13 · Continuous Bedford–Taylor products

21 NAG Tags

13.1 Continuous products and the role of the potential

The mixed product of Theorem 11.1.5 can be understood through two operations: multiplication of a positive current by a continuous function, and distributional differentiation. Their domains explain why the notation d′d′′u∧Td'd''u\wedge T is not an unrestricted wedge product of currents.

A recursive construction with stated regularity

Proposition 13.1.1. (Multiplication and weak convergence). Let TjT_j be ordinary positive currents of one fixed bidegree converging weakly to an ordinary positive current TT. Let uju_j be continuous functions converging locally uniformly to a continuous function uu. Then ujTj⟶uTu_jT_j\longrightarrow uT as currents. The products are defined through the Radon measures obtained by testing the positive currents against smooth complementary forms.

Open proof obligation (RC2). Prove uniform approximation near the test support in the required category; the proof explicitly identifies CLD12 Corollaries 5.4.7–5.4.8 as the density input.

NAG-900257 · Proof status

Definition 13.1.2. (One continuous potential acting on a closed current). Let TT be an ordinary positive current with d′T=d′′T=0d'T=d''T=0, and let u∈C25u\in\mathcal C_{25}. Define d′d′′u∧T:=d′d′′(uT).\begin{equation} d'd''u\wedge T:=d'd''(uT). \end{equation} At this point the right side, and therefore the construction, uses the specified function uu. Dependence only on its curvature current is a separate question.

Proposition 13.1.3. (Positivity and closedness). The current in (13.1.1) is ordinary positive and is killed by both d′d' and d′′d''. If uu is smooth, it is the usual product of the smooth form d′d′′ud'd''u with TT.

Open proof obligation (RC2). The argument uses multiplier convergence, whose native density premise is open.

NAG-900258 · Proof status

This argument uses the continuous generated sheaf directly. It does not require the disputed all-stage equality with C0∩P25C^0\cap\mathcal P_{25}.

Proposition 13.1.4. (Iteration and local uniform stability). Let TT have bidegree (p,p)(p,p) and satisfy the preceding positivity and closedness assumptions. For u1,…,uq∈C25u_1,\ldots,u_q\in\mathcal C_{25}, with p+q≤np+q\leq n, the recursive expression d′d′′(u1d′d′′(u2⋯d′d′′(uqT)⋯ ))\begin{equation} d'd''\bigl(u_1d'd''(u_2\cdots d'd''(u_qT)\cdots)\bigr) \end{equation} is a positive closed current of bidegree (p+q,p+q)(p+q,p+q). If Tj→TT_j\to T weakly through positive closed currents and each ui,j→uiu_{i,j}\to u_i locally uniformly through C25\mathcal C_{25}, the corresponding recursive products converge weakly to (13.1.2).

Open proof obligation (RC2). The induction uses one-step positivity and multiplier convergence. The cited CLD25 recursion is explanatory here and its citation does not discharge either open native premise.

NAG-900259 · Proof status

Open proof obligation (RC3). For ui∈Uu_i\in\mathcal U, the symmetry and independence of smooth approximations in Theorem 11.1.5 remain pending. For smooth approximants on a common smaller neighbourhood, the smooth products are symmetric. Passing this identity to the recursive products requires the local uniform stability in Proposition 13.1.4, which in turn depends on the open multiplier-convergence assertion in Proposition 13.1.1. Strongly positive forms need not span all symmetric forms.

NAG-900104 · Proof status

Independence of the potential is another assertion

Definition 13.1.5. (The continuous Bedford–Taylor property). A positive closed current TT has the continuous Bedford–Taylor property on a specified class of continuous potentials if, locally, d′d′′[u]=d′d′′[v]⟹d′d′′(uT)=d′d′′(vT)\begin{equation} d'd''[u]=d'd''[v]\quad\Longrightarrow\quad d'd''(uT)=d'd''(vT) \end{equation} for all potentials u,vu,v in that class. This is the continuous interpretation of the property singled out in [CLD25] (Definition 8.6.2).

Remark 13.1.6 (A proved case and a missing inference). The integration current [1][1] has this property: u[1]=[u]u[1]=[u], so (13.1.3) is exactly its hypothesis. For an arbitrary positive closed TT, the implication also holds if u−vu-v is smooth with d′d′′(u−v)=0d'd''(u-v)=0, by the Leibniz calculation above. If u−vu-v is merely continuous and its current Hessian vanishes, the same argument is unavailable. This is why the potential-independence condition is stated separately in [CLD25] (§8.6.1).

Similarly, a positive continuous scalar multiplier preserves positivity of a current but need not preserve either closedness or the Bedford–Taylor property. Example A.1.3 disproves the unrestricted multiplication assertion printed in Corollary 8.6.7 of that text.

Metrics, measures, and actual norms

For a line bundle with continuous local weights in U\mathcal U, the products of its curvature currents are defined locally by the preceding construction and glue: changes of frame add the smooth pluriharmonic function −log⁡∣f∣-\log|f| for an analytic unit ff. This is a legitimate use of the smooth case in Remark 13.1.6. The argument concerns currents and their products; it does not forget the normalization of the original metric.

Remark 13.1.7 (Constants are invisible to curvature). If ∥⋅∥′=e−c∥⋅∥\|\cdot\|'=e^{-c}\|\cdot\|, then its weights satisfy φe′=φe+c\varphi'_e=\varphi_e+c, and c1(L,∥⋅∥′)=c1(L,∥⋅∥).c_1(L,\|\cdot\|')=c_1(L,\|\cdot\|). The two norms are nevertheless different for c≠0c\ne0. Equality of currents or of top-degree measures therefore cannot identify actual metrics. This distinction persists when currents are used in Deligne pairings or in adelic intersection theory.

The passage to unbounded potentials requires an integrability domain for products such as uTuT. The continuous recursion above supplies no automatic product when that multiplication is undefined. In particular, it supplies neither a general non-pluripolar product nor a finite-energy theory by notation alone. Those constructions require their own definitions and comparison theorems.

13.2 The estimates and identities behind continuous products

What a Chern–Levine–Nirenberg estimate controls

Fix an ordinary positive closed current TT, a relatively compact test region, and a compactly supported smooth complementary form ω\omega. A local CLN estimate bounds expressions of the shape ∣⟨u0 d′d′′u1∧⋯∧d′d′′uq∧T,ω⟩∣≤C∏i=0q∥ui∥K,\left|\left\langle u_0\,\mathrm d'\mathrm d''u_1\wedge\cdots\wedge\mathrm d'\mathrm d''u_q\wedge T, \omega\right\rangle\right| \leq C\prod_{i=0}^q\|u_i\|_K, for smooth psh functions on a fixed larger neighbourhood, with the usual choice of nested compact sets and cutoffs. The constant depends on the region, the fixed tests and cutoffs, and the relevant local mass of TT. It is not uniform over an unbounded family of background currents.

