OpenMath

Proof source

A compact-source obstruction even in the unrestricted strong completion

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 compact-source obstruction even in the unrestricted strong completion

A compact-source obstruction even in the unrestricted strong completion

This section specializes only to the explicit counterexample background (DO6), with n≥2n\geq2, Q=(−1,1)nQ=(-1,1)^n, and remaining degree q=1q=1. The field, norm, analytic space, and original test category are unchanged. Write U=t−1(Q)U=t^{-1}(Q) and let σ:Q→U\sigma:Q\to U be the joint original Gauss section. The new potential space is precisely E^str1(U)\widehat{\mathcal E}^{1}_{\rm str}(U) of Definition def:r11-strong-first-energy-completion. Its elements need not have an actual common lower barrier or an ambient classical representative. Their currents and multipliers are those already constructed in (IC10)–(IC13).

The frozen inputs are exactly the R6 joint-Gauss convexity and original calibration, and the R10 compact-source convex rigidity listed in the power-class obstruction section. The already proved actual and fixed-barrier obstructions (DO7) and (BV13) are not reproved here. The new issue is whether the energy bound of an arbitrary strongly Cauchy row still forces a finite convex trace in its current limit. The following estimate answers that issue.

Lemma (Local trace compactness from a first-energy bound). Let (uj)⊂ES,∂1(U)(u_j)\subset\mathcal E^1_{S,\partial}(U) be actual classical potentials with sup⁡je1(uj)≤M<∞\sup_j e_1(u_j)\leq M<\infty. Their finite convex Gauss traces gj=uj∘σ≤0g_j=u_j\circ\sigma\leq0 have uniformly bounded L1L^1 norms on every compact real subbox of QQ. They are locally uniformly bounded and locally equicontinuous. Consequently a subsequence converges locally uniformly on QQ to a finite convex function g≤0g\leq0. No common pointwise lower barrier is assumed.

Status: NEW_PROVED in this precise original moment example. The compactness assertion concerns real convex traces, not an asserted compactness of ambient classical psh potentials in their original category.

Proof. The finite-convex-trace argument preceding (DO8) applies separately to every actual uju_j: its compact negative sublevels supply one finite Gauss value, and extended convexity bounded above by zero then gives finiteness everywhere. Fix a compact real subbox B⋐QB\Subset Q and put K=σ(B)K=\sigma(B), a compact ordinary analytic set. The normalized smooth original tester v0=1n∑i=1nti2satisfiesD(v0S)=2σ∗(dy).(SO1)v_0=\frac1n\sum_{i=1}^n t_i^2 \quad\hbox{satisfies}\quad D(v_0S)=2\sigma_*(dy). \tag{SO1} Let ρ\rho be the strict-box barrier from (TC2). Choose A>0A>0 so that Aρ<−2A\rho<-2 near KK and set ψ=max⁡(v0−2,Aρ).(SO2)\psi=\max(v_0-2,A\rho). \tag{SO2} The exact core construction (TC7)–(TC8), with its harmless constant rescaling, shows that ψ\psi is an actual bounded finite-mass member of every positive power class. In particular e1(ψ)<∞e_1(\psi)<\infty, and it agrees with v0−2v_0-2 near KK. Original locality therefore gives 1Kμψ=2σ∗(1Bdy)\mathbf1_K\mu_\psi=2\sigma_*(\mathbf1_Bdy). The global constant-one mixed energy inequality yields ∫B(−gj) dy=12∫K(−uj) dμψ≤12∫U(−uj) dμψ≤12M1/2e1(ψ)1/2.(SO3)\int_B(-g_j)\,dy =\frac12\int_K(-u_j)\,d\mu_\psi \leq\frac12\int_U(-u_j)\,d\mu_\psi \leq\frac12M^{1/2}e_1(\psi)^{1/2}. \tag{SO3} Every multiplier in this equation is the previously justified actual multiplier. This proves the local L1L^1 bound, including the case M=0M=0.

