OpenMath

Proof source

Compatible piecewise-smooth tests and critical tangent pairings

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. Compatible piecewise-smooth tests and critical tangent pairings

Compatible piecewise-smooth tests and critical tangent pairings

The currents in the first part of this section are real supercurrents on an open subset Ω⊂RN\Omega\subset\mathbf R^N, with N≥2N\ge2. They are symmetric, middle-positive, separately d′d'- and d′′d''-closed, and have bidimension (2,2)(2,2). Write β=∑id′xi∧d′′xi,μT=12T∧β2.(PT1)\beta=\sum_i d'x_i\wedge d''x_i, \mu_T=\tfrac12T\wedge\beta^2. \tag{PT1} All coefficient measures are controlled by this trace, as in MT4. No weak/strong positivity identification is made.

Nonduplication status.

The uniform O(ϵ)O(\sqrt\epsilon) piecewise-test approximation mechanism is retained: R6, lem:r6-weighted-pws-error, records it as a direct corollary of R4 Lemmas 71.1–71.2 and Theorem 71.4, even with bounded Borel multipliers. Its real bidimension-two proof is written below to verify precisely the tangent-limit interface used here. The nodal construction likewise makes the retained Courant scalar generation explicit in the homogeneous constant-tensor case. These mechanisms are not counted as new principal theorems. The new composition is their use with MT and LE to give the complete realized critical-tangent pairing in PT11.

A compatible finite-PS test is a compactly supported ambient (2,2)(2,2) form on a finite polyhedral complex covering a neighborhood of its support, with smooth extensions near the closed cells, whose tangential pullbacks agree on every common face. This is a supplied genuine ambient presentation, not independent coefficient data attached to different source labels. Bounded first derivatives on a slightly larger compact neighborhood follow from the finite closed-cell data.

Uniform control of normal factors near a wall

Let H={ℓ=0}H=\{\ell=0\} be an affine hyperplane, with ∣dℓ∣=1|d\ell|=1, and put n′=d′ℓn'=d'\ell, n′′=d′′ℓn''=d''\ell. A (2,2)(2,2) coefficient tensor belongs to its normal-factor subspace if it lies in IH=n′∧Λ1,2+n′′∧Λ2,1.(PT2)I_H=n'\wedge\Lambda^{1,2} +n''\wedge\Lambda^{2,1}. \tag{PT2} This is exactly the kernel of tangential restriction to HH.

Lemma (Normal strip bounds). Fix compact sets K⊂int⁡K′⋐ΩK\subset\operatorname{int}K'\Subset\Omega. For every such TT and every sufficiently small ϵ>0\epsilon>0, (T∧n′∧n′′∧β)(K∩{∣ℓ∣≤ϵ})≤Cϵ μT(K′),∣⟨T,ψ⟩∣≤C∥ψ∥∞ϵ μT(K′)(PT3)\begin{split} (T\wedge n'\wedge n''\wedge\beta) (K\cap\{ |\ell|\le\epsilon\}) &\le C\epsilon\,\mu_T(K'),\\ |\langle T,\psi\rangle| &\le C\|\psi\|_\infty\sqrt\epsilon\, \mu_T(K') \end{split} \tag{PT3} whenever the bounded Borel test ψ\psi is supported in that strip and takes values in IHI_H. The constants depend on the fixed geometry and cutoff, not on TT. The coefficient measures involving at least one normal two-vector index have zero restriction to HH itself. The trace μT\mu_T may nevertheless charge HH.

