{"id":"02758adf-a8e9-4c75-b729-6b9ce7289e8a","arxiv_id":"2602.22162","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The pure weight-2 extension of the universal theta divisor is the Zariski closure twisted by div θinv, where θinv is the difference of the tropical and smooth tropical Riemann theta functions; Moret-Bailly's key formula and a universal Néron–Tate height formula follow.","lead":"A theta divisor over the moduli space of abelian varieties extends to the boundary either by Zariski closure or by a canonical 'pure weight 2' extension. The authors prove the two differ by a tropicalisation of the Riemann theta function, extend Moret-Bailly's key formula to the boundary, and derive a universal formula for Néron–Tate heights of points.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Boundary key formula rests on quoted Berkovich results [JS22, Cor 10.7], [JS24, Cor 5.7/5.8, Prop 5.9] whose hypotheses are unchecked at char-2/non-normal and root-stack strata; if they fail, Thm 8.10 and hence Thm 1.3 lose support. Cor 8.21 also has a sign/factor-2 slip.","rationale":"Reading in good faith: the paper's central claim is the boundary key formula Cor 8.20, and I traced its proof chain. The internal cocycle computations (Lemmas 5.3, 6.4, Prop 6.6, Thm 5.4) are checkable and correct; the derivation from Cor 7.15 and Thm 8.10 to Cor 8.20 is coherent; the root-stack construction, while terse, is plausible (n = 1 for g ≤ 4). I found no circularity and no tuning: θinv is explicitly constructed from the tropical theta function. The single load-bearing risk is exactly what the reader identified: the boundary vanishing-order comparisons in Thm 8.9 depend on [JS22, Cor 10.7] and [JS24, Cor 5.7/5.8, Prop 5.9], and the paper does not state or check the hypotheses of these quoted results in the two regimes it needs them most (char 2, where it admits non-normality; and the g ≥ 5 root-stack case). Since [JS24] is an arXiv preprint by a co-author and [JS22] is published, the former component is not fully externally validated. I also independently confirmed the reader's finding that (81) disagrees with the formal consequence of Cor 8.20: the correct purity relation has coefficient −(n²−1)/2 and the opposite sign. This is a genuine slip, but it lives in a corollary; the isomorphism in Cor 8.20 is unaffected, so it does not by itself move the verdict. Conditional acceptance with a request to fix (81) and to add a hypothesis check for the Berkovich inputs remains the right verdict.","tokens_in":50723,"tokens_out":32222,"duration_ms":294059,"concrete_test":"Compute the characteristic-2 boundary case explicitly: g = 2, R = F₂[[t]], a totally degenerate principally polarised abelian scheme whose special fibre is a product of two Tate curves. Build the tautological model P as the fibre square (71) using the explicit mDV fan of §6, and pull back the 8th-power theta section s (the section f of (77)). Directly compute (i) the order of vanishing of s along each irreducible component of the special fibre of P, and (ii) c = ord_{v,f*L_A}(s) = −log sup_{Sk(A)}∥s∥_L (Prop 8.4). If the orders in (i) are not all equal — as can happen when the special fibre is non-reduced in characteristic 2 — Prop 8.7(ii) fails and Thm 8.10/Thm 1.3 must be restricted to Z[1/2]. (The Cor 8.21 slip is independently settled by re-deriving it from Cor 8.20.)","verdict_should_be":"UNCHANGED","load_bearing_attack":"Cor 8.20 (Thm 1.3) rests on Thm 8.10, 'div f = 8 Θ̄', which is proven via Thm 8.9 and Lemmas 8.12–8.16. The narrowest external premises are Prop 8.4 ([JS22, Cor 10.7]) and Prop 8.7 ([JS24, Cor 5.7, Prop 5.9]): (i) the sup of the canonical norm over the skeleton equals the special-fibre vanishing order of the section in the Moret-Bailly model, and (ii) the vanishing order of s is constant along the irreducible components of the special fibre of the tautological (Alexeev–Nakamura) model. The paper uses these in two settings not visibly covered by the quoted hypotheses: residue characteristic 2 — where §1.7/Rem 8.19 say N^{mDV}_{g,1} is non-normal and the Cartier-to-Weil map fails, so special fibres of the AN model need not be reduced — and the