{"id":"e0c60a13-4dd0-4574-b435-23d1318e7331","arxiv_id":"2411.14388","paper_version":2,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"An effective Ito-Langevin equation whose noise source is determined solely by the steady-state fission rate reproduces the exact non-Poisson covariance of neutron counts in heterogeneous subcritical systems.","lead":"Reactor neutron counting noise deviates from Poisson statistics because fission chains produce correlated counts. This paper shows that a simple Gaussian (Langevin) process with a noise source proportional to the local fission rate reproduces that deviation exactly, potentially making full 3D noise simulations far cheaper than brute-force Monte Carlo.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Exactness of Eq. (36) requires the fission pair-emission kernel to factorize; the paper's own multi-isotope extension violates this, so the 'general heterogeneous' claim is overbroad.","rationale":"The paper is a careful, self-contained derivation, and the point-model limit correctly recovers the Feynman-alpha formula, so the core computational idea is credible. The concern is not with the algebra leading to Eqs. (31)–(33), which is a valid representation of the exact second factorial moments from the PGF. The load-bearing step is the factorization in Eq. (34) that turns the general kernel into the product s(x)χ(ϑ1)χ(ϑ2). This step is asserted, not derived, and it is not a harmless physical detail: it is exactly what allows a single scalar white-noise source to be used, and it is what the efficiency argument in Sec. VII is built on. The multi-isotope extension in Eqs. (49)–(51) is internally inconsistent with the no-cross-terms statement: averaging the spectrum and multiplying by the summed s introduces cross terms that the exact kernel forbids. This demonstrates that the factorization fails even under the paper's own independent-emission model whenever the fissile composition is heterogeneous in spectrum. Therefore the central claim should be stated conditionally: Eq. (36) is exact for a single fissile isotope with independent, identical-spectrum neutron emission; for general mixtures or correlated emission, the exact source is the non-factorized kernel in Eq. (33), and one must either use a multi-channel noise source or a Gaussian field with that kernel as its covariance. This does not invalidate the proposed computational route, but it changes the scope of the headline claim and requires a revision of Sec. VII. The requested change is modest, so conditional acceptance is appropriate rather than rejection.","tokens_in":16314,"tokens_out":27418,"duration_ms":277602,"concrete_test":"Consider a homogeneous subcritical mixture of two fissile isotopes with different fission spectra χ_A≠χ_B and known multiplicities. Compute d_ij exactly from Eq. (31) with the general kernel K(ϑ1,ϑ2)=F_Aν_A(ν_A−1)χ_A(ϑ1)χ_A(ϑ2)+F_Bν_B(ν_B−1)χ_B(ϑ1)χ_B(ϑ2), then compute it from Eq. (36) using s=F_Aν_A(ν_A−1)+F_Bν_B(ν_B−1) and χ=(F_Aχ_A+F_Bχ_B)/(F_A+F_B). Use two detectors with different energy responses, e.g. fast and thermal, over a finite time window. If the two predicted Rossi-alpha cross-correlations differ, the factorization underlying Eq. (36) is falsified. A direct algebraic check is to verify whether K(ϑ1,ϑ2) equals [ΣF_iν_i(ν_i−1)]χ(ϑ1)χ(ϑ2); it does only when χ_A=χ_B, so the test will fail for distinct spectra.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central identity (7) is secured only if the exact pair-production kernel obtained from the PGF, Eq. (33), factorizes as s(x)χ(ϑ1x)χ(ϑ2x). The paper assumes this in Eq. (34): each fission emits ν neutrons independently from a single spectrum χ. This is not true for a general heterogeneous system, and the paper's own multi-isotope generalization shows the problem. With several fissile isotopes, the exact kernel is Σ_i F_i(x) ν_i(ν_i−1) χ_i(ϑ1)χ_i(ϑ2), a sum of factorized terms with no cross terms. Equations (49)–(51), however, define a total s(x) and a flux-weighted average spectrum χ; expanding s(x)χ(ϑ1)χ(ϑ2) produces cross-isotope terms proportional to F_i F_j χ_i(ϑ1)χ_j(ϑ2) for i≠j, which are absent from the exact kernel. Thus Eq. (36) with the averaged χ cannot reproduce the exact covariance unless all fission spectra are identical. Real fission emission also carries energy/angular correlations between neutrons from the same fission (energy and momentum conservation), which breaks Eq. (34) further. The general framework of Eqs. (31)–(33) survives; the headline claim that a single scalar Gaussian white noise with strength √(ν(ν−1)F(x)) is exact for a general heterogeneous system does not.