{"id":"788c8e18-1e6e-49b3-887c-47b2551486a7","arxiv_id":"1908.02526","paper_version":2,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"For every n, the Brauer-Manin obstruction does not obstruct positive integer solutions to the Erdos-Straus equation, yet it gives a universal local condition (a product of Hilbert symbols equal to -1) for all such solutions.","lead":"This paper computes the Brauer-Manin obstruction for the affine cubic surfaces associated with the Erdos-Straus equation 4/n = 1/a + 1/b + 1/c. It shows this obstruction does not block natural-number solutions, but that it does explain a new product-of-Hilbert-symbols condition that all solutions must satisfy.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified; the central argument is sound for n≥2, with a minor n=1 edge-case error in Theorem 1.8.","rationale":"The reader's ACCEPT verdict is correct. I independently verified the arithmetic core: Theorem 1.2's proof rests on the real component sign (Lemma 3.1), the triviality at good odd primes (Lemma 3.5), and the 2-adic computation (Lemma 3.8), each of which is sound. The singular Brauer group result (Theorem 1.7) is used for the geometric interpretation on U_n, but Remark 2.11 explicitly provides a fallback on the desingularisation, so a failure there would not invalidate the arithmetic conclusions. I found one false edge case: n=1 in Theorem 1.8. This does not impact the Erdős–Straus conjecture, which concerns n≥2, but the theorem statement should be corrected to n≥2. The reader's weakest_assumption, while a legitimate point of scrutiny, is not the most load-bearing; the local symbol computations are. Therefore the verdict remains unchanged.","tokens_in":14143,"tokens_out":53698,"duration_ms":557902,"concrete_test":"Verify the n=1 case explicitly: take the adelic point with real component u_∞=(3/4,3/4,3/4) and u_p=(0,0,0) for every finite prime p, then compute the total invariant of α as inv∞(-1,-1)_∞ · ∏_p inv_p(α) = (-1)·1 = -1; if this holds, Theorem 1.8(1.4) is false for n=1 and the statement should be restricted to n≥2.","verdict_should_be":"UNCHANGED","load_bearing_attack":"I find no load-bearing concern for the central claim (Theorem 1.2 and Theorem 1.8 for n≥2). The proof of Theorem 1.2 is an explicit local computation of Hilbert symbols (Lemmas 3.1, 3.5, 3.8) plus reciprocity; I checked the valuations, the 2-adic table, and the good-prime argument and they are internally consistent. The reader's identified weak assumption, Theorem 1.7, is not load-bearing: even if the identification Br U_n = Br \\tilde U_n failed, Remark 2.11 supplies a direct construction on the desingularisation, and the local invariants are unchanged. The only concrete error I found is that Theorem 1.8(1.4) is false for n=1: for n=1, all finite local invariants of alpha are 1 and inv∞ on U_1(R)+ is -1, so every point of U_1(R)+ × ∏ U_1(Z_p) has total invariant -1 and the Brauer kernel is empty. This is outside the Erdős–Straus range n≥2 and does not affect the main results.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the affine cubic surface U_n defined by the Erdős–Straus equation 4u1u2u3 = n(u1u2+u1u3+u2u3), computes its Brauer group via a desingularisation, and determines the resulting Brauer–Manin set for integral points. The main theorems assert: (i) there is no Brauer–Manin obstruction to the existence of natural-number solutions to 4/n = 1/u1 + 1/u2 + 1/u3 (Theorem 1.1, made precise as non-emptiness of the Brauer set in Theorem 1.8(1.4)); (ii) for odd n, every natural solution satisfies an explicit product formula ∏_{p|n}(-u1/u3, -u2/u3)_p = -1 (Theorem 1.2), while non-natural integer solutions satisfy the opposite product = 1 (Theorem 1.5); and (iii) strong approximation for integral points fails, with a Brauer–Manin obstruction to strong approximation (Theorem 1.8(1.5)) but not every failure explained by Brauer–Manin (Theorem 1.9). The proofs go through a desingularisation of U_n, a computation of Br U_n via a new general result on Brauer groups of surfaces with rational singularities (Theorem 1.7), and explicit local Hilbert-symbol computations including a 2-adic case analysis (Lemma 3.8).","tokens_in":14369,"tokens_out":20767,"duration_ms":201090,"significance":"If correct, the paper is a significant contribution to both the arithmetic of log K3 surfaces and the Erdős–Straus problem. It provides one of the few computations of the Brauer group of a singular surface, exhibits a transcendental Brauer–Manin obstruction to strong approximation on a surface of arithmetic interest, and yields an explicit, falsifiable necessary condition (Theorem 1.2) that recovers and unifies earlier results of Yamamoto and Elsholtz–Tao. The main derivation is self-contained