For completeness, the convex compactness step is elementary. Fix a compact real set C⋐Int⁡(B)C\Subset\operatorname{Int}(B) and choose δ>0\delta>0 such that each Euclidean ball B(x,δ)B(x,\delta) with x∈Cx\in C lies in BB. The functions hj=−gjh_j=-g_j are nonnegative and concave. For z∈B(x,δ/2)z\in B(x,\delta/2), write z=(x+y)/2z=(x+y)/2 with y∈B(x,δ)y\in B(x,\delta). Concavity gives hj(z)≥hj(x)/2h_j(z)\geq h_j(x)/2. Integration implies 0≤−gj(x)≤2vol⁡(B(0,δ/2))∫B(−gj) dy.(SO4)0\leq -g_j(x) \leq\frac{2}{\operatorname{vol}(B(0,\delta/2))} \int_B(-g_j)\,dy. \tag{SO4} Thus the traces have uniform two-sided bounds on every compact interior set. Apply this bound on a slightly larger compact set. On any line through a point of the smaller set, the monotone secant slopes of a convex function are then bounded by its oscillation on the larger set divided by the fixed distance to its boundary. This gives a uniform Lipschitz bound on every smaller compact subbox. A finite-net subsequence argument and a countable exhaustion yield local uniform convergence of a subsequence. Convexity and the upper bound zero pass to its finite limit. ◻

Theorem (Strong completion does not create a nonzero compact source). For the precise background just specified, let u∈E^str1(U)u\in\widehat{\mathcal E}^{1}_{\rm str}(U). If the positive original measure D(uS)=N^1(u)D(uS)=\widehat N_1(u) has tropical pushforward carried by a compact subset of QQ, then D(uS)=0.(SO5)D(uS)=0. \tag{SO5} In particular, the nonzero positive compact source μ=σ∗(χ dy)\mu=\sigma_*(\chi\,dy) of (DO9) is not D(uS)D(uS) for any member of this unrestricted strong completion. Thus passing from actual power-class potentials to the complete relative strong first-energy space does not make finite-dual or capacity-controlled data universally solvable.

Status: NEW_PROVED as an extension of the frozen source obstruction to the new strong completion without an actual lower barrier. The universal source-solvability inference being rejected is FALSE.

Proof. Choose an actual strongly Cauchy defining row (uj)(u_j) for uu. Its first energies are bounded by (EV9), or (IC2). The preceding lemma supplies a subsequence of its Gauss traces converging locally uniformly to a finite convex gg. The strong completed calculus gives the original coefficient convergence ujS→uSu_jS\to uS, before applying any tropical test. Consequently, for every φ∈Cc∞(Q)\varphi\in C_c^\infty(Q), ⟨t∗D(uS),φ⟩=lim⁡j⟨ujS,D(φ(t))⟩=lim⁡j∫Qgj Δφ dy=∫Qg Δφ dy=⟨Δg,φ⟩.(SO6)\begin{split} \langle t_*D(uS),\varphi\rangle &=\lim_j\langle u_jS,D(\varphi(t))\rangle\\ &=\lim_j\int_Qg_j\,\Delta\varphi\,dy =\int_Qg\,\Delta\varphi\,dy =\langle\Delta g,\varphi\rangle. \end{split} \tag{SO6} The pullback φ(t)\varphi(t) is an original compact smooth test: inverse images of compact real subboxes in this original torus are compact polyannuli. Each identity uses the already fixed all-original calibration. It does not define the completed current by a selected toric test subfamily.

The positive measure Δg\Delta g has compact support in QQ. The exact frozen R10 convex compact-marginal rigidity theorem, valid here for n≥2n\geq2, gives Δg=0\Delta g=0. Hence t∗D(uS)=0t_*D(uS)=0. Positivity makes the original measure zero: for every compact ordinary L⊂UL\subset U, its mass is at most the mass of t−1(t(L))t^{-1}(t(L)), which is the zero pushforward mass of the compact set t(L)t(L). Radon inner regularity then gives (SO5).

The datum (DO9) is finite, nonzero, compactly supported, dominated by one half of an actual normalized tester measure, and has finite energy-dual size, as already proved there. It contradicts (SO5) if it were a completed source. No claim is made that this special background exhausts arbitrary backgrounds; one allowed background suffices to disprove the unqualified universal source assertion. ◻

The resulting minimal equation gate.

The strong completion solves a precise nonlinear Cauchy problem and retains all-original closed curvature currents. It does not contain every solution of the larger relative linear energy problem even when q=1q=1. The surviving convex trace is an explicit obstruction. A prospective general nonlinear Dirichlet theorem must therefore specify an admissible operator image or a different potential class; neither completion alone nor a finite datum energy norm supplies the missing equation. The analytic PS/PL relation quotient and low-real-degree CTO are separate and are not used in this argument.

