Conjecture Atlas
Frozen public edition v0.5.1. Read the current version of this page.
Programme paper

17 November 2026 · 15 PDF pages

Download standalone PDF

Conjectures and programme

Strong Green Forms and Beilinson–Bloch Heights

A local-to-global research programme

17 November 2026

Abstract

The local problem is to construct an actual non-Archimedean zero-curvature Green form for a homologically trivial cycle, on its prescribed support complement and against all compact actual tests. The global problem is to obtain a Beilinson–Bloch pairing on the full complementary homologically trivial Chow groups. We separate the local construction, the vanishing of its Chow obstruction, and the reciprocity and principal transformation laws needed for the global sum. We give the formal curvature-correction argument and a precise local-to-global criterion, then formulate the remaining analytic and arithmetic problems in dependency order. Correspondence adjunction and comparison with classical heights are additional steps. For products of curves, the resulting height system leads to separate questions of arithmetic Lefschetz, positivity, integral discreteness and fields of definition.

1. The local problem and its arithmetic purpose

On a curve, Green functions turn degree-zero divisors into local intersection pairings. Their global sum is related to the canonical height on the Jacobian. In higher codimension the same construction asks for a Green object of positive bidegree, compatible with rational equivalence, intersection and change of field. An actual differential-form representative provides more than a numerical local pairing: it places that pairing inside a differential calculus in which its support and functoriality can be studied.

We work locally over a finite extension K/QpK/\mathbf Q_p with one fixed absolute value. The absolute value is extended, rather than renormalized, on every finite extension of KK. Let X/KX/K be smooth, projective and geometrically integral of dimension nn. Write D=d′d′′,Bp,q=BXanp,qD=d'd'',\qquad \mathcal B^{p,q}=\mathcal B^{p,q}_{X^{\mathrm{an}}} for the ordinary sheaf of actual δ\delta-forms in the sense of Mihatsch [1]. The underlying topology is the ordinary topology of the Berkovich analytification.

Fix a prime ℓ≠p\ell\ne p. We write CH⁡r(X)Q,hom=ker⁡ ⁣(CH⁡r(X)Q⟶Heˊt2r(XK‾,Qℓ(r))).\operatorname{CH}^r(X)_{\mathbf Q,\mathrm{hom}} =\ker\!\left(\operatorname{CH}^r(X)_\mathbf Q\longrightarrow H^{2r}_{\mathrm{\acute et}}(X_{\overline K},\mathbf Q_\ell(r))\right). Geometric algebraic equivalence defines the subgroup CH⁡r(X)Q,alg\operatorname{CH}^r(X)_{\mathbf Q,\mathrm{alg}}. For a prescribed rational cycle Z=∑iai[Zi]Z=\sum_i a_i[Z_i] of codimension rr, let ∣Z∣=⋃ai≠0Zi,UZ=Xan∖∣Z∣an.|Z|=\bigcup_{a_i\ne0}Z_i,\qquad U_Z=X^{\mathrm{an}}\setminus |Z|^{\mathrm{an}}. The integration current of ZZ retains its coefficients and the cycle multiplicities of the actual integration theory.

Conjecture 1.1 (Strong zero-curvature Green forms). For every such valued field KK, every X/KX/K, every 1≤r≤n1\le r\le n, and every prescribed rational cycle ZZ with [Z]∈CH⁡r(X)Q,hom[Z]\in\operatorname{CH}^r(X)_{\mathbf Q,\mathrm{hom}}, there exists γZ∈Br−1,r−1(UZ)\gamma_Z\in\mathcal B^{r-1,r-1}(U_Z) such that the following integrals are defined and ∫UZγZ∧Dη=−∫Zanη(η∈Bcn−r,n−r(Xan)). \int_{U_Z}\gamma_Z\wedge D\eta =-\int_{Z^{\mathrm{an}}}\eta \qquad \left(\eta\in\mathcal B_c^{n-r,n-r}(X^{\mathrm{an}})\right).(1.1) The form and identity are over the original valued field KK.

The prescribed representative is part of the data. Replacing ZZ by a rationally equivalent cycle may be useful in a global pairing, but it does not change the local target in Conjecture 1.1. Likewise, the test space in (1.1) is the complete compactly supported actual space. A finite detector, when available, is a criterion for constructing this form, rather than a replacement for its test space.

Mihatsch’s theory supplies the actual form category and Green constructions for complete-intersection formal models [1, Theorems 4.5 and 5.5]. Künnemann’s height pairing for cycles on abelian varieties [2] provides an arithmetic comparison point. The purpose of the present programme is the general form-level statement (1.1) and its passage to global heights. The conditional local reductions and the full available construction arguments are collected in [4].

The main implication is the curvature correction of Theorem 2.3: prescribed Green pairs together with vanishing of their actual Bott–Chern classes give strong Green forms. The local-to-global criterion in Proposition 6.3 identifies the further statements needed to obtain a Chow pairing from these forms. These two reductions determine the research programme in Section 6. Its analytic branch concerns existence and descent; its arithmetic branch begins only after the pairing, correspondence adjunction and curve normalization have been established.

2. Curvature and the original Chow-source obstruction

The first construction should produce a curved Green identity. A second argument then removes its curvature. This separates the support problem, which depends on the prescribed cycle, from a vanishing problem depending on its Chow class.

Definition 2.1 (Actual Bott–Chern space). For 1≤r≤n1\le r\le n, set HBC,actr,r(X)={ω∈Br,r(Xan):d′ω=d′′ω=0}DBr−1,r−1(Xan).H^{r,r}_{\mathrm{BC,act}}(X)= \frac{\{\omega\in\mathcal B^{r,r}(X^{\mathrm{an}}):d'\omega=d''\omega=0\}} {D\mathcal B^{r-1,r-1}(X^{\mathrm{an}})}. This is an algebraic quotient of real vector spaces.

Hypothesis 2.2 (Chow obstruction and prescribed Green pairs). For each smooth projective integral KK-variety in the argument, the following constructions exist.