and checkable: the Brauer element is constructed from the defining equation, the local invariants are computed by explicit Hensel lifting and Hilbert-symbol formulae, and no parameters are fitted to data. Theorem 1.7 on rational surface singularities is of independent interest. The only concrete error I found is the n=1 edge case in Theorem 1.8(1.4), which lies outside the Erdős–Straus range n≥2 and is easily corrected.","major_comments":[{"comment":"The assertion (U_n(R)_+ × ∏ U_n(Z_p))^Br ≠ ∅ is false for n=1. For n=1, Lemma 3.5 applies to every odd prime (since p ∤ 2n) and Lemma 3.8 applies to p=2, so inv_p α = 1 for all p and all u ∈ U_1(Z_p); Lemma 3.1 gives inv_∞ α = −1 on U_1(R)_+. Hence for every point of U_1(R)_+ × ∏ U_1(Z_p) the product of local invariants is −1, so no point lies in the right kernel of the Brauer pairing and the Brauer set is empty. This contradicts (1.4). The error is an edge case outside the Erdős–Straus range n ≥ 2; the statement should be restricted to n ≥ 2, and the proof of Theorem 1.8 should be adjusted accordingly.","section":"Theorem 1.8, Eq. (1.4)"}],"minor_comments":[{"comment":"The abstract contains 'stro ng approximation' with an erroneous space; it should read 'strong approximation'.","section":"Abstract"},{"comment":"The table of cases in Lemma 3.8 is asserted without explanation; the rows for (r1,r2) ≡ (5,1) and (7,3) mod 8 are omitted by symmetry, which should be stated, and a brief derivation of how the entries follow from equation (3.4) would aid the reader.","section":"Lemma 3.8"},{"comment":"The first sentence of the proof, 'as U_n(Z)≠∅ and n>0, we have (3.5)≠∅', is terse: for n≥2 one should explicitly use the known integer solution for the p-adic components and the positive real solution u_i = 3n/4 for the real component.","section":"Proof of Theorem 1.8"}],"recommendation":"minor_revision","confidential_remarks":"The n=1 counterexample to Theorem 1.8(1.4) is a local edge-case error and does not affect the paper's central results for n ≥ 2; I would accept after this correction. The worry about the reliance on Theorem 1.7 is mitigated by Remark 2.11, which provides a direct route on the desingularisation. I saw no issues with attribution or fit with the journal."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First thing to know: this is a solid paper, not a breakthrough. It shows that the Brauer-Manin obstruction does not rule out natural solutions to Erdos-Straus, but does obstruct strong approximation. The genuinely new pieces are Theorem 1.2, a universal Hilbert-symbol condition for all odd n, and Theorem 1.7, a surjectivity result for Brauer groups of rational singularities. The paper also computes the full Brauer group of the singular surface, including the transcendental part; that is rare and gives the paper independent value. The derivation is honest: the quaternion algebra is constructed from the equation, and the local computations are explicit. The authors are careful to note where they rely on existing rational-singularity machinery (Theorem 1.7) and supply a fallback via the desingularisation in Remark 2.11, so the central claim does not depend on the most delicate identification. I checked the reciprocity step and the valuation arguments; they hold.\n\nThe only concrete error I find is small and in the statement of Theorem 1.8: for n=1, equation (1.4) is false. The stress-test note is right that for n=1 the Brauer kernel is empty because the invariant at infinity is -1 and all finite invariants are 1. The Erdos-Straus conjecture starts at n>=2, and Theorem 1.8 for n>=2 is correct. This should be a one-line correction in a final version (exclude n=1 or adjust the statement). Also, some local invariant tables in Lemma 3.8 are asserted after a case check rather than shown in full; this is a minor exposition issue, not a gap. The dependence on external results for Theorem 1.7 is real but not load-bearing, given Remark 2.11. The paper cites prior work appropriately, including the Yamamoto and Elsholtz-Tao results it recovers.