Proof. Choose 0≤χ∈Cc∞(int⁡K′)0\le\chi\in C_c^\infty(\operatorname{int}K') equal to one near KK. Choose smooth convex functions uϵu_\epsilon on the real line such that uϵ′′(s)≥ϵ−1(∣s∣≤ϵ),∣uϵ(s)∣≤C(1+∣s∣)(PT4)u_\epsilon''(s)\ge\epsilon^{-1} \quad (|s|\le\epsilon),\qquad |u_\epsilon(s)|\le C(1+|s|) \tag{PT4} on the fixed relevant range, with CC independent of ϵ\epsilon. For instance rescale one convex smoothing of ∣s∣|s| whose second derivative has compact support and is bounded below on the unit interval. Then D(uϵ∘ℓ)=uϵ′′(ℓ)n′∧n′′D(u_\epsilon\circ\ell)=u_\epsilon''(\ell)n'\wedge n''. Positivity and closedness give ϵ−1(T∧n′∧n′′∧β)(K∩{∣ℓ∣≤ϵ})≤⟨T,χD(uϵ∘ℓ)∧β⟩=⟨T,(uϵ∘ℓ)Dχ∧β⟩≤CμT(K′).(PT5)\begin{split} \epsilon^{-1} (T\wedge n'\wedge n''\wedge\beta) (K\cap\{|\ell|\le\epsilon\}) &\le\langle T,\chi D(u_\epsilon\circ\ell)\wedge\beta\rangle\\ &=\langle T,(u_\epsilon\circ\ell)D\chi\wedge\beta\rangle\\ &\le C\mu_T(K'). \end{split} \tag{PT5} The last inequality follows from the bounded scalar function in PT4 and the fixed middle-order bound for Dχ∧βD\chi\wedge\beta. This is a distributional Stokes identity with smooth functions; no current is restricted to the wall in deriving it.

Use an orthonormal frame with first covector dℓd\ell. The cocoefficient matrix of TT is positive semidefinite on two-vectors. Its diagonal entries whose index contains this first covector are bounded by the normal trace in PT5; its other diagonal entries are bounded by the total trace. For a Borel set BB, positivity and Radon polarization give ∣μIJ∣(B)≤μII(B)μJJ(B).(PT6)|\mu_{IJ}|(B) \le\sqrt{\mu_{II}(B)\mu_{JJ}(B)}. \tag{PT6} Each coefficient of an IHI_H-valued tensor has such a normal index in at least one column. Apply PT5 to that diagonal entry and the trace bound to the other. The finite sum of coefficient estimates proves the second inequality of PT3. Letting ϵ↓0\epsilon\downarrow0 proves the assertion on HH. It only removes normal-factor coefficient measures there, not the tangential trace mass. ◻

A finite-wall approximation with all lower faces retained

Lemma (Uniform smooth approximation of compatible tests). Let θ\theta be a compatible finite-PS compact test. There are compactly supported smooth coefficient convolutions θϵ\theta_\epsilon in one fixed compact neighborhood and a constant CθC_\theta such that ∣⟨T,θ−θϵ⟩∣≤Cθϵ μT(K′)(PT7)|\langle T,\theta-\theta_\epsilon\rangle| \le C_\theta\sqrt\epsilon\,\mu_T(K') \tag{PT7} for every current in the stated class. The pairing with θ\theta is independent of ambient representatives of its lower-face values.

Proof. Take the finite list of affine hyperplanes containing the coefficient facets, and refine by their arrangement. Fix one smooth closed-cell extension for each resulting full cell. At a point of a lower face, choose an incident full cell to give a bounded Borel representative. Any two such choices differ by a sum of tensors in IHI_H for hyperplanes containing that face: adjacent incident cells have this property by tangential compatibility, and the local full cell adjacency through facets is connected. Alternatively the kernel of restriction to the face is the sum of the ideals generated by its conormals. PT3 makes all these differences pair to zero on the face. The chosen Borel representative therefore gives the canonical pairing.

Extend by zero outside a fixed compact neighborhood of the support, where the original test already vanishes. Let θϵ\theta_\epsilon be its coefficient convolution with a smooth kernel supported in ∣a∣≤ϵ|a|\le\epsilon. For a generic translation vector aa, traverse the segment from xx to x−ax-a. Within cells, its total smooth variation is at most C∣a∣C|a|. At a crossed facet HH the jump is in IHI_H, because its tangential restriction is zero. All these jump coefficients are uniformly bounded by the finite closed-cell bounds. The hyperplane crossed by the segment has distance at most ∣a∣|a| from xx. A generic segment crosses at most one facet of each of the finitely many hyperplanes; crossings at lower faces can be omitted from the translation integral, since they form a Lebesgue-null set of translations. If xx itself lies on lower faces, the initial choice of incident cell is handled by the same finite incident-cell chain just used above.

