OpenMath

Proof source

A residue-field section is stronger than finite unit lifting

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. A residue-field section is stronger than finite unit lifting

A residue-field section is stronger than finite unit lifting

Let kk be the original finite residue field. Finite surjectivity from a residue local ring, used in ML and UA, does not supply a field section preserving all algebraic relations. The following example stays over kk.

Proposition (A smooth proper divisor without a rational coefficient-field section). There are a smooth projective rational threefold V/kV/k and a smooth proper integral divisor D⊂VD\subset V such that the residue map OV,ηD⟶k(D)(RE1)\mathcal O_{V,\eta_D}\longrightarrow k(D) \tag{RE1} has no kk-algebra section defined on the whole field k(D)k(D).

Proof. Choose the smooth plane cubic E⊂Pk2E\subset\mathbf P^2_k defined by {Y2Z=X3−XZ2,char⁡k≠2,Y2Z+YZ2=X3,char⁡k=2.(RE2)\begin{cases} Y^2Z=X^3-XZ^2,&\operatorname{char}k\ne2,\\ Y^2Z+YZ^2=X^3,&\operatorname{char}k=2. \end{cases} \tag{RE2} In odd characteristic its affine possible singular points have Y=0Y=0 and X∈{0,1,−1}X\in\{0,1,-1\}; the derivative in XX is respectively 1,−2,−21,-2,-2. The point at infinity is smooth because its derivative in ZZ is nonzero. In characteristic two the three partial derivatives are X2,Z2,Y2X^2,Z^2,Y^2, with no common projective zero. Thus EE is smooth in either case, and a smooth plane cubic has genus one.

Put V=Pk2×Pk1,D=E×Pk1.(RE3)V=\mathbf P^2_k\times\mathbf P^1_k, \qquad D=E\times\mathbf P^1_k. \tag{RE3} These are smooth and projective, and DD is a Cartier divisor. Their function fields are k(V)=k(x,y,s)k(V)=k(x,y,s) and k(D)=k(E)(s)k(D)=k(E)(s). If RE1 had a field section, its composition with the inclusion in k(V)k(V), restricted to k(E)k(E), would give a kk-embedding k(E)↪k(x,y,s).(RE4)k(E)\hookrightarrow k(x,y,s). \tag{RE4} It would define a dominant rational map Pk3⇢E\mathbf P^3_k\dashrightarrow E.

Base change to an algebraic closure. Choose two points in its open domain having distinct images, and take the projective line through them. The restricted rational map on that line is nonconstant. It induces an inclusion kˉ(E)⊂kˉ(t)\bar k(E)\subset\bar k(t). The field-theoretic Lüroth theorem, valid in every characteristic, makes this intermediate field rational. This contradicts the genus-one property. Therefore RE4, and hence the proposed section, cannot exist. ◻

The proposition refutes the stronger prescription of choosing all rational residue lifts as one relation-preserving coefficient-field section. Every fixed finite list still lifts individually through RE1, and the ordinary product and finite strict-comparison arguments of CC, UA and UF are unaffected.

Corollary (Two units and one relation need not lift exactly). Let xE,yE∈k(E)×⊂k(D)×x_E,y_E\in k(E)^\times\subset k(D)^\times be the affine coordinates of RE2. There are no units X,Y∈OV,ηD×X,Y\in\mathcal O_{V,\eta_D}^\times reducing to xE,yEx_E,y_E and satisfying the corresponding equation {Y2=X3−X,char⁡k≠2,Y2+Y=X3,char⁡k=2.(RE5)\begin{cases} Y^2=X^3-X,&\operatorname{char}k\ne2,\\ Y^2+Y=X^3,&\operatorname{char}k=2. \end{cases} \tag{RE5}

Proof. Such units would define a homomorphism from the integral affine coordinate ring of EE to k(V)k(V). Its kernel is prime. A nonzero prime in this one-dimensional affine domain is maximal and has residue field algebraic over kk. This is incompatible with the transcendental residues xE,yEx_E,y_E. Hence the homomorphism is injective and extends to k(E)↪k(V)k(E)\hookrightarrow k(V), contradicting the proof above. ◻

The estimate left outside the core fiber.

The direct polynomial comparison in UF has ∣A^−A0∣<1|\widehat A-A_0|<1. In its completed core fiber ∣A0∣=1|A_0|=1, which gives the needed relative strict inequality. On a different ray a chosen surface numerator can instead have order a>0a>0, while the first normal error has order m>0m>0. The comparison then requires a<ma<m; it need not follow when a≥ma\ge m. A finite Taylor cancellation alone supplies only a finite error order; absent an additional divisibility or relative bound, it does not control arbitrary later rays. The proposition explains why a universal rational coefficient section cannot be assumed to remove all such errors. It is not a counterexample to finite controlled correction or to intrinsic descent.

Classical input checked.

The all-characteristic algebraic form of Lüroth’s theorem was checked in M. Mustaţă, Lecture 1. Definitions and first examples, Theorem 2.1 and its algebraic proof on pp. 4–6, https://public.websites.umich.edu/~mmustata/lecture1_rationality.pdf, accessed 2026-11-15. Only that classical field theorem is imported; the divisor-section obstruction above is the displayed deduction.