integral/saturated root-stack base change of §8.4 (needed for flatness when g ≥ 5), whose flatness Lemma 8.12 is a three-line proof citing [Mol21, Thm 2.1.4]. If constancy (Prop 8.7(ii)) fails, the equality c = ord_{v,f*L_A}(s) in Thm 8.9 fails, taking down Lemma 8.16, Thm 8.10, and the boundary part of Thm 1.3. Additionally, Cor 8.21 is demonstrably inconsistent with Cor 8.20: pulling back 8c1(L) = 8[Θpure] − 4π*c1(ω) along [n] and using [n]*L = L^{n²} yields [n]*Θpure = n²Θpure − (n²−1)/2 π*c1(ω), not the printed +(n²−1)π*c1(ω). This is a real slip in a headline-adjacent purity statement; it does not by itself sink Cor 8.20, since the difference is a base pullback either way.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the extension of the universal theta divisor over compactified moduli spaces of principally polarised abelian varieties. It defines a 'pure weight 2' extension Θpure = Θ + div θinv, where θinv is a difference of a piecewise-linear tropical theta function and a smooth tropical function, and proves that this extension satisfies the purity relation up to base pullbacks. The main theorem (Thm 1.3 / Cor 8.20) is an extension of Moret-Bailly's key formula to the boundary: (Lpure)^⊗8 ≅ O(8Θpure) ⊗ π*ω^⊗−4 as adelic/b-line bundles on the mixed Delaunay–Voronoi compactification. The paper also derives arithmetic consequences, including a universal formula for Néron–Tate heights and a description of the degeneration of the Riemann theta function.","tokens_in":51221,"tokens_out":12006,"duration_ms":99250,"significance":"If the main results are correct, they constitute a substantial conceptual advance: they give a precise tropical correction that distinguishes the Zariski closure from the pure weight-2 extension, extend Moret-Bailly's key formula across the boundary, and establish a clean comparison with Yuan–Zhang's invariant adelic line bundles. A particular strength of the paper is its explicit, checkable machinery: the functions θpl, θeq, and θinv are defined by closed formulas with no fitted parameters, and the key cocycle computations (Lemmas 5.3, 6.4, Prop 6.6) are internally coherent and were spot-checked by this referee. The logical chain from the Berkovich input (Props 8.4 and 8.7) through Thm 8.10 to Cor 8.20 is traceable, and the paper is careful to distinguish the different languages of log b-divisors, adelic divisors, and log adelic line bundles.","major_comments":[{"comment":"The proof of the boundary key formula relies on Prop 8.4 ([JS22, Cor 10.7]) and Prop 8.7 ([JS24, Cor 5.7, Prop 5.9]) for the constancy of orders of vanishing along irreducible components of the special fiber of the tautological (Alexeev–Nakamura) model. These results are used in two settings not visibly covered by the quoted hypotheses: residue characteristic 2 (where N^{mDV}_{g,1} is non-normal and the Cartier-to-Weil map fails) and after the integral/saturated root-stack base change of §8.4. If Prop 8.7(ii) fails, the equality c = ord_{v,f*L_A}(s) in Thm 8.9 fails, and with it Lemma 8.16, Thm 8.10, and the boundary part of Thm 1.3. The authors should state the precise hypotheses of [JS24] and explicitly verify them in these settings, or supply a proof of the needed generality.","section":"§8.4 (Thm 8.9–8.10, Props 8.4, 8.7)"},{"comment":"The displayed formula [n]*Θpure ∼ n²Θpure + (n²−1)π*c1(ω) is inconsistent with Cor 8.20. Pulling back the isomorphism L^⊗8 = O(8Θpure) ⊗ π*ω^⊗−4 along [n] and comparing with the n²-th tensor power of the same isomorphism yields [n]*Θpure ∼ n²Θpure − (n²−1)/2 π*c1(ω) (up to sign convention for c1). Both the sign and the factor are incorrect as printed. This does not by itself invalidate Cor 8.20, but the purity statement in Cor 8.21 is false as stated and must be corrected.","section":"§8.5, Cor 8.21"},{"comment":"The representability theorems for N^Σ_g and N^{˜Σ}_{g,1} are justified by one-paragraph sketches ('local models are given by the spectra of the monoid rings'). Since these stacks underlie all subsequent constructions (log b-line bundles, tropical theta functions, the key formula), a rigorous proof or a precise reference is needed. As written, this is a gap in the