The estimate is striking because the left side contains second derivatives of every uiu_i, while the right side contains only their values. Positivity is the reason derivatives can be removed without losing control. It also explains why an estimate for arbitrary signed currents would need a different argument.

The induction mechanism

Choose nested compact neighbourhoods of the support of ω\omega and cutoffs equal to one on the smaller neighbourhoods. Positive-cone domination bounds the coefficients of the test by finitely many positive tests. For one curvature factor, integration by parts moves d′d′′\mathrm d'\mathrm d'' from u1u_1 to a cutoff times the test. The derivatives now fall on fixed smooth data. Their coefficients are bounded on the larger compact set, so ordinary positivity bounds the resulting pairing by a fixed local mass of TT, times ∥u1∥K\|u_1\|_K.

Repeat this step for the remaining factors, enlarging the compact region at each stage. The closedness of the inner mixed current removes derivatives of that current. Positivity allows the same coefficient estimate at the next step. The scalar u0u_0 is controlled directly by its sup norm. This is the structure of the CLN argument in CLD25 Chapter 8; the technical formulation keeps track of the degrees and support sets.

There is a printed degree slip in Proposition 8.4.1: the symbol pp must be read as the bidimension of the initial current for the displayed pairing to have the stated complementary degree. The estimate is used here with explicitly complementary test degrees, as explained in Appendix A.

Symmetry on the continuous generated class

The recursive expression in the preceding section is written in an order. Its symmetry is a theorem. For smooth potentials, it follows from the wedge algebra. For two continuous generated potentials u,vu,v, first fix a smooth uu and pass to the generated continuous class in vv: the identity d′d′′(u d′d′′(vT))=d′d′′(v d′d′′(uT))\mathrm d'\mathrm d''(u\,\mathrm d'\mathrm d''(vT))=\mathrm d'\mathrm d''(v\,\mathrm d'\mathrm d''(uT)) is local and stable under the uniform limits allowed in the definition, by the multiplier-convergence theorem. Now fix such a vv and make the same argument in uu. This proves the two-factor symmetry on the continuous class. Adjacent transpositions give symmetry for any finite number of factors.

This argument does not require replacing every generated function by one globally chosen sequence of original smooth psh functions. It uses the defining closure property of the continuous sheaf. It also does not establish the separate kernel property: if two different potentials have the same ambient Hessian, their actions on an arbitrary background current still need comparison.

Differences of psh functions

CLD25 calls the real vector sheaf generated by its continuous psh functions the dpsh class. Locally one can write a member as a difference of continuous psh functions. Products extend multilinearly by expanding such differences. Independence of the chosen positive decomposition follows from linearity of the scalar multiplication and distributional differentiation in the recursive construction, together with the established symmetry.

For example, a smooth function on a relatively compact real chart can be written as u=(u+C∣x∣2)−C∣x∣2u=(u+C|x|^2)-C|x|^2 with both terms convex when CC dominates the negative part of the Hessian. Pulling back this local argument explains why original smooth functions belong to the dpsh class. PL and G-smooth functions require the corresponding piecewise convex decomposition and the source’s local chart hypotheses.

The word dpsh does not assert positivity. A dpsh curvature is generally a difference of positive currents. It is a useful domain for multilinear calculus, just as differences of semipositive metrics are useful in arithmetic intersection theory.

Projection formulas and the direction of functoriality

For a proper morphism in the category where smooth pullback and current pushforward are defined, the elementary identity f∗(f∗u S)=u f∗Sf_*(f^*u\,S)=u\,f_*S comes from evaluating both sides on the same test and using scalar multiplication. Commuting f∗f_* with the distributional derivatives then gives f∗(d′d′′(f∗u)∧S)=d′d′′u∧f∗Sf_*\bigl(\mathrm d'\mathrm d''(f^*u)\wedge S\bigr) =\mathrm d'\mathrm d''u\wedge f_*S for continuous potentials in the admitted class and positive closed SS. Iteration gives the mixed version. The support/properness hypothesis ensures that pulled-back tests have the support needed to evaluate SS.

This is a pushforward theorem with pulled-back potentials. It is not a general pullback construction for arbitrary currents. The latter would require defining the action on tests in the opposite direction and controlling singular intersections. This distinction is important when interpreting the functorial psh condition of GK and the analytic curve tests of Gubler–Rabinoff.

13.3 Polar sets, integrability, and the boundary of the BT theorem

The precise input in the polar-negligibility theorem

Theorem 8.12.1 assumes the local existence of suitable smooth psh sublevel neighbourhoods. Its mixed measure is constructed from continuous locally approachable potentials u1,…,unu_1,\ldots,u_n. The auxiliary function VV is itself locally approachable psh and locally integrable. The conclusion is that VV is locally integrable for the mixed measure.

The analytic version is Theorem 16.9.2. For the good boundaryless analytic setting used in these notes, the preceding sublevel construction provides the needed local geometry. A formula error in the printed smoothing kernel is corrected in Appendix A; the normalization and domain of the regularized maximum are made explicit there.

These hypotheses are narrower than a theorem about every generated psh function and every positive closed background current. In particular, the theorem is not a full statement of capacity theory for arbitrary discontinuous bounded potentials.

Why integrability implies no polar mass

Suppose V≤0V\leq0 near a compact set and ∫(−V) dμ<∞\int(-V)\,d\mu<\infty. For M>0M>0, M μ{V<−M}≤∫{V<−M}(−V) dμ≤∫(−V) dμ.M\,\mu\{V<-M\}\leq\int_{\{V<-M\}}(-V)\,d\mu \leq\int(-V)\,d\mu. The sets {V<−M}\{V<-M\} decrease to {V=−∞}\{V=-\infty\} as M→∞M\to\infty. Hence the latter has zero μ\mu-mass. This elementary argument is the final step; the mathematical work lies in proving the integrability bound for the mixed measure.

The source obtains that bound by combining CLN estimates with decreasing smooth approximants of VV and uniform smooth approximants of the continuous uiu_i. A compactly supported psh barrier controls the localization. Positivity and monotone convergence then pass the estimate to VV. Each approximation has a specific role; a general weak limit of measures would not by itself preserve the desired integrability bound.

Polar sets and small tropical dimension are different criteria

CLD25 Chapter 16 also proves a support theorem for positive closed currents. Under the nontrivial valuation and boundaryless hypotheses, Theorem 16.8.5 says that a positive closed current of bidimension (p,p)(p,p) does not charge, after pairing with a smooth (p,p)(p,p)-form, a closed set whose image in every relevant tropical chart has zero pp-dimensional Hausdorff measure.

