Conjecture Atlas

Proof source

An irreducible selector whose norms do not extend across every boundary

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. An irreducible selector whose norms do not extend across every boundary

An irreducible selector whose norms do not extend across every boundary

Let KK be the original finite extension of Qp\mathbf Q_p, with its original absolute value, and let kk be its residue field. Work on the original three-dimensional torus with coordinates (A,C,S)(A,C,S). Let yy be its unit-radius Gauss point, so its residue field is k(a,c,s)k(a,c,s). Put P(V)=V2−V−A,Z=C2−C−A,F(E)=P(C+E)=E2+(2C−1)E+Z.(HB1)P(V)=V^2-V-A,\qquad Z=C^2-C-A, \qquad F(E)=P(C+E) =E^2+(2C-1)E+Z. \tag{HB1} The zero-slice divisor D={c2−c−a=0}D=\{c^2-c-a=0\} in the smooth residue model is smooth and irreducible. Indeed its derivative with respect to aa is −1-1. Moreover V2−V−aV^2-V-a is irreducible over k(a,s)k(a,s): a rational root, viewed as a rational function of aa over k(s)k(s), can have no finite pole, since a pole of order mm would give one of order 2m2m in its square minus itself. It is therefore a polynomial in aa, whose square minus itself has even positive degree, or degree zero, and cannot equal aa. The polynomial is separable in every residue characteristic. Thus HB1 is an actual irreducible etale zero-slice presentation at the generic point of DD; its derivative 2c−12c-1 there is nonzero.

Let HF\mathcal H_F be the strict ordinary Hensel domain HN3, and let α=C+e\alpha=C+e be its selected root. Since AA is a unit and the product of the two roots of PP is −A-A, α\alpha is an analytic unit on this domain. At the generic zero-slice point its residue is the nonzero element c∈k(D)c\in k(D), so this is one of the allowed finite surface units.

Proposition (Two ordinary branches of the selected norm). The original ordinary open domains B0={0<∣A∣<1, 0<∣C∣<1},B1={0<∣A∣<1, ∣C−1∣<1}(HB2)\mathcal B_0=\{0<|A|<1,\ 0<|C|<1\},\qquad \mathcal B_1=\{0<|A|<1,\ |C-1|<1\} \tag{HB2} are contained in HF\mathcal H_F. At every actual point of these domains one has ∣α∣=∣A∣on B0,∣α∣=1on B1.(HB3)|\alpha|=|A|\quad\hbox{on }\mathcal B_0, \qquad |\alpha|=1\quad\hbox{on }\mathcal B_1. \tag{HB3} No extension of the ground field or restriction to a moment source locus occurs in these identities.

Proof. On B0\mathcal B_0 one has ∣2C−1∣=1|2C-1|=1 and ∣Z∣<1|Z|<1. On B1\mathcal B_1, writing C=1+WC=1+W, one has ∣2C−1∣=∣1+2W∣=1|2C-1|=|1+2W|=1 and ∣Z∣=∣W+W2−A∣<1|Z|=|W+W^2-A|<1. Thus HN3 holds on both ordinary sets, in every residue characteristic. On the first set, ∣C∣<1|C|<1 and ∣e∣<1|e|<1 imply ∣α∣<1|\alpha|<1. The equation α2−α=A\alpha^2-\alpha=A then gives ∣α∣=∣A∣|\alpha|=|A|. On the second set, ∣α−1∣=∣C−1+e∣<1|\alpha-1|=|C-1+e|<1, so ∣α∣=1|\alpha|=1. ◻

For 0<r<10<r<1, let qrq_r be the actual Gauss point with radii (r,1,1)(r,1,1) in (A,C,S)(A,C,S). For 0<s<10<s<1 let x0(r,s)x_0(r,s) be the Gauss point with radii (r,s,1)(r,s,1) in (A,C,S)(A,C,S), and let x1(r,s)x_1(r,s) be the Gauss point with these same radii in (A,C−1,S)(A,C-1,S). These are points of the original analytic space for all real r,sr,s in the displayed intervals, including commensurable radii. The Gauss maximum formula gives lim⁡s→1−x0(r,s)=qr=lim⁡s→1−x1(r,s),lim⁡r→1−qr=y.(HB4)\lim_{s\to1^-}x_0(r,s)=q_r =\lim_{s\to1^-}x_1(r,s), \qquad \lim_{r\to1^-}q_r=y. \tag{HB4} For the first equality, the closed unit discs in the CC-variable centered at zero and at one have the same Gauss norm; the two formulas agree on every polynomial. Uniform approximation on a fixed affinoid neighborhood extends this convergence to all analytic functions. At qrq_r, the two summands of ZZ of degree one and two in CC have independent Gauss monomials, so ∣Z∣=1|Z|=1, and HN3 fails there.

