{"id":"d2b1eace-689b-4c33-8f24-3985a884b428","arxiv_id":"2607.27282","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A mostly expository paper on the exceptional set of the binary Goldbach problem, appending a smoothed fully explicit major-arc formula and a conditional theorem that a sparse Goldbach-type bound excludes Siegel zeros.","lead":"This paper surveys how number theorists proved that almost all even integers are sums of two primes, and adds two new results: a smoothed 'fully explicit' formula for the main arc contribution, and a conditional proof that a sparse Goldbach-type hypothesis forbids a notorious 'exceptional zero'. A generalist might care because the second result tightens the known web showing such a zero would make prime-sum counts oscillate wildly.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition 7.5, the paper's claimed fully explicit formula, is stated without a proof: convergence and truncation of the infinite zero sums in (7.15)-(7.16) are never established.","rationale":"I read the paper in good faith and found the survey sections (2–6) accurate and informative, as the reader did. The internal arithmetic in the sketches of Section 7 and 8 checks out up to the level of detail given. However, the most load-bearing gap is that Proposition 7.5, the paper's new 'fully explicit formula', is not actually proved. The missing step is not a minor technicality: it is the control of the very objects that make the formula explicit—the infinite sums over zeros in (7.15)-(7.16). The paper provides decay estimates for a single archimedean factor (Lemma 7.6), but never shows that these decay estimates survive summation over characters, over q, and over pairs of zeros, nor that the truncation error is uniformly bounded by the right-hand side of (7.17). Without this, the proposition is a formal identity whose convergence status is unknown. This concern does not by itself prove the result false; it means the paper's central new result is not established as written. Theorem 8.2 is a separate conditional statement that also rests on unproved imports from Pintz's paper, which the reader flagged; I agree that this is a secondary but real risk. The appropriate verdict remains CONDITIONAL: the paper would be acceptable if the missing convergence and truncation estimates are supplied (and if the imports from [20] are verified), but not in its current form. I do not see grounds for outright rejection, as the survey part is solid and the sketches are plausible.","tokens_in":26502,"tokens_out":12219,"duration_ms":116789,"concrete_test":"Prove a uniform tail bound for the infinite sums in (7.15)-(7.16) by summing Lemma 7.6 over the family of characters and zeros: show that for every ε>0 the contribution of zeros with |Im ρ|>T to M(N;R)+Z(N;R) is o(X^{3−ϑ} log^5 X) uniformly in N≤X, or exhibit a counterexample. A computationally cheaper surrogate: fix R=10^3, choose a generic N, and compute the truncated sums M_T and Z_T for increasing T to check convergence and whether the residual matches the claimed bound. If the sums do not converge or the truncation error exceeds the claimed size, Proposition 7.5 is unsubstantiated.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's first headline new result, Proposition 7.5 (Eq. 7.17), is asserted with 'In conclusion we obtain the following proposition' after formal manipulations. The quantities M(N;R) and Z(N;R) in (7.15)-(7.16) are infinite sums over all zeros of every L(s,χ) with conductor ≤R. For the identity to be meaningful, these sums must converge (at least in a specified order) and the error from truncating the zero sums must be controlled uniformly for N≤X. Lemma 7.6 bounds a single archimedean factor I^φ_q, but the paper does not sum this bound over the family of primitive characters, over q≤R, and over all zeros ρ1,ρ2. In particular, the implied constants in Lemma 7.6 depend on the C^∞ seminorms of the cutoffs, and the number of characters of conductor r grows like r, while the decay in (7.19) is only polynomial in (1+(|Im ρ1|+|Im ρ2|)T_q/N) with T_q/N ≤ R^{-5/4}. For q near R the cutoff in |Im ρ| is at most R^{5/4}, so the tail contains O(R^{5/4} log R) zeros per character; no