{"id":"581345bf-1946-4a5c-adaa-3a2c3831446c","arxiv_id":"2502.07028","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For incoherent (speckled) initial waves, the scintillation index in branched flow is given by closed formulas in terms of a single dimensionless parameter X_o, and its maximum depends only on the initial correlation radius, not on the medium correlation length.","lead":"This paper derives equations that predict how the brightness fluctuations (scintillation) of light grow and peak as a partially coherent or speckled beam travels through a weakly disordered medium, the regime that produces branched flow. The formulas have no fitted parameters and match simulations, giving a testable theory for the recent experiments on incoherent branched flow and for extreme wave statistics.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Finite-ε convergence to the diffusion limit is asserted, not demonstrated; the closed-form formulas are asymptotic but validated only at one modest scale separation without error bars.","rationale":"The reader's weakest-assumption identification is correct: the closed-form formulas depend on the diffusion-approximation limit of the ray process. I agree this is the load-bearing premise, but I would phrase the concern more precisely as finite-ε validation rather than doubting the underlying convergence theorem, which is standard and internally consistent with the scaling choices. The derivation in Appendices B-E checks out dimensionally and algebraically: the generator (D1) follows from the chosen scaling, the change of variables in (D9)-(D10) is consistent with V=K1-K2, and the reduction leading to (E8) and hence (17) is plausible. The remaining risk is that the asymptotics have not been tested across ε, and the reported agreement has no error bars, so the magnitude of the finite-ε error is unknown. This supports the reader's CONDITIONAL verdict rather than changing it.","tokens_in":17842,"tokens_out":32962,"duration_ms":302541,"concrete_test":"First, simulate the ray equations (14) at the paper's parameters (ℓc/λ=100, σ²λ²=10^{-4}, ρo/λ=10) and compare the empirical distribution of K_z with the diffusion-limit law (D6) at z=zc and 2zc, using a Kolmogorov-Smirnov test. Second, rerun the full paraxial simulations with Xo fixed and ε halved, e.g. (ℓc/λ, ρo/λ) = (100,10), (200,14.1), (400,20), each with 1000 realizations, and check whether S(pc)_max and S(c)_max converge to max Π(0,0) within statistical error. If the curves shift systematically with ε, the finite-ε agreement does not validate the asymptotic formulas.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central formulas (16)-(19) for the scintillation index and intensity correlations are limits as ε→0 of a diffusion approximation applied to the ray process (App. D, generator (D1)). The paper also drops the O(ε^{(5d-1)/2}) remainder in the Vlasov equation (App. B). Both errors are uncontrolled. The simulations use ε = λ/ℓc ≈ 0.01 with d ≈ 0.5, for which the Vlasov remainder is about ε^{0.75} ≈ 0.03, but the diffusion-approximation error itself is not estimated and no error bars are reported. Because the headline claim that the maximal scintillation index depends only on Xo is a property of the limit PDE (17), the quantitative agreement could in principle reflect a fortuitous match at this single scale separation rather than a validated asymptotic result. The absence of shipped code or data prevents the reader from testing this directly.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a stochastic theory of branched flow for waves governed by the two-dimensional paraxial wave equation, focusing on incoherent (speckled) initial fields. It derives approximate PDEs for the intensity correlation function and the scintillation index in two regimes: a coherent plane-wave initial condition and a partially coherent or coherent speckled initial field. The central results are the expressions S^{(pc)}_z = Pi(0,0)-1 and S^{(c)}_z = 2Pi(0,0)-1, where Pi solves the closed PDE (17), and the prediction that the maximal scintillation index depends only on the dimensionless parameter X_o = sigma^(2/3) rho_o / alpha^(1/3), not on the medium correlation length. The theoretical predictions are compared with direct numerical simulations of the paraxial equation and are reported to agree quantitatively without adjustable parameters.","tokens_in":17952,"tokens_out":10846,"duration_ms":100869,"significance":"If the central asymptotic results are rigorous, the paper makes a substantial contribution to the theory of branched flow: it provides a tractable stochastic description of incoherent branched flow, identifies a striking non-trivial dependence of the scintillation-index maximum on a single combination of parameters, and extends the authors' established fourth-moment program to partially coherent sources. The absence of fitted parameters in the comparisons with simulations is a clear strength, as is the consistency of the simulations across variations of rho_o, sigma^2, and l_c. The computational validation is, however, confined to a single modest scale separation, and the main technical step is the asserted diffusion-approximation