{"id":"be9aa896-6502-4824-be85-0a01f027ad79","arxiv_id":"2501.07833","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The normalizer of the non-split Cartan subgroup of level 27 does not occur as a 3-adic Galois image of a non-CM elliptic curve over Q, completing the 3-adic classification.","lead":"This paper completes the classification of 3-adic Galois images of non-CM elliptic curves over Q, proving the last exceptional subgroup of level 27 cannot occur. It develops a quadratic Chabauty method over number fields and computes the rational points on a genus 3 modular curve to obtain the result.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 1.3 depends on quotient maps f1-f3 provided privately by Rouse; with these maps absent from the paper, the final pullback from X(F) to X_ns^+(27)(F) is unverifiable.","rationale":"The reader identified Assumption 1 (abelian log isomorphism) as the weakest assumption. That is a genuine theoretical concern, but the paper reports in §3.7 that the known F-points generate T ⊗_K T, which implicitly verifies surjectivity of the log map; while the verification is not shown in detail, it is at least a computational check. The more concrete, unresolvable gap as written is the final step from X(F) to X_ns^+(27)(F): the quotient maps f1, f2, f3 and the canonical model are external private data, not included in the paper or the public code. Since the paper's headline theorem is the classification of 3-adic Galois images, and that classification is only obtained by pulling back through these unpublicized maps, a single error or omission in these maps would destroy the main result even if the entire quadratic Chabauty computation on X'_H is correct. The reader's conditional verdict is appropriate: the mathematical core appears sound, but the manuscript should require the missing maps and model to be made available or fully specified before the classification claim can be independently verified.","tokens_in":36975,"tokens_out":7878,"duration_ms":79647,"concrete_test":"Obtain the canonical model of X_ns^+(27) and the rational maps f1, f2, f3 from the authors or by independent computation, then recompute the pullback of the 13 F-points of X(F) under f2 and f3. Verify that the resulting set of F-points on X_ns^+(27) is exactly the claimed set of 10 points (8 rational with CM discriminants -4, -7, -16, -19, -28, -43, -67, -163, and 2 additional points with discriminant -4). If the pullback produces any additional F-points, or if it fails to contain the claimed points, Theorem 1.3 would be invalidated.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In §3.8, the proof of Theorems 1.2 and 1.3 uses 'equations of the canonical model of X_ns^+(27) in P^11 along with explicit formulas for rational maps representing quotient morphisms f1, f2, f3' that were 'provided by Jeremy Rouse' but are not included in the paper, appendix, or (apparently) the public Magma repository. The authors state that f1 is not defined at one point, so they instead pull back via f2 and f3; the computation yields exactly the claimed CM points. If any of these maps or the point identification is mistaken, Theorem 1.3—the headline classification of 3-adic Galois images—does not follow even if Theorem 1.1 is correct. This is a concrete, checkable gap: the maps are finite rational maps between explicitly given curves, so they can be published and independently recomputed. Without them, the central claim is not self-contained.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves that the non-split Cartan modular curve X_ns^+(27) has exactly eight rational points, all CM, with discriminants -4, -7, -16, -19, -28, -43, -67, and -163. Since this was the last open case in the Rouse--Sutherland--Zureick-Brown classification, the paper thereby completes the classification of 3-adic Galois images attached to non-CM elliptic curves over Q. The proof combines a new quadratic Chabauty method over the number field F = Q(ζ_3) with restriction of scalars, Nekovář p-adic heights with multiple idèle class characters, and a computation of local heights away from the chosen prime p = 13. The method is applied to a smooth plane quartic quotient X'_H of X_ns^+(27), proving that X'_H(F) consists of thirteen points, and then pulls back through quotient maps to X_ns^+(27) to obtain Theorems 1.2 