{"id":"eca7e090-7f95-4e94-bd0d-4a721adb715f","arxiv_id":"2502.04084","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"An explicit basis of modular units on X1(p) is constructed, yielding computations of the cuspidal group and its rational subgroup for all primes from 11 to 997.","lead":"This paper gives an explicit list of p-2 building-block functions whose zeros and poles sit only at the cusps of the modular curve X1(p), for every prime p at least 5. The list makes the cuspidal group, a finite abelian group attached to the curve, computable by linear algebra for primes up to 997.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition 19's cohomological equality is asserted as 'clear' but is not a formal consequence of the cyclic Galois action; the rational-cuspidal identification rests on it.","rationale":"The paper's main Theorem 1 is well supported: the constructed functions are modular units, their divisor matrix is computed, and its determinant matches the independent Takagi–Yu cuspidal class number. The determinant computation has a minor typo (a missing square in the displayed det(B)), but the final equality with h1(p) appears correct and can be checked for small p. The genuine soft spot is Proposition 19, exactly as the reader's rationale indicates: the proof compresses a nontrivial lattice-theoretic statement into the word 'clear'. Since the verdict was already CONDITIONAL in part because of this, my stress-test does not move it; it sharpens the required check. If the proposed rational-lattice test passes for representative primes, the conditional can be upgraded with confidence; if it fails, the rational cuspidal subgroup claims would need revision. The external Takagi–Yu formula identified in the reader's weakest_assumption is a published theorem and not, by itself, a correctness risk.","tokens_in":26524,"tokens_out":37081,"duration_ms":383621,"concrete_test":"For p=11 and p=13, compute M = div(F(p)) explicitly from the basis of Theorem 1, and compute the SNF of the two lattices ker(N) ∩ M and (1−σ)M inside the Q-cusp coordinates; if their quotients are nontrivial, Proposition 19 is false. Independently cross-check by computing C1(p)(Q) via Magma's modular symbols cuspidal subgroup and comparing its order with |C∞1(p)| = |CQ1(p)| implied by Corollary 18 and the table; agreement for both primes would indicate the asserted equality holds at least there, and disagreement would confirm the concern.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In §6.3, Proposition 19 claims CQ1(p) = C1(p)(Q). The proof reduces the required vanishing of H1(G, div(F(p))) to the equality {D ∈ div(F(p)) : N D = 0} = (1−σ) div(F(p)), and then says 'this is clear since the Galois action fixes the Pi and permutes the Qi cyclically.' That equality is not a formal property of the cyclic action: for a proper finite-index sublattice M of the full cusp-divisor lattice Λ, the norm-kernel in M can be strictly larger than (1−σ)M. Whether it holds here depends on the explicit generators of div(F(p)) supplied by Theorem 1, and the paper provides no verification. If the equality fails, H1(G, div(F(p))) is nonzero and the map DivQc(p) → C1(p)(Q) need not be surjective, so the computations of C1(p)(Q) in Tables 2–3 and the claim that the rational cuspidal group is generated by P-cusp divisors would be unsupported. This does not affect Theorem 1 itself, but it is the most load-bearing unresolved step in the paper's later claims.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"Let p ≥ 5 be prime. The paper fixes a primitive root α modulo p, defines Siegel-function products E_i, F_i, and constructs explicit functions G_0, …, G_{n-1}, H_1, …, H_{n-1}. Theorem 1 claims that these p − 2 functions form a basis for the group F(p) of modular units on X_1(p) modulo constants. The proof builds the matrix of their divisors, computes its determinant by reducing to circulant matrices and Dirichlet-character eigenvalues, and matches it with the known cuspidal class number h_1(p) of Takagi and Yu. The paper then studies two subgroups: units whose divisors are supported on the rational cusps P_i (Theorem 2) and units with Galois-fixed divisors (Theorem 3). It asserts in Proposition 19 that C∞_1(p) = C^Q_1(p) = C_1(p)(Q), uses this to compute the rational cuspidal group for 11 ≤ p ≤ 997, and derives explicit formulas for the orders of P_0 − Q_0 and P_0 − P_{n−1}. The central, determinant-based argument for Theorem 1 is explicit and checkable; the later rational-subgroup and order computations depend on gaps identified below.","tokens_in":26700,"tokens_out":13056,"duration_ms":141287,"significance":"If Theorem 1 is correct, it is a valuable structural result: it provides a fully explicit basis for the modular units on X_1(p) and reduces the computation of the cuspidal group to the Smith normal form of an explicit divisor matrix. The determinant proof is concrete and uses the independent cuspidal class number formula as an external benchmark, so the argument is not circular. The manuscript also ships code and data for the cuspidal-group