limit in the supplement, so the significance and the strength of the validation both depend on that limit being correct.","major_comments":[{"comment":"The convergence of the ray process (X^epsilon_z, K^epsilon_z) solving Eq. (14) to the Markov diffusion with generator (D1) is asserted by reference to standard diffusion-approximation theory, but no theorem statement, hypothesis check, or error estimate is provided for this specific scaled system. This limit is load-bearing: all closed-form results in the main text, including Eqs. (16)-(19) and the X_o-only dependence of the maximal scintillation index, follow from it. The reader cannot verify that the smooth, stationary, mixing assumptions on V are sufficient for the joint convergence needed in the n=2 application, nor can the size of the finite-epsilon error be assessed. Please state the precise convergence theorem used, verify its hypotheses for Eq. (14), and either give a proof of the generator (D1) or cite a reference where this exact result is established; an estimate of the convergence rate would also address the validation concern raised below.","section":"Supplementary App. D"},{"comment":"The numerical validation of the limit formulas is performed at a single scale-separation value, epsilon = lambda/l_c = 0.01 (l_c/lambda = 100), with rho_o/l_c = 0.1, and no error bars are reported. The paper's own estimate in App. B shows an O(epsilon^{(5d-1)/2}) remainder in the Vlasov equation, which is about 0.03 for d = 0.5, and the diffusion-approximation error is uncontrolled. Consequently, the reported 'excellent quantitative agreement' supports the theory at one modest separation of scales, but it does not by itself establish that the epsilon -> 0 limit is the operative mechanism. Please add a convergence study in epsilon (for example, l_c/lambda = 50, 100, 200, with other parameters adjusted to keep X_o fixed) and report the deviation of the simulation curves from the theory as a function of epsilon, or provide a quantitative bound on the expected finite-epsilon corrections.","section":"Main text, Figs. 3-4 and App. H"}],"minor_comments":[{"comment":"The caption of Fig. 3(b) lists only rho_o/lambda and sigma^2 lambda^2 and does not state that l_c is the parameter being varied; please specify the varying parameter and its values explicitly.","section":"Main text, Fig. 3 caption"},{"comment":"The notation Pi_z(x,y) is used in Eq. (16) before the PDE (17) is introduced, and the reader must infer the scaling; please define Pi_z and its arguments (or the dimensionless variables) explicitly in the text preceding Eq. (16).","section":"Main text, Eq. (16)"},{"comment":"Please add a consistency check for the normalization of Gamma(x) in the generator (D1) and the Fokker-Planck equation (D7), since the factors of 1/2 and 2 in (D1), (D8), and (E8) are easy to get wrong and a reader would benefit from an explicit verification that the final PDE (17) follows with the stated coefficients.","section":"Supplementary App. D, Eq. (D1)"},{"comment":"The phrase 'closed-form equations' may be misread as explicit analytic solutions; the main results are a closed coupled PDE system plus explicit small-propagation-distance expansions. Consider rephrasing to 'a closed system of equations' for precision.","section":"Abstract"}],"recommendation":"major_revision","confidential_remarks":"The paper is a strong candidate for a high-profile applied-mathematics or physics letter, and the authors' prior work in this area is well respected. However, the core technical content is concentrated in the supplement, and the diffusion-approximation step is asserted rather than proved. The current validation at one scale separation is not sufficient to fully reassure a mathematically minded referee. A revision that adds a precise convergence statement and a multi-epsilon simulation study would bring the paper to a publishable level. There is no apparent issue with novelty or scope."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: the genuinely new result is a closed-form statistical description of branched flow from an initially speckled field — Eqs. (16) through (19), the two-scale intensity correlation function, and the prediction that the maximum of the scintillation index depends only on Xo = σ^{2/3}ρo/α^{1/3}, not on the medium correlation length. The plane-wave section is a recap of Garnier-Solna, which is fine; the incoherent part is the contribution.\n\nWhat the paper does well: it carries the multiscale/diffusion-approximation program through to explicit PDEs for the fourth-order moments, closes the speckle statistics with Isserlis' theorem, and then validates the theory by direct simulation of the paraxial equation without fitting parameters. The parameter sweeps over ρo, σ², and ℓc are the right tests, and the agreement is quite good. The two-scale behavior of C^I in the coherent-speckle case is a nice nontrivial prediction, and the fact that it shows up in the numerics is real evidence.