Ravi Vakil 9.2.B is "Suppose that $S \rightarrow R$ is a surjection of graded rings. Show that the induced morphism $\text{Proj }R \rightarrow \text{Proj }S$ is a closed embedding."

I don't even see how to prove that the morphism is affine. The only ways I can think of to do this are to either classify the affine subspaces of Proj S, or to prove that when closed morphisms are glued, one gets a closed morphism.

Are either of those possible, and how can this problem be done?


I think a good strategy could be to verify the statement locally, and then verify that the glueing is successful, as you said. Let us call $\phi:S\to R$ your surjective graded morphism, and $\phi^\ast:\textrm{Proj}\,\,R\to \textrm{Proj}\,\,S$ the corresponding morphism. Note that $$\textrm{Proj}\,\,R=\bigcup_{t\in S_1}D_+(\phi(t))$$ because $S_+$ (the irrelevant ideal of $S$) is generated by $S_1$ (as an ideal), so $\phi(S_+)R$ is generated by $\phi(S_1)$. For any $t\in S_1$ you have a surjective morphism $S_{(t)}\to R_{\phi(t)}$ (sending $x/t^n\mapsto \phi(x)/\phi(t)^n$, for any $x\in S$), which corresponds to the canonical closed immersion of affine schemes $\phi^\ast_t:D_+(\phi(t))\hookrightarrow D_+(t)$. It remains to glue the $\phi^\ast_t$'s.

  • $\begingroup$ Hmmm, is it true that gluing together closed morphisms always gives a closed morphism? $\endgroup$ – only Apr 21 '12 at 22:00
  • $\begingroup$ Oh, I think I can prove it works if there are a finite number of closed morphisms, just by gluing the images and then using finite union of closed subschemes are closed. I don't see how to do it here in general though, because I need to glue together infinitely many... $\endgroup$ – only Apr 21 '12 at 22:07
  • $\begingroup$ There is no problem in the infinite case too. If you have an arbitrary open covering $(X_t)$ of a scheme $X$ and a closed subscheme $U_t\subset X_t$ for every $t$, such that $U_s\cap X_t$ and $U_t\cap X_s$ are the same closed subscheme of $U_t\cap U_s$, then you have a unique closed subscheme $U\subset X$ s.t. $U\cap X_t=U_t$ for any $t$. You may take this property as a synonymous of having a closed immersion. Apply it to $X=\textrm{Proj}\,\,S$ and $X_t=D_+(t)$, $U_t=D_+(\phi(t))$. $\endgroup$ – Brenin Apr 21 '12 at 22:29
  • $\begingroup$ Note that for every $t,s\in S_1$, the morphisms of rings $\phi_t$ and $\phi_s$ coincide on $S_{(ts)}$, the localization corresponding to $D_+(t)\cap D_+(s)$. So the condition in the assumption is satisfied. $\endgroup$ – Brenin Apr 21 '12 at 22:31

Almost 7 years late! Here is my try. Hallo Thorsten!

I call our maps $f \colon \operatorname{Proj} B \to \operatorname{Proj} A$ and $\varphi \colon A \to B$. Surjectivity implies that we actually have a well-defined map $\operatorname{Proj} B \to \operatorname{Proj} A$ and a morphism of schemes in this way.

Being a closed immersion is affine-local on the target. Therefore we can consider some cover of open affines $\bigcup_{j \in J} V_j = \operatorname{Proj} A$ and then check that for each $j \in J$ we have a closed immersion $f \mid_{f^{-1}(V_j)} \colon f^{-1}(V_j) \hookrightarrow V_j$. This is described in Vakil's notes as an exercise.

We have that the collection over all homogeneous $g \in A$ of $D(g) = \{\,p \in \operatorname{Proj} A \mid g \notin p \,\}$ cover $\operatorname{Proj} A$. As $\varphi$ is surjective, we have $f^{-1} (D(g)) = D(\varphi(g))$. We now have \begin{align*} f \mid_{D(\varphi(g))} \colon D(\varphi(g)) & \hookrightarrow D(g) \\ p & \mapsto \varphi^{-1} (p) \, . \end{align*}

These sets are all open affines! For any graded ring $R$, we have for any homogeneous $h \in R$ the identification $D(h) = \operatorname{Spec}(R_h)_0 = \operatorname{Spec}\{\, \frac{x}{h^n} \mid n \in \mathbb N, \, \deg x = \deg h \cdot n \,\}$. (Sometimes, $(R_h)_0$ is confusingly written as $R_{(h)}$.) Our map can then be seen as \begin{align*} f \mid_{\operatorname{Spec} (B_{\varphi(g)})_0} \colon \operatorname{Spec} (B_{\varphi(g)})_0 & \hookrightarrow \operatorname{Spec} (A_g)_0 \\ p & \mapsto \varphi^{-1} (p) \; , \end{align*} which corresponds to the surjective ring homomorphism \begin{align*} \varphi (D(g)) \colon (A_g)_0 & \to (B_{\varphi(g)})_0 \\ \frac{x}{g^n} & \mapsto \frac{\varphi(x)}{\varphi(g)^n} \; , \end{align*} which means that $f \mid_{f^{-1}(D(g))} \colon f^{-1}(D(g)) \hookrightarrow D(g)$ is a closed immersion, concluding that $f$ is a closed immersion.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.