{"id":"c48db5f4-78d0-42bf-b4fc-02c926d8a2ad","arxiv_id":"2501.16723","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"There are ≫ x/(log x)^{5/2} primes p ≤ x of the form m^2+n^2+1 with Ω(p+2) ≤ 11.","lead":"This paper proves that there are infinitely many primes p that can be written as m^2+n^2+1 and for which p+2 has at most 11 prime factors. It demonstrates a new vector sieve that combines a half-dimensional sieve with a linear sieve to handle a sparse prime subset.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The vector-sieve fundamental lemma (Prop 3.10) is the load-bearing step: it is only sketched for mixed half-dimensional and linear sieves with overlapping prime sets, and the leading coefficient of Proposition 4.1, hence the positivity of H in Section 5, depends on it.","rationale":"The paper aims at a genuine mixed-dimension vector sieve lower bound for primes p=m^2+n^2+1 with p+2 having bounded Ω. For that claim to hold, the two sieve estimates Propositions 4.1 and 4.3 must have correct leading constants, and the numerical H must be positive. The reader correctly identifies Proposition 3.10 as the weakest point: its proof is a sketch that delegates the core mixed-dimension main-term estimate to [8] without checking the overlap of the two prime sets or the half-dimensional/linear beta-sieve combination. I do not see an internal contradiction that would force a reject, and the final parameter choice θ1=0.449>1/3 avoids some of the secondary issues in Proposition 4.2. The inconsistent z0 exponents and the lack of rigorous error control in the Matlab search are secondary and likely repairable. However, because the leading constant in the lower bound is exactly what must be positive, the unproved vector-sieve proposition is the single most load-bearing concern. An independent re-derivation of (3.14) for the concrete h and prime sets would settle whether the claimed numerical positivity is trustworthy.","tokens_in":22482,"tokens_out":29910,"duration_ms":269906,"concrete_test":"Independently derive the main term of Proposition 3.10 for the specific axiom data h(d1,d2)=1/φ(d1d2) with the parity conditions of (4.3), P1={p≡3 mod 4}, P2=P, κ1=1/2, κ2=1, by expanding the three sums Σ1,Σ2,Σ3 in the lower-bound proof and evaluating the constrained optimization in (3.14). Then recompute H(0.14,0.23,0.449,0.011) using the resulting f((2θ1)^{-1},(2θ2)^{-1}) and F((1-2α)/(2θ1),(1-2α)/(2θ2)). If the independent computation agrees with the stated formulas, the concern is settled; if it differs by more than a few percent in the resulting H value, the positivity threshold in Section 5 is not supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Proposition 3.10 supplies both the lower bound in Proposition 4.1 and the upper bound in Proposition 4.3. If its main-term functions f(σ1,σ2) or F(σ1,σ2) are wrong by a factor 1+O(1) in the relevant range, the numerical margin H(0.14,0.23,0.449,0.011)=1.2471 in Section 5 is not robust: an O(1) error in the leading constant can change the sign of this coefficient. The proof of Proposition 3.10 is only a sketch. After the pre-sieving step with h*, it states that the sums Σ and errors E1,E2 can be handled as in [8, Section 3.3]. That reference is for a vector sieve of two linear sieves, whereas the present application uses κ1=1/2, κ2=1, with P1={p≡3 mod 4} and P2=all primes. For common primes p∈P1∩P2, axiom (A2) does not imply h(d1,d2)=h1(d1)h2(d2), so the lower-bound combination δ1^-δ2^+ + δ1^+δ2^- - δ1^+δ2^+ needs an independent evaluation. No such verification is given, and the claimed formula (3.14) is not derived. There is also an unexplained mismatch: Proposition 3.10 fixes z0=e^{(log z1 z2)^{1/3}}, while the proof of Proposition 4.1 uses z0=e^{(log z1 z2)^{1/2}}. Because Theorem 1.1 relies directly on Proposition 4.1 and on the weighted upper bound in Proposition 4.3, the central lower bound is not completely established as written.