References requiring source context

Some references in this note depend on its original surrounding manuscript. Their source labels are retained in the reading edition; its TeX is available above.

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 a90c5261fb3c635f1244e9b82e2b2c88e65f9382f5d3a9232c94b2b0b32c158c

Frozen-original SHA-256 a90c5261fb3c635f1244e9b82e2b2c88e65f9382f5d3a9232c94b2b0b32c158c

\section{A compact-source obstruction even in the unrestricted strong completion}
\label{sec:r11-strong-completion-source-obstruction}

This section specializes only to the explicit counterexample background
(DO6), with $n\geq2$, $Q=(-1,1)^n$, and remaining degree $q=1$.
The field, norm, analytic space, and original test category are unchanged.
Write $U=t^{-1}(Q)$ and let $\sigma:Q\to U$ be the joint original
Gauss section.  The new potential space is precisely
$\widehat{\mathcal E}^{1}_{\rm str}(U)$ of
Definition~\ref{def:r11-strong-first-energy-completion}.
Its elements need not have an actual common lower barrier or an
ambient classical representative.  Their currents and multipliers
are those already constructed in (IC10)--(IC13).

The frozen inputs are exactly the R6 joint-Gauss convexity and
original calibration, and the R10 compact-source convex rigidity
listed in the power-class obstruction section.  The already proved
actual and fixed-barrier obstructions (DO7) and (BV13) are not
reproved here.  The new issue is whether the energy bound of an
arbitrary strongly Cauchy row still forces a finite convex trace
in its current limit.  The following estimate answers that issue.

\begin{lemma}[Local trace compactness from a first-energy bound]
\label{lem:r11-energy-bounded-convex-trace-compactness}
Let $(u_j)\subset\mathcal E^1_{S,\partial}(U)$ be actual classical
potentials with $\sup_j e_1(u_j)\leq M<\infty$.
Their finite convex Gauss traces $g_j=u_j\circ\sigma\leq0$ have
uniformly bounded $L^1$ norms on every compact real subbox of $Q$.
They are locally uniformly bounded and locally equicontinuous.
Consequently a subsequence converges locally uniformly on $Q$
to a finite convex function $g\leq0$.
No common pointwise lower barrier is assumed.
\end{lemma}

\noindent\textbf{Status: \texttt{NEW\_PROVED}} in this precise
original moment example.  The compactness assertion concerns real
convex traces, not an asserted compactness of ambient classical psh
potentials in their original category.

\begin{proof}
The finite-convex-trace argument preceding (DO8) applies separately
to every actual $u_j$: its compact negative sublevels supply one
finite Gauss value, and extended convexity bounded above by zero
then gives finiteness everywhere.  Fix a compact real subbox
$B\Subset Q$ and put $K=\sigma(B)$, a compact ordinary analytic
set.  The normalized smooth original tester
\[
                v_0=\frac1n\sum_{i=1}^n t_i^2
           \quad\hbox{satisfies}\quad
                D(v_0S)=2\sigma_*(dy).
 \tag{SO1}
\]
Let $\rho$ be the strict-box barrier from (TC2).
Choose $A>0$ so that $A\rho<-2$ near $K$ and set
\[
                     \psi=\max(v_0-2,A\rho).
 \tag{SO2}
\]
The exact core construction (TC7)--(TC8), with its harmless
constant rescaling, shows that $\psi$ is an actual bounded
finite-mass member of every positive power class.  In particular
$e_1(\psi)<\infty$, and it agrees with $v_0-2$ near $K$.
Original locality therefore gives
$\mathbf1_K\mu_\psi=2\sigma_*(\mathbf1_Bdy)$.
The global constant-one mixed energy inequality yields
\[
 \int_B(-g_j)\,dy
   =\frac12\int_K(-u_j)\,d\mu_\psi
   \leq\frac12\int_U(-u_j)\,d\mu_\psi
   \leq\frac12M^{1/2}e_1(\psi)^{1/2}.
 \tag{SO3}
\]
Every multiplier in this equation is the previously justified
actual multiplier.  This proves the local $L^1$ bound, including
the case $M=0$.

