Lemma 57.13.5. Let $k$ be a field. Let $X$ be a smooth proper scheme over $k$. Let $K \in D_{perf}(\mathcal{O}_{X \times X})$. If the Fourier-Mukai functor $\Phi _ K : D_{perf}(\mathcal{O}_ X) \to D_{perf}(\mathcal{O}_ X)$ is isomorphic to the identity functor, then $K \cong \Delta _*\mathcal{O}_ X$ in $_{perf}(\mathcal{O}_{X \times X})$.
Proof. Let $i$ be the minimal integer such that the cohomology sheaf $H^ i(K)$ is nonzero. Let $\mathcal{E}$ and $\mathcal{G}$ be finite locally free $\mathcal{O}_ X$-modules. Then
which is zero if $i < 0$. On the other hand, we can choose $\mathcal{E}$ and $\mathcal{G}$ such that there is a surjection $\mathcal{E}^\vee \boxtimes \mathcal{G}^\vee \to H^ i(K)$ by Lemma 57.9.1. In this case the left hand side of the equalities is nonzero. Hence we conclude that $H^ i(K) = 0$ for $i < 0$.
Let $i$ be the maximal integer such that $H^ i(K)$ is nonzero. The same argument with $\mathcal{E}$ and $\mathcal{G}$ support of dimension $0$ shows that $i \leq 0$. Hence we conclude that $K$ is given by a single coherent $\mathcal{O}_{X \times X}$-module $\mathcal{K}$ sitting in degree $0$.
Since $R\text{pr}_{2, *}(\text{pr}_1^*\mathcal{F} \otimes \mathcal{K})$ is $\mathcal{F}$, by taking $\mathcal{F}$ supported at closed points we see that the support of $\mathcal{K}$ is finite over $X$ via $\text{pr}_2$. Since $R\text{pr}_{2, *}(\mathcal{K}) \cong \mathcal{O}_ X$ we conclude by Functors and Morphisms, Lemma 56.7.6 that $\mathcal{K} = s_*\mathcal{O}_ X$ for some section $s : X \to X \times X$ of the second projection. Then $\Phi _ K(M) = f^*M$ where $f = \text{pr}_1 \circ s$ and this can happen only if $s$ is the diagonal morphism as desired. $\square$
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Your email address will not be published.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).
All contributions are licensed under the GNU Free Documentation License.
Comments (0)