OpenMath

Proof source

Finite signed diffuse data and an exact relative obstacle problem

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. Finite signed diffuse data and an exact relative obstacle problem

Finite signed diffuse data and an exact relative obstacle problem

Keep the original arbitrary symmetric middle-positive separately closed (n−1,n−1)(n-1,n-1)-current SS, the analytic affinoid neighborhood Ω\Omega, and the ordinary part domain V⋐ΩV\Subset\Omega from the signed diffuse Dirichlet theorem. Write (H,E,cV,G,λ)(H,\mathcal E,c_V,G,\lambda) for its faithful part form and put Az=−D(zS).(OB1)Az=-D(zS). \tag{OB1} Thus Gσ=−uσG\sigma=-u_\sigma for a finite signed energy-diffuse measure σ\sigma, and A(Gσ)=σA(G\sigma)=\sigma on every original test. All functions below use their canonical energy-quasicontinuous representatives on FF. Inequalities mean quasi-everywhere on FF; no pointwise boundary condition or ambient classical psh condition is substituted for the specified part-form boundary class.

Lemma (Positive parts are Green potentials). Let σ\sigma be any finite signed energy-diffuse Radon measure and let z=Gσz=G\sigma. Then z+z^+ is the Green potential of its own finite signed diffuse curvature and Az+=1{z~>0}σ−L0(z;S),z+=G(Az+),L0(z;S)≥0.(OB2)\begin{split} Az^+&=\mathbf1_{\{\widetilde z>0\}}\sigma-L_0(z;S),\\ z^+&=G(Az^+),\qquad L_0(z;S)\ge0. \end{split} \tag{OB2} In particular, 1{z~>0}σ≤0⟹z≤0.(OB3)\mathbf1_{\{\widetilde z>0\}}\sigma\le0 \quad\Longrightarrow\quad z\le0. \tag{OB3} These assertions do not require zz or z+z^+ to belong to HH.

Status: DIRECT_COROLLARY of the full signed exact-level and global curvature-variation theorems, with the Green representation proved explicitly below.

Proof. The signed Kato theorem gives the first line of (OB2), and its level measure is finite and diffuse by (SD10)–(SD13). For M>0M>0 put vM=(TMz)+∈Hv_M=(T_Mz)^+\in H. The actual scalar identity vM=z+−(z−M)+v_M=z^+-(z-M)^+ and the signed level formula give D(vMS)=1{0<z~≤M}D(zS)+L0(z;S)−LM(z;S).(OB4)D(v_MS)=\mathbf1_{\{0<\widetilde z\le M\}}D(zS) +L_0(z;S)-L_M(z;S). \tag{OB4} Consequently ∥D(vMS)−D(z+S)∥TV≤2∣σ∣{z~>M}⟶0.\|D(v_MS)-D(z^+S)\|_{\rm TV} \le2|\sigma|\{\widetilde z>M\}\longrightarrow0. Each vMv_M is an HH vector with finite signed Radon curvature, so Lemma lem:r11-bounded-H-tests-diffuse identifies it with the transposition solution of that curvature. Total-variation to L1(λ)L^1(\lambda) stability then gives vM→uD(z+S)v_M\to u_{D(z^+S)}. Also vM→z+v_M\to z^+ in L1(λ)L^1(\lambda) by dominated convergence, since ∣vM∣≤∣z∣|v_M|\le|z|. This proves the second line of (OB2), including the canonical representative: its bounded clippings agree as HH vectors and hence quasi-everywhere.

Under the hypothesis in (OB3), (OB2) gives Az+≤0Az^+\le0. Order preservation of GG therefore gives z+=G(Az+)≤0z^+=G(Az^+)\le0. Thus z≤0z\le0. No test of an unbounded vector as if it belonged to HH occurs in this proof. ◻

Definition (Obstacles with finite curvature and the diffuse solution class). An allowed obstacle is a vector ψ∈H\psi\in H such that τ:=Aψ\tau:=A\psi is a finite signed Radon measure and ∣τ∣|\tau| has finite energy integral. In particular every real original compact smooth function is an allowed obstacle: its curvature has a bounded density with respect to a faithful reference measure, by the original compact coefficient control.

Fix a finite signed energy-diffuse datum μ\mu. A solution pair (w,ρ)(w,\rho) consists of a finite positive energy-diffuse Radon measure ρ\rho and a signed renormalized zero-boundary solution ww of Aw=μ+ρ(OB5)Aw=\mu+\rho \tag{OB5} such that w≥ψ,ρ{w~>ψ~}=0.(OB6)w\ge\psi,\qquad \rho\{\widetilde w>\widetilde\psi\}=0. \tag{OB6} The renormalized condition in (OB5) means the intrinsic class from Definition def:r11-signed-renormalized-boundary, with sign reversed: w∈L1(λ)w\in L^1(\lambda) is finite and quasicontinuous, Tkw∈HT_kw\in H for every k>0k>0, the original chain identity is D(Φ(w)S)=−Φ′(w~)(μ+ρ)+Φ′′(w~)ΓS(w)(OB7)D(\Phi(w)S)=-\Phi'(\widetilde w)(\mu+\rho) +\Phi''(\widetilde w)\Gamma_S(w) \tag{OB7} for C2C^2 scalar Φ\Phi with compactly supported first derivative, and k−1E(T2kw−Tkw)→0k^{-1}\mathcal E(T_{2k}w-T_kw)\to0. Thus the definition does not prescribe ww by a Green formula.

