Conjecture Atlas

Proof source

Forcing an entire completed divisor fiber into its ordinary product

Complete proof text in the unsigned reading edition.

Download sourceBrowse research records

Reading edition of the complete source note. Original labels and explicit equation tags are retained; references use source titles where compiled numbering is unavailable. Current result status is recorded separately in the project register.

Contents of this source
  1. Forcing an entire completed divisor fiber into its ordinary product

Forcing an entire completed divisor fiber into its ordinary product

Use the algebraic standard etale divisor chart of DT and the fixed surface preparation of UA. This section adds finitely many central coordinate functions. Their purpose is to recognize the selected product by inequalities on the entire moment fiber. No assumption on the moment rank of another point in that fiber is made.

Write T=(T1,T2)T=(T_1,T_2), let Θ\Theta be the selected etale coordinate on the original threefold chart, and let Θ0\Theta_0 be the surface coordinate. The fixed equations are Q(Z,T,Θ)=0,P(T,Θ0)=0,Q(0,T,Θ)=P(T,Θ)R0(T,Θ).(UF1)Q(Z,T,\Theta)=0,\qquad P(T,\Theta_0)=0, \qquad Q(0,T,\Theta)=P(T,\Theta)R_0(T,\Theta). \tag{UF1} All coefficients lie in R[T,G−1]R[T,G^{-1}], after including finitely many selected localizers. The functions TiT_i, Θ0\Theta_0, the relevant localizers, and the selected derivative and complementary factors have norm one at the surface point vv. This is the zero-slice chart used in DT. Passing to a smaller such chart or multiplying a coordinate by an original constant does not change the construction. In a degree-one or trivial etale presentation, if the selected Θ0\Theta_0 has zero residue, replace both Θ0\Theta_0 and Θ\Theta by their sum with one. This integral coordinate change makes their core norms one and preserves UF1 and its derivative conditions.

Rational coordinates which recognize a surface neighborhood

Lemma (A finite rational neighborhood packet). Let WW be an ordinary neighborhood of vv in this surface chart. Let h1,…,hsh_1,\ldots,h_s be finitely many actual surface units of norm one at vv. There are a smaller ordinary neighborhood W′⊂WW'\subset W, rational functions rjr_j with ∣hj∣=∣rj∣on W′,(UF2)|h_j|=|r_j|\quad\hbox{on }W', \tag{UF2} and finitely many pairs of integral polynomial functions Aa(T,Θ0),Ba(T,Θ0)A_a(T,\Theta_0),B_a(T,\Theta_0), all of norm one at vv, such that any surface point at which all the selected chart localizers have norm one and ∣Aa∣=∣Ba∣=1for every a(UF3)|A_a|=|B_a|=1\quad\hbox{for every }a \tag{UF3} belongs to W′W' and satisfies ∣hj∣=1|h_j|=1 for every jj. Denominators involving the fixed localizers are cleared in the displayed polynomials.

Proof. SA, or its single-function residue approximation step, replaces each hjh_j by a rational surface function with the same norm on one ordinary neighborhood of vv. Intersect these finitely many neighborhoods with WW. The topology of an affine algebraic analytification is generated by norms of its regular functions. Choose a finite basic algebraic open neighborhood inside that intersection. Include upper bounds for a finite affine coordinate-generating tuple if a fixed bounded chart is desired. Every nonzero regular function has nonzero value at the generic model point vv; its value is in ∣K×∣|K^\times|. Normalize it by an original constant to have norm one. Retaining its norm-one equality implies all the strict inequalities defining that basic open.

Include these finitely many normalized rational functions and the rjr_j. Each has norm one at the generic special-fiber point. It can be written as a quotient of two functions in a fixed smooth integral affine model ring with nonzero reductions. Indeed cancel the common π\pi-order of its numerator and denominator in the localization at that generic point, and add finitely many nonzero-reduction localizers. Clearing these localizers gives integral polynomial numerator and denominator representatives, each with norm one at vv. Retain both. Their norm-one equalities give the required rational-function equalities, placing the point in the chosen basic open and giving ∣rj∣=1|r_j|=1. Call that open W′W'. Equation UF2 proves the last assertion. ◻

For each polynomial in this packet, define its direct central lift by substituting Θ\Theta for Θ0\Theta_0, without changing its coefficients: A^a=Aa(T,Θ),B^a=Ba(T,Θ).(UF4)\widehat A_a=A_a(T,\Theta),\qquad \widehat B_a=B_a(T,\Theta). \tag{UF4} Their residues are nonzero along DD, so these are norm-one actual units on a common ordinary neighborhood of the original central point.