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper derives an effective Ito-Langevin equation whose Gaussian fluctuations reproduce the super-Poisson part of the two-detector covariance in a subcritical neutron system. Section III computes the covariance of a linear Langevin transport model as an integral of a noise source times two one-particle Green functions; Section IV derives the same covariance from the probability generating function of the physical branching process; Section V equates the two expressions and thereby fixes the noise amplitude and spectrum, leading to the main result Eq. (36), with s(x) = nu(nu-1)F(x) and with chi the fission-neutron spectrum. The point-model reduction in Section VI reproduces the Feynman-alpha formula exactly, and Section VII discusses a multi-isotope extension and a numerical scheme with an estimated computational speedup relative to analogue Monte Carlo.","tokens_in":16577,"tokens_out":14249,"duration_ms":130615,"significance":"If the second-moment equivalence of Eq. (7) holds in the claimed generality, the paper provides a practical and parameter-free route to Feynman-alpha and Rossi-alpha correlations using only first-moment data (the steady fission-rate distribution and one-particle Green functions). The derivation is coherent, the matching of the Langevin expression with the exact PGF expression is a genuine structural result, and the point-model test is an exact check of the method. The efficiency estimate for low-efficiency detection is also plausible and useful. The main caveat is that the exactness is conditional on a factorization assumption for the fission pair-emission kernel; the manuscript currently overstates the range of validity of the single-scalar-noise form.","major_comments":[{"comment":"The equality (7) is established only when the pair-emission kernel factorizes as s(x) chi(ϑ1x) chi(ϑ2x). Equation (33) is general, but the step from (33) to (35) assumes that every fission emits neutrons independently from a single spectrum chi. This is not true for a general heterogeneous system, e.g. a mixture of fissile isotopes with different spectra, nor for real fission in which neutron energies and directions from one fission are correlated. The paper should state this factorization as an explicit assumption and restrict the headline claim of exactness to systems satisfying it, or replace the single-source ansatz with a sum of independent sources when the kernel is a sum of factorized terms.","section":"Section V, Eqs. (33)-(36)"},{"comment":"The multi-isotope extension as written is internally inconsistent. For a mixture of isotopes the exact kernel obtained from Eq. (33) is a sum of factorized terms, Σ_i F_i(x) ν_i(ν_i−1) χ_i(ϑ1)χ_i(ϑ2), with no cross terms between different isotopes. Substituting the proposed total s(x) of Eq. (49) and the flux-weighted average spectrum chi of Eq. (51) into s(x)chi(ϑ1)chi(ϑ2) produces cross-isotope terms proportional to F_iF_jχ_iχ_j for i≠j, which are absent from the exact covariance. Thus Eq. (36) with the averaged chi cannot reproduce the exact second moments unless all fission spectra are identical. The derivation can be repaired by using one independent Langevin source per isotope, but the current text is not a valid generalization.","section":"Section VII, Eqs. (49)-(51)"},{"comment":"Equation (34) is not correctly normalized for its use in Eq. (33). As printed, for j=2 it gives c_i = P(i)χ(ϑ1)χ(ϑ2)C(i,2); combined with the explicit factor i(i−1) in Eq. (33), this would produce an extra combinatorial factor and would not lead to Eq. (35). The final result Eq. (35) corresponds instead to the two-neutron marginal c_i = P(i)χ(ϑ1)χ(ϑ2) for independent emission. This definition needs to be rewritten so that the multiplicity bookkeeping in Eqs. (29), (33), and (35) is consistent.","section":"Section V, Eq. (34)"}],"minor_comments":[{"comment":"The abstract and the text describe the result as valid for a 'general heterogeneous sub-critical neutron system'; given the factorization assumption in Eq. (34), this should be qualified, for example as 'for systems in which fission-neutron emissions are independent and identically distributed'.","section":"Abstract and Section V"},{"comment":"Equation (11) writes both L-dagger acting on primed variables and L acting on unprimed variables as equal to the same delta function; this is standard only if one recalls