Closed

A scalar-measure construction with explicit domination

Theorem. (A scalar-measure construction with explicit domination). Let S:Ep,p(U)→RS:\mathcal E^{p,p}(U)\to\mathbb R be linear and nonnegative on every compact form nonnegative on pp-planes. Then SS is a native current. For every α∈Ep,p(U)\alpha\in\mathcal E^{p,p}(U) there is a unique finite signed Radon measure νS,α\nu_{S,\alpha} with compact support in supp⁡α\operatorname{supp}\alpha, such that ∫h dνS,α=S(hα)for every ordinary smooth scalar h.\begin{equation} \int h\,d\nu_{S,\alpha}=S(h\alpha) \quad\text{for every ordinary smooth scalar }h. \end{equation} It is linear in α\alpha, positive when α\alpha is nonnegative on pp-planes, local under ordinary open restriction, and supported in supp⁡S∩supp⁡α\operatorname{supp}S\cap\operatorname{supp}\alpha.

For each cutoff stage there are a fixed compact form θi\theta_i nonnegative on pp-planes and a constant ci≥0c_i\geq0 such that ∣νS,Qiη∣≤ciptf(V),0(η)νS,θi.\begin{equation} |\nu_{S,Q_i\eta}|\leq c_i p_{t_f(V),0}(\eta)\nu_{S,\theta_i}. \end{equation} The assertions extend linearly and canonically to a linear functional whose ordinary local restrictions are finite signed sums of such SS. In that case the right side is replaced locally by the sum of the absolute coefficients times the corresponding positive measures.

For such an SS and any continuous real vv, the formula (vS)(α)=∫v dνS,α(vS)(\alpha)=\int v\,d\nu_{S,\alpha} defines a native current, independently of its signed presentation, with support contained in supp⁡S∩supp⁡v\operatorname{supp}S\cap\operatorname{supp}v. This conclusion is derived from the stated sign and domination hypotheses; order zero alone is not a premise sufficient for it.

Proof

Proof. Fix a stage with NN moment coordinates. By C.1.60, choose 0≤ρ≤10\leq\rho\leq1 compactly supported in Int⁡top(V,U)\operatorname{Int}_{\rm top}(V,U) and equal to one near CiC_i. Put M=sup⁡∣χ∣M=\sup|\chi| and Θp=1p!(∑k=1Nd′xk∧d′′xk)p,θi=Mext⁡0(ρtf∗Θp).\Theta_p=\frac1{p!}\left(\sum_{k=1}^N d'x_k\wedge d''x_k\right)^p, \qquad \theta_i=M\operatorname{ext}_0(\rho t_f^*\Theta_p). For p=0p=0 put Θ0=1\Theta_0=1; for N<pN<p the stage is zero. On a projected decomposable pp-vector zz, Θp\Theta_p has coefficient ∣z∣2|z|^2 and an arbitrary representative η\eta has coefficient ztAηzz^{\mathsf t}A_\eta z in the positive convention. The antisymmetric part contributes zero. A finite matrix bound gives, on every actual joint refinement, including rank drop, ciptf(V),0(η)θi±Qiη ≥0on p-planes.\begin{equation} c_i p_{t_f(V),0}(\eta)\theta_i\pm Q_i\eta\ \geq0 \quad\text{on }p\text{-planes}. \end{equation} The scalar cutoffs cause no difficulty: ρ=1\rho=1 on CiC_i, ∣χ∣≤M|\chi|\leq M, and QiηQ_i\eta vanishes off CiC_i. Positivity of SS gives ∣S(Qiη)∣≤ciS(θi)ptf(V),0(η)|S(Q_i\eta)|\leq c_iS(\theta_i)p_{t_f(V),0}(\eta). Finality in 8.15.4 proves continuity.

Every compact form α\alpha is a finite sum of cutoff-stage images, by C.1.60 and the ordinary sheaf. Applying (C.1.35) to these fixed summands gives a compact form θ\theta nonnegative on pp-planes with θ±α≥0\theta\pm\alpha\geq0. Let C=supp⁡θ∪supp⁡αC=\operatorname{supp}\theta\cup\operatorname{supp}\alpha. For every ordinary smooth hh, the forms ∥h∥Cθ±hα\|h\|_C\theta\pm h\alpha are nonnegative on pp-planes. Therefore ∣S(hα)∣≤S(θ)∥h∥C.\begin{equation} |S(h\alpha)|\leq S(\theta)\|h\|_C. \end{equation} This is a bound by one scalar supremum on a fixed compact set, stronger than an abstract ambient-stage order-zero statement.

We give the measure argument explicitly. Restrictions of ordinary smooth scalars are uniformly dense in C(C,R)C(C,\mathbb R): cover CC by finitely many opens on which a given continuous function oscillates by less than δ\delta, and combine sample values with the nonnegative finite partition from C.1.60. For a nonnegative target the approximants may be chosen nonnegative. Thus (C.1.36) extends h↦S(hα)h\mapsto S(h\alpha) uniquely to a bounded real linear functional Λ\Lambda on C(C)C(C).

