Elliptic constant cycle curves on Kummer surfaces

Jiexiang Huang

REVIEW 3 major objections 6 minor 2 references

Elliptic constant cycle curves on Kummer surfaces

T0 review · 3 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This paper proves that the order of elliptic constant cycle curves on Kummer surfaces is determined by the parity of the torsion order: it equals $n/2$ for even $n$ and $n$ for odd $n$, and every positive integer occurs.

desk verdict First exact computation of constant cycle curve orders on Kummer surfaces, with a load-bearing 'one can check' lattice-index claim that a referee should verify. read the letter →

arxiv 2506.06260 v1 pith:Q7SGL5EU submitted 2025-06-06 math.AG

Jiexiang Huang This is my paper

classification math.AG MSC 14C2514J2814H52

keywords constantcyclecurvesK3surfacesKummerChowgroupsintermediateJacobiansAbel-Jacobimapselliptictorsionpoints

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Constant cycle curves are curves whose points all represent the same class in the Chow group of the ambient K3 surface; the order measures how far the diagonal is from being decomposable, with rational curves being exactly the order-one case. This paper determines the order for the elliptic fibres that occur on Kummer surfaces built from a product of two elliptic curves: if the fibre lies over a torsion point of order $n>2$, the order is $n/2$ for even $n$ and $n$ for odd $n$, except in a narrow CM case. The result is the first exact computation of orders larger than one for non-rational constant cycle curves via intermediate Jacobians, and it implies that every positive integer is realized as the order of a smooth genus-one constant cycle curve on some K3 surface.

What carries the argument

The computation rests on three objects: the torsion class $[Z'_t]=2([t]-[e_1])\times[\mathrm{id}]$ in $\mathrm{CH}^2(E_1\times E_2\times E_t)$, the transcendental intermediate Jacobian $J^3_{\mathrm{tr}}(X\times E_t)$ (the part of the third intermediate Jacobian coming from the transcendental cohomology), and the theorem that the transcendental Abel–Jacobi map is injective on torsion classes. The class $[Z'_t]$ is a compactification of the torsion class $\kappa'_t$ whose order equals $\mathrm{ord}(E_t)$; its image in the intermediate Jacobian has order $d(n)$ because $[\mathrm{id}]$ is a primitive class with self-intersection $-2$. A structural lemma on product cycles in triple products of elliptic curves then shows that no smaller multiple of $[Z'_t]$ can come from $\mathrm{CH}^1(E_1\times E_2)\otimes\mathrm{CH}^1(E_t)$, except in the excluded CM configuration.

What would settle it

Check the two indices in the proof of Lemma 3.4: the embeddings $\pi^*H^2(X,\mathbb{Z})\subset H^2(\mathrm{Bl}(A),\mathbb{Z})$ and $\pi^*(K)\subset\bigoplus\mathbb{Z}[E_i]$ are claimed to have index $2^{11}$; computing either cokernel and finding a different order would break the identification $\ker(\pi^*)\simeq(\bigoplus\mathbb{Z}[E_i]/\pi^*(K))\otimes H^1(E_t,\mathbb{Z})$. Also, for a non-isomorphic isogenous CM pair with $4\mid n$ outside $n=4$, computing $\mathrm{ord}(E_t)$ would test whether the theorem's excluded case really is the only obstruction.

Watch

Extended reading notes

Core claim

The central claim is Theorem 1.1: let $E_1,E_2$ be elliptic curves and $t$ a torsion point of order $n>2$; then the fibre $E_t$ in the Kummer surface $\mathrm{Kum}(E_1\times E_2)$ has order $d(n)$, equal to $n/2$ when $n$ is even and $n$ when $n$ is odd, except possibly when $E_1,E_2$ are non-isomorphic, isogenous with complex multiplication and $4\mid n$. The proof constructs a torsion class $2([t]-[e_1])\times[\mathrm{id}]$ in the Chow group of the abelian threefold $E_1\times E_2\times E_t$, shows its order is $d(n)$ via the Abel–Jacobi map, and proves that this order is preserved when passing back to the Kummer surface. In the isogenous even case the preservation step uses a second-order test with the addition morphism and the intermediate Jacobian; the excluded case is genuinely exceptional, as an explicit example shows both order $1$ and order $2$ can occur for $n=4$ according to the lattice of the two CM curves.

Load-bearing premise

The step that carries the whole proof is that the order of the curve can be read off from the order of a torsion class on the abelian threefold; this relies on an injectivity statement for transcendental pullback maps whose proof contains a lattice-index computation asserted without details.