computations, which is a reproducible-evidence strength. The explicit bases for the P-supported and Galois-fixed unit groups, if fully justified, give practical access to the rational cuspidal group for large p. However, the later claims about C_1(p)(Q) and the order formulas rely on an unproved cohomological vanishing assertion and on an incomplete integrality argument, so the significance of those sections is conditional on repairs.","major_comments":[{"comment":"The equality C^Q_1(p) = C_1(p)(Q) rests on the asserted equality {D ∈ div(F(p)) : N D = 0} = (1 − σ) div(F(p)). The sentence 'this is clear since the Galois action fixes the Pi and permutes the Qi cyclically' is not a proof. For a proper finite-index Galois-stable sublattice M of a permutation lattice, the norm-kernel in M can be strictly larger than (1 − σ)M, so the vanishing of H^1(G, M) has to be verified for this particular M, not for the ambient lattice. In addition, the displayed cohomology sequence omits the term H^1(G, Div_c(p)); that term is zero because Div_c(p) is a permutation lattice, but this should be stated. Since Theorem 1 gives an explicit basis for div(F(p)), the required equality can be checked by an explicit finite computation, for example by putting the Z[G]-module generated by the columns of the divisor matrix into a form that exhibits the norm-kernel and (1 − σ)M. This is load-bearing: Proposition 19 is what identifies Tables 2–3 and the Section 8 order-of-P_0 − P_{n−1} computation with the rational cuspidal group, so the gap needs a complete repair rather than a stylistic revision.","section":"§8.2, Proposition 23"},{"comment":"The proof that every f ∈ F^∞(p) has integer exponents in its E_i-product is not complete. The argument writes f = (f~)^{1/n} with f~ a product of the E_i with integer exponents and then invokes 'Ogg's lemma from [17, Section 4]' without stating the lemma or checking its hypotheses. The passage from an nth root with bounded-denominator Fourier coefficients to integrality of the exponents is exactly the point at issue. This integrality is used in Proposition 24 and hence in Proposition 25, the formula for the order of P_0 − P_{n−1}; a failure here would invalidate that formula. The paper can repair this either by stating and proving the invoked lemma in the present setting, or by deriving the statement from the explicit basis of Theorem 2, since each basis element I_j is expressed with integer E_i-exponents and the final I_{n−1} has the same property.","section":"§8.2, Proposition 23"}],"minor_comments":[{"comment":"In the displayed formula for ord_{P_k}(f_i), the argument of the third Bernoulli polynomial appears as B_2({α^{i+k}/2}); the denominator should presumably be p, matching the surrounding terms. Please correct this typo.","section":"§6.1"},{"comment":"In the paragraph discussing p = 37, the line 'C1(11)(Q) = ⟨[D′]⟩' should almost certainly read 'C1(37)(Q) = ⟨[D′]⟩'; as written it is inconsistent with the surrounding examples.","section":"§7.2"},{"comment":"The determinant calculation for the matrix B would be much easier to check if the authors stated the dimensions of A and B, the ordering of the rows corresponding to the cusps P_0,…,P_{n−1}, Q_0,…,Q_{n−2}, and the ordering of the columns corresponding to G_0,…,G_{n−1}, H_1,…,H_{n−1}. As printed, the block matrix requires a considerable amount of inference.","section":"§5"},{"comment":"Reference [26] is listed as 'Modular Units and Cuspidal Divisor Class Groups of X1(N), 2007' without journal or article numbers; please supply the full bibliographic data.","section":"References"},{"comment":"There are several typographical errors that should be fixed in revision, including 'divisor s' in the abstract, 'representaives' in Lemma 10, 'the the above inequalities' near the end of Section 6, and 'In this this section' at the start of Section 6.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The main result, Theorem 1, is the real content and it holds up. You get an explicit basis of p-2 Siegel-function products for the modular units on X1(p) modulo constants. The proof that the divisor matrix has determinant equal to the Takagi-Yu cuspidal class number is direct and checkable; the row/column manipulations in the determinant computation are sketched but the ingredients (Lemma 16, the block-matrix argument) are standard and credible. This converts cuspidal group computation for X1(p) into Smith normal form of a known matrix, which is a genuine new tool. Theorems 2 and 3 are natural extensions and their short proofs do what they claim: compare orders of vanishing with Yang's basis and use the function I_n with simple zeros at the Q-cusps. The Section 8 formulas for the orders of P0-Q0 and P0-P_{n-1} look plausible; I have not verified every matrix inversion but the method is standard.