foundational layer of the paper.","section":"§4.3, Thm 4.17 and Thm 4.29"}],"minor_comments":[{"comment":"The line 'By proposition 8.5 we have − log ∥s(x)∥L = − log ∥s(x)∥L + θinv_ρ(a, val(x))' appears to have a typo: the first norm on the right should be ∥s(x)∥_{LP}. The subsequent 'In particular' line should then read − log ∥s(x)∥L = c + θinv_ρ(a, x).","section":"§8.3, proof of Thm 8.9"},{"comment":"The statement that 'all our main theorems listed above remain valid over Z' is asserted, but the proof of Thm 8.10 in characteristic 2 is contingent on the external results discussed in the first major comment. Please reconcile the assertion with the actual hypotheses verified.","section":"§1.7 and Remark 8.19"},{"comment":"In the definition of a log b-line bundle, 'a conical section of L' should specify that c is a section of L^{trop} ⊗_{sPL} CCon, matching the notation used later in the paper.","section":"§2.9, Def 2.28"},{"comment":"The notation π*ω^{⊗−4}_{N_{g,1}/N_g} is ambiguous because ω is used both for the Hodge bundle on N_g and for its pullback. Please clarify the base of the Hodge bundle in the displayed formula.","section":"§8.5, Cor 8.20"},{"comment":"The paper depends heavily on the unpublished preprint [JS24] and the book [YZ26]. Please ensure the exact statements quoted are available in the cited versions and add version/date information.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The central idea is strong and the explicit computations are convincing. The main risk is the unresolved applicability of the quoted Berkovich results in characteristic 2 and after root-stack base change; if the authors can confirm that [JS24] covers these cases, the paper is likely acceptable after a revision. The Corollary 8.21 sign error is easily fixed but must be corrected."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First thing to know: if you work on adelic divisors over moduli of abelian varieties, this is the paper that actually writes down the divisor-side extension. The main new object is Θpure = Θ̄ + div θinv, with θinv = θpl − θeq from a tropical theta function. That is not a reformulation; the construction is explicit, has no fitted parameters, and the cocycle computations check out. The paper also proves a boundary version of Moret-Bailly's key formula and a universal Néron–Tate height formula, pulling together log abelian varieties, toroidal b-divisors, and Yuan–Zhang's invariant extension. I spot-checked Lemma 5.3, Prop 6.6, and the chain 8.10 → 8.18 → 8.20 → 9.6; it is logically traceable, and the abstract matches the theorems.\n\nThe soft spots are real but localized. Cor 8.21 prints [n]*Θpure = n²Θpure + (n²−1)π*c1(ω). Pulling back Cor 8.20 along [n] and using [n]*L = L^{n²} gives instead a minus sign and half the coefficient, up to base pullback. That is a correction or a normalization issue in a headline-adjacent statement, not a collapse of the main comparison; Cor 8.20 itself still stands if the divisor computation holds.\n\nMore serious is what Theorem 8.10 rests on. The equality div f = 8Θ̄ uses Proposition 8.7 on constancy of vanishing orders, quoted from JS24, and Proposition 8.4 from JS22. The paper applies these in two settings not explicitly covered by the quoted hypotheses: residue characteristic 2, where the space is non-normal and the Cartier-to-Weil map fails, and the saturated root-stack base change of §8.4, whose flatness gets a three-line proof. If constancy in Prop 8.7(ii) fails there, Lemma 8.16 and Theorem 8.10 lose support, and the boundary part of Theorem 1.3 with it. I don't see circularity or parameter-fitting; the risk is external and needs the authors to either tighten hypotheses or prove the constancy directly.