|_ U = 0$.

  6. If $H^ i(X, K \otimes _{\mathcal{O}_ X}^\mathbf {L} \mathcal{O}_ x) = 0$ for $i \in \mathbf{Z}$ then $K$ is zero in an open neighbourhood of $x$.

Proof. Observe that $H^ i(X, K \otimes _{\mathcal{O}_ X}^\mathbf {L} \mathcal{O}_ x)$ is equal to $K_ x \otimes _{\mathcal{O}_{X, x}}^\mathbf {L} \kappa (x)$. Hence part (5) follows from More on Algebra, Lemma 15.75.4. Part (6) follows from part (5). Part (1) follows from part (5), Lemma 56.14.2, and the fact that the Matlis dual of $\kappa (x)$ is $\kappa (x)$. Part (2) follows from part (1). Part (3) follows from part (5) and the fact that $\mathop{\mathrm{Ext}}\nolimits ^ i(K, \mathcal{O}_ x) = H^ i(X, K^\vee \otimes _{\mathcal{O}_ X}^\mathbf {L} \mathcal{O}_ x)$ by Cohomology, Lemma 20.47.5. Part (4) follows from part (3) and the fact that $K \cong (K^\vee )^\vee $ by the lemma just cited. $\square$

Lemma 56.14.4. Let $k$ be a field. Let $X$ and $Y$ be proper schemes over $k$. Assume $X$ is regular. Then a $k$-linear exact functor $F : D_{perf}(\mathcal{O}_ X) \to D_{perf}(\mathcal{O}_ Y)$ is fully faithful if and only if for any closed points $x, x' \in X$ the maps

\[ F : \mathop{\mathrm{Ext}}\nolimits ^ i_ X(\mathcal{O}_ x, \mathcal{O}_{x'}) \longrightarrow \mathop{\mathrm{Ext}}\nolimits ^ i_ Y(F(\mathcal{O}_ x), F(\mathcal{O}_{x'})) \]

are isomorphisms for all $i \in \mathbf{Z}$. Here $\mathcal{O}_ x$ is the skyscraper sheaf at $x$ with value $\kappa (x)$.

Proof. By Lemma 56.8.1 the functor $F$ has both a left and a right adjoint. Thus we may apply the criterion of Lemma 56.14.1 because assumptions (2) and (3) of that lemma follow from Lemma 56.14.3. $\square$

reference

Lemma 56.14.5. Let $k$ be a field. Let $X$ be a smooth proper scheme over $k$. Let $F : D_{perf}(\mathcal{O}_ X) \to D_{perf}(\mathcal{O}_ X)$ be a $k$-linear exact functor. Assume for every coherent $\mathcal{O}_ X$-module $\mathcal{F}$ with $\dim (\text{Supp}(\mathcal{F})) = 0$ there is an isomorphism $\mathcal{F} \cong F(\mathcal{F})$. Then $F$ is fully faithful.

Proof. By Lemma 56.14.4 it suffices to show that the maps

\[ F : \mathop{\mathrm{Ext}}\nolimits ^ i_ X(\mathcal{O}_ x, \mathcal{O}_{x'}) \longrightarrow \mathop{\mathrm{Ext}}\nolimits ^ i_ Y(F(\mathcal{O}_ x), F(\mathcal{O}_{x'})) \]

are isomorphisms for all $i \in \mathbf{Z}$ and all closed points $x, x' \in X$. By assumption, the source and the target are isomorphic. If $x \not= x'$, then both sides are zero and the result is true. If $x = x'$, then it suffices to prove that the map is either injective or surjective. For $i < 0$ both sides are zero and the result is true. For $i = 0$ any nonzero map $\alpha : \mathcal{O}_ x \to \mathcal{O}_ x$ of $\mathcal{O}_ X$-modules is an isomorphism. Hence $F(\alpha )$ is an isomorphism too and so $F(\alpha )$ is nonzero. Thus the result for $i = 0$. For $i = 1$ a nonzero element $\xi $ in $\mathop{\mathrm{Ext}}\nolimits ^1(\mathcal{O}_ x, \mathcal{O}_ x)$ corresponds to a nonsplit short exact sequence

\[ 0 \to \mathcal{O}_ x \to \mathcal{F} \to \mathcal{O}_ x \to 0 \]

Since $F(\mathcal{F}) \cong \mathcal{F}$ we see that $F(\mathcal{F})$ is a nonsplit extension of $\mathcal{O}_ x$ by $\mathcal{O}_ x$ as well. Since $\mathcal{O}_ x \cong F(\mathcal{O}_ x)$ is a simple $\mathcal{O}_ X$-module and $\mathcal{F} \cong F(\mathcal{F})$ has length $2$, we see that in the distinguished triangle

\[ F(\mathcal{O}_ x) \to F(\mathcal{F}) \to F(\mathcal{O}_ x) \xrightarrow {F(\xi )} F(\mathcal{O}_ x)[1] \]

the first two arrows must form a short exact sequence which must be isomorphic to the above short exact sequence and hence is nonsplit. It follows that $F(\xi )$ is nonzero and we conclude for $i = 1$. For $i > 1$ composition of ext classes defines a surjection

\[ \mathop{\mathrm{Ext}}\nolimits ^1(F(\mathcal{O}_ x), F(\mathcal{O}_ x)) \otimes \ldots \otimes \mathop{\mathrm{Ext}}\nolimits ^1(F(\mathcal{O}_ x), F(\mathcal{O}_ x)) \longrightarrow \mathop{\mathrm{Ext}}\nolimits ^ i(F(\mathcal{O}_ x), F(\mathcal{O}_ x)) \]

See Duality for Schemes, Lemma 48.15.4. Hence surjectivity in degree $1$ implies surjectivity for $i > 0$. This finishes the proof. $\square$


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 0G23. Beware of the difference between the letter 'O' and the digit '0'.