that the two operators act on different arguments. The notation would be clearer if the adjoint relation were written separately for the backward equation in (ϑ',t') and the forward equation in (ϑ,t).","section":"Section III, Eq. (11)"},{"comment":"The integration range of the source time t in Eq. (13) is left implicit. It is enforced by the retarded Green function, but stating t ≤ min(t1,t2) explicitly would help readers.","section":"Section III, Eq. (13)"},{"comment":"The symbol ν is used for both the random number of fission neutrons and its average in the efficiency estimate; using an overbar, e.g. ν̄, for the average would avoid ambiguity in the order-of-magnitude formulas.","section":"Section VII, Eqs. (56)-(60)"}],"recommendation":"major_revision","confidential_remarks":"The core derivation is sound and the point-model check is convincing; the paper is likely publishable after the factorization assumption is made explicit and the multi-isotope generalization is corrected. The issue in Eqs. (49)-(51) is the main substantive flaw and should be fixed before acceptance, but it does not invalidate the single-isotope or identical-spectrum version of the result."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The genuinely new thing here is the observation that the non-Poisson part of neutron noise in a subcritical system can be reproduced by an effective Gaussian Langevin equation whose noise source depends only on the steady-state fission rate and fission multiplicity. Previous Langevin treatments put all stochastic processes into the source term; this paper shows that the Poisson parts from capture, scattering, and detection cancel, leaving only branching. That is a real and useful insight, and the derivation is mostly clean. The noise amplitude is not fitted; it is obtained by equating two independent expressions—the exact PGF second derivative and the Gaussian-process covariance—and the point-model limit reproduces the Feynman-alpha formula exactly. For a single fissile isotope with independent, identical fission-neutron spectra, the main result (Eq. 36) looks correct on inspection.\n\nThe soft spots are real but not fatal. The biggest one is the claim of generality to arbitrary heterogeneous systems. The equivalence requires the fission pair-emission kernel to factorize as s(x) χ(ϑ1) χ(ϑ2). That factorization holds only when each fission emits neutrons independently from a single spectrum. For a mixture of fissile isotopes with different spectra, the exact kernel is a sum of factorized terms, one per isotope, with no cross terms. The paper's multi-isotope prescription in Eqs. (49)-(51) instead uses a flux-weighted average spectrum, which introduces spurious cross terms between different isotopes. So Eq. (36) with the averaged spectrum is not exact for multi-isotope systems. This is an overreach, though it is confined to the discussion section; the single-isotope derivation itself is honest about its assumption in Eq. (34). A referee should ask the authors to either restrict the claim or generalize the Langevin equation to multiple independent noise sources, one per isotope.\n\nThe second soft spot is the efficiency argument. The claimed orders-of-magnitude speedup is based on point-model estimates, not on any demonstration in a heterogeneous 3D geometry. That is fine for a theory paper, but the language should be softened until a benchmark exists. Minor notational sloppiness in the time-integration limits is not worth dwelling on.\n\nWho should read this: people working on reactor noise, safeguards, and subcritical assemblies. It is a specialized but practically important area, and the result could genuinely change how Feynman-alpha and Rossi-alpha calculations are done. The paper deserves a serious referee. I would send it out, with a request to fix the multi-isotope generalization and to temper the computational claims until they are demonstrated.","headline":"A clever and mostly sound derivation of an effective Langevin equation for reactor noise, but the claimed generality to multi-isotope systems goes beyond what the math supports.","tokens_in":17134,"tokens_out":3123,"would_cite":false,"duration_ms":29569,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["82D75","60H10"],"pacs":[],"model":"deepseek-v4-flash","headline":"The super-Poisson part of detector noise is exactly the second moment of a Gaussian Ito process.","keywords":["reactor noise","Feynman-alpha","Rossi-alpha","Ito-Langevin equation","neutron transport","subcritical reactor","probability