\n\nWho should read it: anyone working on heights, theta divisors, or log compactifications of Ag; the height formula in Thm 9.6 is likely to get used. It deserves a serious referee. I would send it out with a request to fix (81) and to expand §8.4/§8.7 verification or at least state exactly which assumptions from JS22/JS24 are being imported.","headline":"The pure weight-2 theta divisor is explicitly Θ̄ + div θinv and the boundary key formula is mostly solid; Cor 8.21 has a real sign/factor slip, and the char-2/root-stack reliance on quoted Berkovich results is the main thing to press.","tokens_in":51898,"tokens_out":2074,"would_cite":true,"duration_ms":20789,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14K25","14G40","14T90"],"pacs":[],"model":"deepseek-v4-flash","headline":"The two natural compactifications of a universal theta divisor differ exactly by the tropical Riemann theta function, and Moret-Bailly's key formula survives on the boundary.","keywords":["theta divisor","pure weight 2","adelic line bundles","tropical theta function","log abelian varieties","Moret-Bailly key formula","Delaunay–Voronoi compactification","Néron–Tate height"],"falsifier":"Construct an explicit degenerating family of principally polarized abelian varieties over a DVR (e.g., a genus-2 example) where the Alexeev–Nakamura model has a reducible special fiber, and compute the vanishing order of the canonical theta section along each component. If the order differs from the supremum of the canonical metric on the skeleton (as asserted by the quoted results), the key formula would fail. Similarly, in characteristic 2, check whether the Cartier divisor 8Θ defined in §8.4 equals the Zariski closure when pulled back to a smooth stratum; a counterexample would invalidate T","tokens_in":50448,"feed_emoji":"📐","tokens_out":1507,"duration_ms":16446,"temperature":0.7,"pith_summary":"This paper resolves a subtle ambiguity in extending the universal theta divisor to a compactification of the moduli space of principally polarized abelian varieties. The authors show that the two natural choices—the Zariski closure and the unique extension that is 'pure of weight 2'—differ precisely by a tropicalization of the Riemann theta function, a piecewise-linear correction term. They prove that Moret-Bailly's key formula, which relates the eighth power of the polarizing line bundle to the theta divisor and the Hodge bundle, extends as an exact isomorphism of adelic (or b-) line bundles over the entire compactification. If this holds, it gives a uniform, boundary-controlled description of the Néron–Tate height of a point, linking arithmetic heights to tropical theta functions and Alexeev–Nakamura models.","feed_headline":"Tropical theta function pins down theta divisor extension","feed_subtitle":"Two natural compactifications differ by a tropical correction, and Moret-Bailly's key formula survives at the boundary, yielding a universal","key_machinery":"The central object is the invariant tropical theta function θinv = θpl − θeq, the difference between a piecewise-linear theta function (a tropicalization of the classical Riemann theta series) and an 'equilibrium' smooth variant defined by interpolation. This difference descends to a continuous conical function on the tropicalization of the compactified moduli space, and it is used to twist the Zariski closure of the theta divisor to obtain a pure weight-2 extension. The proof relies on the theory of logarithmic abelian varieties (Kajiwara–Kato–Nakayama), log b-line bundles, and a detailed comparison of vanishing orders of sections via Berkovich skeleta and Alexeev–Nakamura models.","core_discovery":"On the mixed Delaunay–Voronoi compactification of the universal abelian variety, the pure weight-2 extension of the universal theta divisor (the Zariski closure twisted by the invariant tropical theta function, Θpure = Θ̄ + div θinv) is exactly the extension that makes the purity relation [n]*Θ ∼ n²Θ hold up to pullbacks from the base. The key formula becomes (Lpure)⊗8 = O(8Θpure) ⊗ π*ω⊗−4 as adelic (equivalently log-b) line bundles, extending Moret-Bailly's classical isomorphism from the interior to the whole compactification. This identifies Yuan–Zhang's invariant adelic extension of the polarizing line bundle with the purely combinatorially defined line bundle Lθpl(θinv), and yields a uni","pith_inferences":["The identity Θpure = Θ̄ + div θinv likely governs the boundary asymptotics of not just Θ but of the whole family of theta line bundles with characteristics, suggesting a tropical correction formula for all symmetric divisors