and 1.3.","tokens_in":37152,"tokens_out":5902,"duration_ms":63223,"significance":"If the proof is correct, the paper resolves a well-known open case in the arithmetic of elliptic curves: it completes the list of possible 3-adic Galois images for non-CM elliptic curves over Q, giving the 47 groups of [RSZ22, Table 3]. This is a major result in the area of Galois images and modular curves. The paper also makes a methodological contribution: it extends explicit quadratic Chabauty to curves over number fields with Mordell--Weil rank equal to d g, using restriction of scalars and multiple idèle class characters, and it contains a careful p-adic precision analysis that corrects an earlier bound in [Bal+23]. The availability of Magma code and the explicit precision statements are strengths; the proof is computational but is accompanied by a certification framework rather than being a bare numerical verification.","major_comments":[{"comment":"The deduction of Theorems 1.2 and 1.3 from Theorem 1.1 depends on the quotient maps f1, f2, f3, which are stated to have been provided by Jeremy Rouse but are not included in the paper, in the appendix, or in the linked Magma repository. This is load-bearing: the pullback via f2 and f3 is the only step connecting the verified set X'_H(F) to the set X_ns^+(27)(F), and hence to the headline classification of 3-adic Galois images. The authors should include the equations of the canonical model of X_ns^+(27) in P^11 and explicit rational maps for f1, f2, f3 (or a reproducible transcript), and specify the image of each of the thirteen F-points, including the point where f1 is not defined.","section":"Section 3.8"},{"comment":"Assumption 1 asserts that the abelian logarithm map (2.1) is an isomorphism, and the paper states that this is satisfied for the example, but no verification is described. This assumption is needed to extend the Nekovář height pairing to a Q_p-bilinear pairing on T and hence to define the functions h_Z and rho on X(F ⊗ Q_p). The statement in Section 3.7 that the elements e_{Z_1}(x) and e_{Z_2}(x) generate T ⊗_K T appears to imply surjectivity of the logarithm on the K-span, but the implication is not spelled out and no computation of the p-adic closure or rank of J(F) is reported. Please provide a direct verification of Assumption 1, or clarify exactly how the generation statement in Section 3.7 constitutes that verification.","section":"Section 2.3"},{"comment":"The matrices Z_1 and Z_2 are obtained by recognizing Q_p-approximations as algebraic numbers via LLL, and the paper explicitly states that exact equality with the true Hecke correspondences is not proved. The authors argue that p-adic accuracy up to an explicit bound suffices because all computations can be performed over Q_p except the Hodge-filtration algorithm of Section 2.4.5, and that the required accuracy is inherited from the certified Frobenius matrices. This reduction is plausible but should be documented in detail: state the numerical precision needed for each step of Algorithm 2.8, explain how the LLL output is certified to that precision, and state what would happen if the recognized algebraic numbers were only approximate. Without this, the cycle input to the height computation is not rigorously fixed.","section":"Section 3.5, Remark 3.2"}],"minor_comments":[{"comment":"The sentence describing the maps f1, f2, f3 would be more informative if it stated the degrees of these maps and identified the single F-point at which f1 is not defined, since that point plays a special role in the pullback argument.","section":"Section 3.8"},{"comment":"The list of thirteen known F-points would be easier to check if it indicated which points are CM and which is the non-CM point with j-invariant j_0, rather than leaving this to the surrounding text.","section":"Section 3.1, equation (3.2)"},{"comment":"The constant c_3 = min_j{v_p(d_j)} is used in (2.24) but is defined only after (2.20); moving the definition immediately before (2.24) would improve readability.","section":"Section 2.7.2"},{"comment":"The proof of Proposition 3.3 relies on the stable model description of Figure 1 and the relation between the minimal regular model and the stable model; a one-sentence reminder that the minimal regular model dominates the stable model and that the relevant component is the