This criterion is geometric and dimension-theoretic. Polar-negligibility is formulated through a potential taking the value −∞-\infty. One should not identify the two notions without another theorem. A set may be small in one chosen tropical image while being visible in another; the support theorem requires all the prescribed charts.

What a complete bounded theory would add

Continuous scalar multiplication is canonical because positive currents have measure coefficients. If a bounded potential is discontinuous, multiplication remains meaningful once a measurable representative is fixed, but the continuity theorem needed for iteration changes. Locally uniform convergence is no longer available. One must control decreasing limits, exceptional sets, and the dependence on the representative.

In complex pluripotential theory, capacities, quasi-continuity, and monotone convergence make these operations compatible. A corresponding non-Archimedean theory must specify its psh class and its test space before claiming the analogy. The toric theory of BGJK and global variational theories provide important established cases, treated in Chapter 21; they should be compared on their actual domains. The continuous arguments above are a foundation for that comparison, not a proof of every bounded or finite-energy statement.

References. [CLD25] (§§8.4–8.12, 16.8–16.9).

13.4 Local mass estimates

The first product theorem proves positivity by approximation. Once positivity is known, integration by parts gives a mass estimate depending on the potential itself, not on derivatives of its approximants. This is the estimate needed for repeated products.

Lemma 13.4.1. For a smooth real function χ\chi, the coefficient matrix of d′d′′χ∧βq−1\mathrm d'\mathrm d''\chi\wedge\beta_{q-1} satisfies ∥Hd′d′′χ∧βq−1∥op≤q∥D2χ∥op.\left\lVert H_{\mathrm d'\mathrm d''\chi\wedge\beta_{q-1}}\right\rVert_{\mathrm{op}} \leq q\left\lVert D^2\chi\right\rVert_{\mathrm{op}}.

Proof. At the point in question, diagonalize the real symmetric Hessian by an orthogonal basis change, with eigenvalues λ1,…,λn\lambda_1,\ldots,\lambda_n. Such a change preserves ω\omega and all the exterior Euclidean norms. The coefficient matrix of the wedge is diagonal; its entry indexed by a qq-element set II is ∑i∈Iλi\sum_{i\in I}\lambda_i. This follows by expanding βq−1\beta_{q-1} and observing that only indices different from the Hessian index survive. The absolute value of each entry is at most qmax⁡i∣λi∣q\max_i|\lambda_i|. ◻

Proposition 13.4.2. (One-step mass bound). Let T,u,R=d′d′′(uT)T,u,R=\mathrm d'\mathrm d''(uT) satisfy Theorem 9.8.1. Choose 0≤χ∈Cc∞(Ω)0\leq\chi\in C_c^\infty(\Omega) equal one near KK, and a compact LL containing supp⁡χ\operatorname{supp}\chi. For any c∈Rc\in\mathbb R, μR(K)≤q∥D2χ∥∞,op∥u−c∥L,∞ μT(L).\begin{equation} \mu_R(K)\leq q\left\lVert D^2\chi\right\rVert_{\infty,\mathrm{op}} \left\lVert u-c\right\rVert_{L,\infty}\,\mu_T(L). \end{equation}

Proof. Since RR is positive and χ=1\chi=1 on KK, μR(K)≤R(χβq−1)=T(u d′d′′χ∧βq−1).\mu_R(K)\leq R(\chi\beta_{q-1}) =T(u\,\mathrm d'\mathrm d''\chi\wedge\beta_{q-1}). Here βq−1\beta_{q-1} is constant and (9.7.1) is valid on the continuous product uTuT. Closedness of TT gives T(d′d′′(χβq−1))=0T(\mathrm d'\mathrm d''(\chi\beta_{q-1}))=0, so uu may be replaced by u−cu-c. Now apply (9.3.5) and Lemma 13.4.1; all coefficients of the test are supported in LL. ◻

The constant in this bound depends on the cutoff and on the bidimension, not on the rate of approximation of uu. Subtracting a constant is legitimate because the background is closed. In applications one may take c=inf⁡Luc=\inf_Lu, replacing the potential norm by its oscillation on LL.

13.5 Mixed products in the approximation class

We now prove iteration and symmetry for the explicit approximation class used above. This result is deliberately stated with that class in its hypotheses; it makes no identification with a larger generated or current-psh class.

Definition. Write U(Ω)\mathcal U(\Omega) for the continuous functions that locally admit smooth approximants with positive Hessian, converging locally uniformly on one fixed neighbourhood.

Theorem 13.5.2. Let TT be positive and closed of bidimension (q,q)(q,q), and let u1,…,uk∈U(Ω)u_1,\ldots,u_k\in\mathcal U(\Omega) with k≤qk\leq q. The recursively defined current R(u1,…,uk;T)=d′d′′(u1 d′d′′(u2⋯d′d′′(ukT)⋯ ))R(u_1,\ldots,u_k;T) =\mathrm d'\mathrm d''\bigl(u_1\,\mathrm d'\mathrm d''(u_2\cdots\mathrm d'\mathrm d''(u_kT)\cdots)\bigr) is positive and closed of bidimension (q−k,q−k)(q-k,q-k), local in the potentials and background, and symmetric in the uiu_i. If Tj⇀TT_j\rightharpoonup T are positive closed currents and ui,j→uiu_{i,j}\to u_i locally uniformly with all ui,j,ui∈U(Ω)u_{i,j},u_i\in\mathcal U(\Omega), then R(u1,j,…,uk,j;Tj)⇀R(u1,…,uk;T).R(u_{1,j},\ldots,u_{k,j};T_j) \rightharpoonup R(u_1,\ldots,u_k;T).

Proof. Existence, positivity and closedness follow inductively from Theorem 9.8.1. Each multiplication and derivative commutes with restriction, giving locality.

We prove continuity before symmetry. For k=0k=0 it is the assumed convergence of TjT_j. Suppose it is proved for k−1k-1. Let Sj=R(u2,j,…,uk,j;Tj)S_j=R(u_{2,j},\ldots,u_{k,j};T_j) and S=R(u2,…,uk;T)S=R(u_2,\ldots,u_k;T). By induction Sj⇀SS_j\rightharpoonup S, and these currents are positive. Theorem 9.6.3 gives u1,jSj⇀u1Su_{1,j}S_j\rightharpoonup u_1S. Differentiation gives the claimed convergence for kk. This argument does not require derivative bounds on ui,ju_{i,j}.