Lemma (The entire fiber has a selected zero-slice point). There is a finite actual central tuple JDJ_D containing ZZ, the chart coordinate and localizer functions, the direct lifts in UF4, and the selected polynomial derivative, with the following property. On a sufficiently outer collar, if an original point xx in their common ordinary chart satisfies tJD(x)=tJD(qD(r)),r<1,(UF5)t_{J_D}(x)=t_{J_D}(q_D(r)), \qquad r<1, \tag{UF5} then it belongs to the selected ordinary product tube and its zero-slice point x0x_0 belongs to W′W' and satisfies ∣hj(x0)∣=1|h_j(x_0)|=1 for all jj. This assertion includes arbitrary moment ranks and all commensurable radial values.

Proof. If P(T,Θ)P(T,\Theta) is not identically zero on the central chart, retain it as another actual unit, after original-constant normalization if necessary. It is nonzero at the central generic point, so its unit domain contains one ordinary central neighborhood. If it is identically zero, no coordinate is needed for it. Retain PΘ(T,Θ)P_\Theta(T,\Theta), and the finite zero-slice and chart localizers, in addition to UF4. Every norm-one function in this list is indeed norm one at qD(r)q_D(r): on the product, ∣Θ−Θ0∣<1|\Theta-\Theta_0|<1, and integral polynomial evaluation preserves every nonzero reduction. Also ∣P(T,Θ)(qD(r))∣<1,∣PΘ(T,Θ)(qD(r))∣=1.(UF6)|P(T,\Theta)(q_D(r))|<1, \qquad |P_\Theta(T,\Theta)(q_D(r))|=1. \tag{UF6} This is the selected DT Hensel comparison. The finite collar merely places all points and unit domains in the original central chart.

Equation UF5 gives exactly these inequalities at xx, together with ∣Z(x)∣=r<1|Z(x)|=r<1, ∣Ti(x)∣=∣Θ(x)∣=1|T_i(x)|=|\Theta(x)|=1, and norm one for every retained localizer. Coefficients of PP therefore have norm at most one. Hensel’s lemma in the complete field H(x)\mathcal H(x) gives the unique root θ0\theta_0 with ∣θ0−Θ(x)∣<1.(UF7)|\theta_0-\Theta(x)|<1. \tag{UF7} If P(T,Θ)P(T,\Theta) is identically zero, take θ0=Θ(x)\theta_0=\Theta(x). This root defines the surface point x0x_0; all selected localizers remain nonzero there.

Integral polynomial evaluation in UF7 gives ∣Aa(T,θ0)−A^a(x)∣<1,∣Ba(T,θ0)−B^a(x)∣<1.(UF8)|A_a(T,\theta_0)-\widehat A_a(x)|<1, \qquad |B_a(T,\theta_0)-\widehat B_a(x)|<1. \tag{UF8} Their central norms at qD(r)q_D(r) are one, so UF5 and UF8 give UF3 at x0x_0. The preceding lemma gives x0∈W′x_0\in W' and ∣hj(x0)∣=1|h_j(x_0)|=1.