In a fixed coordinate basis this proves the decomposition θϵ−θ=Rϵ+∑HJH,ϵ,(PT8)\theta_\epsilon-\theta =R_\epsilon+\sum_H J_{H,\epsilon}, \tag{PT8} where ∥Rϵ∥∞≤Cϵ\|R_\epsilon\|_\infty\le C\epsilon, each JH,ϵJ_{H,\epsilon} is a bounded Borel IHI_H-valued tensor with norm at most CC, and its support lies in a fixed multiple of the ϵ\epsilon-strip around HH. Indeed the conormal of an affine hyperplane is constant, so transporting a jump from its crossing point to xx preserves membership in IHI_H. A fixed linear splitting of this finite-dimensional subspace gives bounded coefficient representatives before integrating over the translations. There are finitely many walls and no accumulating crossing sequence.

The ordinary coefficient trace bound controls the first term in PT8 by CϵμT(K′)C\epsilon\mu_T(K'). PT3 controls each remaining term by CϵμT(K′)C\sqrt\epsilon\mu_T(K'). Their finite sum proves PT7, including the strata where several coefficient walls meet. No simultaneous covector primitive is used or needed. ◻

Theorem (Weak continuity on compatible finite-PS tests). Suppose Tj→T0T_j\to T_0 on all compact smooth real tests and the currents satisfy the stated middle positivity and both closedness equations, with uniformly bounded trace on each compact subset. Then for every fixed compatible finite-PS compact test, ⟨Tj,θ⟩⟶⟨T0,θ⟩.(PT9)\langle T_j,\theta\rangle \longrightarrow \langle T_0,\theta\rangle. \tag{PT9} The limit may charge coefficient walls. No zero-wall-mass hypothesis is required.

Proof. The limit has the same middle positivity and closedness and the same local trace bounds. For fixed ϵ\epsilon the convoluted test is smooth, so its pairings converge. PT7 bounds the two approximation errors uniformly in jj. First pass to the limit in jj, then let ϵ\epsilon tend to zero. This proves PT9. ◻

A constructive ambient extension for homogeneous tensors

Lemma (Nodal extension from a geometric subfan). Let Σ\Sigma be a finite rational fan in RN\mathbf R^N. Suppose a constant (p,q)(p,q) tensor is given on the span of each cone, with identical tangential restrictions on common faces. Then it has a genuine ambient homogeneous coefficient-degree-zero finite-PS extension. If the data are column-symmetric, the extension may be chosen symmetric. This lemma extends regular forms, not their primitives.

Proof. Refine a complete rational hyperplane fan containing all equations for the spans of the supplied cones, extensions of their relative facet equations, and the coordinate hyperplanes, and make a common simplicial refinement. This gives a complete simplicial fan Δ\Delta such that the supplied support is a subfan after subdivision. The coordinate hyperplanes remove lineality before triangulation. Choose a vector vrv_r on every ray and its continuous homogeneous PL nodal function φr\varphi_r: on a cone containing that ray it is the corresponding coordinate, and on other cones it is zero. These formulas agree on shared faces.

Fix a global order of the rays. For strictly increasing ray multi-indices I,JI,J of lengths p,qp,q in that order, consider their union. If its rays span a face τ\tau of Δ\Delta lying in the supplied subfan, let aIJa_{IJ} be the value of the supplied tensor on the corresponding ordered ray vectors in its two columns. The value is independent of a containing supplied cone by compatibility. If their union is not contained in any cone, or its smallest face is not supplied, set aIJ=0a_{IJ}=0. The finite sum E^=∑I,JaIJdP′φI∧dP′′φJ(PT10)\widehat E=\sum_{I,J}a_{IJ} d'_P\varphi_I\wedge d''_P\varphi_J \tag{PT10} with the standard column sign convention defines an actual ambient finite-PS form. On any full simplicial cone the nonzero nodal differentials are its dual ray basis. PT10 therefore has exactly the assigned tensor coefficients. On a supplied cone every relevant smaller ray face is supplied, so its restriction is the original tensor. The same argument proves all shared-face compatibilities. Column symmetry is preserved by making the symmetric coefficient assignments. This explicit scalar construction does not presume an extension of independently chosen source primitives or identify coincident source labels. ◻

Only geometric single-valued face data are covered by this extension lemma. Different actual source labels with incompatible values on the same geometric image do not define its input. LE’s leading energy already has a supplied ambient presentation by the scalar first jets; PT10 is an additional constructive extension option.