Symmetry is local. Around any point, intersect the finitely many approximation neighbourhoods and choose simultaneous smooth approximants ui,ju_{i,j} there. For smooth potentials, Proposition 9.7.2 identifies the recursion with d′d′′u1,j∧⋯∧d′d′′uk,j∧T.\mathrm d'\mathrm d''u_{1,j}\wedge\cdots\wedge\mathrm d'\mathrm d''u_{k,j}\wedge T. All the smooth factors have even total degree, so they commute. Apply the continuity just proved to this expression in either order. The two limits agree on each neighbourhood and hence globally. ◻

Corollary 13.5.3. (Mixed local mass bound). Let K⋐int⁡LK\Subset\operatorname{int}L with L⋐ΩL\Subset\Omega compact. For k≤qk\leq q there is a constant CC depending on K,L,k,q,nK,L,k,q,n and chosen intermediate cutoffs, such that μR(u1,…,uk;T)(K)≤C μT(L)∏i=1kosc⁡L(ui).\mu_{R(u_1,\ldots,u_k;T)}(K) \leq C\,\mu_T(L)\prod_{i=1}^k\operatorname{osc}_L(u_i).

Proof. Choose nested compact sets K=K0⊂int⁡K1⊂⋯⊂int⁡Kk⊂LK=K_0\subset\operatorname{int}K_1\subset\cdots\subset\operatorname{int}K_k\subset L, with smooth cutoffs equal one on Ki−1K_{i-1} and supported in KiK_i. Apply Proposition 13.4.2 first to the outermost product and then to its background, continuing inward. At each step subtract inf⁡Lui\inf_Lu_i. The bidimension factors and the Hessian bounds of the finitely many cutoffs multiply into CC. ◻

The same proof shows that the mixed product is separately additive when the potentials involved and their sums belong to U\mathcal U, and is homogeneous for nonnegative scalars. Positivity remains attached to this admitted class. A proof that a different psh definition belongs to U\mathcal U is a comparison theorem, not part of the formal recursion.

13.6 Intrinsic limits and admitted potentials

Theorem 13.6.1. In the algebraic scope of Section 8.8, let Tj⇀TT_j\rightharpoonup T be ordinary positive currents of fixed bidimension on WW, and let fj→ff_j\to f locally uniformly with all fj,ff_j,f continuous. Then fjTj⇀fTf_jT_j\rightharpoonup fT.

Proof. Fix an actual symmetric compactly supported test η\eta and choose an actual positive control θ\theta and MM such that Mθ±η≥0M\theta\pm\eta\geq 0. For the corresponding signed measures, ∣μTj,η∣≤MμTj,θ,μTj,θ(W)=Tj(θ)⟶T(θ).|\mu_{T_j,\eta}|\leq M\mu_{T_j,\theta},\qquad \mu_{T_j,\theta}(W)=T_j(\theta)\longrightarrow T(\theta). Hence their total variations are uniformly bounded and their supports lie in the fixed compact supp⁡η\operatorname{supp}\eta. For smooth scalar ϕ\phi, weak convergence of currents gives ∫ϕ dμTj,η=Tj(ϕη)⟶T(ϕη).\int\phi\,d\mu_{T_j,\eta}=T_j(\phi\eta)\longrightarrow T(\phi\eta). One may insert a cutoff equal one near supp⁡η\operatorname{supp}\eta if ϕ\phi is not compactly supported. Scalar density on that compact set and the uniform variation bound extend this convergence to every continuous scalar function on a neighbourhood of the compact set. Finally split (fjTj−fT)(η)=∫(fj−f) dμTj,η+∫f d(μTj,η−μT,η).(f_jT_j-fT)(\eta) =\int(f_j-f)\,d\mu_{T_j,\eta} +\int f\,d(\mu_{T_j,\eta}-\mu_{T,\eta}). The first term tends to zero by uniform convergence and the variation bound; the second by the continuous-scalar convergence just proved. Symmetrization treats arbitrary tests. ◻

Definition 13.6.2. Let UX(W)\mathcal U_X(W) denote the continuous functions locally uniformly approximable, on one fixed neighbourhood of each point, by original smooth functions uju_j such that the actual smooth form d′d′′uj\mathrm d'\mathrm d''u_j is ordinary positive.

This is an explicitly admitted class. No equivalence with all continuous functions whose curvature current is positive is asserted. In particular, Proposition 8.12.2 does not prove membership in this class.

Theorem 13.6.3. Let TT be an ordinary positive closed current of bidimension (q,q)(q,q) on WW. For u1,…,uk∈UX(W)u_1,\ldots,u_k\in\mathcal U_X(W), k≤qk\leq q, the recursive product RX(u1,…,uk;T)=d′d′′(u1 d′d′′(u2⋯d′d′′(ukT)⋯ ))R_X(u_1,\ldots,u_k;T) =\mathrm d'\mathrm d''\bigl(u_1\,\mathrm d'\mathrm d''(u_2\cdots\mathrm d'\mathrm d''(u_kT)\cdots)\bigr) is an intrinsic ordinary positive closed current. It is symmetric in the potentials and commutes with locally uniform convergence of admitted potentials and simultaneous weak convergence of positive closed backgrounds of fixed degree.

Proof. The first product is defined by Theorem 8.14.1 and distributional differentiation. If uju_j are smooth positive-curvature approximants on a fixed neighbourhood, Theorem 13.6.1 gives ujT⇀uTu_jT\rightharpoonup uT. Smooth Leibniz on actual tests and closedness of TT give d′d′′(ujT)=d′d′′uj∧T⇀d′d′′(uT).\mathrm d'\mathrm d''(u_jT)=\mathrm d'\mathrm d''u_j\wedge T\rightharpoonup\mathrm d'\mathrm d''(uT). Wedge positivity is checked after a common refinement of the smooth form and the test. It follows from Proposition 7.5.2 face by face. Thus the approximating currents are positive, and their limit is positive. It is closed by the differential identities. This argument works locally; a nonnegative partition of unity gives the global assertion.

Iterate this construction. To prove convergence, induct on the number of potentials. The inner products converge by induction and are positive. Apply Theorem 13.6.1 to their multiplication by the outer potentials, then differentiate. For symmetry, choose simultaneous smooth approximants locally for the finitely many potentials. The smooth product is the wedge of their curvature forms with TT, hence unchanged by permutation. The convergence just proved identifies its limits in any order. All operators commute with restriction, so the local equalities glue. ◻

This theorem supplies the intrinsic version of the Euclidean mixed-product argument in Section 13.5 for UX\mathcal U_X. The continuous-scalar relations are handled by Theorem 8.14.1. Equality of two nonsmooth potentials modulo the curvature kernel is still a different question, and has not been used in the proof.