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper claims Theorem 1.1: #{p ≤ x : p prime, p = m^2+n^2+1, m,n ∈ N, Ω(p+2) ≤ 11} ≫ x/(log x)^{5/2}. The proof combines Iwaniec's semi-linear sieve for the condition that p−1 is a sum of two squares with a linear sieve for p+2, using a new vector sieve (Proposition 3.10) for two beta sieves of mixed dimensions, together with Richert's weighted sieve to control the number of prime factors of p+2. The final arithmetic is reduced to a numerical optimization of four parameters (θ1, θ2, θ, λ) for which H(0.14, 0.23, 0.449, 0.011) = 1.2471 > 0.","tokens_in":22853,"tokens_out":9672,"duration_ms":81879,"significance":"If the result is correct, it is a genuine advance: it gives an unconditional lower bound of the expected order of magnitude for primes of a sparse quadratic form with a bounded shift, extending Iwaniec's theorem on primes p = m^2+n^2+1. The vector sieve for two beta sieves of different dimensions is an interesting technique with potential applications to other sparse prime problems. The paper is clearly written and the overall strategy is standard, following Heath-Brown and Li. However, the load-bearing Proposition 3.10 is only sketched, and there are inconsistencies between its statement and its application that need to be resolved before the argument is complete.","major_comments":[{"comment":"The proof of the lower-bound part of Proposition 3.10 is only sketched and refers to [8, Section 3.3] for the main-term evaluation. That reference treats two linear sieves with a multiplicative density h(d1,d2) = h1(d1)h2(d2), whereas the present application uses κ1 = 1/2 and κ2 = 1 with P1 = {p ≡ 3 mod 4} and P2 = all primes. For p ∈ P1 ∩ P2, axiom (A2) does not imply h(p,p) = h(p,1)h(1,p), so the lower-bound combination δ1^-δ2^+ + δ1^+δ2^- − δ1^+δ2^+ on page 13 requires an independent evaluation. The formula (3.14) for f(σ1,σ2) is asserted without derivation, and Propositions 4.1 and 4.3 depend directly on this formula. This is a load-bearing gap.","section":"Section 3.3, Proposition 3.10"},{"comment":"Proposition 3.10 explicitly assumes log z1 ≍ log z2, but the numerical parameters chosen in Section 5, θ1 = 0.449 and θ2 = 0.011, give log z1 / log z2 = θ1/θ2 ≈ 40.8, which is not bounded by an absolute constant. Propositions 4.1 and 4.3 apply Proposition 3.10 with z1 = x^{θ1} and z2 = x^{θ2}, so the main theorem is obtained outside the stated range of validity. Either the optimization must be restricted to θ1 ≍ θ2, or Proposition 3.10 must be proven without this condition.","section":"Section 3.3 / Section 5"},{"comment":"There is an inconsistency in the choice of z0. Proposition 3.10 states z0 = e^{(log z1 z2)^{1/3}}, but the proof of Proposition 4.1 (after equation (4.4)) uses z0 = e^{(log z1 z2)^{1/2}}. These correspond to different pre-sieving levels, and the error term E2 in the proof of Proposition 3.10 depends on |log z0|. The claimed bound (4.4) is therefore not justified by the stated proposition. The authors should specify the correct z0 and verify that the error terms remain of the claimed size.","section":"Section 4.1, proof of Proposition 4.1"},{"comment":"The lower bound in (4.4) is written with a '+' sign before the error term: '≥ ... + x/(log x)^{10}'. Since the lower-bound vector sieve (3.12) has an error term that is bounded in absolute value, the displayed inequality should have a subtracted error term. As written, the sign is incorrect, although the asymptotic conclusion (4.9) would be unaffected if the error is negative. Please correct the sign or clarify the convention.","section":"Section 4.1, equation (4.4)"}],"minor_comments":[{"comment":"The abstract contains a typo: 'semi-lin ear' should be 'semi-linear'.","section":"Abstract"},{"comment":"The reference to 'Matom¨aki, Radziwi/suppress l/suppress l, and Tao' should read 'Matomäki, Radziwiłł, and Tao'.","section":"Introduction, page 2"},{"comment":"The proof relies on the numerical inequality H(0.14, 0.23, 0.449, 0.011) > 0. To make the proof fully rigorous, the authors should provide a verified computation (e.g., interval arithmetic) or a rigorous analytic bound, rather than a stepwise Matlab search without error bounds.","section":"Section 5, numerical verification"},{"comment":"The constant in the statement is written as '2c1c2^2c3C(θ1)' without parentheses around the product; consider writing it as '(2c1 c2^2 c3 C(θ1) + o(1))' for clarity, matching the proof.","section":"Proposition 4.2 statement"}],"recommendation":"major_revision","confidential_remarks":"The paper is promising and well within the scope of an analytic number theory journal. The central result is conditional on the vector sieve Proposition 3.10, which is not proven for the mixed-dimensional, overlapping-prime-set case needed here. In addition, the numerical parameters in Section 5 violate the stated assumption log z1 ≍ log z2 of that proposition. These are fixable by a fuller proof or a reparameterization, so I recommend major revision rather than rejection. The authors should also correct the z0 mismatch and the sign error in (4.4)."