Theorem (The full finite-diffuse obstacle and contact theorem). For every allowed obstacle ψ\psi and every finite signed energy-diffuse datum μ\mu, there is exactly one solution pair in Definition “Obstacles with finite curvature and the diffuse solution class”. It satisfies w=G(μ+ρ),Aw=μ+ρ,0≤ρ≤(τ−μ)+≤τ++μ−,τ=Aψ.(OB8)\begin{split} w&=G(\mu+\rho),\qquad Aw=\mu+\rho,\\ 0\le\rho&\le(\tau-\mu)^+\le\tau^++\mu^-,\qquad \tau=A\psi. \end{split} \tag{OB8} The second line is the actual measure inequality, including the contact set and every original test; no density of the arbitrary background is assumed. In particular the reaction is finite.

The solution is the least member of the following explicitly nonempty class: Mμ,ψ={G(μ+η): η≥0 finite and energy diffuse, G(μ+η)≥ψ}.(OB9)\mathscr M_{\mu,\psi}= \{G(\mu+\eta):\ \eta\ge0\text{ finite and energy diffuse},\ G(\mu+\eta)\ge\psi\}. \tag{OB9} It is therefore an actual relative envelope, with a finite positive reaction carried by contact. All equations are original current equalities on the ordinary domain, for every compact smooth scalar test and every actual moment refinement.

Status: NEW_PROVED. This extends the finite-energy Hilbert obstacle problem to every finite signed diffuse datum. The obstacle hypothesis and part-form boundary class are retained explicitly; it is not an unrestricted classical obstacle or nonlinear Monge–Ampere existence assertion.

Proof. First consider a finite-energy datum μj\mu_j and the Hilbert obstacle solution wjw_j from Theorem “The original relative obstacle reaction”. Its reaction ρj\rho_j is positive, has finite energy integral, and is carried by {wj=ψ}\{w_j=\psi\}. It need not initially be assumed finite in total mass. Put fj=wj−ψ∈H+f_j=w_j-\psi\in H_+. Its curvature is D(fjS)=τ−μj−ρj.D(f_jS)=\tau-\mu_j-\rho_j. The total variation on the right has finite energy integral, so the HH level theorem (RT2) applies even if its total mass near the boundary is infinite. Since fj+=fjf_j^+=f_j, its zero-level formula is 1{f~j=0}(τ−μj−ρj)≥0.(OB10)\mathbf1_{\{\widetilde f_j=0\}} (\tau-\mu_j-\rho_j)\ge0. \tag{OB10} Combine this with the contact carrier of ρj\rho_j to obtain 0≤ρj≤(τ−μj)+.(OB11)0\le\rho_j\le(\tau-\mu_j)^+. \tag{OB11} This is the required finite-mass bound; it is derived before being used in any finite-data approximation argument.

Now let μj=1Fjμ\mu_j=\mathbf1_{F_j}\mu be an increasing finite-energy nest for ∣μ∣|\mu|, and solve these finite-energy obstacle problems. The signed measures converge in total variation to μ\mu and μj−≤μ−\mu_j^-\le\mu^-. Thus all reactions are dominated by the same finite diffuse measure π=τ++μ−,0≤ρj≤π.(OB12)\pi=\tau^++\mu^-,\qquad 0\le\rho_j\le\pi. \tag{OB12} For two indices put d=wj−wl∈Hd=w_j-w_l\in H. The source equation extends to Tkd∈HT_kd\in H. On {d>0}\{d>0\} the measure ρj\rho_j is zero, because wj>wl≥ψw_j>w_l\ge\psi there. On {d<0}\{d<0\} the measure ρl\rho_l is zero. Therefore E(Tkd)=∫Tkd~ d(μj−μl+ρj−ρl)≤k∥μj−μl∥TV,cV{∣d~∣>k}≤∥μj−μl∥TVk.(OB13)\begin{split} \mathcal E(T_kd) &=\int T_k\widetilde d\,d(\mu_j-\mu_l+\rho_j-\rho_l)\\ &\le k\|\mu_j-\mu_l\|_{\rm TV},\qquad c_V\{|\widetilde d|>k\} \le\frac{\|\mu_j-\mu_l\|_{\rm TV}}{k}. \end{split} \tag{OB13} The first equality follows from the closed clipping rule and the HH source identity. The reaction integral is nonpositive by the two contact observations. The capacity estimate follows by the normal clipped capacity test. Thus wjw_j is Cauchy in capacity.