\n\nThe soft spot is Proposition 19. The paper asserts that C^Q_1(p) = C_1(p)(Q), i.e., every Galois-invariant cuspidal class comes from a Galois-invariant divisor, by claiming H^1(Gal, div(F(p))) = 0 and reducing to the norm-kernel equality {D in div(F(p)) : N D = 0} = (1-sigma) div(F(p)). The text says this is 'clear' from the cyclic permutation of the Q_i. That is not a formal consequence: div(F(p)) is a proper finite-index sublattice of Div_c(p), and for proper sublattices the norm-kernel can strictly contain (1-sigma)M. The cyclic action alone does not buy the equality; you need the explicit shape of div(F(p)) (which Theorem 1 supplies, but the paper does not do the verification). Until that step is filled in, Tables 2-3 and the assertion that the rational cuspidal group is generated by P-cusp divisors are conditional. This does not touch Theorem 1 or the C_1(p) computations.\n\nI do not see circularity: the determinant comparison is against the known cuspidal class number, not the target result. Citations to Takagi, Yu, and Yang are appropriate.\n\nVerdict: worth refereeing; Theorem 1 is a clean, valuable result. Send it out, but insist that Proposition 19 either get a real proof or be demoted to a conjecture with the rational subgroup claims flagged accordingly.","headline":"Solid explicit basis for modular units on X1(p); the main theorem holds, but Proposition 19's cohomology step is a real gap that the rational-cuspidal claims ride on.","tokens_in":27284,"tokens_out":5675,"would_cite":true,"duration_ms":54112,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11G16","11G18"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that the p-2 functions G0,...,G_{n-1},H1,...,H_{n-1} form a basis for modular units on X1(p) modulo constants, making cuspidal group computation a linear-algebra exercise.","keywords":["modular units","cuspidal group","Siegel functions","modular curves","X1(p)","cuspidal class number","Jacobian","Smith normal form"],"falsifier":"For a specific prime, such as p=11, compute the Smith normal form of the integer matrix whose columns are the divisors of G0,...,G_{n-1},H1,...,H_{n-1}; the invariant factors must reproduce C1(11)≈Z/25Z. An independent computation of the cuspidal group by another method returning a different group for any prime would refute Theorem 1.","tokens_in":1801,"feed_emoji":"🔢","tokens_out":1919,"duration_ms":98108,"temperature":0.7,"pith_summary":"The paper establishes an explicit basis for the group of modular units on X1(p), for prime p≥5: with n=(p-1)/2, the p-2 functions G0,...,G_{n-1},H1,...,H_{n-1} generate the group up to constants. Since modular units are precisely the functions whose zeroes and poles lie on the cusps, a basis converts the problem of computing the cuspidal group C1(p), the subgroup of the Jacobian generated by differences of cusps, into a Smith normal form computation for a divisor matrix. The paper also gives explicit bases for the subgroups of units whose divisors are supported on the rational cusps and of units with Galois-fixed divisors, and proves these two subgroups are equal to the rational cuspidal subgroup. A reader should care because, despite the cuspidal class number being known, the additive structure of the cuspidal group had remained largely inaccessible computationally; this basis makes it concrete.","feed_headline":"Explicit basis found for modular units on X1(p)","feed_subtitle":"The p minus 2 functions reduce cuspidal group computation to Smith normal form.","key_machinery":"The load-bearing construction is a fixed integer α of multiplicative order p-1 modulo p, which labels the 2n cusps of X1(p) as P_i=α^i/p and Q_i=p/α^i, for 0≤i<n, and defines the Siegel-function products E_i=∏_{j=0}^{p-2} g_{α^i/p,α^j/p} and F_i=g_{0,α^i/p}. Products f=∏ $E_i^{{e_i}}$ $F_i^{{f_i}}$ are modular units on X1(p) exactly when the exponents satisfy three congruence conditions (Proposition 9). The proof of the basis factors the divisor matrix as M=AB; the determinant of B is evaluated through circulant-matrix eigenvalue calculations using Bernoulli numbers B_{2,χ}, and |det M|=$p^{2}$|∏ B_{2,χ}|^2 is matched with the known cuspidal class number.","core_discovery":"The central claim is Theorem 1: the p-2 functions G0,...,G_{n-1},H1,...,H_{n-1}, constructed as products of Siegel functions attached to a primitive root α modulo p, form a basis for F(p)/C^×, the group of modular units on X1(p) modulo constants. The proof computes the determinant of the matrix representing the divisors of these functions and shows it equals the known cuspidal class number h1(p)= (p ∏_χ (1/4) B_{2,χ})^2, where χ runs over even nonprincipal Dirichlet characters modulo p. Thus the index of the subgroup generated by these divisors in the degree-zero cuspidal divisor group is exactly the order of the cuspidal group, so the listed functions cannot form a proper sublattice. The paper further proves bases for F^∞(p) (divisors supported on the rational cusps) and F^Q(p) (Galois-fixed divisors), and shows the corresponding quotient groups all coincide with the rational cuspidal subgroup C1(p)(Q).","pith_inferences":["The circulant structure of the divisor matrix suggests the same explicit-basis strategy could extend to prime powers p^n or other congruence subgroups, provided a cuspidal class number formula is available; the paper does not