{"id":"5568a710-278f-4144-bd73-4c3aa2c7d43d","arxiv_id":"2411.16108","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For a subvariety of an abelian scheme with the t-th degeneracy locus removed, the intersection with all flat group subschemes of relative dimension at most t has bounded total height.","lead":"This paper proves a bounded height theorem for intersections of subvarieties of abelian schemes with flat group subschemes, generalizing Habegger's theorem on abelian varieties to families. It also derives a higher-dimensional base version of Silverman's specialization theorem and a bounded height step toward Zhang's ICM conjecture.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Geometric bridge Theorem 5.9 is the load-bearing step; Lemma 5.10's extension claim is under-proven but the argument appears plausible.","rationale":"The paper gives a coherent proof of a substantial family version of Habegger's bounded-height theorem. The arithmetic part—continuity, compactness, height lower and upper bounds, and the Noetherian induction—is carefully executed and internally consistent. The geometric bridge Theorem 5.9 is the natural place for a hidden assumption, and the reader correctly identified it. My independent review confirms that Theorem 5.9 uses Lemma 5.10 to make the lifted kernel subbundle algebraic; without this, the Zariski-closure argument giving vertical defect < g−g' does not work. The proof of Lemma 5.10 is terse: the second statement is justified by citing Prop. 3.6, but the implication from weakly-special closure to equality of endomorphism rings over R is not spelled out. This is a genuine soft spot, but not a demonstrated contradiction. The theorem is likely correct, and the proof of Lemma 5.10 can probably be expanded via standard arguments (the graph of an endomorphism is weakly special, its bi-algebraic closure extends the endomorphism). Therefore the reader's ACCEPT verdict remains justified, though a live review should ask the author to provide a detailed proof of Lemma 5.10. I recommend no change to the verdict.","tokens_in":40106,"tokens_out":48107,"duration_ms":408749,"concrete_test":"Verify Lemma 5.10 in the simplest non-bi-algebraic case: let S be a non-bi-algebraic curve in the moduli space A_g, and take an R-endomorphism f of A = A_g|_S. Compute the bi-algebraic closure of the graph of f inside A^biZar × A^biZar and check that the closure is again the graph of an R-endomorphism of A^biZar. If the closure has a larger base projection or is not a graph, the extension f1 used in Lemma 5.10 does not exist and Theorem 5.9 fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem 8.4's height competition relies on converting geometric nondegeneracy into positivity of self-intersections, which is done by the bridge Theorem 5.9. If the Betti volume form ω^d vanishes on X, Theorem 5.9 concludes X = X^deg(g−g') from a fiber F of the Betti-adapted map, using Lemma 5.10 to ensure the Zariski closure of the lifted fiber has relative dimension at most g−g'. The critical sub-lemma is Lemma 5.10, whose second statement asserts that the kernel subbundle ~W = ker(df) is the restriction onto ~S of an algebraic vector bundle over the bi-algebraic closure S^biZar. Its proof says: 'Prop. 3.6 implies that the endomorphism groups of A and A^biZar coincide. In particular, the R-homomorphism A → B can be extended to an R-homomorphism of abelian schemes f1 : A1 → B1 over S^biZar.' However, Proposition 3.6 is only about bi-algebraic subvarieties being weakly special; the claimed equality End(A) = End(A^biZar) and the extension of a real-homomorphism are not directly established. If this extension fails, ~W is not algebraic (even in the weak sense of having an algebraic closure of the same rank), the estimate rel.dim ~F^Zar ≤ g−g' collapses, and Theorem 8.3's positivity [(f^*~L_B)^d]_X > 0 is unsupported. Since every later height bound in Propositions 7.3, 7.5, and Theorem 7.2 depends on this positivity, Lemma 5.10 is the single most load-bearing step.