representing the principal polarization.","Since θinv is a tropical object, the result hints that arithmetic heights of points on degenerating abelian varieties can be approximated by evaluating tropical theta functions on skeleta—a theme that could extend to higher-rank degenerations and to the Zhang–Kawazumi φ-invariants.","The passage through log b-line bundles is a template for comparing other natural extensions (e.g., canonical vs. Zariski closure of cycles) on toroidal compactifications, potentially giving a general 'purity correction' mechanism for other moduli problems.","A testable consequence of the universal height formula is that the local non-archimedean contribution to Néron–Tate height, as a function of the base point, is piecewise-linear with slopes given by the tropical theta function; this could be checked in explicit one-parameter families of elliptic curves."],"forward_implications":["The pure weight-2 extension of the theta divisor is now an explicit, computable object on the compactification, with its boundary behavior controlled by the tropical theta function.","Moret-Bailly's key formula holds over the entire compactification, giving an isomorphism of adelic line bundles and thus a boundary-compatible link between the polarizing bundle, theta divisor, and Hodge bundle.","The universal Néron–Tate height formula makes the height of a point on a principally polarized abelian variety computable from local intersection multiplicities, an invariant tropical theta function, and the classical normalized theta function at archimedean places.","The purity relation [n]*Θpure ∼ n²Θpure + (n²−1)π*c1(ω) holds as a b-divisor identity, showing that pure weight 2 extends naturally from abelian varieties to the universal family.","Over Z, the results remain valid even in characteristic 2, where the moduli stack is not normal, by defining the Cartier divisor 8Θ via a suitable rational section rather than via Zariski closure."],"fun_headline_variants":["Tropical theta function pins down pure divisor extension","Pure theta divisor extension revealed by tropical correction","Key formula survives compactification via pure theta divisor","Tropicalization of Riemann theta explains divisor extension"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The boundary key formula and the purity relation rely on a chain of previously established results about vanishing orders of sections of line bundles on Alexeev–Nakamura models (quotations of [JS22] and [JS24]), particularly the equality of sup norms on Berkovich skeleta with vanishing orders along the special fiber; if any of those results fail in the full generality needed here—e.g., on non-normal strata in characteristic 2 or after the root-stack base change—then the equal","fun_headline_variants_meta":{"raw":{"variants":["Tropical theta function pins down pure divisor extension","Pure theta divisor extension revealed by tropical correction","Key formula survives compactification via pure theta divisor","Tropicalization of Riemann theta explains divisor extension"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000704,"raw_usage":{"total_tokens":2993,"prompt_tokens":705,"completion_tokens":2288,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":449,"completion_tokens_details":{"reasoning_tokens":2239}},"tokens_in":449,"tokens_out":2288,"duration_ms":14827,"temperature":1.0,"reasoning_tokens":2239,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-02T20:50:10.764986+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Construct an explicit degenerating family of principally polarized abelian varieties over a DVR (e.g., a genus-2 example) where the Alexeev–Nakamura model has a reducible special fiber, and compute the vanishing order of the canonical theta section along each component. If the order differs from the supremum of the canonical metric on the skeleton (as asserted by the quoted results), the key formula would fail. Similarly, in characteristic 2, check whether the Cartier divisor 8Θ defined in §8.4 equals the Zariski closure when pulled back to a smooth stratum; a counterexample would invalidate T","supporting_citations":[],"review_version":1}