13.7 Quantitative continuity of mixed products

Weak convergence of currents is convergence of their values on each fixed smooth compactly supported test. It does not measure the distance between their coefficient measures in total variation. A quantitative version of the local continuity theorem makes this distinction precise and explains which derivatives must be controlled.

We first work on an open real domain Ω⊂Rn\Omega\subset\mathbb R^n, with the conventions and trace measures of Chapter 9. Let TT be ordinary positive and closed of bidimension (q,q)(q,q). All potentials belong to U(Ω)\mathcal U(\Omega), and R(u;T)R(\mathbf u;T) denotes the product of Theorem 13.5.2. The Euclidean estimate below is not asserted on an arbitrary analytic space by simply choosing one tropical chart: an analytic test can involve further refinements. Its intrinsic qualitative counterpart is Theorem 13.6.3.

Proposition 13.7.1. (A telescoping identity). For two kk-tuples u,v\mathbf u,\mathbf v, where 1≤k≤q1\leq k\leq q, put Si=R(v1,…,vi−1,ui+1,…,uk;T).S_i=R(v_1,\ldots,v_{i-1},u_{i+1},\ldots,u_k;T). Then, as currents, R(u;T)−R(v;T)=∑i=1kd′d′′((ui−vi)Si).\begin{equation} R(\mathbf u;T)-R(\mathbf v;T) =\sum_{i=1}^k\mathrm d'\mathrm d''\bigl((u_i-v_i)S_i\bigr). \end{equation}

Proof. Replace u1u_1 by v1v_1, then u2u_2 by v2v_2, and continue. At the iith replacement, symmetry moves the changing factor to the outside of the recursion. Linearity of continuous scalar multiplication and of d′d′′\mathrm d'\mathrm d'' identifies the difference with d′d′′((ui−vi)Si)\mathrm d'\mathrm d''((u_i-v_i)S_i). Adding these differences cancels every intermediate product. This proof uses the chosen potentials; it does not replace their differences by abstract curvature classes. ◻

Theorem 13.7.2. (A bound on each fixed test). Fix a smooth compactly supported test η\eta of bidegree (q−k,q−k)(q-k,q-k) and a compact L⋐ΩL\Subset\Omega whose interior contains its support. There is a constant Cη,LC_{\eta,L}, depending also on the chosen intermediate cutoffs, such that ∣⟨R(u;T)−R(v;T),η⟩∣≤Cη,LμT(L)∑i=1k∥ui−vi∥L∏j<iosc⁡L(vj)∏j>iosc⁡L(uj).\begin{align*} &\left|\left\langle R(\mathbf u;T)-R(\mathbf v;T),\eta\right\rangle\right|\\ &\quad\leq C_{\eta,L}\mu_T(L) \sum_{i=1}^k\|u_i-v_i\|_L \prod_{j<i}\operatorname{osc}_L(v_j) \prod_{j>i}\operatorname{osc}_L(u_j). \end{align*} An empty product is one. Only the fixed test and cutoffs, not the varying potentials, are differentiated in this estimate.

Proof. We may symmetrize η\eta, since all currents being paired are symmetric. Choose a compact neighbourhood KK of supp⁡η\operatorname{supp}\eta in int⁡L\operatorname{int}L. The telescoping identity gives ⟨R(u;T)−R(v;T),η⟩=∑iSi((ui−vi)d′d′′η).\left\langle R(\mathbf u;T)-R(\mathbf v;T),\eta\right\rangle =\sum_i S_i\bigl((u_i-v_i)\mathrm d'\mathrm d''\eta\bigr). The form d′d′′η\mathrm d'\mathrm d''\eta has bidegree (q−k+1,q−k+1)(q-k+1,q-k+1), exactly the bidimension of SiS_i. Its symmetric coefficient matrix has a finite operator-norm bound AηA_\eta. The trace-measure estimate of Chapter 9 therefore gives ∣Si((ui−vi)d′d′′η)∣≤Aη∥ui−vi∥L μSi(K).|S_i((u_i-v_i)\mathrm d'\mathrm d''\eta)| \leq A_\eta\|u_i-v_i\|_L\,\mu_{S_i}(K). Apply Corollary 13.5.3 to the k−1k-1 potentials defining SiS_i, using intermediate compact sets between KK and LL. There are finitely many positions ii, so a common maximum of the cutoff constants suffices. This proves the formula. For k=1k=1, there are no inner potentials and the argument is simply the trace estimate for TT. ◻

Corollary 13.7.3. If all the oscillations in the preceding theorem are bounded by MM and μT(L)≤B\mu_T(L)\leq B, then ∣⟨R(u;T)−R(v;T),η⟩∣≤Cη,LB Mk−1∑i∥ui−vi∥L.|\langle R(\mathbf u;T)-R(\mathbf v;T),\eta\rangle| \leq C_{\eta,L}B\,M^{k-1}\sum_i\|u_i-v_i\|_L. For k=1k=1 the factor Mk−1M^{k-1} is one, also when M=0M=0. Thus on a bounded family the product is Lipschitz on each fixed smooth test, with respect to uniform distance of the potentials.

The constant depends on the second derivatives of η\eta and on the localization. It is not uniform over all continuous tests of sup norm one. Consequently this corollary gives no total-variation convergence. Nor does weak convergence Tj→TT_j\to T alone give a numerical rate for the background error. The simultaneous-background theorem is obtained by continuity of scalar multiplication and induction, rather than by assigning a norm distance to arbitrary currents.

13.8 Locality, maxima and contact sets

The maximum operation connects regularization to local comparison arguments. It is useful to record exactly what follows from locality before invoking any comparison principle. Work either in the real setting above or in the intrinsic algebraic setting of Theorem 13.6.3.

Proposition 13.8.1. Let u,vu,v belong to the admitted local uniform approximation class and put w=max⁡(u,v)w=\max(u,v). Then ww belongs to that class. For fixed additional admitted potentials a\mathbf a, R(w,a;T)∣{u>v}=R(u,a;T)∣{u>v},R(w,a;T)∣{v>u}=R(v,a;T)∣{v>u}.\begin{align*} R(w,\mathbf a;T)|_{\{u>v\}}&=R(u,\mathbf a;T)|_{\{u>v\}},\\ R(w,\mathbf a;T)|_{\{v>u\}}&=R(v,\mathbf a;T)|_{\{v>u\}}. \end{align*}

Proof. On a common approximation neighbourhood choose smooth psh approximants uj,vju_j,v_j. Apply the normalized regularized maximum of Chapter 12, with smoothing parameter tending to zero. Convexity and coordinatewise monotonicity preserve positivity of the Hessian. Its uniform error from the ordinary maximum tends to zero, while ∣max⁡(uj,vj)−max⁡(u,v)∣≤max⁡(∣uj−u∣,∣vj−v∣).|\max(u_j,v_j)-\max(u,v)| \leq\max(|u_j-u|,|v_j-v|). This proves membership in the admitted class. On the open set {u>v}\{u>v\} the functions ww and uu agree, so locality of scalar multiplication and distributional differentiation identifies the products. The second equality is the same argument. ◻

