3

I am going through the proof of Proposition 4.17 in Liu's Algebraic Geometry and Arithmetic Curves:

enter image description here

First question (SOLVED):

I can't verify why $V(f\vert_U)$ contains $W\cap U$. If we write $\phi:W\cong\operatorname{Spec}A$ and $\psi:U\cong\operatorname{Spec}B$, then formally $W=V(f\vert_W)$ means that $$ \phi(W)=V({\phi^\#}^{-1}f\vert_W)=V(\mathfrak p). $$ where for notational convenience $a={\phi^\#}^{-1}f\vert_W$. And $W\cap U\subset V(f\vert_U)$ means that $$ \psi(W\cap U)=\psi(W)\cap\operatorname{Spec}B=V({\psi^\#}^{-1}f\vert_V)=V(b) $$ with $b={\psi^\#}^{-1}f\vert_V$. So let $x\in W$. We need to show that $b\in\psi(x)$. We know that $$ a\in\phi(x). $$

Attempt (SOLVED):

Let's say for notational convenience that $W=\operatorname{Spec}A$ (equality as opposed to an isomorphism). We still work with an explicit isomorphism $\psi:U\cong\operatorname{Spec}(B)$. So we want to show that $$ \psi(\psi^{-1}(U)\cap W)=V(b) $$ where $b=(\psi^{-1})^\#f\vert_W$. Let $x\in \psi^{-1}(U)\cap W$. Then $x$ is given by a prime $\mathfrak p$ of $A$, and write $f\vert_W=a\in A$. Then we know that $a\in\mathfrak p$. Write $\mathfrak q=\psi(\mathfrak p)$. Choose some $g\in A$ such that $$ \mathfrak p\in\operatorname{Spec}(A_g)\subset \psi^{-1}(U)\cap W. $$ Note that we have a sheaf map $$ \Psi=\psi\vert_{\operatorname{Spec}(A_g)}:\operatorname{Spec} A_g\to\psi(\operatorname{Spec}A_g)\subset\operatorname{Spec}B, $$ which corresponds to a map of rings $$ G:B\to A_g, $$ i.e., $\Psi=\operatorname{Spec}(G)$. So we have $$ G^{-1}(\mathfrak p)=\mathfrak q. $$ To conclude that $b\in\mathfrak q$, I would need that $G(b)=a$. This is true, if we remember that $b=(\psi^{-1})^\# f\vert_U$ and $a=f\vert_W$ and that $G$ corresponds to $\psi^\#$ (along with restriction). More precisely, we have $$ a/1\in A_g=\mathcal O_{\operatorname{Spec}A}(\operatorname{Spec}A_g), $$ and $$ G:\sigma\mapsto \sigma\vert_{\psi(\operatorname{Spec}A_g)}\mapsto \psi^\#\sigma\vert_{\psi(\operatorname{Spec}A_g)}. $$ so recalling the definition of $b$, we get indeed $$ G:b\mapsto f\vert_{\operatorname{Spec}A_g}=a/1\in A_g. $$ It's clear to me how to adapt this proof for the 'general' case where we have an isomorphism $\phi:U\cong\operatorname{Spec}A$ (instead of equality), which I won't write down.

Second question (SOLVED):

How can I show that $\xi_1$ and $\xi_2$ would be generic points of $U_1\cap U_2$?

Attempt: (SOLVED)

Let $Z_i=\overline{\{\xi_i\}}$. I know that $U_1\cap U_2$ is irreducible, and hence contains a generic point $\xi$ (since $U$ is a scheme), i.e. $$ \operatorname{Cl}_{U_1\cap U_2}(\xi)=U_1\cap U_2. $$ It follows that $$ \operatorname{Cl}_{U_1}(\{\xi\})=\operatorname{Cl}_{U_1}\operatorname{Cl}_{U_1\cap U_2}(\{\xi\})=\operatorname{Cl}_{U_1}(U_1\cap U_2)=U_1, $$ where in the last step I used that $U_1\cap U_2$ is dense in $U_1$ (which follows from irreducibility of $U_1$). Therefore $$ \operatorname{Cl}_{Z_1\cap U_1}(\{\xi\})=\operatorname{Cl}_{U_1}(\{\xi\})\cap U_1\cap Z_1=U_1\cap Z_1. $$ Now using that $U_1\cap Z_1$ is dense in $Z_1$ (by irreducibility of $Z_1$), we conclude that $$ \operatorname{Cl}_{Z_1}(\{\xi_i\})=\operatorname{Cl}_{Z_1}(\operatorname{Cl}_{Z_1\cap U_1}(\{\xi\}))=Z_1. $$ and hence $\xi=\xi_1$ by uniqueness of the generic point. The case for $Z_2$ is similar.

Sha Vuklia
  • 3,960
  • 4
  • 19
  • 37

1 Answers1

2

If $X$ is the spectrum of an integral domain, $\xi$ is the generic point of $X\iff \{\xi\} = \cap_{U: \text{ non-empty open subset of }X }U$. Using this, $\{\xi_1\}=\cap_{\emptyset \neq U \subset U_1\cap U_2}U=\{\xi_2\}$ if $U_1 \cap U_2 \neq \emptyset.$

Acrobatic
  • 1,281
  • 1
  • 3
  • 12
  • I know that if $\xi$ is generic in $X$, then $\xi\in V$ for any non-empty open $ V$. My problem is that $\xi_i$ is just generic in the sense that $\overline{\xi_i}=Z_i$ is an irreducible component. I suppose I can show then that ${\xi_i}=\bigcap_{U\text{ open}:U\cap Z\neq\emptyset}(U\cap Z_i)$, but then it seems I need to show that for any open $U\subset U_1\cap U_2$, we have $U\cap Z_i\neq \emptyset$. So could you clarify why $U_1\cap U_2\cap Z_1\neq\emptyset$? – Sha Vuklia May 23 '23 at 14:34
  • Liu’s definition of a generic point states that if $\xi_1\in U_1$ is a generic point of $X$, it is also a (the) generic point of $U_1$. In particular, $U_1\cap Z_1 =$ the closure of ${\xi_1}$ in $U_1$ = $U_1$ and $U_1\cap U_2\cap Z_1=U_1\cap U_2\neq \emptyset$. – Acrobatic May 23 '23 at 16:34
  • I'm not sure about $U_1\cap U_2\cap Z_1=U_1\cap U_2$ (the latter is open while the former isn't necessarily), but it doesn't matter, because I have arrived at an argument via another route. – Sha Vuklia May 23 '23 at 17:23