Finally the polynomial QQ has a unique selected root within distance less than one of θ0\theta_0. Its Newton inequalities follow from UF1, ∣Z∣<1|Z|<1, the retained derivative and complementary factor norms, and the same integral coefficient bounds. The original Θ(x)\Theta(x) is that root. The values Z(x),T(x),θ0Z(x),T(x),\theta_0 in H(x)\mathcal H(x) specify a joint seminorm, and hence an H(x)\mathcal H(x)-valued point of the analytic product. Its two underlying factor points alone would not specify that product point. Locally these root constructions are analytic: continuity gives strict Newton margins, and iteration converges on an affinoid neighborhood. Root uniqueness glues the resulting inverse analytic map to the product. Composing with the selected product map fixes Z,TZ,T, and fixes Θ\Theta by the preceding unique-root argument, so the composition is the identity. Thus the conclusion is membership in an actual ordinary product domain. No graded-rank count has been used. ◻

The complete central image near the supplied ray

Take HH to be the fixed final protected surface tuple in the proof of UA. Its unit domain is fixed before the annular interval. For each interval II, the NSA construction is valid on the full inverse image of a target neighborhood of tH(v)t_H(v) in that domain. In particular it contains the entire central HH-fiber in any fixed smaller unit domain. This is the final GP complete-image conclusion of NSA, retained in UA’s fixed geometric preparation.

Choose WW inside that fixed unit domain, and apply the packet lemma to every coordinate of HH. Include direct central representatives of the rational functions rjr_j as well as their separate numerators and denominators in JDJ_D. Near every point of the fiber in UF5, their norms equal those of the surface HH-coordinates, by UF2 and the strict polynomial estimates UF8.

Theorem (Completed core-ray fiber localization). For the fixed divisor and fixed central finite-PS tensor in UA, there is one finite central refinement FF, and a protected ordinary central domain, such that for every compact interval II in one sufficiently outer collar the following hold.

  1. The entire FF-moment fiber over every tF(qD(r))t_F(q_D(r)), r∈Ir\in I, is contained in an ordinary domain carrying the transported NSA covector. The same assertion holds for every actual source in these fibers, including points whose earlier moment rank is only two or less.

  2. There is an open target neighborhood OI\mathcal O_I of {tF(qD(r)):r∈I}\{t_F(q_D(r)):r\in I\} whose complete ordinary inverse image lies in that domain. On its complete image the transported finite-PS covector has the full all-face strain identity. It has the fixed-control approximation and all-background conclusions of UA for this interval.

One finite refinement works for a supplied finite list of divisors and their respective core rays. The statement concerns neighborhoods of those completed core rays; it does not assert that they include every new ray which projects to an earlier ray.

Proof. Retain JDJ_D, all original coefficient coordinates, and the finite rational representatives just specified. Apply one central GP on their common original ordinary domain. Call the completed tuple FF and its compact protected closure D0D_0. The protecting coordinates are constant on an ordinary neighborhood of the original central point, so every sufficiently short core ray is retained unchanged. All core values in a sufficiently outer collar lie inside the open protected target box.

Let xx be any point in the entire FF-fiber over a core value. Projection to JDJ_D gives UF5. The preceding lemma places its zero-slice point in the fixed surface unit domain and in the entire central HH-fiber. UA’s fixed-control lemma puts that full protected fiber inside the valid AF product for the chosen radial interval. The final NSA primitive is valid there on the complete image. The direct central lifts and UF8 transfer it on an ordinary neighborhood of xx. The surface covector is a single ambient formula ΛI\Lambda_I; these transferred formulas therefore agree wherever their domains overlap. Take their union as the ordinary domain in the first assertion. This argument has not removed a source according to its rank.

The inverse image in D0D_0 of the compact set of core values for II is compact and, by the first assertion, is contained in that ordinary domain. The closed complement in D0D_0 has compact moment image, disjoint from those core values. An open target neighborhood avoiding this compact image and contained in the original protected box is the required OI\mathcal O_I. Protection ensures that its full inverse image is computed inside D0D_0, and hence lies in the stated ordinary domain.

The coefficient identity was transported on every face, so it holds on this entire inverse image and on every later retaining refinement, including rank degeneration. Apply the same strain smoothing argument as in UA on a smaller protected truncation. For finitely many supplied divisors, take the union of the finite packets before the final central GP and the maximum of their finitely many radial thresholds. Projection to each packet proves its fiber assertion separately. No further iteration of surface preparation is required for this finite list. ◻

Source text and edition identity

This reading copy omits personal attribution and places existing mathematical question marks over their relation symbols. The mathematical content and source status are retained.