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The genuinely new thing here is the problem and the mixed-dimension vector sieve: combining a semi-linear sieve (kappa=1/2, primes 3 mod 4) with a linear sieve (kappa=1, all primes) to handle the simultaneous conditions on p-1 and p+2. Heath-Brown and Li did two linear sieves; this extends the framework to mixed dimensions and applies it to primes of the form m^2+n^2+1 with p+2 almost prime. The target set is sparse (expected order x/(log x)^{5/2}), and the claimed lower bound with at most 11 prime factors is a plausible, nontrivial improvement over what a naive combination of Iwaniec and Chen would give. The numerical optimization (H=1.2471 at the chosen parameters) is standard and not circular; the parameters are chosen to make the main term positive, which is fine.\n\nThe paper does several things well. The structure is clear: the switching-principle upper bound (Prop 4.2) is worked out in detail, the Bombieri-Vinogradov input is cited appropriately, and the treatment of the error terms through Cauchy-Schwarz is standard and looks correct. The explicit Euler products and the Mertens-type estimates in Prop 4.1 and 4.3 are checked carefully.\n\nThe soft spots are real, though I think they are fixable. Proposition 3.10 is the load-bearing step: both the lower bound in Prop 4.1 and the upper bound in Prop 4.3 depend on its main-term functions F and f. The proof is a sketch, and the critical part—the lower-bound combination delta_1^- delta_2^+ + delta_1^+ delta_2^- - delta_1^+ delta_2^+—is simply asserted to work \"as in [8, Section 3.3].\" That reference is for two linear sieves with the same prime set. Here the two sieves have different dimensions and different prime sets, and for common primes p in P1∩P2, axiom (A2) does not give h(d1,d2)=h1(d1)h2(d2). So the evaluation of the sum Sigma, and hence the formula (3.14), needs an independent verification. There's also a small but worrying mismatch: Proposition 3.10 fixes z0=e^{(log z1 z2)^{1/3}}, but the proof of Proposition 4.1 uses z0=e^{(log z1 z2)^{1/2}}. That could be a typo, but it's exactly in the step where the size of the pre-sieving error depends on z0.\n\nI would not desk-reject this. The result is interesting, the strategy is sound, and the gaps are localized. But as it stands, the central lower bound is not fully established: the leading constant of the main term in Prop 4.1 could change sign if the mixed-dimension f and F from Prop 3.10 are wrong by an O(1) factor. The authors need to write out the proof of Proposition 3.10 in full, or state it as a conditional lemma with a clear reference to a complete treatment.\n\nWho is this for? Sieve theorists and anyone working on sparse prime sets. It deserves a serious referee, but the referee should insist on a complete proof of Prop 3.10 before publication.","headline":"A genuine new result in sieve theory, but the load-bearing vector-sieve proposition is only sketched and needs a real proof before I'd trust the constant.","tokens_in":23431,"tokens_out":810,"would_cite":false,"duration_ms":8879,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11N05","11N35","11N36"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that infinitely many primes $p=m^2+n^2+1$ satisfy $\\Omega(p+2)\\le 11$, with count $\\gg x/(\\log x)^{5/2}$.","keywords":["almost primes","sums of two squares","semi-linear sieve","linear sieve","vector sieve","logarithmic sieve weight","mean value theorem for primes in arithmetic progressions","primes represented by quadratic forms"],"falsifier":"Recompute the numerical value of the main-term constant $H(0.14,0.23,0.449,0.011)$ from the displayed formulas; if the computed value is not positive, the proof's parameter search fails. Separately, test Proposition 3.10 on a model two-dimensional sequence with the same local