Since wj=G(μj+ρj)w_j=G(\mu_j+\rho_j) by the finite-energy theorem, positivity of the transposition Green operator and (OB12) give ∣wj∣≤G(2∣μ∣+τ+)=:Bquasi-everywhere.|w_j|\le G(2|\mu|+\tau^+)=:B \quad\hbox{quasi-everywhere}. The fixed potential BB is finite quasi-everywhere and belongs to L1(λ)L^1(\lambda). The usual summable-capacity subsequence construction from (MD7)–(MD10) gives a finite quasicontinuous limit ww and capacity convergence of the whole sequence. Coercivity implies convergence in λ\lambda-measure; the displayed domination then gives wj→ww_j\to w in L1(λ)L^1(\lambda). The inequalities wj≥ψw_j\ge\psi pass outside a polar set to w≥ψw\ge\psi. For completeness, the source identity also gives E(Tkwj)≤k(∥μ∥TV+π(F))\mathcal E(T_kw_j)\le k(\|\mu\|_{\rm TV}+\pi(F)). Thus each sequence of bounded clippings has a weak HH limit, whose L2L^2 representative is TkwT_kw. Its canonical representative is the capacity limit: choose a summable-capacity subsequence, then apply Mazur’s lemma to its tails. The convex combinations still converge uniformly outside the same arbitrarily small-capacity exceptional sets, while their strong HH convergence identifies the canonical limit. Hence Tkw∈HT_kw\in H with the same quasi-everywhere values.

Write ρj=rjπ\rho_j=r_j\pi with 0≤rj≤10\le r_j\le1. The finite Radon space here is second countable, so L1(π)L^1(\pi) is separable. Weak-star sequential compactness gives a subsequence with rj⇀∗rr_j\rightharpoonup^*r in L∞(π)L^\infty(\pi), where 0≤r≤10\le r\le1. Set ρ=rπ\rho=r\pi. It is finite, positive and diffuse. For every f∈L∞(λ)f\in L^\infty(\lambda) the canonical GfGf is bounded and hence belongs to L1(π)L^1(\pi). Passing to the limit in the finite-energy Green identity gives ∫wf dλ=lim⁡j∫Gf~ d(μj+ρj)=∫Gf~ d(μ+ρ).(OB14)\int wf\,d\lambda =\lim_j\int\widetilde{Gf}\,d(\mu_j+\rho_j) =\int\widetilde{Gf}\,d(\mu+\rho). \tag{OB14} The first passage uses L1(λ)L^1(\lambda) convergence; the two datum passages use total variation and weak-star convergence, respectively. Thus w=G(μ+ρ)w=G(\mu+\rho). The signed diffuse theorem now supplies its complete intrinsic renormalized class and every original current equation, so no current-only uniqueness assumption is needed. The preceding clipped representative identification makes this an equality of canonical functions quasi-everywhere as well as in L1L^1.

For contact let g(s)=min⁡(1,s+)g(s)=\min(1,s^+). The functions g(wj−ψ)g(w_j-\psi) converge in capacity to g(w−ψ)g(w-\psi) and are bounded. Finite diffuseness of π\pi gives convergence in L1(π)L^1(\pi). The exact finite-energy contact condition and (OB12) therefore imply ∫g(w~−ψ~) dρ=lim⁡j∫g(w~−ψ~) dρj=lim⁡j∫g(w~j−ψ~) dρj=0.(OB15)\begin{split} \int g(\widetilde w-\widetilde\psi)\,d\rho &=\lim_j\int g(\widetilde w-\widetilde\psi)\,d\rho_j\\ &=\lim_j\int g(\widetilde w_j-\widetilde\psi)\,d\rho_j=0. \end{split} \tag{OB15} The first passage is weak-star convergence against the fixed bounded function. The second is bounded by the L1(π)L^1(\pi) error. Since w≥ψw\ge\psi, this proves the exact contact carrier. In particular the proof does not restrict an uncontrolled weak measure limit to a discontinuous contact set.

For any solution pair in the stated class, signed renormalized uniqueness first gives w=G(μ+ρ)w=G(\mu+\rho). Since ψ=Gτ\psi=G\tau, the difference f=w−ψ=G(μ+ρ−τ)f=w-\psi=G(\mu+\rho-\tau) is a signed diffuse Green potential and is nonnegative. Its signed zero-level Kato formula gives 1{f~=0}(τ−μ−ρ)≥0.\mathbf1_{\{\widetilde f=0\}} (\tau-\mu-\rho)\ge0. The contact carrier yields the second line of (OB8), exactly as in (OB10)–(OB11).