Example 13.8.2. (The contact contribution). On the real line take u(t)=tu(t)=t and v(t)=−tv(t)=-t. Their Hessians vanish, but d′d′′max⁡(u,v)=d′d′′∣t∣=2δ0.\mathrm d'\mathrm d''\max(u,v)=\mathrm d'\mathrm d''|t|=2\delta_0. Thus the two restrictions in Proposition 13.8.1 do not determine the product on {u=v}\{u=v\}. For a test χ\chi, integration by parts separately on the two half-lines gives ∫∣t∣χ′′(t) dt=2χ(0)\int |t|\chi''(t)\,dt=2\chi(0), which proves the formula and its normalization. In a tropical annulus coordinate the same local calculation applies with the relevant lattice calibration.

For a truncation uc=max⁡(u,c)u_c=\max(u,c), locality gives d′d′′uc=d′d′′u\mathrm d'\mathrm d''u_c=\mathrm d'\mathrm d''u on {u>c}\{u>c\} and d′d′′uc=0\mathrm d'\mathrm d''u_c=0 on {u<c}\{u<c\}. The level set {u=c}\{u=c\} can carry mass. Removing that mass by restricting to a strict sublevel complement is an additional operation; it is not the original product. This distinction underlies the separate definitions of truncated and non-pluripolar products in theories where the latter are available.

References. [CLD12] (Section 5.6); [CLD25] (Sections 8.4–8.7). The quantitative estimate combines the local mass bound with the telescoping identity; the maximum statements use regularized maxima and locality. These are consequences in the stated admitted classes, not extensions to arbitrary current-psh potentials.

13.9 Symmetry without choosing simultaneous smooth approximants

The continuous generated sheaf C25\mathcal C_{25} was defined by local uniform closure from smooth psh functions. Its sections need not come with one common one-step approximation. Nevertheless the recursive products are symmetric. The proof uses the defining closure property separately in the two variables of an adjacent interchange.

For the concrete analytic assertions below, use the algebraic scope of Section 8.8: an algebraically closed complete nontrivially valued field and paracompact open subsets of analytifications of separated integral algebraic varieties. The background currents are ordinary positive and closed, and Theorem 13.6.1 supplies multiplication and joint continuity on actual original smooth tests. The closure arguments also apply abstractly in any calculus where these same inputs have been established. For u∈C25u\in\mathcal C_{25} write AuT=d′d′′(uT)A_uT=\mathrm d'\mathrm d''(uT). The preceding construction gives a positive closed current, and uj→u locally uniformly,Tj⇀T⟹AujTj⇀AuT.u_j\to u\text{ locally uniformly},\quad T_j\rightharpoonup T \quad\Longrightarrow\quad A_{u_j}T_j\rightharpoonup A_uT. All currents in this implication have the same degree and are positive closed.

Lemma 13.9.1. Suppose a property of continuous functions is local, holds for smooth psh functions, and is preserved under locally uniform limits. Then the property holds for every section of C25\mathcal C_{25}, provided it is imposed on restrictions to every open subset.

Proof. The functions satisfying the property form a subsheaf of continuous functions, contain the starting sheaf of smooth psh functions, and are closed under the prescribed limit operation. The minimality in the definition of the generated sheaf gives the conclusion. No countable-stage or fixed-neighbourhood representation of an arbitrary section is asserted. ◻

Theorem 13.9.2. For u,v∈C25u,v\in\mathcal C_{25} and an ordinary positive closed current TT of the appropriate degree, AuAvT=AvAuT.A_uA_vT=A_vA_uT. Consequently every iterated product Au1⋯AuqTA_{u_1}\cdots A_{u_q}T is symmetric in the potentials.

Proof. First fix a smooth psh function vv. For smooth psh uu, the identity is the commutativity of the two smooth curvature factors. Consider the property that the identity holds for every positive closed TT on every open restriction. It is local. If uj→uu_j\to u locally uniformly, joint continuity gives AujAvT→AuAvTA_{u_j}A_vT\to A_uA_vT and AvAujT→AvAuTA_vA_{u_j}T\to A_vA_uT. Hence the property is closed under these limits. Lemma 13.9.1 proves it for all u∈C25u\in\mathcal C_{25} and smooth vv.

Now consider the property of vv that the identity holds for every u∈C25u\in\mathcal C_{25} and every such TT, on every open restriction. The first step gives it for smooth psh vv. If vj→vv_j\to v locally uniformly, then AvjT→AvTA_{v_j}T\to A_vT; applying joint continuity with the fixed continuous multiplier uu gives AuAvjT→AuAvTA_uA_{v_j}T\to A_uA_vT. The other side converges to AvAuTA_vA_uT by the same theorem. A second application of the closure principle gives the desired identity for all v∈C25v\in\mathcal C_{25}.

In an iterated product the factors inside an adjacent pair produce another positive closed current. The two-factor identity therefore allows every adjacent interchange. Such interchanges generate all permutations. ◻

This proof distinguishes two types of independence. The product is independent of an approximation used to construct it and of the order of its potential factors. It has not thereby been shown to depend only on the ambient curvature currents d′d′′[ui]\mathrm d'\mathrm d''[u_i]. That latter statement is the kernel property of Chapter 14.

13.10 Chern operators with continuous generated weights

Proposition 13.10.1. Let a metrized line bundle L‾\overline L have local weights in C25\mathcal C_{25}. For every ordinary positive closed current TT, the local currents d′d′′(φeT)\mathrm d'\mathrm d''(\varphi_eT) glue to a global positive closed current, denoted c1(L‾)∧Tc_1(\overline L)\wedge T. These operators commute for different such metrized line bundles.

Proof. For frames e′=gee'=ge, one has φe′−φe=−log⁡∣g∣\varphi_{e'}-\varphi_e=-\log|g|. This difference is original smooth and has zero d′d′′\mathrm d'\mathrm d'' because gg is an analytic unit. The two closedness conditions on TT give d′d′′((φe′−φe)T)=d′d′′(−log⁡∣g∣)∧T=0.\mathrm d'\mathrm d''\bigl((\varphi_{e'}-\varphi_e)T\bigr) =\mathrm d'\mathrm d''(-\log|g|)\wedge T=0. Thus the local currents agree on overlaps. Positivity and closedness are local and were proved for AuTA_uT. Commutativity is Theorem 13.9.2 on a common trivializing neighbourhood. ◻

The gluing uses a very specific kernel: logarithms of analytic units. It therefore remains valid even when an arbitrary-current theorem for all current-harmonic continuous functions is unavailable. This is why line-bundle intersection calculus can be established on a useful class before the general harmonic-kernel problem is settled.