The central scalar as a tangent-current pairing

Return to the protected original setting of MT and LE. Let TT be the full-fiber real background transpose, and let Tδj=δj2(aδj)∗TT_{\delta_j}=\delta_j^2(a_{\delta_j})_*T converge to any critical tangent T0T_0, where aδ(x)=x/δa_\delta(x)=x/\delta. Let EE be a homogeneous coefficient-degree-zero (1,1)(1,1) symbol with a supplied compatible ambient finite-PS presentation. This applies in particular to LE’s actual leading energy Z0Z_0.

The smooth test transpose MT2 also agrees on this regular PS pairing. Indeed its cocoefficient Radon measures are the proper transposes of the corresponding original coefficient measures, by polarization of MT2. Finite Borel multiplication by the PS coefficients, with their actual face values, is precisely the retained regular-PS module action. The normal-wall zero-coefficient conclusion of PT3 removes any ambient representative ambiguity. Thus no higher-degree delta action is being inferred from the smooth transpose alone.

Theorem (Every critical tangent has the same relevant paired value). For the cutoff ρ\rho of LE7, lim⁡jλδj,S(E)=−⟨T0,E∧Dρ⟩,cS(P)=−⟨T0,Z0∧Dρ⟩.(PT11)\begin{split} \lim_j\lambda_{\delta_j,S}(E) &=-\langle T_0,E\wedge D\rho\rangle,\\ c_S(P)&=-\langle T_0,Z_0\wedge D\rho\rangle. \end{split} \tag{PT11} The second equality holds for every critical tangent of the protected background. It is a full tensor pairing, with all mixed entries and wall traces retained.

Proof. Homogeneity of the coefficient functions of EE gives the exact scaling identity ⟨Tδ,E∧Dρ⟩=⟨T,E∧D[ρ(x/δ)]⟩.(PT12)\langle T_\delta,E\wedge D\rho\rangle =\langle T,E\wedge D[\rho(x/\delta)]\rangle. \tag{PT12} Both (1,1)(1,1) factors together contribute δ−4\delta^{-4} under real pullback, and MT’s prefactor δ2\delta^2 leaves precisely the δ−2\delta^{-2} of the displayed Hessian. The original Borel module identification just explained makes the right side equal to ⟨MS(E),Dχδ⟩\langle M_S(E),D\chi_\delta\rangle.

MT supplies the uniform trace bounds of PT9 on each compact real set. The fixed test E∧DρE\wedge D\rho is compatible finite PS. PT9 therefore applies to the left side of PT12, even if the tangent current charges walls. This proves the first equality in PT11. Taking E=Z0E=Z_0 and using the unconditional leading reduction LE9 proves the second. Tangent uniqueness is not asserted or needed. ◻

If two supplied ambient extensions represent the same actual symbol, the original Borel module pairings coincide at every δ\delta. PT9 consequently gives the same tangent pairing for both extensions. This is the precise independence needed here; support containment alone would not justify multiplying an arbitrary discontinuous extension by a tangent current.

Remaining analytic realization gate.

PT11 does not put a critical tangent into the retained analytic weak-background class, does not construct an original analytic background from an arbitrary real cone current, and does not prove its pairing is zero for an actual-zero leading energy. The realized tangent family is constrained by all original refinements. It is not identified with every positive closed real current on a selected fan or with the full old Killing dual. The remaining cancellation is exactly the vanishing of the paired value in PT11 for those realized backgrounds.

Inputs.

The real coefficient inequalities and critical compactness are MT2–MT15, with the middle-positive normalization of PT1. The original-to-real identification uses the retained R6 regular-PS Borel module, GP’s complete proper fibers, and LE’s actual scalar first-jet construction. The finite-wall proof and nodal extension are supplied above. No general current restriction, arbitrary analytic pushforward, or GK test extension is used.