Finally let v=G(μ+η)∈Mμ,ψv=G(\mu+\eta)\in\mathscr M_{\mu,\psi}. Then z=w−v=G(ρ−η)z=w-v=G(\rho-\eta). On {z>0}\{z>0\} one has w>v≥ψw>v\ge\psi, so ρ\rho is zero there and 1{z>0}(ρ−η)≤0\mathbf1_{\{z>0\}}(\rho-\eta)\le0. Lemma “Positive parts are Green potentials” gives z≤0z\le0. This proves leastness. The class is nonempty before solving the obstacle: take η=τ++μ−\eta=\tau^++\mu^-, since G(μ+η)=G(μ++τ+)≥Gτ=ψG(\mu+\eta)=G(\mu^++\tau^+)\ge G\tau=\psi. Two solution pairs are each least, so their potentials coincide; their original equations then identify their reaction measures. This proves uniqueness and independence of the extracted subsequence. ◻

Theorem (Order and quantitative dependence on the signed source). For a fixed allowed obstacle, write wμw_\mu for the solution. For any two finite signed diffuse data μ,ν\mu,\nu one has ∣wμ−wν∣≤G∣μ−ν∣,∥wμ−wν∥L1(λ)≤2∥μ−ν∥TV,cV{∣w~μ−w~ν∣>k}≤∥μ−ν∥TVk.(OB16)\begin{split} |w_\mu-w_\nu|&\le G|\mu-\nu|,\\ \|w_\mu-w_\nu\|_{L^1(\lambda)}&\le2\|\mu-\nu\|_{\rm TV},\\ c_V\{|\widetilde w_\mu-\widetilde w_\nu|>k\} &\le\frac{\|\mu-\nu\|_{\rm TV}}{k}. \end{split} \tag{OB16} If μ≤ν\mu\le\nu, then wμ≤wνw_\mu\le w_\nu. For fixed source the solution also increases when the allowed obstacle increases. These are relative energy orders; no pointwise statement outside the coefficient support is asserted.

Status: NEW_PROVED for the finite signed diffuse obstacle map.

Proof. Put d=wμ−wνd=w_\mu-w_\nu and b=G∣μ−ν∣≥0b=G|\mu-\nu|\ge0. The signed Green potential z=d−bz=d-b has source Az=μ−ν−∣μ−ν∣+ρμ−ρν.Az=\mu-\nu-|\mu-\nu|+\rho_\mu-\rho_\nu. On {z>0}\{z>0\} one has d>b≥0d>b\ge0, so the contact condition gives ρμ=0\rho_\mu=0 there. The remaining restricted source is nonpositive. Apply (OB3) to obtain d≤bd\le b; interchange the two data to obtain −d≤b-d\le b. This is the first line of (OB16). The torsion bound and the finite-datum capacity bound for bb give the remaining two lines. If μ≤ν\mu\le\nu, apply (OB3) directly to dd, using its same contact sign, to get order. For ordered obstacles, the solution of the larger obstacle belongs to the class (OB9) for the smaller one, so leastness proves order. ◻

Corollary (Positive nest data give decreasing finite reactions). If μ≥0\mu\ge0 and μj↑μ\mu_j\uparrow\mu are its finite-energy nest restrictions, then their finite-energy obstacle solutions satisfy wj↑wμ,ρj↓ρμ,∥ρj−ρμ∥TV⟶0,0≤ρj≤τ+.(OB17)w_j\uparrow w_\mu,\qquad \rho_j\downarrow\rho_\mu,\qquad \|\rho_j-\rho_\mu\|_{\rm TV}\longrightarrow0,\qquad 0\le\rho_j\le\tau^+. \tag{OB17} In this positive-data case the limiting reaction itself has finite energy integral. The potential wμw_\mu need not belong to HH.

Status: NEW_PROVED as a stronger reaction convergence statement for positive nested data.

Proof. Order gives wl≥wjw_l\ge w_j for l≥jl\ge j. On the contact set {wl=ψ}\{w_l=\psi\} one also has wj=ψw_j=\psi. The nonnegative HH vector f=wl−wjf=w_l-w_j has curvature D(fS)=−(μl−μj)−(ρl−ρj).D(fS)=-(\mu_l-\mu_j)-(\rho_l-\rho_j). Its zero-level Kato inequality, restricted to the contact set of wlw_l, gives ρl≤ρj\rho_l\le\rho_j there because μl≥μj\mu_l\ge\mu_j. Off that set ρl=0\rho_l=0. Hence the reactions decrease as measures. Their common finite bound τ+\tau^+ from (OB11) gives total-variation convergence. The preceding uniqueness and source convergence identify the limits with wμ,ρμw_\mu,\rho_\mu. Domination by the finite-energy measure τ+\tau^+ gives finite energy integral for the reaction. This does not put the possibly infinite-energy source μ\mu or the whole potential wμw_\mu in HH. ◻

Dependency and scope audit.