Reading-copy SHA-256 ffe79a597ce33f1bda49807d73a79382931cdac5c5fd405291ef6a3284d83481

Frozen-original SHA-256 ffe79a597ce33f1bda49807d73a79382931cdac5c5fd405291ef6a3284d83481

\section{Forcing an entire completed divisor fiber into its ordinary product}
\label{sec:r13-divisor-fiber-product-localization}

Use the algebraic standard etale divisor chart of DT and the
fixed surface preparation of UA.  This section adds finitely
many central coordinate functions.  Their purpose is to
recognize the selected product by inequalities on the entire
moment fiber.  No assumption on the moment rank of another
point in that fiber is made.

Write \(T=(T_1,T_2)\), let \(\Theta\) be the selected etale
coordinate on the original threefold chart, and let
\(\Theta_0\) be the surface coordinate.  The fixed equations
are
\[
 Q(Z,T,\Theta)=0,\qquad P(T,\Theta_0)=0,
 \qquad Q(0,T,\Theta)=P(T,\Theta)R_0(T,\Theta).
\tag{UF1}
\]
All coefficients lie in \(R[T,G^{-1}]\), after including
finitely many selected localizers.  The functions \(T_i\),
\(\Theta_0\), the relevant localizers, and the selected
derivative and complementary factors have norm one at the
surface point \(v\).  This is the zero-slice chart used in
DT.  Passing to a smaller such chart or multiplying a
coordinate by an original constant does not change the
construction.  In a degree-one or trivial etale presentation,
if the selected \(\Theta_0\) has zero residue, replace both
\(\Theta_0\) and \(\Theta\) by their sum with one.  This
integral coordinate change makes their core norms one and
preserves UF1 and its derivative conditions.

\subsection{Rational coordinates which recognize a surface neighborhood}

\begin{lemma}[A finite rational neighborhood packet]
\label{lem:r13-rational-neighborhood-packet}
Let \(W\) be an ordinary neighborhood of \(v\) in this
surface chart.  Let \(h_1,\ldots,h_s\) be finitely many
actual surface units of norm one at \(v\).  There are a
smaller ordinary neighborhood \(W'\subset W\), rational
functions \(r_j\) with
\[
                         |h_j|=|r_j|\quad\hbox{on }W',
\tag{UF2}
\]
and finitely many pairs of integral polynomial functions
\(A_a(T,\Theta_0),B_a(T,\Theta_0)\), all of norm one
at \(v\), such that any surface point at which all the
selected chart localizers have norm one and
\[
                       |A_a|=|B_a|=1\quad\hbox{for every }a
\tag{UF3}
\]
belongs to \(W'\) and satisfies \(|h_j|=1\) for every
\(j\).  Denominators involving the fixed localizers are
cleared in the displayed polynomials.
\end{lemma}

\begin{proof}
SA, or its single-function residue approximation step,
replaces each \(h_j\) by a rational surface function
with the same norm on one ordinary neighborhood of
\(v\).  Intersect these finitely many neighborhoods
with \(W\).  The topology of an affine algebraic
analytification is generated by norms of its regular
functions.  Choose a finite basic algebraic open
neighborhood inside that intersection.  Include upper
bounds for a finite affine coordinate-generating tuple
if a fixed bounded chart is desired.  Every nonzero
regular function has nonzero value at the generic
model point \(v\); its value is in \(|K^\times|\).
Normalize it by an original constant to have norm
one.  Retaining its norm-one equality implies all
the strict inequalities defining that basic open.

Include these finitely many normalized rational
functions and the \(r_j\).  Each has norm one at
the generic special-fiber point.  It can be written
as a quotient of two functions in a fixed smooth
integral affine model ring with nonzero reductions.
Indeed cancel the common \(\pi\)-order of its
numerator and denominator in the localization at
that generic point, and add finitely many
nonzero-reduction localizers.  Clearing these
localizers gives integral polynomial numerator
and denominator representatives, each with norm
one at \(v\).  Retain both.  Their norm-one
equalities give the required rational-function
equalities, placing the point in the chosen basic
open and giving \(|r_j|=1\).  Call that open
\(W'\).  Equation UF2 proves the last assertion.
\end{proof}