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 4dbf5b5ea8036ed190fe0eec4469f919d6b2b16294d5749a5760a744d7a189eb

Frozen-original SHA-256 4dbf5b5ea8036ed190fe0eec4469f919d6b2b16294d5749a5760a744d7a189eb

\section{Compatible piecewise-smooth tests and critical tangent pairings}
\label{sec:r13-compatible-ps-tangent-pairing}

The currents in the first part of this section are real
supercurrents on an open subset $\Omega\subset\mathbf R^N$,
with $N\ge2$.  They are symmetric, middle-positive,
separately $d'$- and $d''$-closed, and have bidimension
$(2,2)$.  Write
\[
             \beta=\sum_i d'x_i\wedge d''x_i,
                      \mu_T=\tfrac12T\wedge\beta^2.
 \tag{PT1}
\]
All coefficient measures are controlled by this trace, as in
MT4.  No weak/strong positivity identification is made.

\paragraph{Nonduplication status.}
The uniform $O(\sqrt\epsilon)$ piecewise-test approximation
mechanism is retained: R6,
\texttt{lem:r6-weighted-pws-error}, records it as a direct
corollary of R4 Lemmas 71.1--71.2 and Theorem 71.4, even
with bounded Borel multipliers.  Its real bidimension-two
proof is written below to verify precisely the tangent-limit
interface used here.  The nodal construction likewise makes
the retained Courant scalar generation explicit in the
homogeneous constant-tensor case.  These mechanisms are not
counted as new principal theorems.  The new composition is
their use with MT and LE to give the complete realized
critical-tangent pairing in PT11.

A \emph{compatible finite-PS test} is a compactly supported
ambient $(2,2)$ form on a finite polyhedral complex covering
a neighborhood of its support, with smooth extensions near
the closed cells, whose tangential pullbacks agree on every
common face.  This is a supplied genuine ambient presentation,
not independent coefficient data attached to different source
labels.  Bounded first derivatives on a slightly larger
compact neighborhood follow from the finite closed-cell data.

\subsection{Uniform control of normal factors near a wall}

Let $H=\{\ell=0\}$ be an affine hyperplane, with
$|d\ell|=1$, and put
$n'=d'\ell$, $n''=d''\ell$.  A $(2,2)$ coefficient
tensor belongs to its \emph{normal-factor subspace} if it
lies in
\[
 I_H=n'\wedge\Lambda^{1,2}
                     +n''\wedge\Lambda^{2,1}.
 \tag{PT2}
\]
This is exactly the kernel of tangential restriction to $H$.

\begin{lemma}[Normal strip bounds]
\label{lem:r13-normal-strip-bounds}
Fix compact sets $K\subset\operatorname{int}K'\Subset\Omega$.
For every such $T$ and every sufficiently small $\epsilon>0$,
\[
 \begin{split}
 (T\wedge n'\wedge n''\wedge\beta)
       (K\cap\{ |\ell|\le\epsilon\})
             &\le C\epsilon\,\mu_T(K'),\\
 |\langle T,\psi\rangle|
             &\le C\|\psi\|_\infty\sqrt\epsilon\,
                              \mu_T(K')
 \end{split}
 \tag{PT3}
\]
whenever the bounded Borel test $\psi$ is supported in that
strip and takes values in $I_H$.  The constants depend on
the fixed geometry and cutoff, not on $T$.  The coefficient
measures involving at least one normal two-vector index have
zero restriction to $H$ itself.  The trace $\mu_T$ may
nevertheless charge $H$.
\end{lemma}

