{"id":"266b5ce1-7170-4478-893b-7fe21a628d0e","arxiv_id":"2412.10340","paper_version":4,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For every non-CM elliptic curve over Q, the index of the adelic Galois image is bounded by 10^21(h_F(E)+40)^4.42, and by h_F(E)^{3+o(1)} as the height grows.","lead":"This paper proves explicit upper bounds, in terms of the Faltings height, on how far the Galois representation attached to a non-CM elliptic curve over the rationals is from being surjective. It improves prior ineffective bounds by Zywina and Lombardo, giving polynomial bounds with explicit constants and a near-optimal asymptotic exponent.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 5.9's displayed bound is numerically false at h_F = -0.75 (approximately 4.64 versus 4.49), so the constants 1454, 1266.4, 21000, and the 10^21 bound are not established as written.","rationale":"The reader's weakest assumption correctly identifies Lemma 5.9 as the point where the explicit constants of the main theorem are least secure. My independent check confirms that the displayed chain in the proof is numerically invalid at the boundary h_F = -0.75, which is inside the claimed range since h_F > -0.75. This is not a vague worry about rigor: the exact constants 1454, 1266.4, 21000, and 14400 propagate through Theorem 5.1 and Theorem 5.17 into the headline bound of Theorem 1.8. The concern is genuinely load-bearing because the central claim includes explicit constants; if Lemma 5.9 requires a larger constant, the final 10^21 coefficient and 4.42 exponent may fail, even if the qualitative polynomial bound survives. I do not see an internal contradiction that would force rejection: the lemma statement may still be true, and a sharper analytical treatment of the implicit inequality could repair the proof. That is exactly why the conditional verdict, pending a corrected verification of Lemma 5.9, remains appropriate. The secondary issues noted by the reader, such as unshipped MAGMA and FindOpenImage computations, are real reproducibility concerns but are not the single most load-bearing point.","tokens_in":59243,"tokens_out":14541,"duration_ms":129596,"concrete_test":"For each h in [-0.75, 0], let T_*(h) be the largest solution of pi T = 3 log T + 6h + 8.66, and compare T_*(h) with 2.29h + 6.21. If T_*(h) exceeds the target at any h, Lemma 5.9 is false and the constants in Theorems 5.1, 5.17, and 7.1 must be recomputed; if not, the lemma survives but the written proof needs a corrected numerical step. Then propagate the corrected constants through Theorem 5.17 to check whether the final bound 10^21 (h_F + 40)^4.42 still holds.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing weakness is in Section 5, Lemma 5.9. For T = (1/[K:Q]) sum_sigma 1/rho(E_sigma,L_sigma)^2, the proof invokes [24, Proposition 3.2] and [57, Lemma B.1] to obtain T < (3/pi) log((12/pi) h_F + 5.52 + 4e^2) + (6/pi) h_F + 2.76, and then asserts this is < 2.29 h_F + 6.21, citing x > 32.2 and log(x)/x <= log(32.2)/32.2. At h_F = -0.75 the log argument is 32.2114, so the displayed upper bound evaluates to about 4.643, whereas 2.29(-0.75) + 6.21 = 4.4925. The quoted monotonicity inequality does not close this gap; in fact the claimed linear bound is not implied by the preceding chain. This lemma is used in inequalities (5.10) and (5.12) to control the average inverse period, and it feeds directly into the constants 1454 and 1266.4 in Theorem 5.1, the constants 21000 and 14400 in Theorem 5.17, and ultimately the 10^21 (h_F + 40)^4.42 bound in Theorem 1.8(1). A correct proof may exist by solving the implicit inequality more sharply, but the proof as printed does not establish the lemma.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves fully explicit, height-dependent upper bounds for the index of the adelic Galois image of a non-CM elliptic curve over Q. The main theorem (Theorem 1.8) gives [GL2(Ẑ):Im ρ_E] < 10^21 (h_F(E)+40)^4.42, an asymptotic bound h_F(E)^{3+o(1)}, and a conductor bound, improving previous work of Zywina and Lombardo. The proof combines a structural classification of p-adic images in the non-split Cartan case (Sections 3 and 6), an effective surjectivity theorem via the Gaudron–Rémond period/isogeny estimates (Section 5), and new entanglement estimates among division fields (Sections 7.1–7.2). The paper also classifies possible images of ρ_{E,p^n} when Im ρ_{E,p} lies in