densities: any concrete violation of its claimed upper or lower inequality would refute the lemma the theorem rests on.","tokens_in":22261,"feed_emoji":"🔢","tokens_out":11710,"duration_ms":109888,"temperature":0.7,"pith_summary":"This paper proves that infinitely many primes $p$ can be written as $p=m^2+n^2+1$ with $m,n\\in\\mathbb{N}$ and, at the same time, have $p+2$ with at most 11 prime factors counted with multiplicity. The quantitative form is $$\\#\\{p\\le x: p \\text{ prime},\\ p=$m^{2}$+$n^{2}$+1,\\ m,n\\in\\mathbb{N},\\ \\$\\Omega$(p+2)\\le 11\\}\\gg \\frac{x}{(\\log x)^{5/2}},$$ which matches the order predicted heuristically even for the stronger condition that $p+2$ be prime. This matters because the family $p=m^2+n^2+1$ is genuinely sparse: an upper-bound sieve already gives only $O(x/(\\log x)^{3/2})$ primes in it, so the almost-prime condition on the shift does not cost an extra logarithmic factor. The proof combines a semi-linear sieve for the condition that $p-1$ be a sum of two squares with a linear sieve for $p+2$, inside a vector sieve with two different dimensions.","feed_headline":"Infinitely many primes m²+n²+1 have p+2 with ≤11 factors","feed_subtitle":"A vector sieve joining a half-dimensional and a linear sieve yields the heuristic order x/(log x)^{5/2} for this sparse prime family.","key_machinery":"The carrying object is the vector-sieve fundamental lemma for two $\\beta$ sieves of different dimensions (Proposition 3.10). A $\\beta$ sieve of dimension $\\kappa$ is a combinatorial sieve whose weights are Möbius functions restricted to specified divisors, with $\\kappa=1/2$ giving the semi-linear sieve and $\\kappa=1$ giving the linear sieve. The lemma counts pairs $(m,n)$ surviving two different prime sets, $\\mathcal P_1$ and $\\mathcal P_2$, by pre-sieving out small primes below $z_0=(\\log z_1z_2)^{1/3}$ and then bounding the doubly-sifted count by the $\\beta$-sieve functions $F_1,F_2$ and $f_1,f_2$, combined through the inf/sup formulas (3.13) and (3.14). The same framework is used in two modes: a lower-bound mode for the main sifting term, and an upper-bound weighted mode in which a weight $1-\\log p/\\log y$ is summed over prime divisors of $p+2$. The remaining error terms are controlled by mean-value estimates for primes in arithmetic progressions.","core_discovery":"The paper's central claim is a lower-bound, sieve-theoretic result: for all large $x$, the number of primes $p\\le x$ of the form $p=m^2+n^2+1$ whose shift $p+2$ has at most 11 prime factors is $\\gg x/(\\log x)^{5/2}$. The argument decomposes the sieving by the set of primes congruent to $3$ modulo $4$: the setup forces $p-1$ to have exactly one factor of $2$, so the representation condition reduces to the absence of $3\\pmod 4$ prime factors in the odd part. A vector-sieve fundamental lemma supplies lower and upper bounds for two simultaneous $\\beta$ sieves of dimensions $1/2$ and $1$; the Buchstab term is handled by a switching principle; and a logarithmic weight function is optimized numerically with the parameters $\\theta_2=0.011$, $\\theta_1=0.449$, $\\theta=0.23$, and $\\lambda=0.14$, producing the bound 11.","pith_inferences":["The value 11 appears to be a computational artifact of a coarse parameter search rather than a structural limit; a finer optimization or a small improvement to the sieve lemma could plausibly reduce it.","The mixed-dimension vector-sieve lemma is likely reusable for other sparse prime sets defined by one representation condition and one almost-prime condition, provided the corresponding local density and prime-distribution estimates are available.","A natural test of the architecture is to compare the proven lower bound with the heuristic constant on finite ranges; if the ratio settles near $x/(\\log x)^{5/2}$, the remaining gap to an asymptotic formula looks mostly technical.","The multi-shift extension discussed in the paper is where the method faces its sharpest test, since a genuine multi-variable vector sieve would require distribution estimates beyond those used here."],"forward_implications":["There are infinitely many primes $p$ with $p=m^2+n^2+1$ and $\\Omega(p+2)\\le 11$.","The lower bound has the same shape $x/(\\log x)^{5/2}$ as the conjectured count for the case $p+2$ prime, so the