This question is a special case of Open subschemes of affine schemes are affine? where it is established that in general, open subschemes of affine schemes are not affine. I was wondering if this was true in a more specific case, i.e. for principal opens. Take an affine scheme $X= Spec A$. My guess would be that for any $f\in A$ $$ (D(f), \mathcal O_{X\vert D(f)})=(Spec A_f,\mathcal O_{Spec A_f})$$ as ringed spaces. To do this I would use the natural map $$\pi: A\rightarrow A_f$$ to construct maps $$\phi:D(f)\rightarrow Spec A_f$$ $$\psi: Spec A_f\rightarrow D(f)$$ with $\phi:\mathfrak p \mapsto (\pi(\mathfrak p))$, the ideal generated by $\pi(\mathfrak p)$, and $\psi: \mathfrak q\mapsto \pi^{-1}(\mathfrak q\cap \pi(A)) $. Would this work? Is there a simpler way?
1 Answers
Considering the map $\pi$ is the right idea. Recall (or try prove) the following lemma.
$\textbf{Lemma:}$ Let $f\colon A\to B$ be a ring homomorphism. Then the continuous map on the spectra $F=\operatorname{Spec}f$ can be extended to a map of schemes $(F,F^\sharp)\colon \operatorname{Spec}B\to \operatorname{Spec}A$ such that on the principal open sets of $\operatorname{Spec}B$ $F^\sharp$ is just the map making the square involving $f$ and the localization maps commute.
So, in this situation we obtain a map of schemes $(F,F^\sharp)\colon \operatorname{Spec}A_f \to \operatorname{Spec}A$, where $F=\operatorname{Spec}\pi$. Clearly, the image of $F$ is contained in, and in fact equal to, the principal open $D(f)$. So $(F,F^\sharp)$ restricts to map of schemes $G=(F',F'^\sharp)\colon \operatorname{Spec}A_f \to (D(f),\mathcal{O}_{\operatorname{Spec}A\vert{D(f)}})$.
We wish to prove that $G$ is an isomorphism. On the topological level that's clearly the case. To see it's an isomorphism on the level of sheaves, it's enough to check this on a basis. A basis for the topology on $D(f)$ is given by the principal opens of $\operatorname{Spec}A$ which are contained in $D(f)$, i.e. those of the form $D(fg)$. Now, as mentioned above, $F'^\sharp_{D(fg)}$ is just the map $A_{fg}\to (A_f)_{(g/1)}$ making the diagram involving $\pi$ and the localization maps $A\to A_{fg}$ and $A_f \to (A_f)_{(g/1)}$ commute, and this is an isomorphism.
- 1,188
- 6
- 11
-
Thank you! This seems to answer my question. Where could I find this lemma? – user306194 Feb 11 '16 at 16:24
-
See Hartshorne, II, Prop. 2.3. – Plankton Feb 11 '16 at 18:10
-
You are welcome. Just to let you know, because you haven't accepted any of the answers to your questions yet, you can accept an answer by clicking on the tick next to it. – Plankton Feb 12 '16 at 10:45