claim such an extension.","The numerical prominence of [P0-Q0] and [P0-P_{n-1}] as generators hints that a structural theorem about which cusp differences generate large factors of C1(p) may be waiting; that would be a natural follow-up.","One immediate testable consequence of the paper's method is that pushing the Smith normal form computation to primes above 997 should reproduce the known cuspidal class number formula exactly, and any discrepancy would signal either a flaw in the basis claim or a limitation of the numerical implementation."],"forward_implications":["Computing the abstract structure of C1(p) for a given prime reduces to taking the Smith normal form of an integer matrix of size p-2 whose entries are the orders of vanishing of the explicit basis functions at the cusps.","The rational cuspidal subgroup is generated by differences of the rational cusps P0,...,P_{n-1}, so its structure is accessible through the smaller basis I1,...,I_{n-1}.","Explicit formulae are obtained for the orders of [P0-Q0] and [P0-P_{n-1}] in J1(p), two elements that appeared as the largest cyclic factors in the computed examples.","Any choice of p-2 functions from the 2n candidates that includes G_{n-1} and H_{n-1} is still a basis, giving flexibility in presentations.","The computed tables give the full decomposition of C1(p) and C1(p)(Q) for every prime 11≤p≤997."],"supporting_citations":[{"why":"Supplies the description of modular units on X(N) as products of Siegel functions, which the constructions start from.","marker":"[9]"},{"why":"Provides the cuspidal class number formula for X1(p) that the determinant computation is compared against.","marker":"[22]"},{"why":"Supplies the order of the subgroup generated by the infinity-cusps, a factor in the class number formula.","marker":"[30]"},{"why":"Provides a basis for the analogous subgroup of units supported over the infinity-cusp and the strategy of using the known class number to verify a basis.","marker":"[26]"},{"why":"Proves finiteness of the cuspidal group, justifying that the quotient by unit divisors has the stated order.","marker":"[12]"},{"why":"Gives the complementary finiteness theorem for the cuspidal group.","marker":"[6]"}],"fun_headline_variants":["Explicit basis for modular units on X1(p) found","p-2 functions give complete modular unit basis","Basis of modular units simplifies cuspidal group study","Modular unit basis exposes large cuspidal quotient","Cuspidal group quotients from explicit unit basis"],"cache_read_input_tokens":29440,"weakest_assumption_plain":"The proof rests on a previously established formula for the size of the cuspidal group: if that formula were incorrect, the equal-index argument would not show that the constructed functions span all modular units.","fun_headline_variants_meta":{"raw":{"variants":["Explicit basis for modular units on X1(p) found","p-2 functions give complete modular unit basis","Basis of modular units simplifies cuspidal group study","Modular unit basis exposes large cuspidal quotient","Cuspidal group quotients from explicit unit basis"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000327,"raw_usage":{"total_tokens":1799,"prompt_tokens":887,"completion_tokens":912,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":503,"completion_tokens_details":{"reasoning_tokens":831}},"tokens_in":503,"tokens_out":912,"duration_ms":8504,"temperature":1.0,"reasoning_tokens":831,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T23:36:20.583960+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For a specific prime, such as p=11, compute the Smith normal form of the integer matrix whose columns are the divisors of G0,...,G_{n-1},H1,...,H_{n-1}; the invariant factors must reproduce C1(11)≈Z/25Z. An independent computation of the cuspidal group by another method returning a different group for any prime would refute Theorem 1.","supporting_citations":[{"cited_title":"1, 3, 3.1, 3.1, 3.2","cited_arxiv_id":null,"evidence_quote":"Supplies the description of modular units on X(N) as products of Siegel functions, which the constructions start from."},{"cited_title":"2, 348–374","cited_arxiv_id":null,"evidence_quote":"Provides the cuspidal class number formula for X1(p) that the determinant computation is compared against."},{"cited_title":"1, 5, 5 Depar tment of Ma thema tics, University College London, Lon don, WC1H 0AY, UK Email address : e.lupoian@ucl.ac.uk 25","cited_arxiv_id":null,"evidence_quote":"Supplies the order of the subgroup generated by the infinity-cusps, a factor in the class number formula."},{"cited_title":"1, 5, 5, 6.1","cited_arxiv_id":null,"evidence_quote":"Provides a basis for the analogous subgroup of units supported over the infinity-cusp and the strategy of using the known class number to verify a basis."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Proves finiteness of the cuspidal group, justifying that the quotient by unit divisors has the stated order."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the complementary finiteness theorem for the cuspidal group."}],"review_version":1}