NAG-001567 · Accepted whole
Descending finite Banach generation
1 Valued fields, Banach algebras and analytic points
Open in the English reader · PDF page 116
Statement and full selected proof · exact TeX
\begin{theorem}[\bi{Descending finite Banach generation}{有限 Banach 生成元的下降}]\label{fnd:freebanachfinitegenerationdescent}
\bi{Let $A$ be a generalized $K$-affinoid algebra, let $M$ be any Banach $A$-module with contractive action, and let $r$ be finite and free over $K$. Put $M_r=M\widehat\otimes_KK_r$. Suppose that there are $u_1,\ldots,u_n\in M_r$ and $C>0$ such that every $y\in M_r$ has an expression}
{设 $A$ 为 generalized $K$-affinoid 代数,$M$ 为作用 contractive 的任意 Banach $A$-模,$r$ 为在 $K$ 上自由的有限元组。令 $M_r=M\widehat\otimes_KK_r$。假设存在 $u_1,\ldots,u_n\in M_r$ 和 $C>0$,使每个 $y\in M_r$ 都有表示}
\[
y=\sum_{i=1}^n b_i u_i,
\qquad b_i\in A_r,\qquad
\max_i\|b_i\|\leq C\|y\|.
\]
\bi{Then there are finitely many $m_1,\ldots,m_t\in M$ and $D>0$ such that every $m\in M$ has an expression $m=\sum_j a_jm_j$ with $\max_j\|a_j\|\leq D\|m\|$. In particular, $M$ is algebraically finite and its given norm is uniformly equivalent to its canonical finite-module norm. If the absolute value on $K_r$ is nontrivial, it is enough to assume that $M_r$ is algebraically finite over $A_r$: the required constant $C$ is then automatic. For trivially valued $K_r$, the displayed lifting hypothesis is retained.}
{则存在有限个 $m_1,\ldots,m_t\in M$ 和 $D>0$,使每个 $m\in M$ 都有表示 $m=\sum_j a_jm_j$,其中 $\max_j\|a_j\|\leq D\|m\|$。特别地,$M$ 在代数意义下有限,且其给定范数与典范有限模范数一致等价。若 $K_r$ 的绝对值非平凡,则只需假设 $M_r$ 在 $A_r$ 上代数有限,所需常数 $C$ 自动存在。对赋值平凡的 $K_r$,仍保留所列提升假设。}
\end{theorem}
\begin{proof}
\bi{If $M_r=0$, the isometric constant embedding gives $M=0$. Otherwise let $Q:A_r^n\twoheadrightarrow M_r$ send the standard generators to $u_i$. Choose $\varepsilon>0$ with $\theta=C\varepsilon<1$. By the coefficient description, choose finite Laurent families $v_i$ with $\|u_i-v_i\|\leq\varepsilon$, and let $Q'$ be the generator map for the $v_i$. Its difference from $Q$ has norm at most $\varepsilon$.}
{若 $M_r=0$,则由常系数嵌入等距可知 $M=0$。否则,令 $Q:A_r^n\twoheadrightarrow M_r$ 将标准生成元送到 $u_i$。选择 $\varepsilon>0$,使 $\theta=C\varepsilon<1$。由系数描述,存在有限 Laurent 族 $v_i$,满足 $\|u_i-v_i\|\leq\varepsilon$。以 $Q'$ 表示 $v_i$ 的生成映射,则 $Q'-Q$ 的范数至多为 $\varepsilon$。}
\bi{The perturbation still has actual lifts with bound $C$. Indeed, given $y$, put $y_0=y$. Choose $b_k\in A_r^n$ with $Qb_k=y_k$ and $\|b_k\|\leq C\|y_k\|$, and set $y_{k+1}=y_k-Q'b_k$. Then}
{扰动后仍有界为 $C$ 的实际提升。事实上,给定 $y$,令 $y_0=y$。选取 $b_k\in A_r^n$,使 $Qb_k=y_k$ 且 $\|b_k\|\leq C\|y_k\|$,并令 $y_{k+1}=y_k-Q'b_k$。于是}
\[
\|y_{k+1}\|\leq\varepsilon\|b_k\|
\leq\theta\|y_k\|,
\qquad \|b_k\|\leq C\theta^k\|y\|.