\begin{proof}
Choose $0\le\chi\in C_c^\infty(\operatorname{int}K')$
equal to one near $K$.  Choose smooth convex functions
$u_\epsilon$ on the real line such that
\[
 u_\epsilon''(s)\ge\epsilon^{-1}
          \quad (|s|\le\epsilon),\qquad
 |u_\epsilon(s)|\le C(1+|s|)
 \tag{PT4}
\]
on the fixed relevant range, with $C$ independent of
$\epsilon$.  For instance rescale one convex smoothing
of $|s|$ whose second derivative has compact support and
is bounded below on the unit interval.  Then
$D(u_\epsilon\circ\ell)=u_\epsilon''(\ell)n'\wedge n''$.
Positivity and closedness give
\[
 \begin{split}
 \epsilon^{-1}
   (T\wedge n'\wedge n''\wedge\beta)
          (K\cap\{|\ell|\le\epsilon\})
 &\le\langle T,\chi D(u_\epsilon\circ\ell)\wedge\beta\rangle\\
 &=\langle T,(u_\epsilon\circ\ell)D\chi\wedge\beta\rangle\\
 &\le C\mu_T(K').
 \end{split}
 \tag{PT5}
\]
The last inequality follows from the bounded scalar function
in PT4 and the fixed middle-order bound for $D\chi\wedge\beta$.
This is a distributional Stokes identity with smooth functions;
no current is restricted to the wall in deriving it.

Use an orthonormal frame with first covector $d\ell$.
The cocoefficient matrix of $T$ is positive semidefinite on
two-vectors.  Its diagonal entries whose index contains this
first covector are bounded by the normal trace in PT5; its
other diagonal entries are bounded by the total trace.
For a Borel set $B$, positivity and Radon polarization give
\[
           |\mu_{IJ}|(B)
                   \le\sqrt{\mu_{II}(B)\mu_{JJ}(B)}.
 \tag{PT6}
\]
Each coefficient of an $I_H$-valued tensor has such a normal
index in at least one column.  Apply PT5 to that diagonal
entry and the trace bound to the other.  The finite sum of
coefficient estimates proves the second inequality of PT3.
Letting $\epsilon\downarrow0$ proves the assertion on $H$.
It only removes normal-factor coefficient measures there,
not the tangential trace mass.
\end{proof}

\subsection{A finite-wall approximation with all lower faces retained}

\begin{lemma}[Uniform smooth approximation of compatible tests]
\label{lem:r13-uniform-compatible-ps-approximation}
Let $\theta$ be a compatible finite-PS compact test.  There
are compactly supported smooth coefficient convolutions
$\theta_\epsilon$ in one fixed compact neighborhood and a
constant $C_\theta$ such that
\[
       |\langle T,\theta-\theta_\epsilon\rangle|
                 \le C_\theta\sqrt\epsilon\,\mu_T(K')
 \tag{PT7}
\]
for every current in the stated class.  The pairing with
$\theta$ is independent of ambient representatives of its
lower-face values.
\end{lemma}

\begin{proof}
Take the finite list of affine hyperplanes containing the
coefficient facets, and refine by their arrangement.  Fix
one smooth closed-cell extension for each resulting full
cell.  At a point of a lower face, choose an incident full
cell to give a bounded Borel representative.  Any two such
choices differ by a sum of tensors in $I_H$ for hyperplanes
containing that face: adjacent incident cells have this
property by tangential compatibility, and the local full
cell adjacency through facets is connected.  Alternatively
the kernel of restriction to the face is the sum of the
ideals generated by its conormals.  PT3 makes all these
differences pair to zero on the face.  The chosen Borel
representative therefore gives the canonical pairing.

Extend by zero outside a fixed compact neighborhood of the
support, where the original test already vanishes.  Let
$\theta_\epsilon$ be its coefficient convolution with a
smooth kernel supported in $|a|\le\epsilon$.
For a generic translation vector $a$, traverse the segment
from $x$ to $x-a$.  Within cells, its total smooth variation
is at most $C|a|$.  At a crossed facet $H$ the jump is in
$I_H$, because its tangential restriction is zero.  All
these jump coefficients are uniformly bounded by the finite
closed-cell bounds.  The hyperplane crossed by the segment
has distance at most $|a|$ from $x$.  A generic segment
crosses at most one facet of each of the finitely many
hyperplanes; crossings at lower faces can be omitted from
the translation integral, since they form a Lebesgue-null
set of translations.  If $x$ itself lies on lower faces,
the initial choice of incident cell is handled by the same
finite incident-cell chain just used above.