unique one dominating the genus 0 component would make the argument easier to follow.","section":"Section 3.6, Lemma 3.6"},{"comment":"The correction to the factorization of j_0 compared with [RSZ22, §9.1] is important; please add a short explanation or reference for the claim that the earlier factorization does not lift to an F-point on X_ns^+(9).","section":"Remark 3.2"}],"recommendation":"major_revision","confidential_remarks":"The missing quotient maps f1--f3 are the main obstacle to independent verification of the classification result. Since the maps are finite rational maps between explicitly given curves, they can be supplied by the authors; this is a fixable gap rather than a fatal error. I would also encourage the authors to archive a specific commit of the Magma repository at the time of the revision, so that the published version refers to a fixed version of the code."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The headline result is real and important: with this paper, the classification of 3-adic Galois images of non-CM elliptic curves over Q is complete. That closes a problem that has been open for years, and it also gives a new proof of the class number one problem. The method is a genuine extension of quadratic Chabauty: it works over the number field Q(ζ3) with Jacobian rank 6, beyond the old d(g−1)=4 bound, using restriction of scalars plus two independent idèle class characters. That is a substantial technical advance, not a repackaging.\n\nThe paper is also honest about its own gaps. Section 2.3 states Assumption 1 (the abelian log map is an isomorphism, so r=dg) and says it is satisfied for X'_H, but does not prove or verify it. That assumption is load-bearing: without it the height pairing cannot be extended to T, and the quadratic Chabauty functions may not be defined on all of X(F⊗Qp). The reader's report flagged this, and I agree it is the biggest structural soft spot. The correction to the precision bound in [Bal+23] is a good sign, not a red flag.\n\nThe soft spot that worries me more, practically, is the endgame. Theorem 1.3 is deduced by pulling back the proven set X(F) through quotient maps f2, f3, but the rational maps representing f1, f2, f3 were provided privately by Jeremy Rouse and are not in the paper or (apparently) the public Magma repository. The proof states that f1 is not defined at one point, so they use f2 and f3 instead, but without the actual equations and maps an independent referee cannot check that the pullback lands exactly on the claimed CM points. This is a concrete, fixable gap: the maps are finite rational maps between explicitly given curves, and they can be published. As written, Theorem 1.3 is not self-contained.\n\nMinor point: the Z_i matrices are recognized via LLL from 344 p-adic digits, without a proof that the recognized algebraic numbers are exactly right. The authors argue this is sufficient because everything runs over Q_p, but the Hodge filtration computation in Algorithm 2.8 requires entries in F. They say the p-adic accuracy suffices; that is plausible but should be stated more carefully.\n\nWho is this for? Any arithmetic geometer working on Chabauty methods or Galois images. It deserves a serious referee: the main theorem is significant, the method is a real advance, and the gaps are concrete and fixable. My recommendation: send it to review with a request that the authors include the Rouse maps in the paper or a permanent public repository. The core mathematics appears sound, but the proof of the headline theorem must be independently verifiable.","headline":"Closes the last open case of the 3-adic classification and pushes quadratic Chabauty over number fields past the previous rank bound, but the final pullback step leans on unpublished maps from Rouse that need to be made public.","tokens_in":37707,"tokens_out":859,"would_cite":true,"duration_ms":10933,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11G18","11G50","14G05","11F80"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that the level-27 non-split Cartan modular curve has exactly eight rational points over Q, all CM, completing the 3-adic Galois image classification.","keywords":["quadratic Chabauty","non-split Cartan modular curve","3-adic Galois images","rational points","elliptic curves over the rationals","Nekovář p-adic