the normaliser of a non-split Cartan subgroup (Theorem 1.9).","tokens_in":59590,"tokens_out":14474,"duration_ms":92281,"significance":"If the proof is completed as intended, this is a substantial contribution: it supplies the first fully explicit polynomial bound in the Faltings height for the adelic index over Q, with the polynomial exponent 4.42 and the asymptotic exponent 3+o(1) going well beyond Lombardo's effective but enormous bounds. The use of prior external results (Gaudron–Rémond, Mazur, RSZB, Le Fourn–Lemos, and the author's earlier work with Lombardo) is honest and non-circular, and the classification in the non-split Cartan case is a useful complement to the RSZB database. The paper is carefully structured and many of the estimates are explicit rather than merely existential.","major_comments":[{"comment":"The proof of Lemma 5.9 does not establish the displayed inequality. After applying [57, Lemma B.1], the proof claims (3/π) log((12/π)h_F + 5.52 + 4e^2) + (6/π)h_F + 2.76 < 2.29 h_F + 6.21, citing x > 32.2 and log x / x ≤ log(32.2)/32.2 for x ≥ 32.2. At h_F = -0.75 the left side evaluates to about 4.643, whereas 2.29(-0.75)+6.21 = 4.4925, so the asserted inequality is false; the cited monotonicity bound does not close this gap. This lemma is load-bearing: it is used in (5.10) and (5.12) to bound the average inverse period and hence feeds into the constants 1454 and 1266.4 in Theorem 5.1, the constants 21000 and 14400 in Theorem 5.17, and ultimately the 10^21 (h_F+40)^4.42 bound in Theorem 1.8(1). The lemma may be salvageable by solving the implicit inequality πT ≤ 3 log T + 6h_F + 8.66 more sharply than the printed chain does, but the proof as written must be repaired and all resulting constants re-verified before the main theorem can be accepted as stated.","section":"Section 5, Lemma 5.9"}],"minor_comments":[{"comment":"Several phrases are repeated or have minor typos, e.g. 'attache d' in the title line, 'the same p roblems' in the introduction, and 'we also give an improved and effective version' missing the object 'of Zywina's bound'; these do not affect the mathematics but should be cleaned up.","section":"Throughout"},{"comment":"The paper relies on the MAGMA function FindOpenImage and on MAGMA computations for finitely many cases (e.g. the four j-invariants in (2.1), the list (7.2), and the RSZB labels). It would help reproducibility if the relevant scripts or explicit outputs were included in an ancillary file; as written the reader cannot independently verify these finite computations without reimplementing them.","section":"Lemma 2.8 and Proposition 7.10"},{"comment":"In the proof of Theorem 5.14, the sentence 'using d h(j)> 2 > e/(1.6n)' is not immediately clear: the last inequality seems to require n > 1, but the case n = 1 is also covered by the surrounding argument; please clarify the range of n in this step.","section":"Section 5.2, Theorem 5.14"},{"comment":"The second assertion of Theorem 7.1 is stated with a function δ(x) whose denominator can be small for x near -0.75; the later proof restricts to h_F > 4·10^15 for that branch, but the statement as written may suggest the bound holds uniformly. Please state the ranges in the displayed asymptotic claim more precisely.","section":"Theorem 7.1, statement"}],"recommendation":"major_revision","confidential_remarks":"The paper is clearly the work of a capable author and the overall architecture is sound, but the numerical gap in Lemma 5.9 is exactly the kind of localized defect that can be fixed. Because all the advertised explicit constants flow through this lemma, I would like the author to repair the proof and then recompute the constants in Theorems 5.1, 5.17, 7.1, and 1.8. I do not see a reason to doubt the qualitative polynomial-with-exponent-4.42 statement if the lemma is replaced by a correct version, but the exact constants must be revalidated. The reliance on MAGMA and FindOpenImage computations is acceptable for this field, but the journal may want to ask for the verification scripts."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this paper is the real thing. It gives the first effective polynomial bound on [GL2(hat Z):Im rho_E] over Q with constants one can actually quote—10^21(h+40)^4.42—and an asymptotic h^{3+o(1)}; it also classifies the non-split Cartan p-adic images that are the last open case of Serre uniformity. The improvement over Zywina (ineffective) and Lombardo (effective but absurd constants) is substantial. The architecture is coherent: a Gaudron-Rémond-style effective surjectivity theorem, a group-theoretic classification at level p^n, and careful entanglement estimates.