NAG-003022 · Accepted whole
Exact stability, with a normality condition for saturation
Proper direct image on historical ordinary compact tests
Open in the English reader · PDF page 1884
Statement · exact TeX
\begin{theorem}[Exact stability, with a normality condition for saturation]\label{thm:localization}
Let $A\subset C\subset F$ be a finitely generated $K$-subalgebra of $F$ containing $A$. Then $M_C=C^\times/K^\times$ is a lattice, and there is an exact sequence
\[
0\longrightarrow M_A\longrightarrow M_C
\longrightarrow C^\times/A^\times\longrightarrow0.
\tag{4}\label{eq:localexact}
\]
If $A$ is normal, the first inclusion is saturated. This applies in particular to $C=A_h$ for $h\in A\setminus\{0\}$ and to
\[
C=A[q_1^{\pm1},\ldots,q_s^{\pm1}],\qquad q_i\in F^\times.
\tag{5}\label{eq:ratadj}
\]
If $A$ is generated by units, so is the ring in (\ref{eq:ratadj}), including $A_h$ as the case $q_1=h$. Thus these rings are very affine. In (\ref{eq:ratadj}) the morphism $\Spec C\to U$ is not asserted to be an open immersion. When one inverts finitely many nonzero numerators and denominators $a_i,b_i\in A$, the ring is the genuine principal-open localization $A_h$, where $h=\prod_i a_i b_i$.
\end{theorem}Full selected proof · exact TeX
\begin{proof}
The ring $C$ is an integral finite-type $K$-algebra, so Theorem~\ref{thm:lattice} applies to it. The inclusion of rings induces an injection of units, and it induces an injection modulo constants because the constant field in both quotients is the same embedded $K$. Taking these successive group quotients gives (\ref{eq:localexact}).
Suppose now that $A$ is normal and that $n[g]\in M_A$ for $g\in C^\times$ and $n>0$. Then $g^n=c u$ for some $c\in K^\times$ and $u\in A^\times$. Both $g$ and $g^{-1}$ are integral over $A$, by the monic equations $X^n-cu$ and $X^n-(cu)^{-1}$. They lie in $F$, so normality puts both in $A$. Therefore $[g]\in M_A$, which is exactly saturation. The generation assertion follows directly by adjoining the specified units and their inverses. The final identification follows because making each nonzero $a_i,b_i$ invertible is equivalent to inverting their product.
\end{proof}Typeset reader excerpt

Typed proof inputs
- NAG-003013 · selected whole (earlier same-packet whole)
Reader cross-references
Proof map
Verified map inputs · NAG-003022
Registered proof input
- NAG-003013 · Boundary construction of the actual unit lattice · accepted whole
- Apply finite freeness over the same algebraically closed K to A and the actual finite-type integral subalgebra C of F. Only the saturation clause uses normality. A ring formed by adjoining Laurent functions need not be an open localization.
NAG-003022 · Exact stability, with a normality condition for saturation · Chapter argument