  1. There is a rational-linear, Chow-invariant map oX,Kr:CH⁡r(X)Q⟶HBC,actr,r(X).\mathfrak o^r_{X,K}:\operatorname{CH}^r(X)_\mathbf Q\longrightarrow H^{r,r}_{\mathrm{BC,act}}(X). It is computed by actual degree-rr model Chern-character forms and is compatible with the genuine analytic pullbacks used in the argument.

  2. For every prescribed rational codimension-rr cycle ZZ, there are GZ∈Br−1,r−1(UZ),ΩZ∈Br,r(Xan),G_Z\in\mathcal B^{r-1,r-1}(U_Z),\qquad \Omega_Z\in\mathcal B^{r,r}(X^{\mathrm{an}}), with d′ΩZ=d′′ΩZ=0d'\Omega_Z=d''\Omega_Z=0, [ΩZ]=oX,Kr([Z])[\Omega_Z]=\mathfrak o^r_{X,K}([Z]), and ∫UZGZ∧Dη+∫Zanη=∫XanΩZ∧η(η∈Bcn−r,n−r(Xan)). \int_{U_Z}G_Z\wedge D\eta+\int_{Z^{\mathrm{an}}}\eta =\int_{X^{\mathrm{an}}}\Omega_Z\wedge\eta \qquad\left(\eta\in\mathcal B_c^{n-r,n-r}(X^{\mathrm{an}})\right).(2.1) All displayed pairings are defined in the actual calculus.

Theorem 2.3 (Reduction to source-image vanishing). Assume Hypothesis 2.2. If oX,Kr(CH⁡r(X)Q,hom)=0, \mathfrak o^r_{X,K}\bigl(\operatorname{CH}^r(X)_{\mathbf Q,\mathrm{hom}}\bigr)=0,(2.2) then Conjecture 1.1 holds for (X,r)(X,r). More precisely, each prescribed cycle ZZ whose class has zero obstruction has a Green form satisfying (1.1) on UZU_Z.

Proof. Zero obstruction means that the given global curvature has a global actual primitive: choose ΨZ∈Br−1,r−1(Xan)\Psi_Z\in\mathcal B^{r-1,r-1}(X^{\mathrm{an}}) with DΨZ=ΩZD\Psi_Z=\Omega_Z. Put γZ=GZ−ΨZ∣UZ.\gamma_Z=G_Z-\Psi_Z|_{U_Z}. Actual Stokes gives ∫XΨZ∧Dη=∫XΩZ∧η\int_X\Psi_Z\wedge D\eta=\int_X\Omega_Z\wedge\eta. Subtracting this equality from (2.1) proves (1.1). The correction is global on XanX^{\mathrm{an}}, so its restriction preserves UZU_Z. ◻

The theorem asks for vanishing on the image of the original Chow group. It leaves the ambient Bott–Chern space free to contain other classes. This distinction also governs the residual-period formulation below: a large obstruction target can coexist with a zero image of homologically trivial cycles.

2.1. Finite descent and structured primitives

A finite cover often supplies a model on which the curvature becomes numerically accessible. Returning the primitive to KK is then part of the construction.

Hypothesis 2.4 (A compatible finite trace). Let q:Y→Xq:Y\to X be finite dominant of native generic degree dd. On the class of forms occurring in a chosen primitive construction, suppose an actual trace is defined, commutes with DD, and satisfies Tr⁡q(q∗Ω)=dΩ\operatorname{Tr}_q(q^*\Omega)=d\Omega for the closed actual form under consideration, with adjunction against all compact actual complementary tests.

Proposition 2.5 (Primitive descent). Under Hypothesis 2.4, if a primitive β\beta in the trace domain satisfies Dβ=q∗ΩD\beta=q^*\Omega, then Ψ=1dTr⁡q(β)satisfiesDΨ=Ω.\Psi=\frac1d\operatorname{Tr}_q(\beta) \quad\text{satisfies}\quad D\Psi=\Omega.

Proof. The stated compatibilities give DΨ=d−1Tr⁡q(Dβ)=d−1Tr⁡q(q∗Ω)=ΩD\Psi=d^{-1}\operatorname{Tr}_q(D\beta)=d^{-1}\operatorname{Tr}_q(q^*\Omega)=\Omega. ◻

The construction draft uses a trace on the actual scalar–curvature module generated locally by u0Du1⋯Duku_0Du_1\cdots Du_k, with piecewise-linear potentials. When a primitive is assembled from several such generators on finite covers, descent is applied to the original generators along the composite maps. For an extension L/KL/K, the denominator is the full degree [L:K][L:K], including ramification and residue degree. These are the form-domain and normalization requirements that a proof of Hypothesis 2.4 must meet.

3. Two geometric sources of vanishing

The obstruction formulation admits two different kinds of geometric input. Algebraic equivalence gives an additive parameter space. Abelian varieties give special-fibre numerical corrections. They lead to distinct reduction arguments.

3.1. Additive parameter spaces

Lemma 3.1 (Compact rigidity). Let AA be a compact topological abelian group and WW a torsion-free abelian group. An additive map φ:A→W\varphi:A\to W is zero if its kernel contains a neighbourhood of the identity.

Proof. Choose an identity neighbourhood U⊆ker⁡φU\subseteq\ker\varphi, and an open identity neighbourhood VV with V−V⊆UV-V\subseteq U. Finitely many translates of VV cover AA. For a∈Aa\in A, two sufficiently many successive multiples of aa belong to the same translate, so ma∈Uma\in U for some integer m>0m>0. Thus mφ(a)=0m\varphi(a)=0, and torsion-freeness gives φ(a)=0\varphi(a)=0. ◻

Hypothesis 3.2 (Rational-fibre rigidity). For smooth separated S/KS/K, a0∈S(K)a_0\in S(K) and an actual form ω\omega on (X×KS)an(X\times_KS)^{\mathrm{an}}, there is an ordinary neighbourhood VV of a0a_0 such that sa∗ω=sa0∗ω(a∈V∩S(K)).s_a^*\omega=s_{a_0}^*\omega\qquad(a\in V\cap S(K)). The obstruction in Hypothesis 2.2 is compatible with these rational-fibre pullbacks.