In a fixed coordinate basis this proves the decomposition
\[
       \theta_\epsilon-\theta
             =R_\epsilon+\sum_H J_{H,\epsilon},
 \tag{PT8}
\]
where $\|R_\epsilon\|_\infty\le C\epsilon$, each
$J_{H,\epsilon}$ is a bounded Borel $I_H$-valued tensor
with norm at most $C$, and its support lies in a fixed
multiple of the $\epsilon$-strip around $H$.  Indeed the
conormal of an affine hyperplane is constant, so transporting
a jump from its crossing point to $x$ preserves membership
in $I_H$.  A fixed linear splitting of this finite-dimensional
subspace gives bounded coefficient representatives before
integrating over the translations.  There are finitely many
walls and no accumulating crossing sequence.

The ordinary coefficient trace bound controls the first term
in PT8 by $C\epsilon\mu_T(K')$.  PT3 controls each remaining
term by $C\sqrt\epsilon\mu_T(K')$.  Their finite sum proves
PT7, including the strata where several coefficient walls
meet.  No simultaneous covector primitive is used or needed.
\end{proof}

\begin{theorem}[Weak continuity on compatible finite-PS tests]
\label{thm:r13-compatible-ps-weak-continuity}
Suppose $T_j\to T_0$ on all compact smooth real tests and
the currents satisfy the stated middle positivity and both
closedness equations, with uniformly bounded trace on each
compact subset.  Then for every fixed compatible finite-PS
compact test,
\[
                     \langle T_j,\theta\rangle
                              \longrightarrow
                     \langle T_0,\theta\rangle.
 \tag{PT9}
\]
The limit may charge coefficient walls.  No zero-wall-mass
hypothesis is required.
\end{theorem}

\begin{proof}
The limit has the same middle positivity and closedness and
the same local trace bounds.  For fixed $\epsilon$ the
convoluted test is smooth, so its pairings converge.  PT7
bounds the two approximation errors uniformly in $j$.
First pass to the limit in $j$, then let $\epsilon$ tend
to zero.  This proves PT9.
\end{proof}

\subsection{A constructive ambient extension for homogeneous tensors}

\begin{lemma}[Nodal extension from a geometric subfan]
\label{lem:r13-nodal-homogeneous-form-extension}
Let $\Sigma$ be a finite rational fan in $\mathbf R^N$.
Suppose a constant $(p,q)$ tensor is given on the span of
each cone, with identical tangential restrictions on common
faces.  Then it has a genuine ambient homogeneous
coefficient-degree-zero finite-PS extension.  If the data
are column-symmetric, the extension may be chosen symmetric.
This lemma extends regular forms, not their primitives.
\end{lemma}

\begin{proof}
Refine a complete rational hyperplane fan containing all
equations for the spans of the supplied cones, extensions
of their relative facet equations, and the coordinate
hyperplanes, and make a
common simplicial refinement.  This gives a complete
simplicial fan $\Delta$ such that the supplied support is
a subfan after subdivision.  The coordinate hyperplanes
remove lineality before triangulation.  Choose a vector
$v_r$ on every ray and its continuous homogeneous PL
nodal function $\varphi_r$: on a cone containing that ray
it is the corresponding coordinate, and on other cones
it is zero.  These formulas agree on shared faces.

Fix a global order of the rays.  For strictly increasing
ray multi-indices $I,J$ of lengths $p,q$ in that order,
consider their union.  If its rays span a face $\tau$
of $\Delta$ lying in the supplied subfan, let $a_{IJ}$
be the value of the supplied tensor on the corresponding
ordered ray vectors in its two columns.  The value is
independent of a containing supplied cone by compatibility.
If their union is not contained in any cone, or its smallest
face is not supplied, set $a_{IJ}=0$.  The finite sum
\[
     \widehat E=\sum_{I,J}a_{IJ}
                  d'_P\varphi_I\wedge d''_P\varphi_J
 \tag{PT10}
\]
with the standard column sign convention defines an actual
ambient finite-PS form.  On any full simplicial cone the
nonzero nodal differentials are its dual ray basis.  PT10
therefore has exactly the assigned tensor coefficients.
On a supplied cone every relevant smaller ray face is
supplied, so its restriction is the original tensor.
The same argument proves all shared-face compatibilities.
Column symmetry is preserved by making the symmetric
coefficient assignments.  This explicit scalar construction
does not presume an extension of independently chosen
source primitives or identify coincident source labels.
\end{proof}

Only geometric single-valued face data are covered by this
extension lemma.  Different actual source labels with
incompatible values on the same geometric image do not
define its input.  LE's leading energy already has a
supplied ambient presentation by the scalar first jets;
PT10 is an additional constructive extension option.