Editorial extensions

If this is right

For every integer $n>1$, any elliptic curve can be embedded as a constant cycle curve of order exactly $n$ into some Kummer surface $\mathrm{Kum}(E\times F)$, so arbitrary orders occur in genus one.
Every positive integer is realized as the order of a smooth genus-one constant cycle curve on a K3 surface, sharpening the previous state where only order one and divisibility bounds were known.
When $4\nmid n$ or the two elliptic curves are non-isogenous, the order is exactly $d(n)$, so the known upper bound $\mathrm{ord}(E_t)\mid n$ is improved by halving it for even $n$.
A concrete $n=4$ example with non-isomorphic isogenous CM curves shows the excluded case is not vacuous: the order can be $1$ or $2$ depending on divisibility conditions on the lattices.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The same intermediate-Jacobian strategy could be applied to other families of constant cycle curves on K3 surfaces, for example the ramification curves of double planes, where only divisibility bounds are currently known.
The parity dichotomy $n/2$ versus $n$ suggests the order may be governed by whether the involution $(-1)$ on the elliptic curve acts trivially on the relevant torsion class; a testable prediction is that any elliptic fibration fibre over a torsion section of order $n$ would satisfy the same dichotomy whenever the fibre has a smooth model on a K3 surface.
One could try to extend the explicit congruence calculation to all $4\mid n$ CM cases and determine exactly when the order drops below $n/2$; the answer is likely a congruence condition on the endomorphism ring and the torsion point.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

You should know two things about this paper. The main theorem is a genuine advance: it computes the exact order of elliptic constant cycle curves on Kummer surfaces, not just the upper bound ord(C) | n from Huybrechts, and it yields the first realization of every integer as an order. The method is also new: this is the first nontrivial use of the transcendental intermediate Jacobian route to pin down orders. The paper is well written and the author is honest about the excluded case, giving an explicit example where the order degenerates.

The soft spot is real. The reduction in Proposition 3.5, which identifies ord(E_t) with ord(kappa'_t), rests on Lemma 3.4. There the paper asserts without proof that two lattice embeddings have index 2^11 and that the cokernel identifies as claimed ('one can check'). That assertion is load-bearing: if either index is off or the cokernel has a hidden kernel, injectivity of pi^*_tr could fail and the lower bound d(n) would not follow. The later computations in Section 4 are conditional on this step. An expert in Kummer lattice theory could likely verify the indices in a few lines, but the paper as written does not include those lines. A referee should ask for them. The congruence computation in Lemma 4.5 is also compressed; it is probably correct but needs checking.

I do not see circularity or invented entities. The cited [CTSS83] is the standard injectivity result, and Huybrechts's lemma is used as a definitional upper bound, not as this author's result. The example in Section 5 is a helpful sanity check rather than a handwave.

Who is this for? People working on Chow groups of K3s and constant cycle curves. It deserves a serious referee: the result is consequential and the strategy is reusable, but the missing lattice verification should be supplied before publication. My own verdict would be conditional accept pending that check. I would bring it to a reading group as a good example of the intermediate Jacobian method, but I would flag Lemma 3.4.

Recommendation: send to peer review. The gap is fillable and localized, and the payoff justifies the referee time.

Referee Report

3 major / 6 minor

Summary. The paper studies the Huybrechts order of elliptic constant cycle curves E_t on Kummer surfaces X = Kum(E_1 × E_2) associated to torsion points t ∈ E_1 of order n > 2. The main theorem (Theorem 1.1) claims ord(E_t) = d(n) := n/2 if 2|n and n otherwise, with a single excluded case (E_1, E_2 non-isomorphic CM isogenous with 4|n). The proof uses a general strategy via the transcendental intermediate Jacobian J^3_tr(X × E_t) (Section 2.2), then reduces to an abelian threefold E_1 × E_2 × E_t: Lemma 3.1 constructs a d(n)-torsion compactification [Z'_t] of the Chow class κ'_t, Lemma 3.4 and Proposition 3.5 aim to show ord(E_t)=ord(κ'_t), and Sections 4.2–4.3 compute ord(κ'_t) using Chow group decompositions and the Abel–Jacobi map. Section 5 gives an example in the excluded case where the order is 1 or 2 depending on numerical conditions. The paper concludes that every positive integer occurs as the order of a smooth genus-one constant cycle curve.