Theorem 3.3 (The algebraically trivial source). Assume Hypothesis 2.2 and Hypothesis 3.2. Then oX,Kr(CH⁡r(X)Q,alg)=0.\mathfrak o^r_{X,K}\bigl(\operatorname{CH}^r(X)_{\mathbf Q,\mathrm{alg}}\bigr)=0. Consequently every prescribed geometrically algebraically trivial cycle satisfies (1.1).

Proof. Choose a finite Galois extension L/KL/K over which the given class zLz_L is a finite rational sum of curve-correspondence images of degree-zero divisor classes, together with the relevant rational-equivalence identities. The Poincaré bundles on the pointed curves and their Jacobians produce an additive family Θ\Theta on XL×JX_L\times J, for a product of Jacobians J/LJ/L, and a point t∈J(L)t\in J(L) with Θt=zL\Theta_t=z_L.

Let A=Res⁡L/KJA=\operatorname{Res}_{L/K}J. Evaluation AL→JA_L\to J, followed by algebraic finite-flat pushforward, gives an additive Chow family ΘN\Theta^N on X×AX\times A. At the point a∈A(K)=J(L)a\in A(K)=J(L) corresponding to tt, proper pushforward and refined fibre pullback give (ΘN)a=[L:K]z.(\Theta^N)_a=[L:K]z. Represent the family obstruction by a model Chern form on (X×A)an(X\times A)^{\mathrm{an}}. By Hypothesis 3.2, the map b⟼oX,Kr((ΘN)b)b\longmapsto\mathfrak o^r_{X,K}((\Theta^N)_b) is locally constant at zero. It is additive, takes zero to zero, and has values in a real vector space. Since A(K)A(K) is compact, Lemma 3.1 makes the map zero. Hence [L:K]oX,Kr(z)=0[L:K]\mathfrak o^r_{X,K}(z)=0, and rational linearity proves the asserted vanishing. Apply Theorem 2.3 to the prescribed representative. ◻

If this vanishing is available, the remaining obstruction factors through CH⁡r(X)Q,hom/CH⁡r(X)Q,alg.\operatorname{CH}^r(X)_{\mathbf Q,\mathrm{hom}}/\operatorname{CH}^r(X)_{\mathbf Q,\mathrm{alg}}. The target remains the image of this quotient under the obstruction map. Establishing that the quotient itself is zero is a substantially different problem.

3.2. Numerical corrections and Chow retractions

A numerical model certificate for a class b∈CH⁡r(V)Qb\in\operatorname{CH}^r(V)_\mathbf Q is a virtual bundle ξ∈K0(V)Q\xi\in K_0(\mathcal V)_\mathbf Q on an integral projective flat model such that ch⁡r(ξ∣V)=b,deg⁡(ch⁡r(ξs)∩[T])=0\operatorname{ch}_r(\xi|_V)=b, \qquad \deg\bigl(\operatorname{ch}_r(\xi_s)\cap[T]\bigr)=0 for every integral vertical rr-cycle TT. On Künnemann’s semistable abelian models, the source argument uses the special-fibre exact sequence to subtract a vertical correction from a Chow extension. The corrected component restrictions yield such a certificate. The analytic numerical detector and finite descent are the required passage from this algebraic certificate to an actual primitive.

Proposition 3.4 (Chow-retraction transfer). Assume the obstruction kernels are stable under the algebraic Chow correspondences used below, including finite algebraic pushforward. Let L/KL/K be finite. Suppose there are smooth projective intermediate varieties Vj/LV_j/L and opposite-degree correspondences uj:XL⇝Vju_j:X_L\rightsquigarrow V_j, vj:Vj⇝XLv_j:V_j\rightsquigarrow X_L such that, on the homologically trivial group in degree rr, ∑j(vj)∗(uj)∗=λid⁡,λ∈Q×.\sum_j(v_j)_*(u_j)_*=\lambda\operatorname{id}, \qquad \lambda\in\mathbf Q^\times. If every (uj)∗zL(u_j)_*z_L has zero obstruction in the native KK-theory, then oX,Kr(z)=0\mathfrak o^r_{X,K}(z)=0.

Proof. Kernel stability and rational linearity annihilate the obstruction of the sum, hence of λzL\lambda z_L. Apply algebraic pushforward along π:XL→X\pi:X_L\to X. The Chow identity π∗π∗=[L:K]\pi_*\pi^*=[L:K] gives zero obstruction for λ[L:K]z\lambda[L:K]z. Divide by this nonzero rational scalar. ◻

For products of pointed curves, the projectors splitting h(C)=1⊕h1(C)⊕Lh(C)=\mathbf1\oplus h^1(C)\oplus\mathbb L factor the diagonal through products of Jacobians with Tate shifts. The companion construction paper separates this algebraic identity from the regular-model construction and from supported numerical correspondence transfer. Its model-sensitive criterion requires a regular projective flat target model after finite extension; that condition remains attached to that version of the argument.

4. Residual periods and the general source

Geometric transfer handles classes generated by a controlled family of correspondences. The general source-image problem asks why a homologically trivial class outside those families should still annihilate the actual obstruction.

Definition 4.1 (A finite original-source detector). For fixed (X,K,r)(X,K,r), a finite original-source detector consists of linear functionals P1,…,Pm:CH⁡r(X)Q,hom⟶RP_1,\ldots,P_m:\operatorname{CH}^r(X)_{\mathbf Q,\mathrm{hom}}\longrightarrow\mathbf R and a comparison with the obstruction construction such that P1(z)=⋯=Pm(z)=0⟹oX,Kr(z)=0. P_1(z)=\cdots=P_m(z)=0 \quad\Longrightarrow\quad \mathfrak o^r_{X,K}(z)=0.(4.1) The maps and their construction data are fixed before zz is chosen. A faithful detector additionally has the reverse implication.

Conjecture 4.2 (Vanishing of the original source periods). For every faithful finite detector associated with the original actual obstruction, with its comparison hypotheses satisfied, all its functionals vanish on CH⁡r(X)Q,hom\operatorname{CH}^r(X)_{\mathbf Q,\mathrm{hom}}.

Corollary 4.3. Under Hypothesis 2.2, the existence of a faithful detector as in Definition 4.1 and Conjecture 4.2 imply Conjecture 1.1 for (X,K,r)(X,K,r).