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves a family version of Habegger's bounded height theorem: for an abelian scheme A/S over a number field K and a subvariety X of A, the intersection of X outside Gao's t-th degeneracy locus with a sufficiently small height neighborhood of the union A_{\\le t} of flat group subschemes of relative dimension at most t is a set of bounded total height. The proof has three main parts: a geometric bridge theorem (Theorem 5.9) using the Ax–Schanuel theorem and Betti forms to convert nondegeneracy into positivity of self-intersections; an arithmetic height-competition argument (Section 7) using Yuan–Zhang intersection theory and a compactness transfer from Habegger's lemma via Masser's specialization theorem; and a Noetherian induction enlarging the Zariski open subset to the complement of the degeneracy locus (Theorem 8.4). The paper gives two applications: a higher-dimensional generalization of Silverman's specialization theorem and a bounded-height result toward Zhang's ICM conjecture.","tokens_in":40410,"tokens_out":15218,"duration_ms":148259,"significance":"If the proof is completed, this is a substantial contribution. It establishes a genuine family version of Habegger's theorem, with applications to specialization of Mordell–Weil groups and to Zhang's conjecture, and it cleanly isolates the required geometric inputs. The paper is honest about its external dependencies: the degeneracy locus is Gao's, the intersection theory is Yuan–Zhang's, and the compactness transfer in Proposition 2.7 relies on Masser's specialization theorem and Habegger's lemma. The overall architecture is coherent: Theorem 7.2 follows from explicit height inequalities, and Theorem 8.4 follows by Noetherian induction. The main weakness is the under-proved Lemma 5.10, which is load-bearing for the geometric bridge and hence for all later height bounds.","major_comments":[{"comment":"The second statement of Lemma 5.10 is not proved. The proof asserts that Proposition 3.6 implies End(A)=End(A^{biZar}) and that an R-homomorphism A→B extends to an R-homomorphism A_1→B_1 over S^{biZar}, but Proposition 3.6 is a characterization of bi-algebraic subvarieties as weakly special subvarieties and does not directly compare endomorphism groups nor provide such an extension. This extension is load-bearing: Theorem 5.9 uses the algebraicity of ~W to conclude rel.dim ~F^{Zar}≤g−g′, and Theorem 8.3 converts that into the positivity [(f^*~L_B)^d]_X>0 used in Propositions 7.3–7.5 and Theorem 7.2. A complete proof of the extension statement, or an alternative argument establishing the needed rank bound on the Zariski closure of the lifted fibers, is required.","section":"§5.2, Lemma 5.10"},{"comment":"The identity A_{\\le t}=∪_{i=1}^r N(B_i) is stated without justification. For a flat group subscheme H of relative dimension ≤t, the argument should specify a surjective homomorphism to a quotient of A that kills H; this requires composing the quotient A→A/H^0 with a sufficiently divisible endomorphism to kill the finite part. The current text, which writes 'the quotient A/NH ... Then H⊆N(B)', does not make clear that the killing homomorphism is the composite [N]∘(A→A/NH). Please spell out this step explicitly, since the main theorem depends on it.","section":"§8.2, proof of Theorem 8.4"}],"minor_comments":[{"comment":"The proof of Proposition 2.7 would be easier to follow if the precise compactness property from Habegger's [27, Lem. 2] were recorded, since the definitions of K_N(A,B) and K_{N_δ}(A,B) in Section 7.1 rely on it.","section":"§2.3, Proposition 2.7"},{"comment":"After applying Corollary 3.9, the notation ~F is reused for the lift of F to C^g×H_g; please state the lifting construction explicitly, since it is not literally the same as the fiber component defined earlier in the paragraph.","section":"§5.2, proof of Theorem 5.9"},{"comment":"The sentence asserting H_μ⊆B_{\\le g−1} after noting H_μ∩B is strict should clarify that this holds for the flat group-subscheme components; possible vertical components over finitely many fibers should be handled explicitly or excluded by a bounded-height observation.","section":"§9.2, proof of Theorem 9.5"},{"comment":"Theorems A and B are restated in Section 9 as Theorem 9.4 and Theorem 9.5; adding explicit cross-references at the statements in the introduction would help the reader.","section":"Introduction, Theorems A and B"}],"recommendation":"major_revision","confidential_remarks":"This is a strong paper that fits the journal's scope. The main theorem is significant and the overall strategy is convincing, but the proof of Lemma 5.10 is a genuine gap in the geometric bridge. I would not reject the paper, but I would not accept it without a complete proof of that lemma and a clarification of the A_{\\le t} decomposition in Theorem 8.4."