13.11 Proper pushforward and projection

Let f:Y→Xf:Y\to X be a proper map between the chosen analytic opens induced by an algebraic morphism over the same field, in the scope just specified. Both spaces are pure dimensional. We index the next statement by bidimension, which is preserved by current pushforward. If TT has bidimension (q,q)(q,q) on YY, then f∗Tf_*T acts on (q,q)(q,q)-tests on XX.

Lemma 13.11.1. For an ordinary positive current TT and a continuous function uu on XX, f∗((u∘f)T)=u f∗T.f_*((u\circ f)T)=u\,f_*T.

Proof. For smooth uu, evaluate both sides on a compactly supported test η\eta and use f∗(uη)=(u∘f)f∗ηf^*(u\eta)=(u\circ f)f^*\eta. For continuous uu, uniformly approximate it by smooth functions near the compact support of η\eta. Properness makes the inverse image of that support compact. The order-zero estimates on both spaces pass the smooth identity to the limit. This also proves the identity when ff is merely proper over the support involved. ◻

Theorem 13.11.2. If TT is ordinary positive closed and u1,…,us∈C25(X)u_1,\ldots,u_s\in\mathcal C_{25}(X) with s≤qs\leq q, then f∗(Au1∘f⋯Aus∘fT)=Au1⋯Aus(f∗T).f_*\bigl(A_{u_1\circ f}\cdots A_{u_s\circ f}T\bigr) =A_{u_1}\cdots A_{u_s}(f_*T).

Proof. Pullback preserves smooth psh functions and locally uniform convergence; the closure principle therefore gives ui∘f∈C25(Y)u_i\circ f\in\mathcal C_{25}(Y). Pullback preserves ordinary positive test forms, so f∗Tf_*T is ordinary positive. It is closed because the differentials commute with pullback on tests. For one factor, commutation of current pushforward with distributional differentiation and Lemma 13.11.1 give f∗d′d′′((u∘f)T)=d′d′′f∗((u∘f)T)=d′d′′(uf∗T).f_*\mathrm d'\mathrm d''((u\circ f)T)=\mathrm d'\mathrm d''f_*((u\circ f)T) =\mathrm d'\mathrm d''(u f_*T). Apply this successively to the positive closed inner products to obtain the formula for ss factors. ◻

For a finite flat same-field map of degree dd, apply the theorem with T=[Y]T=[Y] and the calibrated identity f∗[Y]=d[X]f_*[Y]=d[X]. It gives the degree-dd projection formula for these mixed products. If instead one integrates a scalar extension over its new native ground field, Chapter 23 gives p∗[XL]L=[X]Kp_*[X_L]_L=[X]_K and the corresponding formula has no extra degree factor. The recursive argument uses the supplied calibration identity and does not choose that normalization itself.

This theorem is a statement about currents. It does not say that proper pushforward of an actual delta-form is an actual delta-form, or that an alteration preserves an exact support complement. Those are the distinct representative questions discussed in Chapter 23.

13.12 A monotone consequence and its limit

Proposition 13.12.1. Suppose ui,j∈C25u_{i,j}\in\mathcal C_{25} decrease pointwise to finite continuous functions uiu_i, for 1≤i≤s1\leq i\leq s. On each compact subset, convergence is uniform. The limits belong locally to C25\mathcal C_{25}, and the mixed products converge weakly, also with a weakly convergent sequence of ordinary positive closed background currents.

Proof. For a compact set KK and ϵ>0\epsilon>0, the open sets {ui,j−ui<ϵ}\{u_{i,j}-u_i<\epsilon\} increase with jj and cover KK. A finite subcover is contained in one of them. This is Dini’s argument and proves uniform convergence. Local compactness and the generated-sheaf closure give the membership assertion. The product convergence is then the locally uniform stability theorem already proved. ◻

Continuity of the limit is essential to this argument. Without it, the sets used in the compactness proof need not be open. A general bounded-potential monotone theorem requires more than relabelling pointwise convergence as uniform convergence.

Example 13.12.2. (Pointwise multipliers do not suffice). On R\mathbb R, let Tj=δ1/jT_j=\delta_{1/j} and T=δ0T=\delta_0. Choose continuous bumps uju_j of height one at 1/j1/j, supported in intervals of radius 1/(10j2)1/(10j^2) around 1/j1/j. Then uj→0u_j\to 0 pointwise, whereas ujTj=Tj⇀δ0u_jT_j=T_j\rightharpoonup\delta_0. Thus multiplication does not converge to 0T0T. The bumps are not psh; this is a counterexample to the bare pointwise multiplier argument, not to a separately proved psh monotone theorem.

References. The generated-class construction is compared with [CLD25] (Sections 8.5–8.7). The proofs above use the ordinary-positive estimates and current operations established in Chapters 8–9; no full strong-cone spanning statement is used.

13.13 Comparison on a continuous global domain

A comparison principle can be separated from the harder problem of defining products for all bounded psh functions. We prove the continuous statement from locality, positivity, and the total mass formula. This identifies a useful consequence of an established calculus without claiming the full bounded-potential theory.

Let XX be compact and let P⊂C(X)\mathcal P\subset C(X) be a specified class of continuous potentials in one fixed background class θ\theta. Assume that P\mathcal P is stable under addition of constants and finite maxima. Suppose its Monge–Ampère operator assigns positive Radon measures of the same finite mass VV, is unchanged by adding constants to a potential, and is local on open sets: if two potentials agree on an open set, their measures agree there. These hypotheses are part of the statement. They must be checked in the chosen analytic theory.

Theorem 13.13.1. For u,v∈Pu,v\in\mathcal P, ∫{u<v}MA⁡(v)≤∫{u<v}MA⁡(u).\int_{\{u<v\}}\operatorname{MA}(v)\leq\int_{\{u<v\}}\operatorname{MA}(u).