The finite-energy obstacle/contact theorem is used only for the nest approximants. Its level-zero Kato consequence proves the uniform finite reaction bound before extracting measures. The signed renormalized theorem supplies both the limiting boundary class and the positive-part Green representation used in comparison; the unbounded potentials are never tested as if they were HH vectors. The common finite diffuse domination is what passes contact through the signed approximation. No ambient regularization of an arbitrary upper envelope, no current pullback under an auxiliary field, and no analytic PS/PL quotient is used. Removing the finite-curvature obstacle hypothesis, imposing ordinary pointwise boundary values, or realizing these envelopes by ambient classical psh functions requires additional arguments and is not asserted here.

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 7db4141cca2586fc8ab89b7c2db92443a3d8fdcb108cb4036aea40b110dbe6be

Frozen-original SHA-256 7db4141cca2586fc8ab89b7c2db92443a3d8fdcb108cb4036aea40b110dbe6be

\section{Finite signed diffuse data and an exact relative obstacle problem}
\label{sec:r11-diffuse-obstacle-contact}

Keep the original arbitrary symmetric middle-positive separately
closed $(n-1,n-1)$-current $S$, the analytic affinoid neighborhood
$\Omega$, and the ordinary part domain $V\Subset\Omega$ from the
signed diffuse Dirichlet theorem.  Write $(H,\mathcal E,c_V,G,\lambda)$
for its faithful part form and put
\[
                              Az=-D(zS).
 \tag{OB1}
\]
Thus $G\sigma=-u_\sigma$ for a finite signed energy-diffuse measure
$\sigma$, and $A(G\sigma)=\sigma$ on every original test.
All functions below use their canonical energy-quasicontinuous
representatives on $F$.  Inequalities mean quasi-everywhere on $F$;
no pointwise boundary condition or ambient classical psh condition
is substituted for the specified part-form boundary class.

\begin{lemma}[Positive parts are Green potentials]
\label{lem:r11-diffuse-positive-part-Green}
Let $\sigma$ be any finite signed energy-diffuse Radon measure and
let $z=G\sigma$.  Then $z^+$ is the Green potential of its own finite
signed diffuse curvature and
\[
 \begin{split}
 Az^+&=\mathbf1_{\{\widetilde z>0\}}\sigma-L_0(z;S),\\
 z^+&=G(Az^+),\qquad L_0(z;S)\ge0.
 \end{split}
 \tag{OB2}
\]
In particular,
\[
        \mathbf1_{\{\widetilde z>0\}}\sigma\le0
                              \quad\Longrightarrow\quad z\le0.
 \tag{OB3}
\]
These assertions do not require $z$ or $z^+$ to belong to $H$.
\end{lemma}

\noindent\textbf{Status: \texttt{DIRECT\_COROLLARY}} of the full
signed exact-level and global curvature-variation theorems, with the
Green representation proved explicitly below.

\begin{proof}
The signed Kato theorem gives the first line of (OB2), and its level
measure is finite and diffuse by (SD10)--(SD13).  For $M>0$ put
$v_M=(T_Mz)^+\in H$.  The actual scalar identity
$v_M=z^+-(z-M)^+$ and the signed level formula give
\[
 D(v_MS)=\mathbf1_{\{0<\widetilde z\le M\}}D(zS)
                                    +L_0(z;S)-L_M(z;S).
 \tag{OB4}
\]
Consequently
\[
 \|D(v_MS)-D(z^+S)\|_{\rm TV}
                   \le2|\sigma|\{\widetilde z>M\}\longrightarrow0.
\]
Each $v_M$ is an $H$ vector with finite signed Radon curvature, so
Lemma~\ref{lem:r11-bounded-H-tests-diffuse} identifies it with the
transposition solution of that curvature.  Total-variation to
$L^1(\lambda)$ stability then gives
$v_M\to u_{D(z^+S)}$.  Also $v_M\to z^+$ in $L^1(\lambda)$ by
dominated convergence, since $|v_M|\le|z|$.  This proves the second
line of (OB2), including the canonical representative: its bounded
clippings agree as $H$ vectors and hence quasi-everywhere.

Under the hypothesis in (OB3), (OB2) gives $Az^+\le0$.
Order preservation of $G$ therefore gives $z^+=G(Az^+)\le0$.
Thus $z\le0$.  No test of an unbounded vector as if it belonged
to $H$ occurs in this proof.
\end{proof}

\begin{definition}[Obstacles with finite curvature and the diffuse solution class]
\label{def:r11-diffuse-obstacle-class}
An allowed obstacle is a vector $\psi\in H$ such that
$\tau:=A\psi$ is a finite signed Radon measure and $|\tau|$ has
finite energy integral.  In particular every real original compact
smooth function is an allowed obstacle: its curvature has a bounded
density with respect to a faithful reference measure, by the
original compact coefficient control.