\n\nWhere I would push back: the reader's stress-test note on Lemma 5.9 lands. The displayed chain at h_F=-0.75 gives ~4.643 on the left vs 4.4925 claimed, and the quoted log x/x <= log32.2/32.2 is not enough to close it. But I do not think the lemma is false. Replacing that bound by the tangent-line bound log x <= log32.2 + (x-32.2)/32.2 (concavity) gives 2.023 h + 6.160 < 2.29 h + 6.21, so the lemma holds with a one-line repair. The constants 1454 etc. should survive; the proof as printed needs a correction.\n\nOther soft spots are minor. The paper invokes a fair amount of black-box computation: MAGMA checks in Lemma 2.9, FindOpenImage for finitely many j-invariants, RSZB tables, LMFDB lookups. That is standard practice in this area and the finite lists are small, but no code is shipped, so a referee cannot machine-verify those steps without re-doing them. The author is honest about the mistake in [1] and gives conditional statements around Claim 2.6. Self-citation of [22] is legitimate: it is an input theorem, not a surrogate for the target result.\n\nVerdict: send to a serious referee. A careful expert should check the Lemma 5.9 repair and the few computational classifications, but the main result is important and likely correct. I would cite it.","headline":"Strong, mostly sound effective bounds for adelic Galois images; Lemma 5.9 has a repairable numerical gap in the printed proof, not a fatal flaw.","tokens_in":60055,"tokens_out":6610,"would_cite":true,"duration_ms":56402,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11G05","11F80","11G18","11R32"],"pacs":[],"model":"deepseek-v4-flash","headline":"For a non-CM elliptic curve over $\\mathbb{Q}$, the adelic Galois image index is less than $10^{21}(h_{\\mathcal{F}}(E)+40)^{4.42}$.","keywords":["elliptic curves","Galois representations","adelic image","open image theorem","Faltings height","non-split Cartan subgroups","effective bounds","torsion fields"],"falsifier":"Run the one-variable check behind Lemma 5.9 across $x\\in[-0.75,0]$: compare $\\frac{3}{\\pi}\\log\\left(\\frac{12}{\\pi}x+5.52+4e^2\\right)+\\frac{6}{\\pi}x+2.76$ with $2.29x+6.21$. If the first expression exceeds the second anywhere in that interval, the displayed proof does not establish the lemma's constant, and the constants depending on it (1454, 1266.4, 21000, 2488320) need revision; otherwise the chain is consistent.","tokens_in":59048,"feed_emoji":"🔢","tokens_out":19347,"duration_ms":162518,"temperature":0.7,"pith_summary":"The open image theorem for elliptic curves guarantees that a non-CM elliptic curve over $\\mathbb{Q}$ has adelic Galois image of finite index in $\\operatorname{GL}_2(\\widehat{\\mathbb{Z}})$, but it does not reveal how large that index can be. This paper answers the quantitative version for rational curves: the index $[\\operatorname{GL}_2(\\widehat{\\mathbb{Z}}):\\operatorname{Im}\\rho_E]$ is less than $10^{21}(h_{\\mathcal{F}}(E)+40)^{4.42}$, where $h_{\\mathcal{F}}(E)$ is the stable Faltings height, a logarithmic measure of the curve's arithmetic size, and for large heights the bound improves to $h_{\\mathcal{F}}(E)^{3+o(1)}$. To prove this, the paper classifies the possible $p^n$-level images when the mod $p$ image sits in the normaliser of a non-split Cartan subgroup, and shows that the product of the corresponding prime powers grows at most like a power of the height. A companion bound in terms of the conductor makes the estimate effective in practice. If the main theorem is right, the missing portion of the Galois image is controlled by a computable function of the curve's size rather than by an unknown curve-dependent constant.","feed_headline":"10^21(h+40)^4.42 bounds Galois image index","feed_subtitle":"Every non-CM elliptic curve over Q has adelic Galois image index below this height polynomial.","key_machinery":"The main working object is the N-Cartan lift: a closed subgroup $G<\\operatorname{GL}_2(\\mathbb{Z}_p)$ whose reduction modulo $p$ is contained in the normaliser of a Cartan subgroup, is not contained in the Cartan, has surjective determinant, and contains a non-scalar element from the Cartan. For such a group the graded pieces $g_n=(G_n/G_{n+1})\\hookrightarrow\\mathfrak{gl}_2(\\mathbb{F}_p)$ obey a forced decomposition $\\mathfrak{gl}_2(\\mathbb{F}_p)=V_1\\oplus V_2\\oplus V_3$ (Lemma 3.7), and the dimension dynamics of these pieces force the level-$p^n$ image to be either the full normaliser $C^+_{ns}(p^n)$ or one of a few exceptional groups (Propositions 3.13 and Theorem 3.14). The second machinery is an effective surjectivity theorem (Theorems 5.1 and 5.17): from the Cartan subgroups one builds a quotient of $E\\times E$ whose degree records the prime powers contributing, and the period-theoretic isogeny theorem used in Section 5 bounds that degree by the Faltings height, giving $\\Lambda<21000(h_{\\mathcal{F}}(E)+40)^{1.308}$ for rational curves. The third machinery is entanglement control: in the non-split Cartan case the prime $p$ is almost totally ramified in $\\mathbb{Q}(E[p^n])$, while at primes not dividing the conductor and $p$ the ramification is small, so the overlap of division fields at different primes is small; Lemma 7.16 converts the product of the $p$-adic indices into the adelic index with an additional factor at most $1536\\cdot 6^\\alpha$.","core_discovery":"The central claim is that for a non-CM elliptic curve over $\\mathbb{Q}$ the size of the missing part of the full adelic Galois image is polynomially controlled by the stable Faltings height. Theorem 1.8 states $[\\operatorname{GL}_2(\\widehat{\\mathbb{Z}}):\\operatorname{Im}\\rho_E]<10^{21}(h_{\\mathcal{F}}(E)+40)^{4.42}$, and as $h_{\\mathcal{F}}(E)\\to\\infty$, $[\\operatorname{GL}_2(\\widehat{\\mathbb{Z}}):\\operatorname{Im}\\rho_E]<h_{\\mathcal{F}}(E)^{3+o(1)}$; an explicit conductor version bounds the same index by $2488320\\,(51N(1+\\log\\log N)^{1/2})^{3\\omega(N)}$. The route is to show that the only primes that can contribute seriously are those for which $\\operatorname{Im}\\rho_{E,p}$ lies in the normaliser of a non-split Cartan subgroup, and that at level $p^n$ the image is almost always the full normaliser $C^+_{ns}(p^n)$. The product of these prime powers is then bounded by an effective surjectivity theorem in terms of $h_{\\mathcal{F}}(E)$, and the entanglement between division fields at different primes is shown to cost only a small multiplicative factor. The paper also classifies the possible (conjecturally non-existent) images of $\\rho_{E,p^n}$ whenever $\\operatorname{Im}\\rho_{E,p}$ is contained in a non-split Cartan normaliser.","pith_inferences":["Going beyond the paper: the same three-step decomposition—residual-image classification, a height-based product bound, and entanglement control—should produce analogous effective index bounds for other compatible families of 2-dimensional Galois representations such as modular forms, whenever their residual images are classified.","Going beyond the paper: the explicit constants invite a finite numerical test: compute the actual adelic index for all non-CM curves up to a moderate height threshold and compare it with $10^{21}(h_{\\mathcal{F}}(E)+40)^{4.42}$; the paper contains no such data, and the comparison would indicate how far the constant is from the truth.","Going beyond the paper: the asymptotic statement $h_{\\mathcal{F}}(E)^{3+o(1)}$ is more robust than the displayed constant $10^{21}$, because it does not depend on the precise constant in Lemma 5.9; optimizing the constant would therefore focus on Lemma 5.9 and on the $1536\\cdot 6^\\alpha$ entanglement factor."],"forward_implications":["For every non-CM elliptic curve over $\\mathbb{Q}$, the adelic Galois image index is bounded by a fixed polynomial in the stable Faltings height with explicit constants.","For sufficiently large heights the bound becomes $h_{\\mathcal{F}}(E)^{3+o(1)}$; in particular, for every $\\varepsilon>0$, the index is eventually smaller than $h_{\\mathcal{F}}(E)^{3+\\varepsilon}$.","The conductor bound $[\\operatorname{GL}_2(\\widehat{\\mathbb{Z}}):\\operatorname{Im}\\rho_E]<2488320(51N(1+\\log\\log N)^{1/2})^{3\\omega(N)}$ gives an effective, lower-exponent replacement for the previous conductor bound.","When $\\operatorname{Im}\\rho_{E,p}$ is contained in the normaliser of a non-split Cartan subgroup, the possible images of $\\rho_{E,p^n}$ form a short classified list: the full normaliser, listed $p=3$ and $p=5$ exceptional groups, or one specific level-$p^2$ semidirect-product group.","If that specific level-$p^2$ group can be excluded for all curves, the asymptotic height