\subsection{The central scalar as a tangent-current pairing}

Return to the protected original setting of MT and LE.
Let $T$ be the full-fiber real background transpose, and
let $T_{\delta_j}=\delta_j^2(a_{\delta_j})_*T$ converge
to any critical tangent $T_0$, where
$a_\delta(x)=x/\delta$.  Let $E$ be a homogeneous
coefficient-degree-zero $(1,1)$ symbol with a supplied
compatible ambient finite-PS presentation.  This applies
in particular to LE's actual leading energy $Z_0$.

The smooth test transpose MT2 also agrees on this regular
PS pairing.  Indeed its cocoefficient Radon measures are
the proper transposes of the corresponding original
coefficient measures, by polarization of MT2.  Finite
Borel multiplication by the PS coefficients, with their
actual face values, is precisely the retained regular-PS
module action.  The normal-wall zero-coefficient conclusion
of PT3 removes any ambient representative ambiguity.  Thus
no higher-degree delta action is being inferred from the
smooth transpose alone.

\begin{theorem}[Every critical tangent has the same relevant paired value]
\label{thm:r13-leading-tangent-pairing}
For the cutoff $\rho$ of LE7,
\[
 \begin{split}
 \lim_j\lambda_{\delta_j,S}(E)
       &=-\langle T_0,E\wedge D\rho\rangle,\\
 c_S(P)&=-\langle T_0,Z_0\wedge D\rho\rangle.
 \end{split}
 \tag{PT11}
\]
The second equality holds for every critical tangent of
the protected background.  It is a full tensor pairing,
with all mixed entries and wall traces retained.
\end{theorem}

\begin{proof}
Homogeneity of the coefficient functions of $E$ gives the
exact scaling identity
\[
 \langle T_\delta,E\wedge D\rho\rangle
       =\langle T,E\wedge D[\rho(x/\delta)]\rangle.
 \tag{PT12}
\]
Both $(1,1)$ factors together contribute $\delta^{-4}$
under real pullback, and MT's prefactor $\delta^2$
leaves precisely the $\delta^{-2}$ of the displayed
Hessian.  The original Borel module identification just
explained makes the right side equal to
$\langle M_S(E),D\chi_\delta\rangle$.

MT supplies the uniform trace bounds of PT9 on each
compact real set.  The fixed test $E\wedge D\rho$ is
compatible finite PS.  PT9 therefore applies to the left
side of PT12, even if the tangent current charges walls.
This proves the first equality in PT11.  Taking $E=Z_0$
and using the unconditional leading reduction LE9 proves
the second.  Tangent uniqueness is not asserted or needed.
\end{proof}

If two supplied ambient extensions represent the same actual
symbol, the original Borel module pairings coincide at every
$\delta$.  PT9 consequently gives the same tangent pairing
for both extensions.  This is the precise independence needed
here; support containment alone would not justify multiplying
an arbitrary discontinuous extension by a tangent current.

\paragraph{Remaining analytic realization gate.}
PT11 does not put a critical tangent into the retained analytic
weak-background class, does not construct an original analytic
background from an arbitrary real cone current, and does not
prove its pairing is zero for an actual-zero leading energy.
The realized tangent family is constrained by all original
refinements.  It is not identified with every positive closed
real current on a selected fan or with the full old Killing
dual.  The remaining cancellation is exactly the vanishing
of the paired value in PT11 for those realized backgrounds.

\paragraph{Inputs.}
The real coefficient inequalities and critical compactness are
MT2--MT15, with the middle-positive normalization of PT1.
The original-to-real identification uses the retained R6
regular-PS Borel module, GP's complete proper fibers, and
LE's actual scalar first-jet construction.  The finite-wall
proof and nodal extension are supplied above.  No general
current restriction, arbitrary analytic pushforward, or GK
test extension is used.