Proof. Apply (4.1) to each class and then Theorem 2.3. The latter theorem gives the form on the prescribed complement and the identity on all actual tests. ◻

The historical finite-period programme is recorded in [9]. Its comparison data select a residual quotient and preserve the numerical periods of the original Chow source. The decisive next step is a source-sensitive vanishing theorem. A comparison with de Rham, étale, syntomic or Abel–Jacobi data is useful here only when it identifies the same source class and the same pairing. An abstract cohomological isomorphism does not by itself compute Pj(z)P_j(z).

The tropical route asks for the corresponding comparison through compatible finite charts and refinements. Its principal tasks are descent of actual forms, compatibility of DD with refinement, and preservation of integration. A topological inverse-limit presentation motivates the construction; the needed assertion is a differential and pairing comparison. The source class must acquire a primitive in the ordinary actual category on the original complement.

5. The global pairing and a weaker vanishing target

Let FF be a number field and X/FX/F smooth, projective and geometrically integral of dimension nn. Write Vr(X)=CH⁡r(X)Q,hom.V^r(X)=\operatorname{CH}^r(X)_{\mathbf Q,\mathrm{hom}}. At finite places use the absolute values extending ∣p∣p=p−1|p|_p=p^{-1}, and at archimedean places the usual real or complex absolute value. Put wv=[Fv:Qv][F:Q],w_v=\frac{[F_v:\mathbf Q_v]}{[F:\mathbf Q]}, where the archimedean local degrees are 11 and 22. Then ∑vwvlog⁡∣a∣v=0\sum_vw_v\log|a|_v=0 for a∈F×a\in F^\times.

Conjecture 5.1 (Global Beilinson–Bloch pairing). For every number field FF, every such X/FX/F, and r,s≥1r,s\ge1 with r+s=n+1r+s=n+1, there is a canonical rational-bilinear real-valued pairing hX:Vr(X)×Vs(X)⟶Rh_X:V^r(X)\times V^s(X)\longrightarrow\mathbf R with the following properties.

  1. It is symmetric in complementary codimensions.

  2. For a finite extension E/FE/F, absolute normalization gives hXE(zE,wE)=hX(z,w)h_{X_E}(z_E,w_E)=h_X(z,w).

  3. For an algebraic correspondence Γ:Y⇝X\Gamma:Y\rightsquigarrow X and complementary classes for which the operations are defined, hX(Γ∗z,w)=hY(z,tΓ∗w)h_X(\Gamma_*z,w)=h_Y(z,{}^t\Gamma_*w).

  4. For disjoint representatives, it is a well-defined sum of local Green pairings, hX([Z],[W])=∑vwvλv(Zv,Wv),λv(Zv,Wv)=∫WvanγZv(v∤∞),h_X([Z],[W])=\sum_vw_v\lambda_v(Z_v,W_v), \qquad \lambda_v(Z_v,W_v)=\int_{W_v^{\mathrm{an}}}\gamma_{Z_v} \quad(v\nmid\infty), with the matching archimedean convention. The sum and the induced Chow pairing are independent of the choices and moving operations.

  5. It agrees with the classical Beilinson–Bloch constructions on their common domains. On a pointed curve, its negative is the positive Néron–Tate form in the corresponding Poincaré convention.

The construction and the axioms of a pairing are followed by different arithmetic questions: the sign on primitive classes, the radical, and bounded-height finiteness. Zhang’s arithmetic standard-conjecture formulation provides the relevant setting for those properties [3]. The present local-to-global problem concerns the domain and construction of the pairing itself.

5.1. The fixed global-source domain

The recovered H16 construction in [5] first works on the intersection of the local obstruction kernels over all finite completions. The finite-detector formulation recorded in [9, §1.3] uses an additional fixed-witness comparison. One fixes a finite set SS of places, the detector spaces, and that comparison before choosing the cycles, and writes the admitted domain as Hr(X)=ker⁡ ⁣(Vr(X)→PXr⨁v∈SVv,r∨). \mathcal H^r(X)=\ker\!\left( V^r(X)\xrightarrow{P_X^r}\bigoplus_{v\in S}V_{v,r}^{\vee}\right).(5.1) Here PXrP_X^r is applied to localizations of global classes. The good-place and comparison conditions in this description are part of the fixed global witness. The good-model vanishing statement alone does not identify the all-completion kernel with a chosen finite tester kernel.

Hypothesis 5.2 (Height assembly on the kernel domain). For the fixed witness in (5.1), a normalized real-valued pairing is constructed on Hr(X)×Hn+1−r(X),\mathcal H^r(X)\times\mathcal H^{n+1-r}(X), with the required local-sum, choice-independence and comparison properties. Any functoriality used on this domain is included among the assembly hypotheses.

Conjecture 5.3 (Global-source period vanishing). For every fixed global witness satisfying the assembly and comparison conditions, the maps PXrP_X^r vanish on Vr(X)V^r(X) in every relevant codimension.

Theorem 5.4 (Enlarging the global domain). Assume Hypothesis 5.2 for complementary codimensions r,sr,s. If PXr=0P_X^r=0 and PXs=0P_X^s=0, its pairing has domain Vr(X)×Vs(X)V^r(X)\times V^s(X) and retains the properties established by the assembly theorem.

Proof. The two vanishing assertions make the two kernels in (5.1) equal to their complete source groups. The same pairing therefore has the asserted domain. Its construction and the proofs of its properties are unchanged. ◻

This gives two paths to the complete global domain. General local source vanishing controls every class over every completion. Global-source period vanishing controls only classes coming from X/FX/F, but that is enough for Theorem 5.4. In both paths, local existence must be combined with the assembly statement; the finite local Green identity alone does not establish convergence or independence of a global height.

6. A long-term proof programme

The local and global targets have different quantifiers. Conjecture 1.1 concerns every homologically trivial cycle over a local field. The global construction only uses localizations of cycles over one number field. Within either target, existence of a Green form, independence of a local intersection number, and descent of a global sum to rational equivalence are distinct assertions. The following problems isolate the additional mathematics at each passage.