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Tangli Ge proves a family version of Habegger's bounded height theorem for abelian schemes, and that is a real result. For a subvariety of an abelian scheme, outside Gao's degeneracy locus, intersection with all flat group subschemes of relative dimension at most t inside an epsilon-height neighborhood has bounded total height. That is the right relative statement, and the applications to higher-dimensional Silverman specialization and to Zhang's ICM conjecture are natural and substantive. The novelty is real: previous bounded height results were for a point or curve base. The overall strategy, using Ax–Schanuel as the geometric bridge and Yuan–Zhang adelic intersection theory for the height competition, is coherent and well organized. The heavy external inputs are clearly cited and independent, and there is no circularity or fitting of parameters.\n\nThe main soft spot is Lemma 5.10, which is load-bearing for Theorem 5.9 and hence for the positivity of self-intersections in Theorem 8.3. The second statement claims the kernel subbundle ~W is the restriction of an algebraic vector bundle over the bi-algebraic closure S^biZar. The proof says that Prop. 3.6 implies End(A)=End(A^biZar), but Prop. 3.6 is a classification of bi-algebraic subvarieties and does not directly give that equality. The equality may be true by Hodge-theoretic arguments about weakly special closures, but it is not established in the text. If that extension fails, the whole height competition would not get off the ground. This needs a detailed proof, not a one-sentence citation. Proposition 2.7 is also a bit quick in its use of Masser's specialization theorem, but that looks fixable and is less central.\n\nThe applications in Section 9 check out modulo the main theorem, and the paper is honest about what is new and what is imported. Minor typos in the applications do not affect the mathematics. This is a paper for arithmetic geometers working on unlikely intersections and specialization phenomena. A serious referee should be assigned, with the explicit instruction to scrutinize Lemma 5.10. My verdict: send to peer review; the result is important and likely correct, but the proof of the algebraicity lemma must be expanded.","headline":"Family version of Habegger's bounded height theorem is a genuine advance; the proof is credible, but Lemma 5.10 needs a proper proof of the algebraicity step.","tokens_in":623,"tokens_out":773,"would_cite":true,"duration_ms":144982,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11G50","14G40","14K05","11G10"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that in every family of abelian varieties, algebraic points of a non-anomalous subvariety that lie close to small flat group subschemes have bounded height.","keywords":["abelian schemes","bounded height","degeneracy locus","weakly special subvarieties","adelic line bundles","Ax–Schanuel","unlikely intersections","specialization of Mordell–Weil groups"],"falsifier":"Find an abelian scheme over a curve, a subvariety $X$ with $\\tau(X) \\ge 0$, and a sequence of algebraic points $p_n \\in (X \\setminus X^{\\mathrm{deg}(t)}) \\cap C(1/n, A_{\\le t})$ with total height $h_M(p_n) \\to \\infty$; such a sequence would directly contradict Theorem 8.4. Equivalently, computing $[(f^*\\widetilde{L}_B)^d]_X$ for a single surjective R-homomorphism $f$ of relative dimension $\\le t$ on a $t$-nondegenerate $X$ and finding zero would falsify the positivity criterion behind the proof.","tokens_in":39860,"feed_emoji":"📐","tokens_out":11069,"duration_ms":93288,"temperature":0.7,"pith_summary":"This paper proves a family version of the bounded-height theorem for subvarieties of abelian varieties. Working with an abelian scheme over a normal quasi-projective base, the author shows that after removing the t-th degeneracy locus, every algebraic point lying within a sufficiently small height neighborhood of a flat group subscheme of relative dimension at most t has bounded total height. The result recovers the known single-abelian-variety theorem when the base is a point, and it yields two Diophantine applications: a specialization theorem for Mordell–Weil groups over