Theorem (Failure of an entire central continuous extension). Let UU be any original ordinary neighborhood of yy, and let GG be any finite tuple of actual analytic units on UU. There is no continuous function Ψ:tG(U)⟶R(HB5)\Psi:t_G(U)\longrightarrow\mathbf R \tag{HB5} such that Ψ(tG(x))=−log⁡∣α(x)∣(x∈U∩HF).(HB6)\Psi(t_G(x))=-\log|\alpha(x)| \qquad(x\in U\cap\mathcal H_F). \tag{HB6} In particular no finite actual refinement supplies a continuous piecewise affine extension on the entire central moment image for this selected norm.

Proof. Choose r<1r<1 sufficiently close to one that qr∈Uq_r\in U. By HB4, both xi(r,s)x_i(r,s) belong to UU for all ss sufficiently close to one. They lie in Bi\mathcal B_i, respectively. The actual moment map is continuous, so both moment families converge to tG(qr)t_G(q_r). Equations HB3 and HB6 would give the two different limiting values −log⁡r>0-\log r>0 and zero for the continuous function HB5 at this same point. This is impossible. The argument applies to every finite tuple GG, not only to the coordinates used to exhibit HB4. ◻

All coefficients of PP have norm at most one at yy, and ∣Res⁡(P,V)(y)∣=∣−A(y)∣=1.(HB7)|\operatorname{Res}(P,V)(y)| =|-A(y)|=1. \tag{HB7} Thus the all-conjugate resultant hypotheses of HN14 hold. The selected logarithm nevertheless lacks the entire extension in HB5. There is no contradiction: its value tends to zero along every approach to yy in the selected domain, whereas HB4 gives two different limits at each nearby excluded point qrq_r. Protected continuity on the realized selected domain, and continuity after adjoining the single original center, are the precise surviving statements. This example refutes the stronger extension shortcut; it does not obstruct an intrinsic differential construction using actual local root domains or a separately justified singular pairing.

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 970def090f6de38e3d639f48d0ed5cced1ed56fd0d6cbc814a0f3e72480fd87e

Frozen-original SHA-256 970def090f6de38e3d639f48d0ed5cced1ed56fd0d6cbc814a0f3e72480fd87e

\section{An irreducible selector whose norms do not extend across every boundary}
\label{sec:r13-hensel-boundary-nonextension}

Let \(K\) be the original finite extension of \(\mathbf Q_p\),
with its original absolute value, and let \(k\) be its residue
field.  Work on the original three-dimensional torus with
coordinates \((A,C,S)\).  Let \(y\) be its unit-radius Gauss
point, so its residue field is \(k(a,c,s)\).  Put
\[
 P(V)=V^2-V-A,\qquad Z=C^2-C-A,
 \qquad F(E)=P(C+E)
       =E^2+(2C-1)E+Z.
\tag{HB1}
\]
The zero-slice divisor \(D=\{c^2-c-a=0\}\) in the smooth
residue model is smooth and irreducible.  Indeed its derivative
with respect to \(a\) is \(-1\).  Moreover \(V^2-V-a\) is
irreducible over \(k(a,s)\): a rational root, viewed as a
rational function of \(a\) over \(k(s)\), can have no finite
pole, since a pole of order \(m\) would give one of order
\(2m\) in its square minus itself.  It is therefore a
polynomial in \(a\), whose square minus itself has even
positive degree, or degree zero, and cannot equal \(a\).
The polynomial is separable in every residue characteristic.
Thus HB1 is an actual irreducible etale zero-slice presentation
at the generic point of \(D\); its derivative \(2c-1\) there
is nonzero.