6.1. Supported Green forms and finite trace

Problem 6.1 (The prescribed-cycle construction). Establish Hypothesis 2.2 for every (X,K,r)(X,K,r) of Conjecture 1.1. For the chosen supported-complex construction, prove independence of the model Chern class, construct GZG_Z on the whole open set UZU_Z, and establish (2.1) on all compact actual tests of XanX^{\mathrm{an}}, with the prescribed cycle coefficients.

The proposed construction uses a perfect complex supported on the prescribed cycle and a decomposition after a proper generically finite map. On the complement of the support, the map must be finite and the complex must admit its asserted line-complex decomposition. Scalar Poincaré–Lelong then produces the Green form upstairs; finite trace supplies its actual representative downstairs. The numerical detector addresses curvature, while the supported complex determines the residue. These two roles require different compatibility statements.

For the finite-trace part, let q:Y→Xq:Y\to X have degree dd and let Mq\mathcal M_q be the stated module of forms to which the construction is applied. The required identities are DTr⁡qβ=Tr⁡qDβ,Tr⁡q(q∗Ω)=dΩ,∫XTr⁡qβ∧η=∫Yβ∧q∗η. \begin{split} D\operatorname{Tr}_q\beta&=\operatorname{Tr}_q D\beta,\qquad \operatorname{Tr}_q(q^*\Omega)=d\Omega,\\ \int_X\operatorname{Tr}_q\beta\wedge\eta&=\int_Y\beta\wedge q^*\eta. \end{split}(6.1) The last equality is required for every compact actual test for which the displayed degrees are complementary. The operator must be independent of a presentation of β\beta in Mq\mathcal M_q. For the restricted scalar–curvature construction, a relation among sums of u0Du1⋯Duku_0Du_1\cdots Du_k must therefore remain a relation after trace. This is the precise descent question; an operator defined on a free list of presentations does not yet define an operator on actual forms.

The supported calculation also has to survive the normal testing maps used by rational equivalences. For a normal integral rr-fold a:M→Xa:M\to X with a−1∣Z∣a^{-1}|Z| finite, the needed formula is ∫M∖a−1∣Z∣a∗GZ∧Dη+∫Ia(Z)η=∫Ma∗ΩZ∧η,η∈Bc0,0(Man), \int_{M\setminus a^{-1}|Z|}a^*G_Z\wedge D\eta +\int_{I_a(Z)}\eta =\int_M a^*\Omega_Z\wedge\eta, \qquad \eta\in\mathcal B_c^{0,0}(M^{\mathrm{an}}),(6.2) where Ia(Z)I_a(Z) is the refined zero-dimensional intersection. Its coefficients include generic Artinian lengths and residue-field degrees. The finite-flat base-change and supported-character arguments in [5, Section 3] are directed precisely at this formula.

The logical order is to establish the actual trace and the supported identity, then their base-change and point-slice compatibilities. Once DΨZ=ΩZD\Psi_Z=\Omega_Z is known, subtracting ΨZ\Psi_Z gives both the strong Green identity and the corresponding zero-curvature version of (6.2). For a finite extension L/KL/K, the primitive descends by Proposition 2.5 with denominator [L:K][L:K] and the extended original absolute value.

6.2. Homological triviality and the curvature obstruction

Problem 6.2 (Vanishing of the curvature obstruction). Assume Hypothesis 2.2. For every (X,K,r)(X,K,r), prove oX,Kr(CH⁡r(X)Q,hom)=0.\mathfrak o^r_{X,K}\bigl(\operatorname{CH}^r(X)_{\mathbf Q,\mathrm{hom}}\bigr)=0.

A solution to Problem 6.1 and Problem 6.2 proves strong G by Theorem 2.3. This is a sufficient route to strong G. Its remaining vanishing assertion can be approached in three ways.

First, use geometric relations in the Chow group. Under rational-fibre rigidity, Theorem 3.3 handles the algebraically trivial subgroup. The remaining map factors through the Griffiths quotient CH⁡r(X)Q,hom/CH⁡r(X)Q,alg.\operatorname{CH}^r(X)_{\mathbf Q,\mathrm{hom}}/\operatorname{CH}^r(X)_{\mathbf Q,\mathrm{alg}}. Where a rational Chow decomposition of the diagonal is available, Proposition 3.4 reduces the obstruction to the corresponding groups of the intermediate varieties. The equality of correspondences must hold in the rational Chow group, with the stated finite-extension and model hypotheses. The task on a residual summand is to kill its image under o\mathfrak o, rather than to prove the whole Griffiths quotient zero.

Second, construct and evaluate a finite detector. For fixed (X,K,r)(X,K,r), Definition 4.1 requires a comparison proving that vanishing of the chosen periods implies vanishing of this actual obstruction. Only after that comparison is proved does period vanishing solve the problem. An arithmetic Abel–Jacobi map may carry nonzero information on a homologically trivial cycle; a comparison with such a map must determine which part the actual periods measure. Geometric homological triviality alone is not a vanishing theorem for an arithmetic Abel–Jacobi class.

Third, seek primitives by tropical or Aeppli methods. A suitable comparison would have to pass from compatible local primitives through refinements to a global Ψ∈Br−1,r−1(Xan)\Psi\in\mathcal B^{r-1,r-1}(X^{\mathrm{an}}) satisfying DΨ=ΩD\Psi=\Omega. Its mathematical obligations are descent of actual representatives, compatibility with d′d' and d′′d'', and equality of the original integrals. The obstruction to gluing local primitives must be computed in the actual complex. A topological inverse-limit description supplies the spaces and maps, but not this differential comparison. This route may replace the numerical-detector argument if it produces the same global actual primitive.

For the global application there is a weaker target: prove oXFv,Fvr(zFv)=0for every z∈Vr(X) and every finite v. \mathfrak o^r_{X_{F_v},F_v}(z_{F_v})=0 \quad\text{for every }z\in V^r(X) \text{ and every finite }v.(6.3) It concerns only localizations of global classes. A proof in both complementary degrees can complete the obstruction-domain construction without resolving Problem 6.2 for arbitrary local classes.