estimate shows the product of the Gauss-sum factors and the archimedean factor is summable to give O(X^{3−ϑ} log^5 X) after squaring. Thus the central explicit-formula result is unsupported. This is load-bearing because Proposition 7.5 is presented as the paper's main new formula, and the survey's narrative relies on it. A secondary concern, already noted by the reader, is that Proposition 8.1 (and hence Theorem 8.2) imports three quantitative facts from Pintz [20] without proof; if any of these is misstated, the oscillation argument collapses. The manuscript flags in §7.5 that the saving 'is not very strong', but it does not flag the missing convergence proof.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper is a survey of the exceptional set in the binary Goldbach problem, covering the Hardy–Littlewood circle method, Siegel–Walfisz and Vinogradov, the Montgomery–Vaughan power saving, and Pintz's refinement. It then claims two new results: Proposition 7.5, a mean-square explicit formula for a smoothed Goldbach count r_φ(N) in terms of N S(N) plus explicit sums M(N;R), Z(N;R) over all zeros of all Dirichlet L-functions of conductor at most R, with error O((X^{3−ϑ}+X^{13/5})(log X)^5); and Theorem 8.2, asserting that a sparse Hardy–Littlewood-type lower and upper bound on r_2(N) along multiples of r̃ rules out a real zero β̃ > 1 − c/log X for the corresponding quadratic character. The survey portion is informative and largely accurate; the new results are not yet supported by the arguments as written.","tokens_in":26739,"tokens_out":14505,"duration_ms":131380,"significance":"If Proposition 7.5 and Theorem 8.2 could be made fully rigorous, they would be interesting contributions: the former would provide a smooth major-arc formula keeping all zeros explicit, and the latter would show that a very sparse Goldbach condition at conductor-multiples is incompatible with a Siegel zero. The survey component is genuinely useful and well organized, and the paper correctly credits the earlier literature. However, the new proofs contain load-bearing gaps in exactly the places where the claims go beyond existing results, so the current version is not ready for publication as a research article.","major_comments":[{"comment":"The proof of the second-moment bound uses Parseval to replace ∑_{N≤X} |∫ S_φ(α)^2(1−b̂_R(α))e(−Nα)dα|^2 by ∫ |S_φ|^4 |1−b̂_R|^2 dα. But S_φ(α)=∑ Λ(n)φ(n/N)e(nα) is defined with the same N that is being summed, so these integrals are not Fourier coefficients of a single fixed function. The displayed Parseval identity is therefore invalid. Relatedly, Lemma 7.2 requires 10R^2≤ηN, which cannot hold for the small N in the range ∑_{N≤X}; this part of the range is not controlled. This invalidates the derivation of (7.8) and hence Proposition 7.5.","section":"§7.2, Lemma 7.3 (Eq. 7.8)"},{"comment":"Lemma 7.4 is quoted as 'standard' with no proof or reference, but the application requires more than absolute convergence for a single fixed test function. The sums M(N;R) and Z(N;R) are infinite over all zeros of every primitive character of conductor ≤R. Lemma 7.6 bounds one archimedean factor with decay (1+(|Im ρ1|+|Im ρ2|)T_q/N)^{−A}, and T_q/N ≤ R^{−1/4}; it is never summed over the roughly R^2 characters and over the relevant zeros. In particular, zeros with |Im ρ| of size R^{1/4} are not damped, and no estimate shows that the double zero sums contribute within the claimed O((X^{3−ϑ}+X^{13/5})(log X)^5). The central explicit-formula result is therefore unsupported.","section":"§7.4, Lemma 7.4 and Prop 7.5 (Eqs. 7.15–7.17)"},{"comment":"Theorem 8.2 rests on Proposition 8.1, whose proof sketch imports three quantitative facts from [20] without exact statements or verification: the Deuring–Heilbronn range (8.2), the closed form S(χ̃,χ̃,N)=χ̃(−1)S(N) in (8.1), and the error terms in (8.3), including O(S(N)X(δ̃L)^{c1}). These are load-bearing because the final oscillation argument needs the secondary main term to have the stated size and sign after subtracting the errors. If any of these inherited estimates fails in the stated ranges, the contradiction with (8.4) collapses. Please either prove these facts or quote exact theorem/lemma numbers with the