Fix a finite signed energy-diffuse datum $\mu$.  A solution pair
$(w,\rho)$ consists of a finite positive energy-diffuse Radon measure
$\rho$ and a signed renormalized zero-boundary solution $w$ of
\[
                               Aw=\mu+\rho
 \tag{OB5}
\]
such that
\[
                     w\ge\psi,\qquad
                     \rho\{\widetilde w>\widetilde\psi\}=0.
 \tag{OB6}
\]
The renormalized condition in (OB5) means the intrinsic class from
Definition~\ref{def:r11-signed-renormalized-boundary}, with sign
reversed: $w\in L^1(\lambda)$ is finite and quasicontinuous,
$T_kw\in H$ for every $k>0$, the original chain identity is
\[
 D(\Phi(w)S)=-\Phi'(\widetilde w)(\mu+\rho)
                         +\Phi''(\widetilde w)\Gamma_S(w)
 \tag{OB7}
\]
for $C^2$ scalar $\Phi$ with compactly supported first derivative,
and $k^{-1}\mathcal E(T_{2k}w-T_kw)\to0$.
Thus the definition does not prescribe $w$ by a Green formula.
\end{definition}

\begin{theorem}[The full finite-diffuse obstacle and contact theorem]
\label{thm:r11-diffuse-obstacle-contact}
For every allowed obstacle $\psi$ and every finite signed
energy-diffuse datum $\mu$, there is exactly one solution pair in
Definition~\ref{def:r11-diffuse-obstacle-class}.  It satisfies
\[
 \begin{split}
 w&=G(\mu+\rho),\qquad Aw=\mu+\rho,\\
 0\le\rho&\le(\tau-\mu)^+\le\tau^++\mu^-,\qquad \tau=A\psi.
 \end{split}
 \tag{OB8}
\]
The second line is the actual measure inequality, including the
contact set and every original test; no density of the arbitrary
background is assumed.  In particular the reaction is finite.

The solution is the least member of the following explicitly
nonempty class:
\[
 \mathscr M_{\mu,\psi}=
 \{G(\mu+\eta):\ \eta\ge0\text{ finite and energy diffuse},\
                         G(\mu+\eta)\ge\psi\}.
 \tag{OB9}
\]
It is therefore an actual relative envelope, with a finite positive
reaction carried by contact.  All equations are original current
equalities on the ordinary domain, for every compact smooth scalar
test and every actual moment refinement.
\end{theorem}

\noindent\textbf{Status: \texttt{NEW\_PROVED}}.  This extends the
finite-energy Hilbert obstacle problem to every finite signed diffuse
datum.  The obstacle hypothesis and part-form boundary class are
retained explicitly; it is not an unrestricted classical obstacle or
nonlinear Monge--Ampere existence assertion.

\begin{proof}
First consider a finite-energy datum $\mu_j$ and the Hilbert obstacle
solution $w_j$ from Theorem~\ref{thm:r11-relative-energy-obstacle}.
Its reaction $\rho_j$ is positive, has finite energy integral, and is
carried by $\{w_j=\psi\}$.  It need not initially be assumed finite
in total mass.  Put $f_j=w_j-\psi\in H_+$.  Its curvature is
\[
                           D(f_jS)=\tau-\mu_j-\rho_j.
\]
The total variation on the right has finite energy integral, so the
$H$ level theorem (RT2) applies even if its total mass near the
boundary is infinite.  Since $f_j^+=f_j$, its zero-level formula is
\[
        \mathbf1_{\{\widetilde f_j=0\}}
                         (\tau-\mu_j-\rho_j)\ge0.
 \tag{OB10}
\]
Combine this with the contact carrier of $\rho_j$ to obtain
\[
                          0\le\rho_j\le(\tau-\mu_j)^+.
 \tag{OB11}
\]
This is the required finite-mass bound; it is derived before being
used in any finite-data approximation argument.

Now let $\mu_j=\mathbf1_{F_j}\mu$ be an increasing finite-energy
nest for $|\mu|$, and solve these finite-energy obstacle problems.
The signed measures converge in total variation to $\mu$ and
$\mu_j^-\le\mu^-$.  Thus all reactions are dominated by the same
finite diffuse measure
\[
                         \pi=\tau^++\mu^-,\qquad 0\le\rho_j\le\pi.
 \tag{OB12}
\]
For two indices put $d=w_j-w_l\in H$.  The source equation extends
to $T_kd\in H$.  On $\{d>0\}$ the measure $\rho_j$ is zero,
because $w_j>w_l\ge\psi$ there.  On $\{d<0\}$ the measure
$\rho_l$ is zero.  Therefore
\[
 \begin{split}
 \mathcal E(T_kd)
   &=\int T_k\widetilde d\,d(\mu_j-\mu_l+\rho_j-\rho_l)\\
   &\le k\|\mu_j-\mu_l\|_{\rm TV},\qquad
 c_V\{|\widetilde d|>k\}
                         \le\frac{\|\mu_j-\mu_l\|_{\rm TV}}{k}.
 \end{split}
 \tag{OB13}
\]
The first equality follows from the closed clipping rule and the
$H$ source identity.  The reaction integral is nonpositive by the
two contact observations.  The capacity estimate follows by the
normal clipped capacity test.  Thus $w_j$ is Cauchy in capacity.