6.3. Reciprocity and descent to a global Chow pairing

The starting domain of the H16 construction is Hr(X)=⋂v∤∞ker⁡ ⁣(Vr(X)⟶HBC,actr,r(XFv),z⟼oXFv,Fvr(zFv)). \mathcal H^r(X)= \bigcap_{v\nmid\infty} \ker\!\left(V^r(X)\longrightarrow H^{r,r}_{\mathrm{BC,act}}(X_{F_v}),\quad z\longmapsto\mathfrak o^r_{X_{F_v},F_v}(z_{F_v})\right).(6.4) Good-reduction vanishing makes only finitely many of these maps possibly nonzero. To replace the remaining obstruction spaces by finite-dimensional tester spaces, as in (5.1), one needs the additional fixed-witness comparison. Thus the all-completion domain is defined before any finite detector is selected.

The following criterion states exactly what the global sum needs. It applies to the obstruction-domain construction and to a direct construction from strong G.

Proposition 6.3 (A local-to-global criterion). Fix a smooth projective geometrically integral X/FX/F of dimension nn, complementary integers r+s=n+1r+s=n+1, and rational subspaces Hr⊆Vr(X)H^r\subseteq V^r(X) and Hs⊆Vs(X)H^s\subseteq V^s(X). Suppose that every pair of classes has disjoint rational cycle representatives. For every such pair (Z,W)(Z,W), suppose that real local numbers λv(Z,W)\lambda_v(Z,W) have been constructed, with the following properties.

  1. Each local number is independent of the chosen Green forms. It is rational-bilinear whenever the cycles are in common disjoint position, and λv(Z,W)=λv(W,Z).\lambda_v(Z,W)=\lambda_v(W,Z).

  2. For every fixed disjoint pair, λv(Z,W)=0\lambda_v(Z,W)=0 for all but finitely many vv.

  3. If Z0,Z1Z_0,Z_1 are rationally equivalent and both avoid WW, there is a rational-equivalence chain over FF, with a single element a∈F×a\in F^\times and a positive integer NN after clearing the rational coefficients of the chain and of WW, such that λv(Z1,W)−λv(Z0,W)=−N−1log⁡∣a∣v\lambda_v(Z_1,W)-\lambda_v(Z_0,W) =-N^{-1}\log|a|_v at every place, with the same aa and NN. The same assertion is required with the two complementary degrees interchanged. The moving statement supplies these chains over FF and permits a representative of a class in HrH^r (respectively HsH^s) to avoid any finite list of codimension-ss (respectively codimension-rr) cycles.

Then hX(z,w)=∑vwvλv(Z,W) h_X(z,w)=\sum_v w_v\lambda_v(Z,W)(6.5) is a well-defined rational-bilinear pairing on Hr×HsH^r\times H^s, symmetric in complementary degrees.

If the subspaces and local constructions are given compatibly over finite extensions, and λu(ZEu,WEu)=λv(ZFv,WFv)(u∣v)\lambda_u(Z_{E_u},W_{E_u})=\lambda_v(Z_{F_v},W_{F_v}) \qquad(u\mid v) for every finite E/FE/F, then the pairing is invariant under finite extension in the absolute normalization.

Proof. The second hypothesis makes (6.5) a finite sum. The first gives independence of Green choices. Summing the third equality and using the weighted product formula proves independence of the representative of the first class while the second is fixed. Symmetry gives the same assertion for the second class.

For two disjoint representative pairs (Z0,W0)(Z_0,W_0) and (Z1,W1)(Z_1,W_1), choose W′W' avoiding both Z0Z_0 and Z1Z_1. Compare successively (Z0,W0),(Z0,W′),(Z1,W′),(Z1,W1).(Z_0,W_0),\quad(Z_0,W'),\quad(Z_1,W'),\quad(Z_1,W_1). The preceding independence assertions identify their heights. Moving a finite list of cycles into common disjoint position now proves rational bilinearity from the first hypothesis. Local symmetry gives global symmetry.

For finite E/FE/F, group the finite sum over the places uu above each vv. The local equality and ∑u∣v[Eu:Qv][E:Q]=[Fv:Qv][F:Q]\sum_{u\mid v}\frac{[E_u:\mathbf Q_v]}{[E:\mathbf Q]} =\frac{[F_v:\mathbf Q_v]}{[F:\mathbf Q]} give the asserted invariance. ◻

For the obstruction-domain route, the supported point-slice identity supplies the principal transformation law and [5, Theorem 3.6] gives the pairing on (6.4), under its standing analytic hypotheses. The remaining task for the full domain is (6.3) in degrees rr and ss.

For the direct route, supply strong Green forms for all local cycles in those degrees and establish the principal-push construction on normal carriers. The latter construction supplies the exact norm evaluation needed in part (3) of Proposition 6.3. Disjoint-polar Stokes supplies part (1), and the good-model construction supplies part (2). This is the route of [5, Theorem 7.5], with its actual-calculus, supported-complex, finite-trace, projective-bundle and immersion-germ hypotheses. It does not require the reverse implication from an arbitrary strong Green solution to vanishing of the selected obstruction.

Problem 6.4 (Completion of the height construction). Establish the hypotheses of Proposition 6.3 on Vr(X)×Vn+1−r(X)V^r(X)\times V^{n+1-r}(X) for every smooth projective geometrically integral X/FX/F and every admissible rr. In the obstruction-domain route, prove (6.3) in both degrees. In the direct route, prove strong G together with the stated principal-push and good-place inputs.

The delicate point in the principal law is an exact global norm, not merely a local constant. For a normal carrier of a higher-Chow one-chain, the refined intersection coefficients and the omitted parameter fibre must give one element of F×F^\times whose absolute value is evaluated at every place. The product formula then performs the Chow descent. This explains why reciprocity, refined point-slice formulas and field normalization are prerequisites for the height construction rather than consequences of a local existence assertion.

6.4. Correspondence adjunction and comparison with classical heights

Problem 6.5 (The functorial and normalized height system). For the pairings constructed in Problem 6.4, prove correspondence adjunction hX(Γ∗z,w)=hY(z,tΓ∗w)h_X(\Gamma_*z,w)=h_Y(z,{}^t\Gamma_*w) on all the stated complementary domains, compatibility with isomorphisms and conjugation, and the curve normalization in Conjecture 5.1. Identify the pairing with each classical construction on its common domain and with the same absolute normalization.