heights","restriction of scalars","Mordell-Weil rank"],"falsifier":"Compute the dimension of the p-adic closure of $J(\\mathbb{Q}(\\zeta_3))$ in the tangent space at $p=13$: a dimension below 6 would break the extension of the height pairing and invalidate the common-zero computation. Alternatively, a single rational point on $X_{\\mathrm{ns}}^+(27)(\\mathbb{Q})$ outside the eight listed points, or a non-CM elliptic curve over $\\mathbb{Q}$ whose 3-adic image is the level-27 non-split Cartan normaliser, would refute the main theorem.","tokens_in":36792,"feed_emoji":"8️⃣","tokens_out":11459,"duration_ms":97736,"temperature":0.7,"pith_summary":"This paper completes the classification of 3-adic Galois images of elliptic curves over $\\mathbb{Q}$ by proving that the normaliser of the non-split Cartan subgroup of level 27 cannot occur as such an image. The central numerical statement is that $X_{\\mathrm{ns}}^+(27)(\\mathbb{Q})$ has exactly eight points, all complex-multiplication (CM) points, with discriminants $-4,-7,-16,-19,-28,-43,-67,-163$. To reach it, the authors compute the full set of $\\mathbb{Q}(\\zeta_3)$-points on a genus-3 plane quartic quotient $X'_H$ whose Jacobian has Mordell--Weil rank 6, a case beyond earlier quadratic Chabauty implementations because the curve is not defined over $\\mathbb{Q}$ and the rank is too large. They develop quadratic Chabauty over number fields via restriction of scalars and Nekov\\'a\\v{r} p-adic heights with independent id\\`ele class characters, combined with a local-height analysis above 3. If the proof is correct, the 3-adic image of every non-CM elliptic curve over $\\mathbb{Q}$ is one of 47 explicit subgroups of $\\operatorname{GL}_2(\\mathbb{Z}_3)$.","feed_headline":"Eight rational points finish the 3-adic image classification","feed_subtitle":"All eight points on the level-27 non-split Cartan curve are CM, closing the last open case for elliptic curves over Q.","key_machinery":"The mechanism is quadratic Chabauty over a number field, built from the abelian logarithm, restriction of scalars, and Nekov\\'a\\v{r} p-adic heights. For $X'_H$ over $F=\\mathbb{Q}(\\zeta_3)$, the paper uses two trace-zero correspondences $Z_i$ and two independent id\\`ele class characters to build four p-adic Coleman functions $\\rho_{i,j}$ on $X'(F\\otimes\\mathbb{Q}_{13})$; an isomorphism assumption on the abelian logarithm extends the height pairing to the full tangent space, the local heights away from 13 vanish by a semistable-reduction argument, and the common zeros are computed to precision 135 in good residue polydiscs. Two independent cycles and characters supply more constraints than the single height function of earlier work, which is what makes rank 6 tractable.","core_discovery":"The paper's central claim is that $X_{\\mathrm{ns}}^+(27)(\\mathbb{Q})$ has exactly eight rational points, all CM points with discriminants $-4,-7,-16,-19,-28,-43,-67,-163$, so no non-CM elliptic curve over $\\mathbb{Q}$ has 3-adic Galois image equal to the normaliser of the non-split Cartan subgroup of level 27. Together with the existing classification, this forces the 3-adic image of every non-CM elliptic curve over $\\mathbb{Q}$ to be one of 47 explicit subgroups of $\\operatorname{GL}_2(\\mathbb{Z}_3)$. The proof passes through a degree-3 quotient: a smooth plane quartic $X'_H$ over $F=\\mathbb{Q}(\\zeta_3)$ whose $F$-points are shown to be exactly thirteen, pulling back to ten $F$-points on $X_{\\mathrm{ns}}^+(27)$, eight of which are rational. The point count is obtained from a quadratic Chabauty method over number fields that handles Jacobian rank 6 on a genus-3 curve by using all primes above a split prime $p$ and multiple id\\`ele class characters.","pith_inferences":["Beyond the paper: if the abelian-logarithm isomorphism is later proved unconditionally for $X'_H$, the exclusion of extra common zeros becomes fully rigorous; until then the point count rests on that stated isomorphism.","Beyond the paper: running the same computation at a second split prime, say $p\\neq 13$, should reproduce the same thirteen points on $X'_H$, giving an independent numerical check of the theorem.","Beyond the paper: the framework suggests a template for the remaining open non-split Cartan curves of levels 25, 49, and 121: locate a low-genus quotient over a small number