higher-dimensional bases, and a bounded-height result for a determinant height that points toward an ICM conjecture on small specializations. A sympathetic reader should care because it turns a qualitative rarity statement into a quantitative height bound in a relative, non-compact setting.","feed_headline":"Small flat subgroups meet non-anomalous points in bounded height","feed_subtitle":"Family version controls algebraic points near flat subgroup schemes outside the anomalous locus.","key_machinery":"The engine is the $t$-th degeneracy locus $X^{\\mathrm{deg}(t)}$, defined as the union of all positive-dimensional subvarieties $Y \\subseteq X$ whose vertical defect $\\delta_v(Y) = \\operatorname{rel.dim}\\langle Y \\rangle - \\dim Y$ is less than $t$, where $\\langle Y \\rangle$ is the smallest weakly special subvariety (a translate of a group subscheme by a constant section) containing $Y$. The argument has three gears: a nondegeneracy criterion, proved with a weak Ax–Schanuel inequality, showing that vanishing of the pullback volume form along $X$ forces $X \\subseteq X^{\\mathrm{deg}(g-g')}$; a compactness statement for surjective R-homomorphisms, reducing all flat group subschemes to finitely many isogeny types; and intersection theory of adelic line bundles on quasi-projective varieties, which converts positivity of self-intersections $[(f^*\\widetilde{L}_B)^d]_X$ into a uniform height upper bound that competes with a lower bound for points near $B$-null loci. Choosing $\\epsilon$ small makes the lower bound beat the upper bound and confines the total height to a bounded interval.","core_discovery":"Let $\\pi: A \\to S$ be an abelian scheme over a number field with a fibre-wise Néron–Tate height and a total height $h$. For $t \\in \\mathbb{N}$, denote by $A_{\\le t}$ the union of all flat group subschemes of relative dimension at most $t$ and by $X^{\\mathrm{deg}(t)}$ the $t$-th degeneracy locus of a subvariety $X \\subseteq A$. The central claim (Theorem 8.4) is that there are constants $\\epsilon, c > 0$ such that every algebraic point of $X \\setminus X^{\\mathrm{deg}(t)}$ lying in the $\\epsilon$-height neighborhood $C(\\epsilon, A_{\\le t})$ satisfies $h(p) \\le c$. Equivalently, on the non-anomalous part of $X$, points that are very close to small group subschemes cannot escape to infinite height. The proof shows that this bounded-height statement follows from positivity of self-intersections of pulled-back adelic line bundles, which in turn follows from a nondegeneracy criterion: if a pulled-back invariant volume form vanishes on $X$, then $X$ is contained in the corresponding degeneracy locus.","pith_inferences":["The paper only handles flat group subschemes that come from the generic fibre, namely the union $A_{\\le t}$; if the same height bound held for the larger union $A(\\le t)$ of group subschemes inside all fibres, Conjecture 9.6 would be true, and the natural missing ingredient is a relative compactness statement for fibre-wise homomorphisms.","One testable consequence of Theorem 9.5 is that for a non-isotrivial elliptic surface over $\\mathbb{P}^1$ with a finitely generated group of sections, the small-value set $\\{s : h_\\Lambda(s) < \\epsilon\\}$ should have bounded height; a computer search for explicit $\\epsilon$ and height bounds in a concrete family would provide numerical evidence.","The nondegeneracy criterion turns a transcendental analytic condition into an algebraic one; this suggests that the degeneracy locus, and hence the height bound, could in principle be computed for explicit families of abelian schemes."],"forward_implications":["When $S$ is a point, Theorem 8.4 reduces to the original bounded-height theorem for a single abelian variety, so the relative statement is a genuine generalization rather than an analogue.","Under the maximal-variation and dimension assumptions, the set of closed points where specialization fails to be injective on a finitely generated Mordell–Weil group is contained in a strict Zariski closed set together with a set of bounded height (Theorem 9.4).","For a non-constant abelian scheme over a curve, the set of points where the determinant height $h_\\Lambda(s)$ is smaller than some $\\epsilon > 0$ is a set of bounded height; in particular, only finitely many bounded-degree points can have $h_\\Lambda(s) < \\epsilon$ (Theorem 9.5).","The