hypotheses needed here.","section":"§8, Prop 8.1 and Thm 8.2"}],"minor_comments":[{"comment":"The text refers to 'Pintz's first paper [20]' and then to 'Pintz's second paper [20]'; one of these should presumably be [21]. The reference labels in the introduction also alternate between [20] and [21] in a confusing way.","section":"§6.2"},{"comment":"The phrase 'fully explicit formula' is stronger than what is proved: (7.17) is a mean-square bound for a smoothed count, with an O(error) and with infinite zero sums whose convergence is not established. The wording should reflect this.","section":"Abstract and §7"},{"comment":"There is a typo: 'pbRep' should be 'b̂_R'.","section":"Eq. (7.10)"},{"comment":"The text contains many typographical artifacts, such as missing words ('zeros', 'contribute', 'bad') and malformed mathematical expressions. These should be corrected in a final version.","section":"General presentation"}],"recommendation":"major_revision","confidential_remarks":"The survey part of this paper is solid and could be published with relatively little additional work. The new results, however, are not in a publishable state: the Parseval step in §7.2 is not valid as written, and the convergence/uniformity of the zero sums in Proposition 7.5 is not established. Theorem 8.2 may be salvageable, but it depends on several unverified imported estimates. I would recommend inviting a revision that either proves Proposition 7.5 properly or clearly separates the survey from the new results and substantially rewrites the proof of the new claims."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: read Sections 2–6 as a survey; treat Section 7 and 8 as announcements with sketches. The historical account checks out, and the arithmetic in the new sections is correct as far as it goes. But the paper's first new result is presented without the convergence analysis its own formulas require.\n\nWhat's genuinely new: Theorem 8.2 is a real interpolation between the dense-set results (Fei, Bhowmik–Halupczok, Jia, Goldston–Suriajaya) and Matomäki–Merikoski's single-multiple test. The scale N ~ r̃^A with A > 5/2 and a power-sized exception set is not in the literature, and the oscillation argument—showing a real zero forces r_2(N) to flip between ~0 and ~2S(N)N on multiples of the conductor—is clean and not circular. Proposition 7.5 would also be a notable formula if fully proved.\n\nSoft spots. The biggest one is the convergence and truncation of the infinite zero sums in (7.15)–(7.16). The paper asserts Lemma 7.4 with \"converges absolutely\" and then writes down Proposition 7.5 by fiat. Lemma 7.6 bounds a single archimedean factor, but that does not sum over all characters and all zeros to the stated X^{3−ϑ} + X^{13/5} mean-square error. The stress-test note is right: this is load-bearing, and the paper itself only flags that the saving \"is not very strong\"—not that the tail is uncontrolled. Theorem 8.2 also rides on three quantitative facts lifted from Pintz [20]: the Deuring–Heilbronn range, the S(χ̃,χ̃,N) evaluation, and the error terms. These are probably fine, but they are black boxes and a referee needs to check the stated ranges.\n\nThe survey is genuinely good. Sections 2–6 give an accurate, well-referenced path from Hardy–Littlewood to Montgomery–Vaughan to Pintz; I checked several attributions and the derivations that are included (e.g., the η=R^{-1/4} optimization) and they are correct. That part deserves publication on its own.\n\nWho it's for: a graduate student wanting a map of the Goldbach exceptional-set literature will get a lot out of Sections 2–6. A specialist will want Section 8 but should demand a complete proof of Proposition 7.5 or a revised statement.