Conjectures and programme
Arithmetic Conjectures for Beilinson–Bloch Heights
Central classes, positivity, and integral Northcott
17 November 2026
Abstract
We formulate an arithmetic programme for height pairings on the homologically trivial Chow groups of products of curves. Exact Chow projectors reduce arithmetic hard Lefschetz to off-central vanishing and reduce primitive integral Northcott to two central assertions: injectivity of the algebraic Abel–Jacobi map and Northcott on a residual Griffiths quotient. The latter assertion contains real positivity but also requires integral and varying-field information. We separate the proposed arithmetic inputs into scalar detection by continuous étale cohomology, a common rational lattice, and ramification control for the minimal fields of cycle classes. The stated implications specify where each input enters.
1. Introduction
The expected positivity of Beilinson–Bloch heights is an arithmetic Hodge–Riemann principle. A Northcott property is stronger: it asks for finiteness of integral cycle classes of bounded degree and height. Even a positive quadratic form does not give this finiteness unless the integral classes are arithmetically discrete. Allowing the field of definition to vary adds a further descent problem.
For products of curves, tensor Chow projectors separate these questions. The all- part of a subproduct carries a central height precisely when its dimension is in codimension . The other parts are responsible for arithmetic Lefschetz. Within the central part, curve correspondences separate a Néron–Tate summand and its residual Griffiths quotient. This paper formulates the resulting conjectures and organizes the arithmetic inputs to their study. The complete reduction proofs are given in the companion manuscript [1].
These conjectures are expressed within the Bloch–Beilinson and arithmetic standard-conjecture framework. The arithmetic Lefschetz and primitive positivity statements are those recorded in [6, Conjecture 2.3.1]. The purpose of the formulation here is to specify the residual objects, their integral images and the exact implications connecting the different arithmetic questions.
1.1. The fixed product and height data
Let be a number field, and let where the curves are smooth, projective and geometrically connected. For a finite extension , write
Fix an algebraic closure of and an embedding into . Homological triviality is defined by geometric rational Betti cohomology. All Chow groups below have rational coefficients unless otherwise indicated. We regard as a subspace of : restriction is injective because a rational equivalence descends to a finite extension and trace composed with restriction multiplies by its degree.
Definition 1.1. A curve-normalized height system consists of pairings on all smooth projective varieties used below, including the subproducts and the source curves of correspondences. For such a variety of dimension , these are over every finite extension , satisfying:
bilinearity over and symmetry under exchange of the complementary degrees;
invariance under extension in the absolute normalization, and covariance under isomorphisms, including Galois conjugation;
correspondence adjunction
on a pointed curve , the form on is the positive Néron–Tate form for the canonical principal polarization, in the corresponding Poincaré normalization.
These hypotheses will be understood in every statement involving . A real positive form means positive definiteness after tensoring the rational group over a fixed number field with . For codimension on , the symbol denotes the complex Abel–Jacobi map with target .
Let be the image in of the integral-coefficient cycles over with zero geometric rational cycle class. Thus torsion and rational equivalence are removed before counting. For a fixed rational projector , we use its projected integral image ; for an integral projector this equals .
Definition 1.2. For compatible integral images in one geometric rational group and a quadratic form , bounded-degree Northcott means that, for every and integer , is finite. Fixed-field Northcott means the same assertion with fixed. Bounded denominators mean the union of for , with fixed.
Northcott for integral images implies the bounded-denominator version by scaling. A fixed initial extension is harmless for bounded-degree assertions: after replacing by , , the integral image remains integral, and the absolute height is unchanged.
1.2. Central cores and the residual pairing
For a subset , put Here . The empty subproduct has zero homologically trivial Chow groups. The core is central if .
Fix such an odd subproduct. Let The maximal curve-visible construction in [1, Section 3] gives an actual self-transpose Chow projector , with . Choose one finite extension over which these projectors are defined for all odd subproducts. For every finite , for a fixed abelian variety . The decomposition is height-orthogonal, and the form on the first summand is a positive Néron–Tate form. Moreover, For , in , define The radical equality makes this pairing well defined, and the projector construction proves its independence of the maximal curve-visible lift. These facts hold for every height system in Definition 1.1.