\n\nThe soft spots: the convergence to the diffusion limit in App. D is asserted rather than demonstrated. The remainder in the Vlasov equation is dropped at O(ε^{(5d-1)/2}), and the diffusion-approximation error is not estimated at all. The simulations sit at one scale separation (ℓc/λ = 100, ε = 0.01, d ≈ 0.5), with no error bars and no convergence study. So the quantitative agreement could be partly fortuitous, though the theory is standard enough that I would bet on it. The absence of shipped code or data makes it hard to check directly, and I would want error bars or a convergence panel in a revision.\n\nAlso minor: the paper itself notes the \"rather limited separation of scales\" and then uses that as cover for not checking the asymptotic error. If the claim is that the formulas are already good at ε = 0.01, that is a stronger claim needing a numerical convergence check — e.g., run at ε = 0.02 and 0.005 and show Smax approaches the limit. The small-z expansions in App. F are consistent and give a useful check.\n\nWho this is for: applied mathematicians and physicists working on branched flow, random media, and optical experiments with rotating diffusers. The Xo-only dependence is a crisp prediction that an experiment could test.\n\nVerdict: send it to peer review. The central argument holds up; the missing convergence diagnostics are fixable. A serious referee should ask for error bars and a finite-ε check, not for a new theory.","headline":"New closed-form statistics for incoherent branched flow, backed by no-parameter numerics; the main weakness is an asserted diffusion limit with no convergence check or error bars, but the paper deserves serious refereeing.","tokens_in":18501,"tokens_out":2707,"would_cite":true,"duration_ms":25657,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35Q41","35R60","60F17","78A45"],"pacs":["42.25.Dd","05.40.-a"],"model":"deepseek-v4-flash","headline":"Closed-form equations for the scintillation index of coherent and incoherent branched flow show that, for speckled initial fields, the peak intensity fluctuation depends only on $X_o=\\sigma^{2/3}\\rho_o/\\alpha^{1/3}$.","keywords":["branched flow","scintillation index","paraxial wave equation","Wigner transform","diffusion approximation","partial coherence","intensity correlation function","random media"],"falsifier":"Fix $\\sigma$, $\\rho_o$, and $\\alpha$, and vary the medium correlation length $\\ell_c$: the theory predicts that the maximum of the scintillation index is unchanged, so a statistically significant change in the peak value would falsify the $X_o$-only dependence. A second check is the predicted relation $S_z^{(c)}=2S_z^{(pc)}+1$ between coherent-speckle and partially coherent settings with identical parameters.","tokens_in":17611,"feed_emoji":"🌊","tokens_out":11061,"duration_ms":91343,"temperature":0.7,"pith_summary":"Branched flow, the focusing of waves into branching filaments by weak smooth disorder, is usually studied with fully coherent waves, but recent experiments use incoherent light. This paper develops a stochastic theory of the paraxial wave equation that covers coherent plane waves, coherent speckle fields, and partially coherent speckle fields. Its central result is a closed-form evolution equation for the intensity correlation function and the scintillation index, from which the value and location of the maximum intensity fluctuation can be read off. The theory shows that for speckled initial fields the maximal scintillation index depends only on the dimensionless parameter $X_o=\\sigma^{2/3}\\rho_o/\\alpha^{1/3}$ and not on the medium correlation length. Direct numerical simulations match the predictions without adjustable parameters.","feed_headline":"Peak branched-flow intensity depends on one parameter only","feed_subtitle":"Closed-form equations predict scintillation maxima for coherent and partially coherent waves, with no free parameters.","key_machinery":"The argument runs through the Wigner transform, the phase-space representation of the field's two-point correlation, which obeys a scaled Vlasov equation in the regime where the initial correlation radius sits between the wavelength and the medium correlation length. The characteristics of this kinetic equation are the ray equations (14), and a diffusion-approximation theorem (App. D) shows that, in the $\\varepsilon\\to0$ limit, the joint ray process converges to a Markov diffusion with generator $L^{(n)}=\\sum_{j=1}^n 2\\alpha K_j\\partial_{X_j}+\\frac12\\sum_{j,j'=1}^n\\Gamma(X_j-X_{j'})\\partial_{K_j}\\partial_{K_{j'}}$, with $\\Gamma(x)=\\int_{-\\infty}^{\\infty}\\mathbb{E}[\\partial_x V(0,0)\\partial_x V(z,x)]\\,dz$. Combining this generator with the Gaussian fourth-moment factorization of the initial speckle field reduces the fourth-order intensity statistics to the closed equation (17) for $\\tilde\\Pi$. The dimensionless parameter $X_o$ appears only through the initial condition and controls the peak scintillation.","core_discovery":"For an initial speckle field with Gaussian statistics and correlation radius $\\rho_o$, the scintillation index is $S_z^{(pc)}=\\tilde\\Pi_{z/z_c}(0,0)-1$ in the partially coherent case and $S_z^{(c)}=2\\tilde\\Pi_{z/z_c}(0,0)-1$ in the coherent-speckle case, where $\\tilde\\Pi$ solves the closed PDE (17); the only memory of the initial field enters through the dimensionless