Since $w_j=G(\mu_j+\rho_j)$ by the finite-energy theorem, positivity
of the transposition Green operator and (OB12) give
\[
                  |w_j|\le G(2|\mu|+\tau^+)=:B
                                        \quad\hbox{quasi-everywhere}.
\]
The fixed potential $B$ is finite quasi-everywhere and belongs to
$L^1(\lambda)$.  The usual summable-capacity subsequence construction
from (MD7)--(MD10) gives a finite quasicontinuous limit $w$ and
capacity convergence of the whole sequence.  Coercivity implies
convergence in $\lambda$-measure; the displayed domination then
gives $w_j\to w$ in $L^1(\lambda)$.  The inequalities $w_j\ge\psi$
pass outside a polar set to $w\ge\psi$.
For completeness, the source identity also gives
$\mathcal E(T_kw_j)\le k(\|\mu\|_{\rm TV}+\pi(F))$.
Thus each sequence of bounded clippings has a weak $H$ limit, whose
$L^2$ representative is $T_kw$.  Its canonical representative is the
capacity limit: choose a summable-capacity subsequence, then apply
Mazur's lemma to its tails.  The convex combinations still converge
uniformly outside the same arbitrarily small-capacity exceptional
sets, while their strong $H$ convergence identifies the canonical
limit.  Hence $T_kw\in H$ with the same quasi-everywhere values.

Write $\rho_j=r_j\pi$ with $0\le r_j\le1$.  The finite Radon
space here is second countable, so $L^1(\pi)$ is separable.
Weak-star sequential compactness gives a subsequence with
$r_j\rightharpoonup^*r$ in $L^\infty(\pi)$, where $0\le r\le1$.
Set $\rho=r\pi$.  It is finite, positive and diffuse.  For every
$f\in L^\infty(\lambda)$ the canonical $Gf$ is bounded and hence
belongs to $L^1(\pi)$.  Passing to the limit in the finite-energy
Green identity gives
\[
 \int wf\,d\lambda
       =\lim_j\int\widetilde{Gf}\,d(\mu_j+\rho_j)
       =\int\widetilde{Gf}\,d(\mu+\rho).
 \tag{OB14}
\]
The first passage uses $L^1(\lambda)$ convergence; the two datum
passages use total variation and weak-star convergence, respectively.
Thus $w=G(\mu+\rho)$.  The signed diffuse theorem now supplies its
complete intrinsic renormalized class and every original current
equation, so no current-only uniqueness assumption is needed.
The preceding clipped representative identification makes this an
equality of canonical functions quasi-everywhere as well as in $L^1$.

For contact let $g(s)=\min(1,s^+)$.  The functions
$g(w_j-\psi)$ converge in capacity to $g(w-\psi)$ and are bounded.
Finite diffuseness of $\pi$ gives convergence in $L^1(\pi)$.
The exact finite-energy contact condition and (OB12) therefore imply
\[
 \begin{split}
 \int g(\widetilde w-\widetilde\psi)\,d\rho
 &=\lim_j\int g(\widetilde w-\widetilde\psi)\,d\rho_j\\
 &=\lim_j\int g(\widetilde w_j-\widetilde\psi)\,d\rho_j=0.
 \end{split}
 \tag{OB15}
\]
The first passage is weak-star convergence against the fixed bounded
function.  The second is bounded by the $L^1(\pi)$ error.  Since
$w\ge\psi$, this proves the exact contact carrier.  In particular
the proof does not restrict an uncontrolled weak measure limit to a
discontinuous contact set.

For any solution pair in the stated class, signed renormalized
uniqueness first gives $w=G(\mu+\rho)$.  Since $\psi=G\tau$,
the difference $f=w-\psi=G(\mu+\rho-\tau)$ is a signed diffuse
Green potential and is nonnegative.  Its signed zero-level Kato
formula gives
\[
                  \mathbf1_{\{\widetilde f=0\}}
                              (\tau-\mu-\rho)\ge0.
\]
The contact carrier yields the second line of (OB8), exactly as in
(OB10)--(OB11).