The integral structures are The projector is integral. The residual projector is rational, with one fixed denominator; that denominator is independent of the cycle and of . Since algebraic triviality is geometric, all these rational quotients embed in their counterparts over .
1.3. The three central assertions
Conjecture 1.3 (Off-central vanishing). For every subproduct , every finite , and every with ,
Conjecture 1.4 (Central algebraic Abel–Jacobi injectivity). For every and finite , the map is injective.
Conjecture 1.5 (Residual integral Northcott). For every , every , and every integer , the set of geometric classes is finite.
These assertions address, respectively, the degrees in which the height pairing should be nondegenerate, its radical on the algebraically trivial part, and the arithmetic size of the remaining integral quotient.
2. From the central assertions to the whole product
2.1. Arithmetic Lefschetz and the primitive form
For , put The primitive integral image is .
Theorem 2.1 (Equivalence of the arithmetic programmes). For the fixed pointed product, positive integers , and a height system satisfying Definition 1.1, the following conditions are equivalent.
- Conjecture 1.3 and Conjecture 1.4 and Conjecture 1.5
hold for all subproducts of .
For every finite and every , and has bounded-degree Northcott on the primitive integral images.
Either condition implies real positive definiteness of all primitive forms and nondegeneracy, also after real scalar extension, of the complementary height pairings over every fixed number field.
The full proof is the principal reduction theorem of [1], completed in its Section 7. Its two mechanisms are the exact central-block identity and the central radical decomposition. We recall them because they determine which conjecture is needed at each step.
The Chow block indexed by is with acting as multiplication by . Writing , the identification sends to , where and inserts on and on . The lowering operator is The arithmetic hard Lefschetz maps are isomorphisms exactly when the noncentral cores vanish. This equivalence is independent of the choice of the positive weights , with the points fixed.
In a central block let , , , and . For , primitivity means ; the signed height is then (2.1) This central-block identity uses only adjunction and the square-zero calculation. Off-central vanishing enters when all of is reduced to these central blocks. If is integral, the single integer clears the denominators of every .
Finally, central integral Northcott is equivalent to Indeed The visible summand has classical bounded-degree Northcott through fixed correspondences with bounded denominators. A nonzero algebraic radical class would have infinitely many integral multiples of height zero. After this radical is removed, the pair detects .
2.2. Residual positivity and its precise role
Conjecture 2.2 (Real residual positivity). For every central subproduct and every finite , extends to a positive definite quadratic form on .
This is a weaker consequence of Conjecture 1.5. More generally, if a subgroup spans a rational vector space , finiteness of all height balls in forces the real extension of to be positive definite. A negative direction gives a negative rational vector and its integer multiples. A real null direction, in a finite-dimensional rational subspace, gives infinitely many integral vectors of bounded height by rounding its scalar multiples. This is the proof of the real-positivity lemma in [1, Section 5.1].
The converse is an integral problem. For example, has a positive real form and a finite free local completion at every prime. Nevertheless . The extra input needed for finiteness is one common rational lattice, rather than a separate local bound at each prime.
With Conjecture 1.4 and Conjecture 2.2, the central form is real positive definite by its orthogonal splitting. Adding Conjecture 1.3 propagates this positivity and nondegeneracy to the product through (2.1). The associated primitive bilinear form then satisfies The finiteness assertion in Conjecture 1.5 requires the additional arithmetic inputs below.
3. Arithmetic detection and the integral lattice
Fix one central subproduct of dimension , and abbreviate to .
3.1. Scalar detection by a residual regulator
For every prime , let The actual projector has geometric cohomology in one degree. Consequently, its part of Hochschild–Serre has one row and gives a canonical identification The continuous cycle class of annihilates . It therefore defines a residual arithmetic Abel–Jacobi map compatible with restriction and trace and independent of the curve-visible lift under the common cohomological identification. The construction is proved in [1, Section 4.1], in the continuous-cohomology framework of [4; 2].