generating function","Monte Carlo efficiency"],"falsifier":"Take a heterogeneous subcritical assembly with a fission model in which neutrons from a single fission are correlated in direction or energy, compute the two-detector covariance matrix exactly from the second derivative of the probability generating function, or from a high-statistics analogue Monte Carlo run, and compare it with the covariance obtained by solving the Langevin equation with the same average fission-rate distribution and Green functions; any statistically significant mismatch in $d_{ij}$ would disprove the factorization premise and, with it, the exactness of the simple source term.","tokens_in":16070,"feed_emoji":"⚛️","tokens_out":9018,"duration_ms":83413,"temperature":0.7,"pith_summary":"This paper tries to establish that the hard part of neutron detector noise, the super-Poisson correlation caused by fission chains, obeys a simple equivalent Gaussian law. In a heterogeneous subcritical system, the covariance matrix of counts at two capture detectors deviates from Poisson statistics exactly as the second moment of an Ito-Langevin process would, provided the noise source is $\\sqrt{\\nu(\\nu-1)F(x)}$ times the fission spectrum. If the equivalence holds, a notoriously expensive problem, simulating full stochastic branching histories to get Feynman-$\\alpha$ and Rossi-$\\alpha$ correlations, becomes a first-moment calculation: solve for the average flux and Green functions, then sample a white-noise-driven Gaussian field. The payoff is that spatial noise calculations for research reactors, safeguards assays, and cross-section validation become computationally practical.","feed_headline":"A single Gaussian source reproduces reactor noise correlations","feed_subtitle":"Feynman-alpha and Rossi-alpha correlations follow from average fission rates and Green functions, not full branching histories.","key_machinery":"The object carrying the argument is the Ito-Langevin ansatz $\\hat{L}_{\\vartheta}\\delta n=\\sqrt{s(x)}\\,\\chi(\\vartheta)\\,\\epsilon(x,t)$ with $s(x)=\\nu(\\nu-1)F(x)$, where $\\hat{L}_{\\vartheta}$ is the neutron transport operator, $F(x)$ is the mean fission-rate density, and $\\chi$ is the fission spectrum. The key simplification is that all non-branching interactions contribute only Poisson noise, which cancels in $d_{ij}$; only fission, with its nonzero second factorial moment $\\nu(\\nu-1)$, survives. Technically, the proof rests on the one-particle Green function $g_1$ and the probability-generating-function transport equation of Bell, and the comparison of the two resulting double-Green-function expressions forces the source term to be $\\sqrt{\\nu(\\nu-1)F(x)}\\,\\chi(\\vartheta)$.","core_discovery":"The paper's central claim is identity (7): for any two capture-based detectors and any counting window, the matrix $d_{ij}$ of deviations from Poisson statistics in a heterogeneous subcritical system equals the time-integrated second moment of a Gaussian Ito process $\\delta n(\\vartheta,t)$ obeying the Langevin equation $\\hat{L}_{\\vartheta}\\delta n=\\sqrt{\\nu(\\nu-1)F(x)}\\,\\chi(\\vartheta)\\,\\epsilon(x,t)$, where $F(x)$ is the steady-state fission-rate density, $\\nu$ is the fission neutron multiplicity, $\\chi$ is the fission-emission spectrum, and $\\epsilon$ is space-time white noise. The derivation compares two double-Green-function expressions: one obtained by transporting Gaussian noise through the one-particle Green function of the transport operator, and one obtained from the second derivative of the probability generating function for the real branching process. The equality fixes the noise amplitude and spectrum uniquely, and in the point-model limit the equation reduces to the classical Feynman-$\\alpha$ formula.","pith_inferences":["An extension the paper leaves implicit: when fission emission is correlated, Eq. (33) still gives the exact pair-emission kernel, so one could simulate the covariance by replacing the independent white noise with a correlated Gaussian field built from that kernel.","Because the effective process is Gaussian, only second moments are reproduced; any experiment sensitive to third or higher cumulants of the count distribution would need a treatment beyond what this paper provides.","The same cancellation argument should transfer to any multiplicative branching process, such as photomultiplier cascades, cosmic-ray air showers, or epidemic trees, where the non-Poisson part is controlled by the second factorial moment of the