For completeness, the convex compactness step is elementary.
Fix a compact real set $C\Subset\operatorname{Int}(B)$ and
choose $\delta>0$ such that each Euclidean ball
$B(x,\delta)$ with $x\in C$ lies in $B$.
The functions $h_j=-g_j$ are nonnegative and concave.
For $z\in B(x,\delta/2)$, write $z=(x+y)/2$ with
$y\in B(x,\delta)$.  Concavity gives
$h_j(z)\geq h_j(x)/2$.  Integration implies
\[
 0\leq -g_j(x)
       \leq\frac{2}{\operatorname{vol}(B(0,\delta/2))}
                       \int_B(-g_j)\,dy.
 \tag{SO4}
\]
Thus the traces have uniform two-sided bounds on every compact
interior set.  Apply this bound on a slightly larger compact
set.  On any line through a point of the smaller set, the
monotone secant slopes of a convex function are then bounded
by its oscillation on the larger set divided by the fixed
distance to its boundary.  This gives a uniform Lipschitz
bound on every smaller compact subbox.  A finite-net
subsequence argument and a countable exhaustion yield local
uniform convergence of a subsequence.  Convexity and the
upper bound zero pass to its finite limit.
\end{proof}

\begin{theorem}[Strong completion does not create a nonzero compact source]
\label{thm:r11-strong-completed-compact-source-obstruction}
For the precise background just specified, let
$u\in\widehat{\mathcal E}^{1}_{\rm str}(U)$.
If the positive original measure $D(uS)=\widehat N_1(u)$
has tropical pushforward carried by a compact subset of $Q$,
then
\[
                              D(uS)=0.
 \tag{SO5}
\]
In particular, the nonzero positive compact source
$\mu=\sigma_*(\chi\,dy)$ of (DO9) is not $D(uS)$ for
any member of this unrestricted strong completion.
Thus passing from actual power-class potentials to the complete
relative strong first-energy space does not make finite-dual or
capacity-controlled data universally solvable.
\end{theorem}

\noindent\textbf{Status: \texttt{NEW\_PROVED}} as an extension
of the frozen source obstruction to the new strong completion
without an actual lower barrier.  The universal source-solvability
inference being rejected is \texttt{FALSE}.

\begin{proof}
Choose an actual strongly Cauchy defining row $(u_j)$ for $u$.
Its first energies are bounded by (EV9), or (IC2).
The preceding lemma supplies a subsequence of its Gauss traces
converging locally uniformly to a finite convex $g$.
The strong completed calculus gives the original coefficient
convergence $u_jS\to uS$, before applying any tropical test.
Consequently, for every $\varphi\in C_c^\infty(Q)$,
\[
 \begin{split}
 \langle t_*D(uS),\varphi\rangle
   &=\lim_j\langle u_jS,D(\varphi(t))\rangle\\
   &=\lim_j\int_Qg_j\,\Delta\varphi\,dy
     =\int_Qg\,\Delta\varphi\,dy
     =\langle\Delta g,\varphi\rangle.
 \end{split}
 \tag{SO6}
\]
The pullback $\varphi(t)$ is an original compact smooth test:
inverse images of compact real subboxes in this original torus
are compact polyannuli.  Each identity uses the already fixed
all-original calibration.  It does not define the completed
current by a selected toric test subfamily.

The positive measure $\Delta g$ has compact support in $Q$.
The exact frozen R10 convex compact-marginal rigidity theorem,
valid here for $n\geq2$, gives $\Delta g=0$.
Hence $t_*D(uS)=0$.  Positivity makes the original measure zero:
for every compact ordinary $L\subset U$, its mass is at most
the mass of $t^{-1}(t(L))$, which is the zero pushforward
mass of the compact set $t(L)$.  Radon inner regularity
then gives (SO5).

The datum (DO9) is finite, nonzero, compactly supported,
dominated by one half of an actual normalized tester measure,
and has finite energy-dual size, as already proved there.
It contradicts (SO5) if it were a completed source.
No claim is made that this special background exhausts arbitrary
backgrounds; one allowed background suffices to disprove the
unqualified universal source assertion.
\end{proof}

\paragraph{The resulting minimal equation gate.}
The strong completion solves a precise nonlinear Cauchy problem
and retains all-original closed curvature currents.  It does
not contain every solution of the larger relative linear
energy problem even when $q=1$.  The surviving convex trace
is an explicit obstruction.  A prospective general nonlinear
Dirichlet theorem must therefore specify an admissible operator
image or a different potential class; neither completion alone
nor a finite datum energy norm supplies the missing equation.
The analytic PS/PL relation quotient and low-real-degree CTO
are separate and are not used in this argument.