The representation has weight . For one fixed set of places of , containing the bad-reduction places, inflation identifies with archimedean places included. This whole target is finite-dimensional over .
Conjecture 3.1 (Residual scalar detection). For every finite and every prime , the scalar-extended map is injective.
The word scalar specifies the strength needed here. Injectivity of the -linear map alone would not imply finite rational rank: a finite-dimensional -space is infinite-dimensional over . In contrast, scalar-extended injectivity at one prime gives
For every fixed , the integral cycle map and one denominator clearing place in a finitely generated -module. Under scalar-extended injectivity, The proof uses the integral image theorem of [2, Theorem 2.4]; it is given in [1, Section 4.3].
3.2. A common lattice across coefficient primes
Conjecture 3.2 (Residual integral discreteness). For every finite , the space is finite-dimensional, and there are a rational basis and an integer such that
For a finite-rank torsion-free group and a real positive definite quadratic form, the following conditions are equivalent: finite height balls, a positive height gap, finite generation, a common denominator bound in a rational basis, and relative compactness of the diagonal image in the finite adeles. The proof in [1, Section 5.2] is an elementary lattice argument. It follows that Conjecture 3.2 and Conjecture 2.2 imply fixed-field Northcott for every . Conversely, fixed-field Northcott together with finite rational rank implies Conjecture 3.2 and real positivity. Finite rank is an explicit premise in this converse.
A route to Conjecture 3.2 is an almost-everywhere saturation comparison. Fix , put , and suppose that is finite-dimensional. Seek a full rational lattice and full local lattices such that the scalar-extended regulators are injective and (3.1) Then the inverse lattices at the finitely many exceptional primes can be absorbed into one integer . Thus (3.1) proves Conjecture 3.2. The fixed-prime local bounds above supply the first type of control; the equality with one common rational lattice is the comparison still required.
Integral detection is formulated on the stated image in a rational quotient. The unrestricted integral continuous cycle class map has torsion counterexamples over number fields in codimension three [5, Theorems 1.3–1.4]. The torsion kernels in those statements and the torsion-free group are different arithmetic objects.
3.3. The height-gap route
There is a second route that can start from one coefficient prime. Suppose is the norm of a real positive quadratic form, and, for some integer , The quantitative descent criterion of [1, Section 5.4] gives finite generation and fixed-field Northcott. If , , and , then The reason is that two classes in the ball congruent modulo would differ by with . The usual height descent then generates the group from a finite ball and representatives modulo .
Scalar detection at one prime supplies a finite quotient with , through the local integral bound. A genuine positive height gap is the additional input. Real positive definiteness alone does not supply this gap on a subgroup with unbounded denominators.
4. Minimal class fields and bounded-degree Northcott
4.1. Trace and the class field
View the residual groups inside the geometric quotient with its discrete -action. For a class , define If , its minimal class field is contained in , and trace gives (4.1) Indeed the pushforward of an integral representative has geometric class equal to the sum over the -embeddings of , and every summand is . Thus the multiplier is at most when . This concerns the field of the class; a particular cycle representative need not descend to .
Conjecture 4.1 (Residual class-field ramification). For every and integer , there is a finite set of finite places of such that the fields , for are unramified outside .
Under fixed-field Northcott for every finite extension, Conjecture 4.1 is equivalent to bounded-degree Northcott. Hermite–Minkowski leaves finitely many possible class fields of degree at most . Equation (4.1) then reduces the classes to the finitely many fixed-field balls The converse follows by taking the union of ramification sets of the finitely many classes in a ball. The full proof is in [1, Section 6.2]; the finite-ramification form of Hermite–Minkowski is [3, Theorem 10.2.2].