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 e4887a70bff5c5c2ece9371c0128ee2bc814fa29537edc59571295dc4a40a62e

Frozen-original SHA-256 e4887a70bff5c5c2ece9371c0128ee2bc814fa29537edc59571295dc4a40a62e

\section{A residue-field section is stronger than finite unit lifting}
\label{sec:r13-residue-field-section-obstruction}

Let \(k\) be the original finite residue field.  Finite
surjectivity from a residue local ring, used in ML and UA,
does not supply a field section preserving all algebraic
relations.  The following example stays over \(k\).

\begin{proposition}[A smooth proper divisor without a rational coefficient-field section]
\label{prop:r13-no-rational-residue-field-section}
There are a smooth projective rational threefold \(V/k\)
and a smooth proper integral divisor \(D\subset V\) such
that the residue map
\[
             \mathcal O_{V,\eta_D}\longrightarrow k(D)
\tag{RE1}
\]
has no \(k\)-algebra section defined on the whole field
\(k(D)\).
\end{proposition}

\begin{proof}
Choose the smooth plane cubic \(E\subset\mathbf P^2_k\)
defined by
\[
 \begin{cases}
 Y^2Z=X^3-XZ^2,&\operatorname{char}k\ne2,\\
 Y^2Z+YZ^2=X^3,&\operatorname{char}k=2.
 \end{cases}
\tag{RE2}
\]
In odd characteristic its affine possible singular
points have \(Y=0\) and \(X\in\{0,1,-1\}\);
the derivative in \(X\) is respectively \(1,-2,-2\).
The point at infinity is smooth because its derivative
in \(Z\) is nonzero.  In characteristic two the three
partial derivatives are \(X^2,Z^2,Y^2\), with no
common projective zero.  Thus \(E\) is smooth in
either case, and a smooth plane cubic has genus one.

Put
\[
 V=\mathbf P^2_k\times\mathbf P^1_k,
 \qquad D=E\times\mathbf P^1_k.
\tag{RE3}
\]
These are smooth and projective, and \(D\) is a
Cartier divisor.  Their function fields are
\(k(V)=k(x,y,s)\) and \(k(D)=k(E)(s)\).
If RE1 had a field section, its composition with
the inclusion in \(k(V)\), restricted to \(k(E)\),
would give a \(k\)-embedding
\[
                         k(E)\hookrightarrow k(x,y,s).
\tag{RE4}
\]
It would define a dominant rational map
\(\mathbf P^3_k\dashrightarrow E\).

Base change to an algebraic closure.  Choose two
points in its open domain having distinct images,
and take the projective line through them.  The
restricted rational map on that line is nonconstant.
It induces an inclusion
\(\bar k(E)\subset\bar k(t)\).  The field-theoretic
L\"uroth theorem, valid in every characteristic,
makes this intermediate field rational.  This
contradicts the genus-one property.  Therefore
RE4, and hence the proposed section, cannot exist.
\end{proof}

The proposition refutes the stronger prescription
of choosing all rational residue lifts as one
relation-preserving coefficient-field section.
Every fixed finite list still lifts individually
through RE1, and the ordinary product and finite
strict-comparison arguments of CC, UA and UF are
unaffected.

\begin{corollary}[Two units and one relation need not lift exactly]
\label{cor:r13-finite-relation-no-exact-lift}
Let \(x_E,y_E\in k(E)^\times\subset k(D)^\times\)
be the affine coordinates of RE2.  There are no units
\(X,Y\in\mathcal O_{V,\eta_D}^\times\) reducing to
\(x_E,y_E\) and satisfying the corresponding equation
\[
 \begin{cases}
 Y^2=X^3-X,&\operatorname{char}k\ne2,\\
 Y^2+Y=X^3,&\operatorname{char}k=2.
 \end{cases}
\tag{RE5}
\]
\end{corollary}

\begin{proof}
Such units would define a homomorphism from the
integral affine coordinate ring of \(E\) to \(k(V)\).
Its kernel is prime.  A nonzero prime in this
one-dimensional affine domain is maximal and has
residue field algebraic over \(k\).  This is
incompatible with the transcendental residues
\(x_E,y_E\).  Hence the homomorphism is injective
and extends to \(k(E)\hookrightarrow k(V)\),
contradicting the proof above.
\end{proof}

\paragraph{The estimate left outside the core fiber.}
The direct polynomial comparison in UF has
\(|\widehat A-A_0|<1\).  In its completed core
fiber \(|A_0|=1\), which gives the needed relative
strict inequality.  On a different ray a chosen
surface numerator can instead have order \(a>0\),
while the first normal error has order \(m>0\).
The comparison then requires \(a<m\); it need
not follow when \(a\ge m\).  A finite Taylor
cancellation alone supplies only a finite error
order; absent an additional divisibility or
relative bound, it does not control arbitrary
later rays.  The proposition
explains why a universal rational coefficient
section cannot be assumed to remove all such
errors.  It is not a counterexample to finite
controlled correction or to intrinsic descent.

\paragraph{Classical input checked.}
The all-characteristic algebraic form of L\"uroth's
theorem was checked in M.~Musta\c t\u a,
\emph{Lecture 1. Definitions and first examples},
Theorem~2.1 and its algebraic proof on pp.~4--6,
\url{https://public.websites.umich.edu/~mmustata/lecture1_rationality.pdf},
accessed 2026-11-15.  Only that classical field theorem
is imported; the divisor-section obstruction above
is the displayed deduction.