\]
\bi{The series $b=\sum_{k\geq0}b_k$ converges in $A_r^n$, has norm at most $C\|y\|$, and satisfies $Q'b=y$ by the telescoping identity. No linear choice of lifts is needed.}
{级数 $b=\sum_{k\geq0}b_k$ 在 $A_r^n$ 中收敛,范数至多为 $C\|y\|$,并由望远镜恒等式满足 $Q'b=y$。这里不需要线性地选择提升。}
\bi{Write $v_i=\sum_{\nu\in F_i}\iota_M(m_{i\nu})T^\nu$ with finite $F_i$, omitting zero coefficients. The union of these lists is nonempty because $Q'$ surjects onto $M_r\ne0$. These finitely many coefficients will generate $M$. Thus $H=\max_{i,\nu\in F_i}r^\nu$ is finite and positive. For $m\in M$, choose a lift $b=(b_i)$ of $\iota_M(m)$ through $Q'$ with norm at most $C\|m\|$. Taking the zeroth coefficient gives, by Lemma~\ref{fnd:freecoefficientretraction},}
{写成 $v_i=\sum_{\nu\in F_i}\iota_M(m_{i\nu})T^\nu$,其中 $F_i$ 有限,并省去零系数。由于 $Q'$ 满射到 $M_r\ne0$,这些列表的并非空。这些有限个系数将生成 $M$,因此 $H=\max_{i,\nu\in F_i}r^\nu$ 有限且为正。对 $m\in M$,通过 $Q'$ 为 $\iota_M(m)$ 选择范数至多为 $C\|m\|$ 的提升 $b=(b_i)$。由 Lemma~\ref{fnd:freecoefficientretraction},取零次系数得到}
\[
m=\sum_{i,\nu\in F_i}p_0^A(b_iT^\nu)m_{i\nu},
\qquad
\|p_0^A(b_iT^\nu)\|
\leq r^\nu\|b_i\|\leq HC\|m\|.
\]
\bi{This proves finite generation with bound $D=HC$. The generator map from the corresponding finite free $A$-module is bounded, while this bound controls its quotient norm from above by the given norm. The two norms are therefore uniformly equivalent. Theorem~\ref{fnd:generalizedfinitemodules} identifies the quotient norm as canonical and complete.}
{这给出有限生成及界 $D=HC$。相应有限自由 $A$-模的生成映射有界,而上述提升界将其商范数从上方控制为给定范数的常数倍。因此两个范数一致等价。Theorem~\ref{fnd:generalizedfinitemodules} 说明该商范数为典范且完备的范数。}
\bi{Finally, if $K_r$ is nontrivially valued and $M_r$ is algebraically finite, choose algebraic generators. The resulting surjection $A_r^n\to M_r$ is bounded by the module inequality. Both sides are Banach $K_r$-spaces with their homogeneous norms by Lemma~\ref{fnd:freecoefficientretraction} and Theorem~\ref{fnd:freepolyradiuscoefficients}. Lemma~\ref{fnd:nonarchopenmapping}, applied over $K_r$, supplies the constant $C$. This use of the open-mapping theorem is confined to the stated nontrivial valuation case. The proof never replaced $M_r$ by $A_r\otimes_A M$ before proving that $M$ is finite.}
{最后,若 $K_r$ 的赋值非平凡且 $M_r$ 代数有限,则选取代数生成元。由模作用不等式,所得满射 $A_r^n\to M_r$ 有界。由 Lemma~\ref{fnd:freecoefficientretraction} 和 Theorem~\ref{fnd:freepolyradiuscoefficients},两端都是具有齐次范数的 Banach $K_r$-空间。在 $K_r$ 上应用 Lemma~\ref{fnd:nonarchopenmapping},即得常数 $C$。这个开映射定理的使用仅限于所述非平凡赋值情形。在证明 $M$ 有限之前,整个论证从未将 $M_r$ 替换为 $A_r\otimes_A M$。}
\end{proof}
Typeset reader excerpt

Typed proof inputs
- NAG-001480 · selected whole
- NAG-001480 · selected whole (requires)
- NAG-001549 · selected whole
- NAG-001549 · selected whole (requires)
- NAG-001559 · selected whole
- NAG-001559 · selected whole (requires)
- NAG-001561 · selected whole
- NAG-001561 · selected whole (requires)
Reader cross-references
Proof map
Verified map inputs · NAG-001567
AND · joint registered proof inputs
- NAG-001480 · Bounded lifting for Banach surjections · accepted whole
- NAG-001549 · The coefficient description of completed scalar extension · accepted whole
- NAG-001559 · Coefficient retractions and constant matrices · accepted whole
- NAG-001561 · Canonical norms and exactness for finite modules · accepted whole
NAG-001567 · Descending finite Banach generation · Chapter argument