proof establishes a practical criterion: $t$-nondegeneracy of $X$ is equivalent to positivity of the self-intersection $[(f^*\\widetilde{L}_B)^d]_X$ for every surjective R-homomorphism $f: A \\to B$ with $\\dim B \\ge \\dim A - t$ (Theorem 8.3 and Corollary 6.14)."],"supporting_citations":[{"why":"Establishes the bounded-height theorem for subvarieties of a single abelian variety, the statement this paper lifts to families.","marker":"[27]"},{"why":"Introduces the $t$-th degeneracy locus and its algebraicity, which the paper adapts to general abelian schemes.","marker":"[19]"},{"why":"Provides the weak Ax–Schanuel inequality for the universal abelian variety used in the nondegeneracy criterion.","marker":"[20]"},{"why":"Develops intersection theory of adelic line bundles on quasi-projective varieties, the toolbox for the height estimates.","marker":"[50]"},{"why":"Proves the curve-base specialization theorem that the paper generalizes to higher-dimensional bases.","marker":"[47]"},{"why":"Poses the small-determinant-height conjecture whose non-constant curve case the paper solves in bounded-height form.","marker":"[51]"},{"why":"Gives finite generation of Mordell–Weil groups, used in the specialization applications.","marker":"[33]"}],"fun_headline_variants":["Points close to small group subschemes stay bounded in height","Flat subgroup neighbors have bounded height outside degeneracy","Family Habegger theorem: height bounded near flat subgroups","Proximity to flat subgroups controls height on non-anomalous part"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is the geometric fact that a subvariety on which a pulled-back invariant volume form vanishes must already lie in the corresponding degeneracy locus; if that bridge failed, the positivity of self-intersections that drives the height comparison would not get started.","fun_headline_variants_meta":{"raw":{"variants":["Points close to small group subschemes stay bounded in height","Flat subgroup neighbors have bounded height outside degeneracy","Family Habegger theorem: height bounded near flat subgroups","Proximity to flat subgroups controls height on non-anomalous part"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00093,"raw_usage":{"total_tokens":3959,"prompt_tokens":898,"completion_tokens":3061,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":514,"completion_tokens_details":{"reasoning_tokens":2992}},"tokens_in":514,"tokens_out":3061,"duration_ms":18276,"temperature":1.0,"reasoning_tokens":2992,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T13:34:38.045449+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Find an abelian scheme over a curve, a subvariety $X$ with $\\tau(X) \\ge 0$, and a sequence of algebraic points $p_n \\in (X \\setminus X^{\\mathrm{deg}(t)}) \\cap C(1/n, A_{\\le t})$ with total height $h_M(p_n) \\to \\infty$; such a sequence would directly contradict Theorem 8.4. Equivalently, computing $[(f^*\\widetilde{L}_B)^d]_X$ for a single surjective R-homomorphism $f$ of relative dimension $\\le t$ on a $t$-nondegenerate $X$ and finding zero would falsify the positivity criterion behind the proof.","supporting_citations":[{"cited_title":"Habegger","cited_arxiv_id":null,"evidence_quote":"Establishes the bounded-height theorem for subvarieties of a single abelian variety, the statement this paper lifts to families."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces the $t$-th degeneracy locus and its algebraicity, which the paper adapts to general abelian schemes."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the weak Ax–Schanuel inequality for the universal abelian variety used in the nondegeneracy criterion."},{"cited_title":"Yuan and S.-W","cited_arxiv_id":null,"evidence_quote":"Develops intersection theory of adelic line bundles on quasi-projective varieties, the toolbox for the height estimates."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Proves the curve-base specialization theorem that the paper generalizes to higher-dimensional bases."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Poses the small-determinant-height conjecture whose non-constant curve case the paper solves in bounded-height form."},{"cited_title":"Lang and A","cited_arxiv_id":null,"evidence_quote":"Gives finite generation of Mordell–Weil groups, used in the specialization applications."}],"review_version":1}