Finally let $v=G(\mu+\eta)\in\mathscr M_{\mu,\psi}$.
Then $z=w-v=G(\rho-\eta)$.  On $\{z>0\}$ one has
$w>v\ge\psi$, so $\rho$ is zero there and
$\mathbf1_{\{z>0\}}(\rho-\eta)\le0$.
Lemma~\ref{lem:r11-diffuse-positive-part-Green} gives $z\le0$.
This proves leastness.  The class is nonempty before solving the
obstacle: take $\eta=\tau^++\mu^-$, since
$G(\mu+\eta)=G(\mu^++\tau^+)\ge G\tau=\psi$.
Two solution pairs are each least, so their potentials coincide;
their original equations then identify their reaction measures.
This proves uniqueness and independence of the extracted subsequence.
\end{proof}

\begin{theorem}[Order and quantitative dependence on the signed source]
\label{thm:r11-diffuse-obstacle-stability}
For a fixed allowed obstacle, write $w_\mu$ for the solution.
For any two finite signed diffuse data $\mu,\nu$ one has
\[
 \begin{split}
 |w_\mu-w_\nu|&\le G|\mu-\nu|,\\
 \|w_\mu-w_\nu\|_{L^1(\lambda)}&\le2\|\mu-\nu\|_{\rm TV},\\
 c_V\{|\widetilde w_\mu-\widetilde w_\nu|>k\}
                              &\le\frac{\|\mu-\nu\|_{\rm TV}}{k}.
 \end{split}
 \tag{OB16}
\]
If $\mu\le\nu$, then $w_\mu\le w_\nu$.
For fixed source the solution also increases when the allowed
obstacle increases.  These are relative energy orders; no pointwise
statement outside the coefficient support is asserted.
\end{theorem}

\noindent\textbf{Status: \texttt{NEW\_PROVED}} for the finite
signed diffuse obstacle map.

\begin{proof}
Put $d=w_\mu-w_\nu$ and $b=G|\mu-\nu|\ge0$.
The signed Green potential $z=d-b$ has source
\[
            Az=\mu-\nu-|\mu-\nu|+\rho_\mu-\rho_\nu.
\]
On $\{z>0\}$ one has $d>b\ge0$, so the contact condition gives
$\rho_\mu=0$ there.  The remaining restricted source is
nonpositive.  Apply (OB3) to obtain $d\le b$; interchange the two
data to obtain $-d\le b$.  This is the first line of (OB16).
The torsion bound and the finite-datum capacity bound for $b$
give the remaining two lines.  If $\mu\le\nu$, apply (OB3)
directly to $d$, using its same contact sign, to get order.
For ordered obstacles, the solution of the larger obstacle belongs
to the class (OB9) for the smaller one, so leastness proves order.
\end{proof}

\begin{corollary}[Positive nest data give decreasing finite reactions]
\label{cor:r11-positive-obstacle-reaction-TV}
If $\mu\ge0$ and $\mu_j\uparrow\mu$ are its finite-energy nest
restrictions, then their finite-energy obstacle solutions satisfy
\[
       w_j\uparrow w_\mu,\qquad
       \rho_j\downarrow\rho_\mu,\qquad
       \|\rho_j-\rho_\mu\|_{\rm TV}\longrightarrow0,\qquad
       0\le\rho_j\le\tau^+.
 \tag{OB17}
\]
In this positive-data case the limiting reaction itself has finite
energy integral.  The potential $w_\mu$ need not belong to $H$.
\end{corollary}

\noindent\textbf{Status: \texttt{NEW\_PROVED}} as a stronger
reaction convergence statement for positive nested data.

\begin{proof}
Order gives $w_l\ge w_j$ for $l\ge j$.  On the contact set
$\{w_l=\psi\}$ one also has $w_j=\psi$.
The nonnegative $H$ vector $f=w_l-w_j$ has curvature
\[
                         D(fS)=-(\mu_l-\mu_j)-(\rho_l-\rho_j).
\]
Its zero-level Kato inequality, restricted to the contact set of
$w_l$, gives $\rho_l\le\rho_j$ there because $\mu_l\ge\mu_j$.
Off that set $\rho_l=0$.  Hence the reactions decrease as measures.
Their common finite bound $\tau^+$ from (OB11) gives total-variation
convergence.  The preceding uniqueness and source convergence identify
the limits with $w_\mu,\rho_\mu$.  Domination by the finite-energy
measure $\tau^+$ gives finite energy integral for the reaction.
This does not put the possibly infinite-energy source $\mu$ or the
whole potential $w_\mu$ in $H$.
\end{proof}

\paragraph{Dependency and scope audit.}
The finite-energy obstacle/contact theorem is used only for the nest
approximants.  Its level-zero Kato consequence proves the uniform
finite reaction bound before extracting measures.  The signed
renormalized theorem supplies both the limiting boundary class and
the positive-part Green representation used in comparison; the
unbounded potentials are never tested as if they were $H$ vectors.
The common finite diffuse domination is what passes contact through
the signed approximation.  No ambient regularization of an arbitrary
upper envelope, no current pullback under an auxiliary field, and no
analytic PS/PL quotient is used.  Removing the finite-curvature
obstacle hypothesis, imposing ordinary pointwise boundary values, or
realizing these envelopes by ambient classical psh functions requires
additional arguments and is not asserted here.