field, control local heights at the bad prime, and solve the resulting multivariate height system."],"forward_implications":["The 3-adic Galois image of every non-CM elliptic curve over $\\mathbb{Q}$ is now explicitly known: one of 47 subgroups of $\\operatorname{GL}_2(\\mathbb{Z}_3)$.","The eight rational points give a new proof of the class-number-one problem, since their CM discriminants are exactly the imaginary quadratic fields of class number one.","The number-field quadratic Chabauty framework extends explicit rational-point computations to curves over imaginary quadratic fields whose Jacobian rank equals twice the genus, beyond the range of earlier methods.","The computation gives a checkable certificate that the thirteen listed points exhaust $X'_H(\\mathbb{Q}(\\zeta_3))$, so no extra point on the quotient can hide an exceptional 3-adic image.","Because all local heights away from 13 vanish, the finite set $\\Upsilon$ is $\\{0\\}$, so the rational points are cut out by a system of p-adic equations rather than by a nontrivial finite value set."],"supporting_citations":[{"why":"Establishes the prior 3-adic classification up to the single exceptional subgroup and constructs the quotient quartic X'_H with its equation.","marker":"[RSZ22]"},{"why":"Introduces explicit quadratic Chabauty for modular curves using Nekovář heights, the foundation this paper extends.","marker":"[Bal+19]"},{"why":"Supplies the quadratic Chabauty algorithms and precision analysis for modular curves that the present implementation adapts to number fields.","marker":"[Bal+23]"},{"why":"Gives explicit quadratic Chabauty over number fields with restriction of scalars and the multivariate Hensel root-finding strategy.","marker":"[Bal+21a]"},{"why":"Computes the semistable reduction of X'_H at 3, used to prove that all local heights away from p vanish.","marker":"[Oss23]"},{"why":"Constructs the global p-adic height pairings that define the quadratic Chabauty functions rho.","marker":"[Nek93]"},{"why":"Provides the local-height decomposition and the vanishing of local heights away from p for good reduction.","marker":"[BD18]"},{"why":"Together with [KL89], determines the Mordell-Weil rank of the Jacobian as 6.","marker":"[GZ86]"},{"why":"Supplies the finiteness and rank computation used with Gross-Zagier to fix rk J(F)=6.","marker":"[KL89]"},{"why":"Supplies the rigid-cohomology reduction algorithm used to compute Frobenius structures and Coleman integrals.","marker":"[Tui16; Tui17]"}],"fun_headline_variants":["Eight CM points close the last 3-adic Galois gap","Quadratic Chabauty over number fields seals 3-adic classification","Level-27 nonsplit Cartan ruled out for non-CM elliptic curves","Last open 3-adic Galois image case resolved: eight CM points"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the p-adic logarithm map from the Mordell--Weil group of the Jacobian to its tangent space is an isomorphism, which forces the rank to equal $2\\cdot 3=6$; the paper states this holds for $X'_H$ but gives no proof or computational verification.","fun_headline_variants_meta":{"raw":{"variants":["Eight CM points close the last 3-adic Galois gap","Quadratic Chabauty over number fields seals 3-adic classification","Level-27 nonsplit Cartan ruled out for non-CM elliptic curves","Last open 3-adic Galois image case resolved: eight CM points"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001339,"raw_usage":{"total_tokens":5530,"prompt_tokens":1118,"completion_tokens":4412,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":734,"completion_tokens_details":{"reasoning_tokens":4332}},"tokens_in":734,"tokens_out":4412,"duration_ms":30749,"temperature":1.0,"reasoning_tokens":4332,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T20:34:19.852353+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the dimension of the p-adic closure of $J(\\mathbb{Q}(\\zeta_3))$ in the tangent space at $p=13$: a dimension below 6 would break the extension of the height pairing and invalidate the common-zero computation. Alternatively, a single rational point on $X_{\\mathrm{ns}}^+(27)(\\mathbb{Q})$ outside the eight listed points, or a non-CM elliptic curve over $\\mathbb{Q}$ whose 3-adic image is the level-27 non-split Cartan normaliser, would refute the main theorem.","supporting_citations":[],"review_version":1}