Here is the elementary representation construction, so no abstract measure assertion is being inferred from the word “current.” For a positive bounded functional PP on C(C)C(C) define a(O)=sup⁡{P(f):0≤f≤1, supp⁡f⊂O},a∗(E)=inf⁡O⊃E opena(O).a(O)=\sup\{P(f):0\leq f\leq1,\ \operatorname{supp}f\subset O\},\qquad a^*(E)=\inf_{O\supset E\text{ open}}a(O). Continuous compact cutoffs and finite partitions on CC are obtained by restricting the native cutoffs just used. Partitions on the compact support of ff prove countable subadditivity of aa on opens; disjoint opens give the reverse finite inequality by summing disjointly supported ff. Hence a∗a^* is an outer measure. To check that an open OO is measurable, choose V⊃EV\supset E with a(V)≤a∗(E)+δa(V)\leq a^*(E)+\delta and 0≤g≤10\leq g\leq1 supported in V∩OV\cap O, with P(g)≥a(V∩O)−δP(g)\geq a(V\cap O)-\delta. Write F=supp⁡gF=\operatorname{supp}g. Tests supported in V∖FV\setminus F have disjoint support from gg, so a(V∖F)≤a(V)−P(g),a∗(E∩O)+a∗(E∖O)≤a∗(E)+2δ.a(V\setminus F)\leq a(V)-P(g),\qquad a^*(E\cap O)+a^*(E\setminus O)\leq a^*(E)+2\delta. Let δ\delta decrease to zero. The reverse inequality is subadditivity. The sets satisfying this splitting identity form a sigma-algebra: for disjoint such sets, iterate the identity finitely, then use the supremum of the finite sums and outer subadditivity for the countable union. The same argument proves countable additivity on that sigma-algebra. Thus a∗a^* restricts to a finite Borel measure μ\mu on CC.

If ff is an admissible test for a(O)a(O), then P(f)≤μ(supp⁡f)P(f)\leq\mu(\operatorname{supp}f) by taking the infimum over open neighborhoods of its support. It follows that μ(O)=a(O)\mu(O)=a(O) is the supremum of μ(F)\mu(F) over compact F⊂OF\subset O. Outer regularity is built into a∗a^*; inner regularity on all Borel sets follows by complements and finiteness. Hence μ\mu is Radon. To verify P(u)=∫u dμP(u)=\int u\,d\mu for continuous u≥0u\geq0, use lower step sums δ∑k=1mμ{u>kδ}\delta\sum_{k=1}^{m}\mu\{u>k\delta\} when u≤mδu\leq m\delta. For the upper estimate use the continuous functions vk=min⁡(1,max⁡(0,(u−kδ)/δ))v_k=\min(1,\max(0,(u-k\delta)/\delta)) with u=δ∑k=0m−1vku=\delta\sum_{k=0}^{m-1}v_k. The resulting upper and lower bounds differ from the corresponding integral by at most 2δμ(C)2\delta\mu(C). Let δ↓0\delta\downarrow0. Linearity gives the representation, and compact cutoffs give uniqueness.

For a bounded real Λ\Lambda, put Λ+(u)=sup⁡0≤v≤uΛ(v)\Lambda^+(u)=\sup_{0\leq v\leq u}\Lambda(v) for u≥0u\geq0. This is bounded, positive and additive: split v≤u1+u2v\leq u_1+u_2 into min⁡(v,u1)\min(v,u_1) and its remainder for one inequality, and add near maximizing choices for the other. Extend linearly; both Λ+\Lambda^+ and Λ+−Λ\Lambda^+-\Lambda are positive. Their Radon representations give a signed representation of Λ\Lambda. Positive uniqueness implies signed uniqueness. Extending by zero from CC gives (C.1.33) on UU.

Tests supported outside supp⁡α\operatorname{supp}\alpha, or in an open where SS vanishes, have zero scalar pairing. Smooth density and Radon regularity on compact subsets of such opens prove the claimed support. The same uniqueness on a compact union of supports proves linearity, locality and independence from CC or θ\theta. If α≥0\alpha\geq0 on pp-planes, nonnegative smooth scalar tests give a positive measure. Apply this positivity to (C.1.35); the measure inequalities imply (C.1.34). Indeed −aλ≤ν≤aλ-a\lambda\leq\nu\leq a\lambda implies ∣ν∣≤aλ|\nu|\leq a\lambda by summing over finite Borel partitions.

Finite signed sums inherit scalar measures, the variation bound, and uniqueness. For local signed presentations, choose a fixed finite partition near each compact test support; summing the measures for the partitioned tests gives the same representation and hence the same measure on overlaps. For continuous vv, compactness and (C.1.34) give on each stage ∣(vS)(Qiη)∣≤cisup⁡Ci∣v∣ νS,θi(U)ptf(V),0(η)|(vS)(Q_i\eta)|\leq c_i\sup_{C_i}|v|\,\nu_{S,\theta_i}(U) p_{t_f(V),0}(\eta) in the positive case, and the corresponding finite sum in the signed case. A fixed local partition gives the same kind of bound. This proves native continuity, and locality proves the stated support. ◻

Citation

Non-Archimedean Geometry Foundations, Tag NAG-003174.

Proof in chapter · Fixed-edition record