Significance. If the main theorem is correct, this is a significant contribution: it gives the first exact computations of the order of non-rational constant cycle curves on K3 surfaces and shows that all orders can be realized. The strategy of using the transcendental intermediate Jacobian together with CTSS83 injectivity is natural and well executed in principle. The paper also provides an explicit example illustrating the subtle excluded CM case. The main result is concrete and falsifiable, and the corollaries are striking. However, several verification steps (the lattice index assertions in Lemma 3.4 and the congruence computation in Lemma 4.5) are not fully demonstrated, so the current version is not yet acceptable.

major comments (3)

[Section 3.3, Lemma 3.4] The proof of Lemma 3.4 contains two unchecked lattice-theoretic assertions that are load-bearing for Proposition 3.5. It is asserted that the embeddings π^*H^2(X,Z) ⊂ H^2(Bl(A),Z) and π^*(K) ⊂ ⊕ Z[E_i] both have index 2^11, and that these induce a natural isomorphism ⊕Z[E_i]/π^*(K) ≃ H^2(Bl(A),Z)/π^*H^2(X,Z), whence ker(π^*) ≃ (⊕Z[E_i]/π^*(K)) ⊗ H^1(E_t,Z). No computation or reference is supplied for these indices. If either index is incorrect, the identification of ker(π^*) with a subgroup lying in the algebraic part may fail, and the equality ord(E_t)=ord(κ'_t) in Proposition 3.5 would be replaced by an inequality, so the lower bound d(n) in Theorem 1.1 would not follow. Please provide a complete verification (e.g., a Gram-matrix computation for the Kummer lattice K and its pullback) or a precise reference that contains the computation.
[Section 4.3, Lemma 4.5] In the proof of Lemma 4.5, the passage from equation (22) to the displayed system of congruence relations is not shown. The coefficients a_{ij} in the second congruence are never defined, and the reduction modulo 2H^3 requires a nontrivial expansion of the cycle representatives. Since Lemma 4.5 is the key input for Proposition 4.4 (the isogenous case, including the no-CM case), this omission is load-bearing. Please expand the computation in full or provide a lemma that states the congruence system with explicit definitions of all coefficients.
[Remark 2.9] Remark 2.9 claims that the injectivity of the transcendental Abel–Jacobi map on torsion subgroups, proved in Lemma 2.6 for K3 surfaces, holds for any surface S with H^1(S,Z) ≠ 0 by 'the same argument'. The argument in Lemma 2.6 uses that Φ^alg_X is an isomorphism; for a surface with H^1 ≠ 0 (e.g., an abelian surface A), the map Φ^alg_S is only surjective with a nontrivial kernel, so the lifting of a torsion class to a torsion class in CH^1(S)⊗CH^1(C)_hom is not immediate. Since the injectivity of Φ^tr_A and Φ^tr_{Bl(A)} is used in Proposition 3.5, please provide a proof (e.g., using divisibility of the kernel) or a reference.

minor comments (6)

[Corollary 1.2] The statement 'Any elliptic curve E can be embedded ... into any Kummer surface X = Kum(E × F)' is confusing, since the fibres of Kum(E×F) are isomorphic to F, not to E. If the intended meaning is that every elliptic curve appears as a fibre for a suitable choice of the Kummer surface, please rephrase.
[Definition 2.2] The word 'integeral' should be 'integral'.
[Section 5] The word 'repectively' should be 'respectively'.
[Proof of Lemma 3.4] The displayed identity 'D⊗θ = 2D⊗α = π_*(π^*(D))⊗α = π^*_alg(D⊗α)' contains a type error: π^*(D) is not defined for D ∈ NS(Bl(A)); the correct pull-push relation is π^*(π_*(D))⊗α = π^*_alg(π_*D⊗α). As written, the reader cannot follow the argument.
[Proposition 4.3(b)] The expression '4d1([t] - [e1]) × [E1]' should be '4d1([t] - [e1]) × [E2]'.
[Lemma 4.5] The notation [id] is used both for the identity correspondence on E_2×E_2 and for the class in Hom(J(E_1),J(E_2)); please clarify the identification.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: all load-bearing inputs are external results (CTSS83 Abel-Jacobi injectivity, Huybrechts' definitions and bounds, standard lattice facts), and the target order d(n) is never used to define or fit itself.

full rationale

The claimed derivation is not circular. The order ord(E_t) is never used to define, fit, or select any object that enters the computation. In Lemma 3.1, the test class [Z'_t] = 2([t]-[e1]) × [id] has order d(n) because t has order n, which is a direct group-theoretic fact independent of ord(E_t). Proposition 3.5 transfers the computation to the abelian threefold using injectivity of the transcendental Abel-Jacobi maps (Lemma 2.6, from CTSS83) and the lattice comparison in Lemma 3.4; the latter contains an asserted but uncomputed index 2^11, which is a missing verification, not a circular reliance on the conclusion. Sections 4.2-4.3 establish the lower bound by showing that no proper divisor d1 of d(n) makes d1[Z'_t] decomposable, using independent facts about product cycles on products of elliptic curves and Abel-Jacobi images; they do not assume the theorem. Reliance on [Huy14] is for definitions and the a priori bound ord(E_t) | n, and [Huy14] is not by the present author; even if treated as a closely related source, it supplies external definitions and a bound rather than carrying the main claim through self-citation. The paper also openly flags the excluded CM case and the nontriviality of ker(pi*) in Remark 3.2, so it is not presenting a fitted exception as a prediction. The skeptical concern about the 'one can check' lattice-index statements in Lemma 3.4 is a genuine correctness risk, but it is not circularity: a failure there would break the reduction ord(E_t)=ord(kappa'_t), not make the theorem true by construction.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