offspring distribution."],"forward_implications":["The two-detector Feynman-alpha and Rossi-alpha correlations can be computed from the steady-state fission-rate distribution and one-particle Green functions, so a full branching-history simulation is not needed for second-moment noise.","The noise integral decomposes into fuel segments, so the contribution of each spatial region to the detector correlation becomes visible, which can guide detector placement and spent-fuel assay design.","In the point-model limit the Langevin result reduces to the classical Feynman-alpha formula, providing a benchmark validation of the method.","For low detection efficiency the required Monte Carlo histories scale as $1/\\mu$ rather than $1/\\mu^2$; the paper's estimates for research-reactor parameters are about $10^8$--$10^9$ histories versus $10^{11}$--$10^{13}$ for analogue simulation."],"supporting_citations":[{"why":"Supplies the Diven factor $\\nu(\\nu-1)$ that becomes the amplitude of the effective noise source.","marker":"[1]"},{"why":"Defines the Feynman-alpha variance-to-mean deviation and gives the point-model formulas that the paper's test case reproduces.","marker":"[9, 10]"},{"why":"Supplies the probability-generating-function transport equation whose second derivatives yield the exact detector covariance on the physical side of the identity.","marker":"[37]"},{"why":"Provides a prior Langevin model for the point model, against which the paper's point-model result is benchmarked.","marker":"[33]"},{"why":"Earlier work showing that point-model noise formulas fail for a real heterogeneous core, motivating a position-dependent treatment.","marker":"[30]"},{"why":"Supplies the analogue Monte Carlo baseline used to estimate the computational cost that the new method reduces.","marker":"[25]"}],"fun_headline_variants":["Gaussian source reproduces reactor noise correlations","Ito-Langevin model cuts reactor noise simulation cost","Reactor noise from one Gaussian process","Single Gaussian source explains Feynman-alpha correlations"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"Each fission is assumed to emit its $\\nu$ neutrons independently, all with the same energy-angle spectrum, so that the two-neutron emission probability is a product of single-neutron spectra.","fun_headline_variants_meta":{"raw":{"variants":["Gaussian source reproduces reactor noise correlations","Ito-Langevin model cuts reactor noise simulation cost","Reactor noise from one Gaussian process","Single Gaussian source explains Feynman-alpha correlations"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000337,"raw_usage":{"total_tokens":1828,"prompt_tokens":869,"completion_tokens":959,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":485,"completion_tokens_details":{"reasoning_tokens":901}},"tokens_in":485,"tokens_out":959,"duration_ms":7340,"temperature":1.0,"reasoning_tokens":901,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T15:14:28.104405+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a heterogeneous subcritical assembly with a fission model in which neutrons from a single fission are correlated in direction or energy, compute the two-detector covariance matrix exactly from the second derivative of the probability generating function, or from a high-statistics analogue Monte Carlo run, and compare it with the covariance obtained by solving the Langevin equation with the same average fission-rate distribution and Green functions; any statistically significant mismatch in $d_{ij}$ would disprove the factorization premise and, with it, the exactness of the simple source term.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the Diven factor $\\nu(\\nu-1)$ that becomes the amplitude of the effective noise source."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the probability-generating-function transport equation whose second derivatives yield the exact detector covariance on the physical side of the identity."},{"cited_title":"Dubi and R","cited_arxiv_id":null,"evidence_quote":"Provides a prior Langevin model for the point model, against which the paper's point-model result is benchmarked."},{"cited_title":"Gabrieli, I","cited_arxiv_id":null,"evidence_quote":"Earlier work showing that point-model noise formulas fail for a real heterogeneous core, motivating a position-dependent treatment."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the analogue Monte Carlo baseline used to estimate the computational cost that the new method reduces."}],"review_version":1}