For each polynomial in this packet, define its direct
central lift by substituting \(\Theta\) for
\(\Theta_0\), without changing its coefficients:
\[
       \widehat A_a=A_a(T,\Theta),\qquad
       \widehat B_a=B_a(T,\Theta).
\tag{UF4}
\]
Their residues are nonzero along \(D\), so these
are norm-one actual units on a common ordinary
neighborhood of the original central point.

\begin{lemma}[The entire fiber has a selected zero-slice point]
\label{lem:r13-entire-fiber-zero-slice}
There is a finite actual central tuple \(J_D\)
containing \(Z\), the chart coordinate and localizer
functions, the direct lifts in UF4, and the selected
polynomial derivative, with the following property.
On a sufficiently outer collar, if an original point
\(x\) in their common ordinary chart satisfies
\[
                    t_{J_D}(x)=t_{J_D}(q_D(r)),
                             \qquad r<1,
\tag{UF5}
\]
then it belongs to the selected ordinary product
tube and its zero-slice point \(x_0\) belongs to
\(W'\) and satisfies \(|h_j(x_0)|=1\) for all \(j\).
This assertion includes arbitrary moment ranks and
all commensurable radial values.
\end{lemma}

\begin{proof}
If \(P(T,\Theta)\) is not identically zero on the
central chart, retain it as another actual unit,
after original-constant normalization if necessary.
It is nonzero at the central generic point, so
its unit domain contains one ordinary central
neighborhood.  If it is identically zero, no
coordinate is needed for it.  Retain
\(P_\Theta(T,\Theta)\), and the finite zero-slice
and chart localizers, in addition to UF4.  Every
norm-one function in this list is indeed norm
one at \(q_D(r)\): on the product,
\(|\Theta-\Theta_0|<1\), and integral polynomial
evaluation preserves every nonzero reduction.
Also
\[
 |P(T,\Theta)(q_D(r))|<1,
 \qquad |P_\Theta(T,\Theta)(q_D(r))|=1.
\tag{UF6}
\]
This is the selected DT Hensel comparison.  The
finite collar merely places all points and unit
domains in the original central chart.

Equation UF5 gives exactly these inequalities
at \(x\), together with \(|Z(x)|=r<1\),
\(|T_i(x)|=|\Theta(x)|=1\), and norm one for
every retained localizer.  Coefficients of \(P\)
therefore have norm at most one.  Hensel's lemma
in the complete field \(\mathcal H(x)\) gives
the unique root \(\theta_0\) with
\[
                       |\theta_0-\Theta(x)|<1.
\tag{UF7}
\]
If \(P(T,\Theta)\) is identically zero, take
\(\theta_0=\Theta(x)\).  This root defines
the surface point \(x_0\); all selected
localizers remain nonzero there.

Integral polynomial evaluation in UF7 gives
\[
 |A_a(T,\theta_0)-\widehat A_a(x)|<1,
 \qquad
 |B_a(T,\theta_0)-\widehat B_a(x)|<1.
\tag{UF8}
\]
Their central norms at \(q_D(r)\) are one,
so UF5 and UF8 give UF3 at \(x_0\).  The
preceding lemma gives \(x_0\in W'\) and
\(|h_j(x_0)|=1\).

Finally the polynomial \(Q\) has a unique
selected root within distance less than one
of \(\theta_0\).  Its Newton inequalities
follow from UF1, \(|Z|<1\), the retained
derivative and complementary factor norms,
and the same integral coefficient bounds.
The original \(\Theta(x)\) is that root.
The values \(Z(x),T(x),\theta_0\) in
\(\mathcal H(x)\) specify a joint seminorm,
and hence an \(\mathcal H(x)\)-valued
point of the analytic product.  Its two
underlying factor points alone would not
specify that product point.  Locally these
root constructions are analytic: continuity
gives strict Newton margins, and iteration
converges on an affinoid neighborhood.  Root
uniqueness glues the resulting inverse
analytic map to the product.  Composing
with the selected product map fixes
\(Z,T\), and fixes \(\Theta\) by the
preceding unique-root argument, so the
composition is the identity.  Thus
the conclusion is membership in an actual
ordinary product domain.  No graded-rank
count has been used.
\end{proof}

\subsection{The complete central image near the supplied ray}

Take \(H\) to be the fixed final protected surface
tuple in the proof of UA.  Its unit domain is fixed
before the annular interval.  For each interval
\(I\), the NSA construction is valid on the full
inverse image of a target neighborhood of
\(t_H(v)\) in that domain.  In particular it
contains the entire central \(H\)-fiber in any
fixed smaller unit domain.  This is the final
GP complete-image conclusion of NSA, retained
in UA's fixed geometric preparation.