\n\nRecommendation: don't desk-reject. Send it out; the survey is publishable and the new ideas justify referee time. The referee report should insist on fixing the convergence issue and on verifying the Pintz imports before the new theorems are accepted.","headline":"The survey half of this paper is genuinely good; the two appended results are credible but under-supported—Proposition 7.5 lacks a proof of convergence for its zero sums, and Theorem 8.2 depends on unverified black-box imports from Pintz.","tokens_in":27596,"tokens_out":2401,"would_cite":true,"duration_ms":27003,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11P32","11M26","11M41"],"pacs":[],"model":"deepseek-v4-flash","headline":"A sparse Hardy–Littlewood bound on Goldbach representations rules out Siegel zeros, and a new smoothed explicit formula exposes every L-function zero in the major arcs.","keywords":["Goldbach problem","exceptional set","Hardy-Littlewood conjecture","circle method","Siegel zeros","Dirichlet L-functions","generalized singular series","explicit formula"],"falsifier":"Test the configuration directly: fix a primitive real character χ̃ mod r̃ and suppose it has a real zero β̃ > 1 − c/log X with X = r̃^A, A > 5/2. Compute r_2(N) for all even multiples N of r̃ in [X/2,X] and check whether δS(N)N ≤ r_2(N) ≤ (2−δ)S(N)N holds for all but X^{3/5} of them. If the bounds hold, Theorem 8.2 asserts the zero cannot exist; if the bounds fail, the theorem is moot. A more targeted check: verify the oscillation pattern r_2(N) ≈ (1 + χ̃(−1)B(β̃,β̃)N^{−2δ'})S(N)N across multiples of r̃, which the proof of Theorem 8.2 predicts must occur whenever β̃ > 1 − c/log X.","tokens_in":26116,"feed_emoji":"🧮","tokens_out":5134,"duration_ms":47269,"temperature":0.7,"pith_summary":"This paper is mostly a survey of the exceptional set in the binary Goldbach problem, tracing how the circle method evolved from Hardy–Littlewood through Montgomery–Vaughan to Pintz's power saving. Its first new contribution is a fully explicit major-arc formula: using a smooth major-arc weight, the smoothed count of representations of N as a sum of two primes is shown, in mean square over N ≤ X, to equal the singular-series main term plus an explicit sum over all zeros of all Dirichlet L-functions of conductor at most X^ϑ, with error (X^{3−ϑ}+X^{13/5})(log X)^5. Its second new contribution is a theorem: if a sparse form of the Hardy–Littlewood conjecture holds for even multiples of a conductor r̃ at scale X = r̃^A with A > 5/2, allowing only X^{3/5} exceptions, then the associated primitive real character has no real zero above 1 − c/log X. A sympathetic reader would care because the second result shows that a plausible average statement about prime sums, even with a power-sized exceptional set, is incompatible with the existence of exceptional zeros—one of the central obstructions in the subject.","feed_headline":"Rare Goldbach checks rule out exceptional zeros","feed_subtitle":"Even a power-sized exception set cannot hide a real zero near 1 if sparse Goldbach bounds hold.","key_machinery":"The load-bearing mechanism is the smooth major-arc weight b_R(n) = Σ_{q≤R} c_q(n) G_{T_q}(n), built from Ramanujan sums c_q(n) and a compactly supported cutoff G with scale T_q = η X q/R. Its Fourier transform is approximately an oscillatory sum of translated copies of a smooth window; it is 1 + O(η²) on major arcs and O(1) on minor arcs. This lets the smoothed exponential sum S_φ(α) be expanded, via character orthogonality and the explicit formula, term by term, producing the archimedean factors I_φ_q(ρ_1,ρ_2) and the generalized singular series S(χ_1,χ_2,N), and bringing all zeros of all L-functions of conductor ≤ R explicitly into the approximation. In Section 8, the same structure, combi","core_discovery":"The paper establishes two new results. First, Proposition 7.5: for R = X^ϑ with 0 < ϑ < 4/9, the smoothed Goldbach count r_φ(N) satisfies Σ_{N≤X} |r_φ(N) − N S(N) − M(N;R) − Z(N;R)|² ≪ (X^{3−ϑ} + X^{13/5})(log X)^5, where M and Z collect, explicitly, the contribution of every zero of every Dirichlet L-function of conductor at most R. Second, Theorem 8.2: if for some A > 5/2 and δ ∈ (0,1) the sparse Hardy–Littlewood bounds δS(N)N ≤ r_2(N) ≤ (2−δ)S(N)N hold for all but at most X^{3/5} even multiples N of r̃ in [X/2,X], with X = r̃^A, then no primitive real character χ̃ mod r̃ has a real zero β̃ > 1 − c/log X. The proof of the second result runs by contradiction: if such a zero existed, the Deu","pith_inferences":["The smooth-weight construction may be adaptable to Heath-Brown's variant of the circle method; the authors themselves note that such ideas could circumvent the interval shortening caused