\n\nWho is this for? Arithmetic geometers working on Brauer-Manin and integral points, and number theorists interested in Erdos-Straus. The paper deserves a serious referee. I would accept it for peer review with a request to fix the n=1 statement. The proof of Theorem 1.7 is a bit condensed but the argument is standard. Overall, the paper is exactly what a good arithmetic geometry paper should look like: it does not overclaim, the main theorem is explicit and checkable, and the unified derivation of known conditions is a real service.","headline":"Solid Brauer-Manin analysis of Erdos-Straus surfaces: new necessary condition, full Brauer group computation, and a minor n=1 edge case in Theorem 1.8.","tokens_in":14883,"tokens_out":2392,"would_cite":true,"duration_ms":24018,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14G05","11D68","11D25","14F22"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that the Brauer–Manin obstruction leaves the Erdős–Straus conjecture open but blocks strong approximation.","keywords":["Erdős–Straus conjecture","Brauer–Manin obstruction","strong approximation","Hilbert symbol","unit fractions","log K3 surfaces","rational singularities","quaternion algebras"],"falsifier":"Enumerate all natural-number solutions for a fixed odd composite $n$, such as $n=45$ or $n=225$, and compute $\\prod_{p\\mid n}(-u_1/u_3,-u_2/u_3)_p$ using the Hilbert-symbol laws in the paper; a single tuple with product $+1$ would refute Theorem 1.2, while a computation showing the Brauer–Manin set on the positive real component is empty would refute Theorem 1.8.","tokens_in":13963,"feed_emoji":"➗","tokens_out":13502,"duration_ms":123200,"temperature":0.7,"pith_summary":"The Erdős–Straus conjecture asks whether every rational number $4/n$ can be written as $1/u_1+1/u_2+1/u_3$ with positive integers $u_i$. The paper approaches this through the affine cubic surface $U_n$ cut out by $4u_1u_2u_3=n(u_1u_2+u_1u_3+u_2u_3)$ and proves that its Brauer group is generated over $\\mathbb{Q}$ by a single quaternion algebra. Using that algebra, it shows that no Brauer–Manin obstruction prevents natural-number solutions from existing, yet a Brauer–Manin obstruction does prevent strong approximation: for odd $n$, every natural solution satisfies $\\prod_{p\\mid n}(-u_1/u_3,-u_2/u_3)_p=-1$, while signed solutions give the opposite sign. The explicit product recovers several previously known necessary conditions as special cases. If the paper is right, the Erdős–Straus conjecture can only be proved by a mechanism genuinely different from a Brauer–Manin obstruction.","feed_headline":"Brauer-Manin blocks approximation but not Erdős-Straus solutions","feed_subtitle":"One quaternion algebra forces a Hilbert-symbol product of -1 over the primes dividing n for every natural solution.","key_machinery":"The load-bearing object is the quaternion algebra $\\alpha=(-u_1/u_3,-u_2/u_3)$, whose local evaluation is the Hilbert symbol $(-u_1/u_3,-u_2/u_3)_v$. The identity (2.5), $-\\frac{u_i}{u_j}=1/(1+u_j/u_k-4u_j/n)$, is what lets $\\alpha$ descend to $\\mathbb{Q}$ by killing the residues along the boundary lines. The identification $\\operatorname{Br}U_n\\cong\\operatorname{Br}\\widetilde{U}_n$ comes from a general theorem for surfaces with rational singularities, so the entire obstruction reduces to a single reciprocity product over the primes dividing $n$.","core_discovery":"On the singular affine surface $U_n$, the paper computes $\\operatorname{Br} U_n/\\operatorname{Br} \\mathbb{Q} \\cong \\mathbb{Z}/2\\mathbb{Z}$, generated by the quaternion algebra $\\alpha=(-u_1/u_3,-u_2/u_3)$, and shows that this class is transcendental rather than algebraic. The local invariant of $\\alpha$ at a place $v$ is the Hilbert symbol $(-u_1/u_3,-u_2/u_3)_v$, and global Hilbert reciprocity turns a global point into a single product over primes. Evaluating the invariant at infinity separates $U_n(\\mathbb{R})_+$, where all coordinates are positive, from the other real component. The result is an exact dichotomy: natural solutions have product $-1$ over the primes dividing $n$, non-natural integer solutions have product $1$, and the Brauer–Manin set for natural solutions is always non-empty but strictly smaller than the full relevant adelic set. In particular, the paper establishes new cases of a transcendental Brauer–Manin obstruction on a log K3 surface.","pith_inferences":["The explicit $-1$ product could be used as a fast sieve: search algorithms for unit-fraction decompositions can discard any tuple failing the product identity before more expensive arithmetic checks.","Because all $U_n$ are $\\mathbb{Q}$-isomorphic to $U_1$, the sharp difference between models suggests that strong approximation for this class of affine surfaces depends delicately on the chosen integral model; analogous rescalings of other surfaces might be worth testing.","The general theorem on Brauer groups of surfaces with rational singularities has potential use elsewhere: an affine surface whose desingularisation has a computable Brauer group can inherit transcendental classes without a separate purity calculation at the singular locus."],"forward_implications":["The Erdős–Straus conjecture cannot be settled by a Brauer–Manin obstruction; any successful proof must find an obstruction or descent mechanism outside this framework.","For odd $n$, every natural solution must satisfy the explicit Hilbert-symbol product condition $\\prod_{p\\mid