Choose \(W\) inside that fixed unit domain,
and apply the packet lemma to every coordinate
of \(H\).  Include direct central representatives
of the rational functions \(r_j\) as well as
their separate numerators and denominators in
\(J_D\).  Near every point of the fiber in
UF5, their norms equal those of the surface
\(H\)-coordinates, by UF2 and the strict
polynomial estimates UF8.

\begin{theorem}[Completed core-ray fiber localization]
\label{thm:r13-completed-ray-fiber-localization}
For the fixed divisor and fixed central finite-PS
tensor in UA, there is one finite central
refinement \(F\), and a protected ordinary
central domain, such that for every compact
interval \(I\) in one sufficiently outer
collar the following hold.
\begin{enumerate}
\item The entire \(F\)-moment fiber over every
\(t_F(q_D(r))\), \(r\in I\), is contained
in an ordinary domain carrying the transported
NSA covector.  The same assertion holds for
every actual source in these fibers, including
points whose earlier moment rank is only two
or less.
\item There is an open target neighborhood
\(\mathcal O_I\) of
\(\{t_F(q_D(r)):r\in I\}\) whose complete
ordinary inverse image lies in that domain.
On its complete image the transported finite-PS
covector has the full all-face strain identity.
It has the fixed-control approximation and
all-background conclusions of UA for this
interval.
\end{enumerate}
One finite refinement works for a supplied
finite list of divisors and their respective
core rays.  The statement concerns neighborhoods
of those completed core rays; it does not
assert that they include every new ray which
projects to an earlier ray.
\end{theorem}

\begin{proof}
Retain \(J_D\), all original coefficient
coordinates, and the finite rational
representatives just specified.  Apply one
central GP on their common original ordinary
domain.  Call the completed tuple \(F\)
and its compact protected closure \(D_0\).
The protecting coordinates are constant on
an ordinary neighborhood of the original
central point, so every sufficiently short
core ray is retained unchanged.  All core
values in a sufficiently outer collar lie
inside the open protected target box.

Let \(x\) be any point in the entire
\(F\)-fiber over a core value.  Projection
to \(J_D\) gives UF5.  The preceding lemma
places its zero-slice point in the fixed
surface unit domain and in the entire
central \(H\)-fiber.  UA's fixed-control
lemma puts that full protected fiber inside
the valid AF product for the chosen radial
interval.  The final NSA primitive is valid
there on the complete image.  The direct
central lifts and UF8 transfer it on an
ordinary neighborhood of \(x\).  The
surface covector is a single ambient
formula \(\Lambda_I\); these transferred
formulas therefore agree wherever their
domains overlap.  Take their union as the
ordinary domain in the first assertion.
This argument has not removed a source
according to its rank.

The inverse image in \(D_0\) of the
compact set of core values for \(I\)
is compact and, by the first assertion,
is contained in that ordinary domain.
The closed complement in \(D_0\) has
compact moment image, disjoint from those
core values.  An open target neighborhood
avoiding this compact image and contained
in the original protected box is the
required \(\mathcal O_I\).  Protection
ensures that its full inverse image is
computed inside \(D_0\), and hence lies
in the stated ordinary domain.

The coefficient identity was transported
on every face, so it holds on this entire
inverse image and on every later retaining
refinement, including rank degeneration.
Apply the same strain smoothing argument
as in UA on a smaller protected truncation.
For finitely many supplied divisors, take
the union of the finite packets before
the final central GP and the maximum of
their finitely many radial thresholds.
Projection to each packet proves its
fiber assertion separately.  No further
iteration of surface preparation is
required for this finite list.
\end{proof}