exponent improves from $3+o(1)$ to $2+o(1)$."],"supporting_citations":[{"why":"Supplies the period-theoretic isogeny estimate that powers the effective surjectivity theorem bounding the product of Cartan-level prime powers.","marker":"[24]"},{"why":"Provides the original effective surjectivity theorem and the quotient construction of $E\\times E$ that Theorem 5.1 generalizes.","marker":"[32]"},{"why":"Contributes the previous non-split Cartan uniformity results, the small-height improvements, and the height-to-$j$-invariant comparison used throughout.","marker":"[22]"},{"why":"Gives the Cartan-lift group lemmas and the earlier height and conductor bounds that this paper improves and makes effective.","marker":"[62]"},{"why":"Classifies the possible mod $p$ images, including the non-split Cartan cases, which Theorem 1.9 extends to level $p^n$.","marker":"[63]"},{"why":"Classifies $p$-adic images outside the non-split Cartan case, used to enumerate the exceptional small-prime groups.","marker":"[51]"},{"why":"Supplies the open image theorem guaranteeing that the adelic image has finite index, the starting point of the whole question.","marker":"[54]"},{"why":"Supplies the analytic inequality used in Lemma 5.9 to obtain the numerical constant $2.29$.","marker":"[57]"}],"fun_headline_variants":["Explicit bound: adelic Galois image index < 10^21(h+40)^4.42","Height polynomial controls Galois image index for non-CM elliptic curves","Non-CM elliptic curves: Galois index bounded by (h+40)^4.42","Galois image index: sharp effective bound via Faltings height","10^21(h+40)^4.42 caps adelic Galois image index"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The explicit constants in the main theorems depend on Lemma 5.9, a numerical inequality comparing an average of inverse squared periods to 2.29 times the Faltings height plus 6.21; if that inequality is not valid as stated, the displayed constants must be revised even though the polynomial form is likely to survive.","fun_headline_variants_meta":{"raw":{"variants":["Explicit bound: adelic Galois image index < 10^21(h+40)^4.42","Height polynomial controls Galois image index for non-CM elliptic curves","Non-CM elliptic curves: Galois index bounded by (h+40)^4.42","Galois image index: sharp effective bound via Faltings height","10^21(h+40)^4.42 caps adelic Galois image index"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001665,"raw_usage":{"total_tokens":6661,"prompt_tokens":1052,"completion_tokens":5609,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":668,"completion_tokens_details":{"reasoning_tokens":5499}},"tokens_in":668,"tokens_out":5609,"duration_ms":33593,"temperature":1.0,"reasoning_tokens":5499,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T15:56:47.358474+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the one-variable check behind Lemma 5.9 across $x\\in[-0.75,0]$: compare $\\frac{3}{\\pi}\\log\\left(\\frac{12}{\\pi}x+5.52+4e^2\\right)+\\frac{6}{\\pi}x+2.76$ with $2.29x+6.21$. If the first expression exceeds the second anywhere in that interval, the displayed proof does not establish the lemma's constant, and the constants depending on it (1454, 1266.4, 21000, 2488320) need revision; otherwise the chain is consistent.","supporting_citations":[{"cited_title":"Gaudron and G","cited_arxiv_id":null,"evidence_quote":"Supplies the period-theoretic isogeny estimate that powers the effective surjectivity theorem bounding the product of Cartan-level prime powers."},{"cited_title":"Le Fourn","cited_arxiv_id":null,"evidence_quote":"Provides the original effective surjectivity theorem and the quotient construction of $E\\times E$ that Theorem 5.1 generalizes."},{"cited_title":"Bounds for Serre's open image theorem","cited_arxiv_id":"1102.4656","evidence_quote":"Gives the Cartan-lift group lemmas and the earlier height and conductor bounds that this paper improves and makes effective."},{"cited_title":"Rouse, A","cited_arxiv_id":null,"evidence_quote":"Classifies $p$-adic images outside the non-split Cartan case, used to enumerate the exceptional small-prime groups."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the open image theorem guaranteeing that the adelic image has finite index, the starting point of the whole question."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the analytic inequality used in Lemma 5.9 to obtain the numerical constant $2.29$."}],"review_version":1}