Proof.
To prove (1) we use the spectral sequence
\[ H^ p(X, \mathop{\mathcal{E}\! \mathit{xt}}\nolimits ^ q(\mathcal{F}, \mathcal{G})) \Rightarrow \mathop{\mathrm{Ext}}\nolimits ^{p + q}_ X(\mathcal{F}, \mathcal{G}) \]
of Cohomology, Section 20.43. Let $x \in X$. We have
\[ \mathop{\mathcal{E}\! \mathit{xt}}\nolimits ^ q(\mathcal{F}, \mathcal{G})_ x = \mathop{\mathcal{E}\! \mathit{xt}}\nolimits ^ q_{\mathcal{O}_{X, x}}(\mathcal{F}_ x, \mathcal{G}_ x) \]
see Cohomology, Lemma 20.51.4 (this also uses that $\mathcal{F}$ is pseudo-coherent by Derived Categories of Schemes, Lemma 36.10.3). Set $d_ x = \dim (\mathcal{O}_{X, x})$. Since $\mathcal{O}_{X, x}$ is regular the ring $\mathcal{O}_{X, x}$ has global dimension $d_ x$, see Algebra, Proposition 10.110.1. Thus $\mathop{\mathcal{E}\! \mathit{xt}}\nolimits ^ q_{\mathcal{O}_{X, x}}(\mathcal{F}_ x, \mathcal{G}_ x)$ is zero for $q > d_ x$. It follows that the modules $\mathop{\mathcal{E}\! \mathit{xt}}\nolimits ^ q(\mathcal{F}, \mathcal{G})$ have support of dimension at most $d - q$. Hence we have $H^ p(X, \mathop{\mathcal{E}\! \mathit{xt}}\nolimits ^ q(\mathcal{F}, \mathcal{G})) = 0$ for $p > d - q$ by Cohomology, Proposition 20.20.7. This proves (1).
Proof of (2). We may use induction on the number of nonzero cohomology sheaves of $K$ and $L$. The case where these numbers are $0, 1$ follows from (1). If the number of nonzero cohomology sheaves of $K$ is $> 1$, then we let $i \in \mathbf{Z}$ be minimal such that $H^ i(K)$ is nonzero. We obtain a distinguished triangle
\[ H^ i(K)[-i] \to K \to \tau _{\geq i + 1}K \]
(Derived Categories, Remark 13.12.4) and we get the vanishing of $\mathop{\mathrm{Hom}}\nolimits (K, L)$ from the vanishing of $\mathop{\mathrm{Hom}}\nolimits (H^ i(K)[-i], L)$ and $\mathop{\mathrm{Hom}}\nolimits (\tau _{\geq i + 1}K, L)$ by Derived Categories, Lemma 13.4.2. Similarly if $L$ has more than one nonzero cohomology sheaf.
$\square$
Comments (3)
Comment #6479 by Noah Olander on
Comment #6480 by Johan on
Comment #6545 by Johan on