Proof. Fix ε>0\varepsilon>0 and put w=max⁡(u,v−ε)w=\max(u,v-\varepsilon). On the disjoint open sets Aε={u<v−ε}A_\varepsilon=\{u<v-\varepsilon\} and Bε={u>v−ε}B_\varepsilon=\{u>v-\varepsilon\}, locality gives MA⁡(w)=MA⁡(v)\operatorname{MA}(w)=\operatorname{MA}(v) and MA⁡(w)=MA⁡(u)\operatorname{MA}(w)=\operatorname{MA}(u), respectively. Positivity and equality of total masses therefore imply MA⁡(v)(Aε)=MA⁡(w)(Aε)≤V−MA⁡(w)(Bε)=V−MA⁡(u)(Bε)=MA⁡(u)({u≤v−ε}).\begin{align*} \operatorname{MA}(v)(A_\varepsilon) &=\operatorname{MA}(w)(A_\varepsilon)\leq V-\operatorname{MA}(w)(B_\varepsilon)\\ &=V-\operatorname{MA}(u)(B_\varepsilon) =\operatorname{MA}(u)(\{u\leq v-\varepsilon\}). \end{align*} As ε\varepsilon decreases to zero through a sequence, both sets in the final inequality increase to {u<v}\{u<v\}. Continuity from below for finite positive measures proves the result. No assertion about the measure of the equality set {u=v}\{u=v\} is needed. ◻

Remark 13.13.2. This proof applies, for example, after the three stated properties have been established for a continuous semipositive metric class on a proper variety. It does not derive those properties from the single condition d′d′′[u]+θ≥0\mathrm d'\mathrm d''[u]+\theta\geq 0 on original smooth tests. In particular, the maximum must belong to the admitted domain and its measure must be local there. The theorem should not be cited as a regularization theorem or as an extension of the domain of MA⁡\operatorname{MA}.

Why finite maxima preserve a generated class

For the continuous generated class used in Chapter 13, maximum stability follows by the same two-variable closure argument as symmetry. The starting step uses a smooth convex regularized maximum, nondecreasing in each argument. Its Hessian chain rule is d′d′′F(u,v)=F1d′d′′u+F2d′d′′v+∑i,j=12Fij d′ui∧d′′uj,(u1,u2)=(u,v).\mathrm d'\mathrm d''F(u,v)=F_1\mathrm d'\mathrm d''u+F_2\mathrm d'\mathrm d''v +\sum_{i,j=1}^2F_{ij}\,d'u_i\wedge d''u_j, \quad (u_1,u_2)=(u,v). For smooth psh inputs all three contributions are positive: Fi≥0F_i\geq 0 and the Hessian matrix (Fij)(F_{ij}) is positive semidefinite. The regularized maxima converge uniformly to the maximum. Keeping one input fixed and passing through local uniform closure, then doing the same for the other input, proves maximum stability on the generated class. The estimate ∥max⁡(u,v)−max⁡(u~,v~)∥∞≤max⁡(∥u−u~∥∞,∥v−v~∥∞).\|\max(u,v)-\max(\widetilde u,\widetilde v)\|_\infty \leq\max(\|u-\widetilde u\|_\infty, \|v-\widetilde v\|_\infty). For metrics on a fixed line bundle, carry out this argument on local weights. Changing a frame adds the same pluriharmonic function to both weights; maximum commutes with that common addition, so the resulting local metric weights glue. One must still use the global degree theorem to identify their total masses.

13.14 A capacity estimate on the same domain

Assume additionally that 0∈P0\in\mathcal P, that convex combinations remain in P\mathcal P, and that symmetric multilinear positive mixed measures are defined, with the usual expansion for θ+d′d′′(tv)\theta+\mathrm d'\mathrm d''(tv). These assumptions hold only on the specified domain. For a Borel set E⊂XE\subset X, define Cap⁡P(E)=sup⁡v∈P, −1≤v≤0∫EMA⁡(v).\operatorname{Cap}_{\mathcal P}(E) =\sup_{v\in\mathcal P,\,-1\leq v\leq0}\int_E\operatorname{MA}(v). This definition records its testing class in the subscript. Different classes may give different capacities, even when their operators agree on the intersection.

Proposition 13.14.1. For u∈Pu\in\mathcal P with u≤0u\leq 0, s≥0s\geq 0 and 0<t≤10<t\leq 1, tnCap⁡P({u<−s−t})≤∫{u<−s}MA⁡(u),t^n\operatorname{Cap}_{\mathcal P}(\{u<-s-t\}) \leq\int_{\{u<-s\}}\operatorname{MA}(u), where nn is the number of factors in the Monge–Ampère operator.

Proof. For an admitted vv with −1≤v≤0-1\leq v\leq 0, put w=tv−sw=tv-s. Since 0<t≤10<t\leq 1, convexity and constant stability make ww admitted. Its curvature is t(θ+d′d′′v)+(1−t)θt(\theta+\mathrm d'\mathrm d''v)+(1-t)\theta. The mixed expansion and positivity give MA⁡(w)≥tnMA⁡(v)\operatorname{MA}(w)\geq t^n\operatorname{MA}(v). Moreover, {u<−s−t}⊆{u<w}⊆{u<−s}.\{u<-s-t\}\subseteq\{u<w\}\subseteq\{u<-s\}. The comparison theorem therefore yields tn∫{u<−s−t}MA⁡(v)≤∫{u<w}MA⁡(w)≤∫{u<w}MA⁡(u)≤∫{u<−s}MA⁡(u).t^n\int_{\{u<-s-t\}}\operatorname{MA}(v) \leq\int_{\{u<w\}}\operatorname{MA}(w) \leq\int_{\{u<w\}}\operatorname{MA}(u) \leq\int_{\{u<-s\}}\operatorname{MA}(u). Take the supremum over the admitted test potentials vv. ◻

Monotonicity, countable subadditivity and continuity along increasing Borel sets follow directly from the definition: each test supplies a positive measure, and increasing suprema commute. Also Cap⁡P(X)=V\operatorname{Cap}_{\mathcal P}(X)=V, since v=0v=0 is an allowed test and all test measures have mass VV. Continuity along decreasing compact sets does not follow from this argument.

Example 13.14.2. On [0,1][0,1], the set function c(E)=sup⁡k≥1δ1/k(E)c(E)=\sup_{k\geq 1}\delta_{1/k}(E) is a supremum of positive measures of mass one. Yet for Kj=[0,1/j]K_j=[0,1/j] one has c(Kj)=1c(K_j)=1 for every jj, while c(⋂jKj)=c({0})=0c(\bigcap_jK_j)=c(\{0\})=0. Thus the measure properties of individual tests do not by themselves make their supremum a Choquet capacity. This example concerns the abstract inference, not a counterexample to a capacity theorem proved for a particular psh class.

The continuous comparison and sublevel estimate are therefore available consequences once their explicit inputs have been verified. Quasicontinuity, pluripolar-set characterizations, monotone convergence for discontinuous bounded potentials, and a Dirichlet existence theorem require further results. The elementary arguments above do not remove those obligations.

References. The proofs combine maximum locality and the fixed-mass identity, followed by the positive mixed expansion. Their analytic inputs are discussed in Chapters 13, 19 and 20; compare the model-based pluripotential framework of [BFJ15].