Suppose $\phi:S_\bullet \rightarrow R_\bullet$ is a morphism of graded rings that has degree $d$, i.e. $\phi$ maps $S_n$ to $R_{dn}$ for all $n$. Then $\phi$ induces a morphism \begin{equation} \Phi:\text{Proj}~R_\bullet \setminus V(\phi(S_+)) \rightarrow \text{Proj}~S_\bullet \end{equation}

If $V(\phi(S_+))$ is empty then $\Phi$ would be a morphism between two projective schemes. If both $R_\bullet$ and $S_\bullet$ are finitely generated in degree $1$, then does the line bundle $\mathcal{O}_{\text{Proj}S_\bullet}(1)$ pulls back to $\mathcal{O}_{\text{Proj}R_\bullet}(d)$?

If it is not true generally, is it true when $S_0=R_0=A$, i.e. they are both graded $A$-algebra? If it is still not true, what about we let $A=k$ is a field?

edit: Suppose with some new conditions, $\mathcal{O}_{\text{Proj}S_\bullet}(1)$ pulls back to $\mathcal{O}_{\text{Proj}R_\bullet}(d)$, then the global sections of $\mathcal{O}_{\text{Proj}S_\bullet}(1)$ pulls back to the global sections of $\mathcal{O}_{\text{Proj}R_\bullet}(d)$, is this the same as the map $S_1 \rightarrow R_d$?


What is $O_{Proj S_*}(n) = O_S(n)$? Whatever it is, you can describe it and thus work with it by knowing the following facts:

  1. $O_S(n)$ is isomorphic to the structure sheaf on each open set of the cover $D(s)$, where $s \in S_n$, via some $\phi_s: O(n)|_{D(s)} \to O|_{D(s)}$. In fact the structure sheaf over $D(s)$ is just $S[s^{-1}]_0$, and $O_S(n)$ is just $S[s^{-1}]_n$. So $\phi_s$ can be taken to be multiplication by $s$, which is an isomorphism on this patch.

  2. The transition function $\phi_s \circ \phi_t^{-1}$ is precisely multiplication by $s/t$. (This follows immediately from 1.)

Main point of this: To understand a line bundle, we just need to know an open cover on which it trivializes, and the transition functions on the intersections.

We know how to pull back a line bundle, also.

Explicitly: If $O_V \to L$ is a trivialization over $V$, then the pullback by $\phi^*$ of this map gives a trivialization of $O_{\phi^{-1}(V)} \cong \phi^* O_V \to \phi^* L$.

The map $\psi: O_{\phi^{-1}(V)} \cong \phi^* O_V$ is the restriction of a canonical globally defined sheaf isomorphism between $\phi^* O_S$ and $O_T$, which is locally just "multiplication" $A \otimes_A B \to B$ - you can probably replace it with an equality without loss, but I personally am confused by that level of sloppiness - at least at this point in my mathematical life. Anyway it is not too difficult to keep track of it.

So let's just apply our understanding of the pullback to $O_S(n)$ and see what happens.

1) $\phi^{-1}(D(s)) = D(\phi(s))$.

So far so good - your condition that $V(\phi(S_+)) = \emptyset$ implies that this gives an open cover of $Proj T$ on which we know that $\phi^* O(n)$ trivializes.

(It may not be the case that $S_n$ surjects onto $S_{dn}$. That is okay, we still get a cover from the $D(\phi(s))$, as $s \in S_n$. Assuming your condition, which is that $V(\phi(S+)) = \emptyset$, for any homogeneous prime in $Proj(T)$ there is some $k$ and $s$ so that $\phi(s)^k \not \in P$. So this implies that $\phi(s) \not \in P$.)

2) Also, the trivialization $T_s: O_S|_{D(s)} \to O_S(n) |_{D(s)}$ becomes some $\phi^*(T_s) : \phi^* O_S \to \phi^*(O_S(n))$ over $D(\phi(s))$. Combing this with the isomorphism $\psi : O_T \to \phi^* O_S$, our trivialization over this chart is $\phi^*(T_s) \circ \psi|_{D(\phi(s)}$.

3) Claim: The transition functions become $\phi(s/t) = \phi(s) / \phi(t)$. Proof:

First we compute: $(\phi^*(T_t) \circ \psi))^{-1} \circ \phi^*(T_s) \circ \psi = \psi^{-1} \phi^* (T_s \circ T_t^{-1}) \psi = \psi^{-1} \_ \times \phi(s/t) \psi = \psi^{-1} (\_ \times \phi(s)/\phi(t)) \psi$. Then note: the isomorphism $\psi$ commutes with multiplication by elements of the structure sheaf (it is an isomorphism of $O_X$ algebras, in particular $O_X$ modules), so in the end we get that our transition function is just multiplication by $\phi(s) / \phi(t)$.

If you look back at the first paragraph, this is some cover and transition maps that describe $O_T(dn)$. So apparently the pullback is isomorphic as a line bundle to $O_T(dn)$.

I think your claim is correct - but maybe I am overlooking something subtle! This stuff is confusing!


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.