Let \(\mathcal H_F\) be the strict ordinary Hensel domain
HN3, and let
\(\alpha=C+e\) be its selected root.  Since \(A\) is a
unit and the product of the two roots of \(P\) is \(-A\),
\(\alpha\) is an analytic unit on this domain.  At the
generic zero-slice point its residue is the nonzero element
\(c\in k(D)\), so this is one of the allowed finite
surface units.

\begin{proposition}[Two ordinary branches of the selected norm]
\label{prop:r13-hensel-two-ordinary-branches}
The original ordinary open domains
\[
 \mathcal B_0=\{0<|A|<1,\ 0<|C|<1\},\qquad
 \mathcal B_1=\{0<|A|<1,\ |C-1|<1\}
\tag{HB2}
\]
are contained in \(\mathcal H_F\).  At every actual point
of these domains one has
\[
             |\alpha|=|A|\quad\hbox{on }\mathcal B_0,
       \qquad |\alpha|=1\quad\hbox{on }\mathcal B_1.
\tag{HB3}
\]
No extension of the ground field or restriction to a moment
source locus occurs in these identities.
\end{proposition}

\begin{proof}
On \(\mathcal B_0\) one has \(|2C-1|=1\) and
\(|Z|<1\).  On \(\mathcal B_1\), writing \(C=1+W\),
one has \(|2C-1|=|1+2W|=1\) and
\(|Z|=|W+W^2-A|<1\).  Thus HN3 holds on both ordinary
sets, in every residue characteristic.  On the first set,
\(|C|<1\) and \(|e|<1\) imply \(|\alpha|<1\).
The equation \(\alpha^2-\alpha=A\) then gives
\(|\alpha|=|A|\).  On the second set,
\(|\alpha-1|=|C-1+e|<1\), so \(|\alpha|=1\).
\end{proof}

For \(0<r<1\), let \(q_r\) be the actual Gauss point
with radii \((r,1,1)\) in \((A,C,S)\).  For \(0<s<1\)
let \(x_0(r,s)\) be the Gauss point with radii
\((r,s,1)\) in \((A,C,S)\), and let \(x_1(r,s)\) be
the Gauss point with these same radii in \((A,C-1,S)\).
These are points of the original analytic space for all
real \(r,s\) in the displayed intervals, including
commensurable radii.  The Gauss maximum formula gives
\[
 \lim_{s\to1^-}x_0(r,s)=q_r
       =\lim_{s\to1^-}x_1(r,s),
       \qquad \lim_{r\to1^-}q_r=y.
\tag{HB4}
\]
For the first equality, the closed unit discs in the
\(C\)-variable centered at zero and at one have the
same Gauss norm; the two formulas agree on every
polynomial.  Uniform approximation on a fixed affinoid
neighborhood extends this convergence to all analytic
functions.  At \(q_r\), the two summands of \(Z\) of
degree one and two in \(C\) have independent Gauss
monomials, so \(|Z|=1\), and HN3 fails there.

\begin{theorem}[Failure of an entire central continuous extension]
\label{thm:r13-hensel-no-entire-central-continuous-extension}
Let \(U\) be any original ordinary neighborhood of
\(y\), and let \(G\) be any finite tuple of actual
analytic units on \(U\).  There is no continuous
function
\[
                     \Psi:t_G(U)\longrightarrow\mathbf R
\tag{HB5}
\]
such that
\[
             \Psi(t_G(x))=-\log|\alpha(x)|
                     \qquad(x\in U\cap\mathcal H_F).
\tag{HB6}
\]
In particular no finite actual refinement supplies a
continuous piecewise affine extension on the entire
central moment image for this selected norm.
\end{theorem}

\begin{proof}
Choose \(r<1\) sufficiently close to one that
\(q_r\in U\).  By HB4, both \(x_i(r,s)\) belong
to \(U\) for all \(s\) sufficiently close to one.
They lie in \(\mathcal B_i\), respectively.  The
actual moment map is continuous, so both moment
families converge to \(t_G(q_r)\).  Equations
HB3 and HB6 would give the two different limiting
values \(-\log r>0\) and zero for the continuous
function HB5 at this same point.  This is impossible.
The argument applies to every finite tuple \(G\),
not only to the coordinates used to exhibit HB4.
\end{proof}

All coefficients of \(P\) have norm at most one
at \(y\), and
\[
                         |\operatorname{Res}(P,V)(y)|
                              =|-A(y)|=1.
\tag{HB7}
\]
Thus the all-conjugate resultant hypotheses of HN14
hold.  The selected logarithm nevertheless lacks the
entire extension in HB5.  There is no contradiction:
its value tends to zero along every approach to
\(y\) in the selected domain, whereas HB4 gives
two different limits at each nearby excluded point
\(q_r\).  Protected continuity on the realized
selected domain, and continuity after adjoining the
single original center, are the precise surviving
statements.  This example refutes the stronger
extension shortcut; it does not obstruct an intrinsic
differential construction using actual local root
domains or a separately justified singular pairing.