The paper rests on standard theorems from the algebraic cycles literature rather than new postulates. The only questionable point is the unproved lattice index computation in Lemma 3.4, which is a specific claim within an established theory rather than a new entity.

assumptions (4)

standard math Abel-Jacobi maps on codimension-two torsion classes are injective (Colliot-Thélène, Sansuc, Soulé [CTSS83]).
Invoked repeatedly (Lemma 2.6, Lemma 3.1(b), Lemma 4.5, Proposition 5.1) to identify the order of torsion cycles in Chow groups with their order in intermediate Jacobians.
standard math Decomposition CH1(E1×E2) ≃ CH1(E1) ⊕ CH1(E2) ⊕ Hom(J(E1),J(E2)) (Eq. (9)).
Used in Section 3.1 and Section 4 to separate base-point, curve, and isogeny contributions to one-cycles on products of elliptic curves.
standard math Kummer lattice facts: K ⊂ H^2(X,Z) has rank 16 and discriminant 2^6; the embeddings in Lemma 3.4 have index 2^11.
Used in Lemma 3.4 to identify ker(π^*) and prove injectivity of π^*_tr; the index assertion is stated without a full proof.
standard math Every point of a constant cycle curve represents the Beauville-Voisin class c_X (Voisin [Voi15]).
Used in Section 3.2 to assert [x0] = c_X and to relate the modified diagonal class to the torsion class κ_t.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Elliptic constant cycle curves on Kummer surfaces." pith.science (2026). https://pith.science/paper/Q7SGL5EU

@misc{pith2026250606260,
  author       = {Pith},
  title        = {Pith review of: Elliptic constant cycle curves on Kummer surfaces},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Q7SGL5EU}},
  note         = {Machine review of arXiv:2506.06260}
}

abstract

The order of a constant cycle curve $C \subset X$ on a K3 surface, defined by Huybrechts, is a positive integer that measures the obstruction to decomposing the diagonal class $\Delta_C$ in the Chow group $\mathrm{CH}^2(X \times C)$. In this paper, we compute the order of elliptic constant cycle curves that naturally arise on Kummer surfaces, by passing to the transcendental intermediate Jacobian $J_{\mathrm{tr}}^3(X \times C)$. As a consequence, every $n \in \mathbb{N}$ can be realized as the order of a constant cycle curve on a K3 surface.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

2 extracted references · 2 canonical work pages

[1998]

Curves and cycles on K3 surfaces

[Huy14] Daniel Huybrechts. “Curves and cycles on K3 surfaces”. Algebraic Geometry (Jan. 2014), 69–106. With an appendix by Claire Voisin. [Voi15] Claire Voisin. “Rational equivalence of 0-cycles on K3 surfaces and conjectures of Huybrechts and O’Grady”. Recent Advances in Algebraic Geometry 417 (2015), 422–36

work page 2015
[2010]

Torsion dans le groupe de Chow de codimension deux

[CTSS83] Jean-Louis Colliot-Th´ el` ene, Jean-Jacques Sansuc, and Christophe Soul´ e. “Torsion dans le groupe de Chow de codimension deux”. Duke Mathematical Journal 50.3 (1983), 763–801. [Ful98] William Fulton. Intersection Theory. 2nd ed. Springer Book Archive. New York, NY: Springer-Verlag Berlin Heidelberg,

work page 1983

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [1998]

Curves and cycles on K3 surfaces

[Huy14] Daniel Huybrechts. “Curves and cycles on K3 surfaces”. Algebraic Geometry (Jan. 2014), 69–106. With an appendix by Claire Voisin. [Voi15] Claire Voisin. “Rational equivalence of 0-cycles on K3 surfaces and conjectures of Huybrechts and O’Grady”. Recent Advances in Algebraic Geometry 417 (2015), 422–36

work page 2015

[2] [2010]

Torsion dans le groupe de Chow de codimension deux

[CTSS83] Jean-Louis Colliot-Th´ el` ene, Jean-Jacques Sansuc, and Christophe Soul´ e. “Torsion dans le groupe de Chow de codimension deux”. Duke Mathematical Journal 50.3 (1983), 763–801. [Ful98] William Fulton. Intersection Theory. 2nd ed. Springer Book Archive. New York, NY: Springer-Verlag Berlin Heidelberg,

work page 1983