by η = R^{−1/4}, potentially extending the range of ϑ in Proposition 7.5.","Neighbouring results require only a single multiple of the conductor to exclude a Siegel zero, whereas Theorem 8.2 needs a power-sized family; the oscillatory mechanism here suggests the sparse hypothesis could be substantially relaxed, and the X^{3/5} exception threshold is likely an artifact of the current minor-arc technology rather than a fundamental barrier.","A concrete, testable prediction of Theorem 8.2 is that if a real zero β̃ > 1 − c/log X existed for a primitive character χ̃ mod r̃, then among the even multiples N of r̃ near X = r̃^A, the count r_2(N) would alternate between values near 0 and near 2S(N)N according to the sign χ̃(−1)—an oscillation that numerical computation at small r̃ and modest A could in principle detect.","Because the main term in Proposition 7.5 is a mean-square identity, it may be used to derive higher moments of the Goldbach representation function, potentially yielding exceptional-set bounds via distributional arguments rather than pointwise major-arc estimates."],"forward_implications":["If Proposition 7.5 is correct, the smoothed Goldbach count admits a mean-square approximation in which every zero of every L-function of conductor at most X^ϑ appears explicitly, with a power saving of the form (X^{3−ϑ} + X^{13/5})(log X)^5.","If Theorem 8.2 is correct, a sparse version of the Hardy–Littlewood conjecture—even one allowing X^{3/5} exceptions—would rule out Siegel zeros entirely, eliminating a major source of ineffectivity in analytic number theory.","The explicit formula generalizes Pintz's earlier finite-over-zero formula to an infinite sum over all zeros, with the smooth weight introducing an explicit archimedean factor B_φ(ρ_1,ρ_2) that reduces to the classical beta factor when the zeros are close to the real axis.","The restriction ϑ ≥ 2/5 in the proof of Theorem 8.2 is forced by the Vinogradov minor-arc bound, suggesting that improvements to that bound would directly improve the allowable scale A > 5/2 in the sparse Hardy–Littlewood hypothesis."],"fun_headline_variants":["Explicit major arcs formula for Goldbach problem","Sparse Goldbach bounds rule out exceptional zeros","No exceptional zeros if sparse Goldbach bounds hold","Goldbach: explicit major arcs, no exceptional zeros","Sparse Goldbach kills exceptional zeros"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The proof of Theorem 8.2 depends on imported quantitative estimates from Pintz's explicit formula—the Deuring–Heilbronn zero-spacing bound, the closed-form evaluation S(χ̃,χ̃,N) = χ̃(−1)S(N), and the error terms in (8.3)—which are cited without proof here; if any of these inherited estimates fails in the stated ranges, the oscillation argument that forces the contradiction collapses.","fun_headline_variants_meta":{"raw":{"variants":["Explicit major arcs formula for Goldbach problem","Sparse Goldbach bounds rule out exceptional zeros","No exceptional zeros if sparse Goldbach bounds hold","Goldbach: explicit major arcs, no exceptional zeros","Sparse Goldbach kills exceptional zeros"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000837,"raw_usage":{"total_tokens":3481,"prompt_tokens":736,"completion_tokens":2745,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":480,"completion_tokens_details":{"reasoning_tokens":2675}},"tokens_in":480,"tokens_out":2745,"duration_ms":19237,"temperature":1.0,"reasoning_tokens":2675,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T10:33:12.674431+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Test the configuration directly: fix a primitive real character χ̃ mod r̃ and suppose it has a real zero β̃ > 1 − c/log X with X = r̃^A, A > 5/2. Compute r_2(N) for all even multiples N of r̃ in [X/2,X] and check whether δS(N)N ≤ r_2(N) ≤ (2−δ)S(N)N holds for all but X^{3/5} of them. If the bounds hold, Theorem 8.2 asserts the zero cannot exist; if the bounds fail, the theorem is moot. A more targeted check: verify the oscillation pattern r_2(N) ≈ (1 + χ̃(−1)B(β̃,β̃)N^{−2δ'})S(N)N across multiples of r̃, which the proof of Theorem 8.2 predicts must occur whenever β̃ > 1 − c/log X.","supporting_citations":[],"review_version":1}