4.2. The extra information needed as the field varies
A fixed set for controls ramification over . It does not bound the ramification of the class field . The distinction persists even when fixed-field estimates are uniform in the degree. For example, choose infinitely many distinct quadratic extensions , with characters , and set These integral images are compatible with trace. Their ranks, height gaps and quotients modulo any integer have degree-dependent uniform bounds, and all fixed-field balls are finite. Yet the infinitely many have degree two and height one. This example isolates the field-control hypothesis in the ramification criterion.
A quantitative strengthening asks for constants , depending only on the fixed residual data and , such that (4.2) This is a research question stronger than Conjecture 4.1. A bound of the displayed form would imply the required finite ramification set at every bounded height.
5. A proof programme and its arithmetic obstacles
The reduction theorem provides two complementary routes. One seeks the three central assertions directly. The other constructs arithmetic detectors for the residual quotient and then proves integral and field-of-definition control.
5.1. The geometric and central steps
The first target is Conjecture 1.3. The tensor Chow decomposition gives an exact test: a nonzero off-central core would produce a misplaced Lefschetz string. A proof must therefore eliminate the core in the rational Chow group, rather than merely its cohomological realization.
The next target is the algebraic kernel in Conjecture 1.4. The projector construction already identifies it as , and adjunction places it in the full central height radical. Establishing its vanishing is a separate arithmetic cycle statement. Once it vanishes, the visible Néron–Tate factor has all the required fixed-field and bounded-degree finiteness properties.
For the residual factor, real positivity is an initial target. It supplies Cauchy–Schwarz and, together with the algebraic kernel statement, central nondegeneracy. The primitive sum-of-squares identity transfers these consequences uniformly to every degree of the product.
5.2. A detector-to-Northcott route
For each fixed finite , the sequence of inputs is:
Prove scalar-extended injectivity of one residual regulator. The finite-dimensional cohomological target then gives finite rational rank.
Prove one common denominator bound for the integral image, for example through the all-prime saturation condition (3.1). Alternatively, combine a finite Kummer quotient with a positive height gap.
Establish real residual positivity. Together with the common lattice, this gives fixed-field Northcott.
Control the ramification of minimal class fields in bounded degree and height. The trace criterion then gives bounded-degree Northcott.
The alternatives in the second step have different arithmetic content. All-prime saturation compares integral cohomology across coefficient primes. The height-gap route instead requires a uniform positive lower bound on nonzero integral classes over the fixed field.
Under real positivity, scalar detection and the almost-everywhere lattice comparison for every finite , the first three steps establish fixed-field Northcott. Adding Conjecture 4.1 proves Conjecture 1.5. Alternatively, a finite Kummer quotient and a genuine gap over every fixed field, together with real positivity and Conjecture 4.1, give the same conclusion without a preliminary finite-rank assumption. Finally, Conjecture 1.3 and Conjecture 1.4 and Theorem 2.1 give the product programme.
The local-to-global construction of a height system and these arithmetic properties meet at Definition 1.1. A construction of local Green forms must supply the global pairing, adjunction and curve normalization before the arithmetic reductions apply. The present programme starts from that explicitly stated height system; its subsequent tasks concern vanishing, positivity, integral discreteness and fields of cycle classes.
References
[1] Arithmetic Lefschetz and integral heights: proofs and reductions on products of curves, companion research manuscript, 17 November 2026.
[2] B. Kahn, On the injectivity and non-injectivity of the -adic cycle class maps, manuscript dated 10 March 2024. Author-hosted PDF.
[3] K. S. Kedlaya, Weil cohomology in practice, Chapter 10, Comparing Galois representations: the Faltings–Serre method, lecture of 30 October 2019. Author-hosted notes.
[4] W. Raskind, Higher -adic Abel–Jacobi mappings and filtrations on Chow groups, arXiv:alg-geom/9412001.
[5] F. Scavia and F. Suzuki, Non-injectivity of the cycle class map in continuous -adic cohomology, arXiv:2208.06713.
[6] S.-W. Zhang, Standard conjectures and height pairings, arXiv:2009.07089v3. The numbering cited here refers to the version of 29 April 2022.