The correspondence argument requires a height projection formula for properly positioned cycles and preservation of the pairing domains under the induced Chow maps. Resolving a correspondence over the number field gives two morphisms; moving with respect to both morphisms allows the projection formula to be applied twice. This is the mechanism of [5, Section 9]. It is additional to Proposition 6.3, which only uses the three local laws stated there.

For comparison with an arithmetic-model height, one fixes a regular projective arithmetic model and numerically admissible rational lifts of both cycles. The local comparison identifies the Green integral with the finite Euler intersection of these supplied lifts. Summing with the weights wvw_v and the matching archimedean convention gives the arithmetic intersection height; see [5, Section 8]. The existence of numerically admissible lifts for an arbitrary cycle is a further problem. A comparison theorem conditional on such lifts does not itself produce them.

Determinant, Deligne-pairing and Poincaré-metric comparisons are a related way to identify the normalized height on families. Their bad-reduction extension calls for admissible vertical corrections. These comparisons can be developed in parallel with the domain problem, but enter the arithmetic applications only on the domains where the resulting identities have been established.

6.5. Arithmetic Lefschetz and positivity on products of curves

The arithmetic branch begins with a height system on the varieties used by the correspondences. Its hypotheses are rational bilinearity and symmetry, invariance under finite extension and conjugation, correspondence adjunction, and the positive Néron–Tate normalization of −hC-h_C on a pointed curve. These are the four height axioms of [6, Section 1.1]. They must hold on all subproducts, source curves and finite extensions in the following discussion.

Fix a product X=∏i=1nCiX=\prod_{i=1}^nC_i of pointed smooth projective geometrically connected curves (Ci,oi)(C_i,o_i) over FF, with polarization ℓ=∑i=1naipi∗[oi],ai∈Z>0.\ell=\sum_{i=1}^n a_i p_i^*[o_i],\qquad a_i\in\mathbf Z_{>0}. For a nonempty subset SS, put YS=∏i∈SCi,ΠS=⨂i∈S(ΔCi−oi×Ci−Ci×oi),USc(E)=ΠS∗Vc(YS,E).Y_S=\prod_{i\in S}C_i,\qquad \Pi_S=\bigotimes_{i\in S} (\Delta_{C_i}-o_i\times C_i-C_i\times o_i),\qquad U_S^c(E)=\Pi_{S*}V^c(Y_{S,E}). Fix an algebraic closure F‾\overline F and an embedding F‾↪C\overline F\hookrightarrow\mathbf C. Choose one finite F0/FF_0/F over which the central projectors of [6] are defined. For ∣S∣=2c−1|S|=2c-1, these are self-transpose rational Chow projectors eS≤ΠSe_S\le\Pi_S and RS=ΠS−eSR_S=\Pi_S-e_S. The eSe_S-summand is represented by a fixed abelian variety and carries its Néron–Tate form. Put AS,E=USc(E)∩CH⁡c(YS,E)Q,alg,GS,E=USc(E)/AS,E.A_{S,E}=U_S^c(E)\cap\operatorname{CH}^c(Y_{S,E})_{\mathbf Q,\mathrm{alg}}, \qquad G_{S,E}=U_S^c(E)/A_{S,E}. The central radical computation gives RSAS,E=ker⁡(AJQ∣AS,E)=AS,E∩rad⁡(hYS∣USc(E)).R_SA_{S,E}=\ker(\mathrm{AJ}_\mathbf Q|_{A_{S,E}}) =A_{S,E}\cap\operatorname{rad}(h_{Y_S}|_{U_S^c(E)}). It therefore defines the residual form by hG,S([z],[w])=hYS(RSz,RSw),qG,S(ξ)=(−1)chG,S(ξ,ξ).h_{G,S}([z],[w])=h_{Y_S}(R_Sz,R_Sw),\qquad q_{G,S}(\xi)=(-1)^ch_{G,S}(\xi,\xi). These constructions and their independence statements use the height axioms above; their proofs are in [7, Section 3]. All the following targets are required for every finite E/F0E/F_0.

Problem 6.6 (Off-central Chow groups and central heights). Prove the following statements for all subproducts.

  1. If ∣S∣≠2c−1|S|\ne2c-1, then USc(E)=0U_S^c(E)=0.

  2. If ∣S∣=2c−1|S|=2c-1, the map AJQ:AS,E⟶Jc(YS,C)(C)⊗ZQ\mathrm{AJ}_\mathbf Q:A_{S,E}\longrightarrow J^c(Y_{S,\mathbf C})(\mathbf C)\otimes_{\mathbf Z}\mathbf Q is injective.

  3. The residual form qG,Sq_{G,S} is positive definite on GS,E⊗QRG_{S,E}\otimes_\mathbf Q\mathbf R.

The first assertion is equivalent, by the Chow-block calculation in [7], to arithmetic hard Lefschetz for the pointed product. The second removes the algebraic height radical from each central block. Together with the third it gives central positivity. Under the first assertion, only central blocks remain, so the primitive sum-of-squares identity propagates this positivity to the whole product. In particular, for Pℓr(XE)=ker⁡ ⁣(ℓn+2−2r:Vr(XE)→Vn+2−r(XE)),P_\ell^r(X_E)=\ker\!\left( \ell^{n+2-2r}:V^r(X_E)\to V^{n+2-r}(X_E)\right), the target is positivity of qℓ,r(z)=(−1)rhX(z,ℓn+1−2rz),1≤r≤⌊n+12⌋.q_{\ell,r}(z)=(-1)^r h_X(z,\ell^{n+1-2r}z), \qquad 1\le r\le\left\lfloor\frac{n+1}{2}\right\rfloor. These are consequences of the central assertions and the height axioms; the construction of a pairing supplies the height to be studied, not the Chow vanishing or its sign.

6.6. Integral discreteness and bounded-degree Northcott

Fix a central subproduct and abbreviate GS,EG_{S,E} and qG,Sq_{G,S} to GEG_E and qGq_G. Let ΛG(E)\Lambda_G(E) be the image in GEG_E of integral-coefficient homologically trivial cycles after applying ΠS\Pi_S. The quotients are viewed inside one geometric rational group, and the residual projector has one fixed denominator. The finiteness problem counts these classes, not their cycle representatives.