almost-prime condition does not cost an extra logarithmic factor.","Under a strong equidistribution conjecture for primes in arithmetic progressions, the argument can be adapted to force $\\Omega(p+2)\\le 3$.","The same vector-sieve framework extends to products such as $(p+2)(p+6)$, and a multi-variable version would handle simultaneous almost-prime conditions on several shifts with count $\\gg x/(\\log x)^{7/2}$ for suitable optimizable bounds.","Numerical optimization of the parameters $\\theta,\\theta_1,\\theta_2,\\lambda$ could lower the number 11 within the same proof framework."],"supporting_citations":[{"why":"Supplies the beta-sieve upper and lower bounds and the fundamental lemma of sieves used throughout the proof.","marker":"[4]"},{"why":"Supplies the vector-sieve framework and the mean-value estimate for products of primes in progressions that controls the leading error terms and shapes Proposition 3.10.","marker":"[8]"},{"why":"Supplies the semi-linear sieve lower bound for primes of the form $m^2+n^2+1$ and the switching principle used to estimate the Buchstab term.","marker":"[10]"},{"why":"Supplies the asymptotic for the auxiliary multiplicative sum needed in the switched estimate for the number of representations by sums of two squares.","marker":"[13]"},{"why":"Supplies the mean-value theorem for primes weighted by a bounded function, used to dispose of the error term in the switching step.","marker":"[20]"}],"fun_headline_variants":["Infinitely many primes m²+n²+1 with p+2 ≤11 prime factors","Primes m²+n²+1: infinitely many with p+2 ≤11 factors","Infinite primes m²+n²+1 with p+2 ≤11 factors","Vector sieve: infinite primes m²+n²+1 with p+2 ≤11 factors"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof relies on a vector-sieve lemma that joins a half-dimensional sieve with a linear sieve, and the paper only sketches why that combination works; if the mixed-dimension lemma fails, the lower bound collapses.","fun_headline_variants_meta":{"raw":{"variants":["Infinitely many primes m²+n²+1 with p+2 ≤11 prime factors","Primes m²+n²+1: infinitely many with p+2 ≤11 factors","Infinite primes m²+n²+1 with p+2 ≤11 factors","Vector sieve: infinite primes m²+n²+1 with p+2 ≤11 factors"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000769,"raw_usage":{"total_tokens":3341,"prompt_tokens":813,"completion_tokens":2528,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":429,"completion_tokens_details":{"reasoning_tokens":2435}},"tokens_in":429,"tokens_out":2528,"duration_ms":20029,"temperature":1.0,"reasoning_tokens":2435,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T11:12:14.128560+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Recompute the numerical value of the main-term constant $H(0.14,0.23,0.449,0.011)$ from the displayed formulas; if the computed value is not positive, the proof's parameter search fails. Separately, test Proposition 3.10 on a model two-dimensional sequence with the same local densities: any concrete violation of its claimed upper or lower inequality would refute the lemma the theorem rests on.","supporting_citations":[{"cited_title":"Friedlander and H","cited_arxiv_id":null,"evidence_quote":"Supplies the beta-sieve upper and lower bounds and the fundamental lemma of sieves used throughout the proof."},{"cited_title":"Heath-Brown and X","cited_arxiv_id":null,"evidence_quote":"Supplies the vector-sieve framework and the mean-value estimate for products of primes in progressions that controls the leading error terms and shapes Proposition 3.10."},{"cited_title":"Iwaniec, Primes of the type φ (x, y ) +a where φ is a quadratic form","cited_arxiv_id":null,"evidence_quote":"Supplies the semi-linear sieve lower bound for primes of the form $m^2+n^2+1$ and the switching principle used to estimate the Buchstab term."},{"cited_title":"Martin, An asymptotic formula for the number of smooth values of a pol ynomial","cited_arxiv_id":null,"evidence_quote":"Supplies the asymptotic for the auxiliary multiplicative sum needed in the switched estimate for the number of representations by sums of two squares."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the mean-value theorem for primes weighted by a bounded function, used to dispose of the error term in the switching step."}],"review_version":1}