n}(-u_1/u_3,-u_2/u_3)_p=-1$, a new local necessary condition that can be checked prime by prime.","Known necessary conditions for solutions, including those for odd primes and odd squares, follow uniformly from the Brauer–Manin formalism rather than from ad hoc reciprocity computations.","The failure of strong approximation is genuine but incomplete: rational points are not dense in the Brauer–Manin set, so further obstructions beyond Brauer–Manin must control strong approximation on these surfaces."],"supporting_citations":[{"why":"Gives the Hilbert-symbol formulas and global reciprocity law that turn the Brauer element into the explicit product in Theorem 1.2.","marker":"[21]"},{"why":"Provides the Brauer–Manin pairing and local-invariant setup used to define the obstruction on adelic points.","marker":"[20]"},{"why":"Supplies the rational-singularity theory used to prove that the Brauer group of the desingularisation maps onto that of the singular surface.","marker":"[16]"},{"why":"Provides Grothendieck purity and the vanishing of Brauer groups of $\\mathbb{P}^1$, used to identify transcendental classes on the desingularisation.","marker":"[9]"},{"why":"Computes the Brauer group of $\\mathbb{G}_m^2$, giving the explicit cyclic-algebra description of the transcendental Brauer group.","marker":"[4]"},{"why":"Supplies the lemma on dense images of rational points used to show that the Brauer–Manin obstruction does not explain all failures of strong approximation.","marker":"[6]"}],"fun_headline_variants":["Brauer-Manin forces product -1 for natural Erdős-Straus","Erdős-Straus: Brauer-Manin exacts a product dichotomy","Transcendental Brauer-Manin splits Erdős-Straus solutions by sign","Brauer-Manin selects natural Erdős-Straus points exactly"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the Brauer group of the singular surface $U_n$ is faithfully captured by the Brauer group of its desingularisation; if that identification failed, the Hilbert-product identity would still make sense on the smooth model but would no longer be an obstruction on $U_n$ itself.","fun_headline_variants_meta":{"raw":{"variants":["Brauer-Manin forces product -1 for natural Erdős-Straus","Erdős-Straus: Brauer-Manin exacts a product dichotomy","Transcendental Brauer-Manin splits Erdős-Straus solutions by sign","Brauer-Manin selects natural Erdős-Straus points exactly"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000974,"raw_usage":{"total_tokens":4058,"prompt_tokens":780,"completion_tokens":3278,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":396,"completion_tokens_details":{"reasoning_tokens":3194}},"tokens_in":396,"tokens_out":3278,"duration_ms":27755,"temperature":1.0,"reasoning_tokens":3194,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:41:17.876217+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Enumerate all natural-number solutions for a fixed odd composite $n$, such as $n=45$ or $n=225$, and compute $\\prod_{p\\mid n}(-u_1/u_3,-u_2/u_3)_p$ using the Hilbert-symbol laws in the paper; a single tuple with product $+1$ would refute Theorem 1.2, while a computation showing the Brauer–Manin set on the positive real component is empty would refute Theorem 1.8.","supporting_citations":[{"cited_title":"Serre, A course in arithmetic , Graduate Texts in Mathematics, No","cited_arxiv_id":null,"evidence_quote":"Gives the Hilbert-symbol formulas and global reciprocity law that turn the Brauer element into the explicit product in Theorem 1.2."},{"cited_title":"Poonen, Rational points on varieties, Graduate Studies in Mathematics, vol","cited_arxiv_id":null,"evidence_quote":"Provides the Brauer–Manin pairing and local-invariant setup used to define the obstruction on adelic points."},{"cited_title":"Lipman, Rational singularities, with applications to algebraic surfaces and unique fac- torization","cited_arxiv_id":null,"evidence_quote":"Supplies the rational-singularity theory used to prove that the Brauer group of the desingularisation maps onto that of the singular surface."},{"cited_title":"Grothendieck, Le groupe de Brauer, III","cited_arxiv_id":null,"evidence_quote":"Provides Grothendieck purity and the vanishing of Brauer groups of $\\mathbb{P}^1$, used to identify transcendental classes on the desingularisation."},{"cited_title":"Colliot-Thélène, A","cited_arxiv_id":null,"evidence_quote":"Computes the Brauer group of $\\mathbb{G}_m^2$, giving the explicit cyclic-algebra description of the transcendental Brauer group."},{"cited_title":"Colliot-Thélène, D","cited_arxiv_id":null,"evidence_quote":"Supplies the lemma on dense images of rational points used to show that the Brauer–Manin obstruction does not explain all failures of strong approximation."}],"review_version":1}