Problem 6.7 (Residual integral Northcott). For every real BB and positive integer DD, prove finiteness of {ξ∈⋃[E:F0]≤DΛG(E):qG(ξ)≤B}.\left\{\xi\in\bigcup_{[E:F_0]\le D}\Lambda_G(E): q_G(\xi)\le B\right\}.

This is stronger than real residual positivity: finite integral height balls imply real positivity by the finite-dimensional rounding argument of [7, Section 5.1]. The programme for the converse direction has two separate parts.

Over a fixed EE, one route seeks a rational basis e1,…,ete_1,\ldots,e_t of GEG_E and one integer NE≥1N_E\ge1 such that dim⁡QGE=t<∞,ΛG(E)⊆NE−1∑j=1tZej. \dim_\mathbf QG_E=t<\infty,\qquad \Lambda_G(E)\subseteq N_E^{-1}\sum_{j=1}^t\mathbf Z e_j.(6.6) Real positivity then makes every bounded-height ball finite. Scalar-extended injectivity of a residual regulator GE⊗QQp→H1(E,Vp)G_E\otimes_\mathbf Q\mathbf Q_p\to H^1(E,V_p), with the finite-dimensional target specified in [6, Section 3.1], is a proposed route to finite rank. The remaining task is one common denominator bound across all coefficient primes. Separate integral bounds at each prime do not yield (6.6). An alternative route uses a finite quotient ΛG(E)/mΛG(E)\Lambda_G(E)/m\Lambda_G(E) for an integer m≥2m\ge2, together with a genuine positive lower bound for qGq_G on the nonzero elements of ΛG(E)\Lambda_G(E). Under real positivity, the height-descent argument in [7, Section 5.4] then gives fixed-field Northcott.

As EE varies, fixed-field Northcott must be combined with control of fields of classes. For a geometric residual class ξ\xi, put F0(ξ)=F‾Stab⁡Gal⁡(F‾/F0)(ξ).F_0(\xi)=\overline F^{\operatorname{Stab}_{\operatorname{Gal}(\overline F/F_0)}(\xi)}. The required field-theoretic statement is that, for fixed B,DB,D, all these extensions of degree at most DD and height at most BB are unramified outside one finite set of places of F0F_0. Hermite–Minkowski then leaves finitely many possible class fields KK. If ξ∈ΛG(E)\xi\in\Lambda_G(E) and K=F0(ξ)K=F_0(\xi), the trace identity gives [E:K]ξ∈ΛG(K),1≤[E:K]≤D.[E:K]\xi\in\Lambda_G(K),\qquad 1\le [E:K]\le D. Fixed-field Northcott applied to the finite collection of balls of height at most D2BD^2B proves Problem 6.7 for B≥0B\ge0; positivity handles negative BB. The full criterion and its converse are proved in [7, Section 6].

Finally, residual integral Northcott together with assertions (1) and (2) of Problem 6.6 gives the product programme of arithmetic hard Lefschetz and primitive integral Northcott, by [7, Section 7]. Thus the hierarchy is: local and global constructions produce the pairing; adjunction and curve normalization permit the central reductions; central Chow vanishing and positivity govern its nondegeneracy; integral discreteness and field control supply the additional finiteness. This order keeps each consequence with the hypotheses that actually imply it.

References

[1] A. Mihatsch, δ\delta-Forms on Lubin–Tate spaces, arXiv:2112.10018, version 4 (2024), https://arxiv.org/abs/2112.10018v4.

[2] K. Künnemann, Height pairings for algebraic cycles on abelian varieties, Ann. Sci. École Norm. Sup. (4) 34 (2001), 503–523, https://www.numdam.org/article/ASENS_2001_4_34_4_503_0.pdf.

[3] S.-W. Zhang, Standard conjectures and height pairings, arXiv:2009.07089, version 3 (2022), https://arxiv.org/abs/2009.07089v3.

[4] Local Green Forms and Chow Correspondences: Conditional Reductions and Construction Draft, companion manuscript, 17 November 2026. Based on the frozen local construction sources of 16 November 2026.

[5] Global Height Pairings from Admitted Green Classes, companion manuscript, 17 November 2026, Theorem 3.6 and the subsequent proper-intersection, reciprocity and comparison arguments; https://dongkaicheng.github.io/math-research-atlas/projects/strong-green/editions/v0.5.1/papers/global-height-pairing/.

[6] Arithmetic Conjectures for Beilinson–Bloch Heights on Products of Curves, companion programme, 17 November 2026; https://dongkaicheng.github.io/math-research-atlas/projects/strong-green/editions/v0.5.1/papers/arithmetic-conjectures/.

[7] Arithmetic Lefschetz and Integral Heights: Proofs and Reductions on Products of Curves, companion manuscript, 17 November 2026; https://dongkaicheng.github.io/math-research-atlas/projects/strong-green/editions/v0.5.1/papers/arithmetic-reductions/.

[8] Algebraic Equivalence and Chow–Motivic Transfer for Actual Non-Archimedean Green Forms, revised internal draft, 16 November 2026, algebraic-transfer-conditional.tex.

[9] Cross-chat research synthesis for the strong Green and Beilinson–Bloch programme, 16 November 2026, cross-chat-synthesis.tex, especially §§1.1–1.3. This is a source and version synthesis; its H16 entry refers to the corresponding frozen assembly proof.

PDF and source

Download the paper

The PDF contains the full article, numbered statements, proofs, and references. The LaTeX download is self-contained, with the included proof sources expanded.

Edition and evidence

Conjectures and conditional formal reductions. Analytic interfaces retain the evidence status of their source records.

Source and PDF fingerprints

8 proof environments retained from the source. This checks document preservation, not mathematical validity.

Expanded LaTeX SHA-256

e35fc18f8a5ad3d8aa41b28c075a30127aafa79212353a9cbee5d12ec59f091f

PDF SHA-256

4df5f2ad6c54b20508a07f1fcc0451704cfe8d0d036045ab5f978b9d2db8ea13

Permanent records and reader corrections

These links retain individual claim identities and revision-bound feedback. They are the reference index for this paper and its associated open problems.