0

Let $f:X\to Y$ be a morphism of schemes. Prove that $f$ is finite if and only if for any affine open subscheme $V=\mathrm{Spec}B$ of $Y$, $f^{-1}(V)=\mathrm{Spec}A$ for some $B$-algebra $A$ which is finitely generated as a $B$-module.

However, I only show that $V$ can be covered by finitely many affine open subschemes $D(b_i)$ of $V$ for $b_i\in B$ such that $A_{a_i}$ is a $B_{b_i}$-algebra which is finitely generated as a $B_{b_i}$-module, where $a_i$ is the image of $b_i$ under $f^\sharp(V):B\to A$.

  • You need to assume $Y$ is separated. Then can you show that $V\cap \mathrm{Spec} A_{a_i}$ is affine and then finish the proof? – Mohan Jun 12 '17 at 02:17
  • Considering ${f^{ - 1}}\left( V \right)$ as $V{ \times _Y}X$ and then just drawing a bunch of pullback diagrams using the cover may help. –  Jun 12 '17 at 02:34

1 Answers1

0

It should be clear that you only need to show that $f^{-1}(V)$ is affine, so you might as well take $X = f^{-1}(V)$ and $Y = \operatorname{Spec}B$. Then to show that $X$ is affine, you can use the criterion for affineness (Hartshorne exercise II.2.17b):

A scheme $X$ is affine if and only if there is a finite set of elements $f_1,\dots, f_r\in A = \Gamma(X,\mathcal{O}_X)$ such that the open subsets $X_{f_i}$ are affine and $(f_1,\dots, f_r) = A$.

You can show that you can take your $f_i$ to be the images of the $b_i$ under the map $f^\sharp : B\to\mathcal{O}_X(X)$ (show that $X_{f^\sharp(b_i)} = \operatorname{Spec} A_{a_i}$). You already know that the $a_i$ generate the unit ideal in $\mathcal{O}_X(X)$ because the $\operatorname{Spec}B_{b_i}$ cover $Y = \operatorname{Spec} B$. I'll leave this part of the proof to you.

Now, to show that $\mathcal{O}_X(X)$ is finitely generated as a $B$-module: let $s\in\mathcal{O}_X(X)$. Then for any $i,$ there exists a set $\{x_{i,j}\}_{j = 1}^{n_i}\subseteq\mathcal{O}_X(X)$ which generate $\mathcal{O}_X(X)\left[\frac{1}{f^\sharp(b_i)}\right] = f^{-1}(B_{b_i}) = \mathcal{O}_{X_{f^\sharp(b_i)}}(X_{f^\sharp(b_i)}) = A_{a_i}$ is a finitely generated $B_{b_i}$-module, so $$ \frac{s}{1} = \sum_{j = 1}^{n_i} f^\sharp(\beta_j)\frac{x_{i,j}}{f^\sharp(b_i)^{r_{i,j}}}\in\mathcal{O}_X(X)\left[\frac{1}{f^\sharp(b_i)}\right] $$ for some $r_{i,j}\in\Bbb N_0$, $\beta_j\in B$. Multiply through by an appropriate power $f^\sharp(b_i)^{r_i}$ to clear denominators and get $$ f^\sharp(b_i)^{r_i} s = \sum_{j = 1}^{n_i} f^\sharp(\gamma_j)x_{i,j} $$ by absorbing the excess $f^\sharp(b_i)$'s in any one term on the right hand side into $f^\sharp(\beta_j)$ to get $\gamma_j = \beta_j\cdot b_i^{?}$ (now an equality in $\mathcal{O}_X(X)$).

Next, you know that $f^\sharp(b_i)^{r_i} = f^\sharp(b_i^{r_i})$ and that $(b_1,\dots, b_n) = B$, which implies that $(b_1^{r_1},\dots, b_n^{r_n}) = B$ (by a standard argument in commutative algebra, you can adapt Ian's proof here if you're not familiar with the proof). Then there exist $c_i\in B$ such that $\sum_{i = 1}^n c_i b_i^{r_i} = 1$, which implies $$ s = \sum_{i = 1}^n f^\sharp(c_i)f^\sharp(g_i^{r_i})s = \sum_{i = 1}^n\left(f^\sharp(c_i)\sum_{j = 1}^{n_i} f^\sharp(\gamma_j)x_{i,j}\right). $$ Thus, the (finite) set $\{x_{i,j}\}_{i,j}$ generates $\mathcal{O}_X(X)$ as a $B$-module.

Stahl
  • 23,212
  • You can also play these games more generally and formally. See here for some details on how to deduce that a property defined "on an affine open cover" holds if and only if it holds for the restriction to any affine open in the target. – Stahl Jun 12 '17 at 04:15