parameter $X_o=\\sigma^{2/3}\\rho_o/\\alpha^{1/3}$. From this the authors derive that the maximum of the scintillation index depends only on $X_o$ and increases with it, while the distance at which the maximum occurs scales as $z_c=\\ell_c/(2\\sigma^{2/3}\\alpha^{2/3})$. They also obtain the two-scale intensity correlation function $C_z^{I,(c)}(x)=\\tilde\\Pi_{z/z_c}(x/\\ell_c,0)+\\tilde\\Pi_{z/z_c}(x/\\ell_c,X_o x/\\rho_o)-1$ for coherent speckle and $C_z^{I,(pc)}(x)=\\tilde\\Pi_{z/z_c}(x/\\ell_c,0)-1$ for partially coherent fields, and verify all predictions against parameter-free simulations.","pith_inferences":["The same closed-form structure should apply to any paraxial system, so ocean-wave statistics over random currents could be predicted from measured effective parameters $\\Gamma$ and $X_o$ without resolving individual caustics.","For non-Gaussian initial fields, the Gaussian fourth-moment factorization fails; the first visible signature should be a distortion of the small-scale term $\\tilde\\Pi(x/\\ell_c,X_o x/\\rho_o)$ in the coherent-speckle correlation function, making that term a measurable probe of non-Gaussian speckle statistics.","Feeding the predicted intensity correlation function into a nonlinear Schr\\\"odinger simulation would provide a quantitative test of the authors' suggestion that linear branched-flow focusing seeds extreme nonlinear events such as freak waves."],"forward_implications":["For speckled initial fields, the maximal intensity fluctuation is set by the initial correlation radius $\\rho_o$ and the medium strength $\\sigma$, but not by the medium correlation length $\\ell_c$; two media with the same $\\sigma$ and $\\rho_o$ produce the same peak scintillation even if their disorder is smoother or rougher.","A coherent-speckle source and a partially coherent source with the same parameters are linked by $S_z^{(c)}=2S_z^{(pc)}+1$, so a diffuser switched between static and rotating modes gives a direct experimental test of coherence effects.","In the coherent-speckle case the intensity correlation function has a two-scale structure: rapid decorrelation on the initial speckle scale $\\rho_o$ and slow, medium-induced variations on the scale $\\ell_c$; a partially coherent source shows only the slow scale and satisfies the energy-conservation identity $\\int C_z^{(pc)}(x)\\,dx=0$.","Early propagation is universal: the partially coherent scintillation index grows as $(\\tilde\\gamma_4/6)(z/z_c)^3$, independent of $\\rho_o$ and of the initial correlation function, matching the plane-wave early growth.","For large propagation distances the field statistics relax to Gaussian, and the scintillation index tends to 1, so branched-flow intensity enhancements are a finite-distance phenomenon."],"supporting_citations":[{"why":"Reports the incoherent branched-flow experiments this paper explains; supplies the rotating and static diffuser setup and the qualitative coherence effects.","marker":"[42]"},{"why":"Derives the white-noise paraxial scintillation equation for plane waves that the present theory generalizes to speckled initial fields.","marker":"[50]"},{"why":"Provides the diffusion-approximation theorems for random ordinary differential equations used to pass from the ray equations to the Markov generator.","marker":"[60]"},{"why":"The split-step Fourier scheme used to solve the closed PDE for $\\tilde\\Pi$ and generate the theoretical curves.","marker":"[53]"},{"why":"Argues that ray and caustic theory cannot predict intensity values because interference matters; the gap the present paper fills.","marker":"[41]"},{"why":"Establishes universal branched-flow statistics and identifies the $\\ell_c/\\sigma^{2/3}$ scale behind the definition of $z_c$.","marker":"[37]"},{"why":"The diffusion-approximation chapter underlying Appendix D's convergence proof for the ray process.","marker":"[S7]"},{"why":"Supplies the Gaussian fourth-moment factorization used to close the intensity statistics for speckled initial conditions.","marker":"[S8]"}],"fun_headline_variants":["One parameter controls branched-flow intensity peaks","Closed-form equations fix branched-flow scintillation maxima","Branched flow peak intensity depends on a single number","Branched flow peaks: one number sets the maximum"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole closed-form theory rests on the assumption that, in the limit where the wavelength is much smaller than the medium correlation length, the random ray motion converges to a Markov diffusion; if that convergence fails or is too slow in the parameter regime of interest, the predicted scintillation indices and their $X_o$-only dependence would not hold.","fun_headline_variants_meta":{"raw":{"variants":["One parameter controls branched-flow intensity peaks","Closed-form equations fix branched-flow scintillation maxima","Branched flow peak intensity depends on a single number","Branched flow peaks: one number sets the maximum"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001373,"raw_usage":{"total_tokens":5580,"prompt_tokens":980,"completion_tokens":4600,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":596,"completion_tokens_details":{"reasoning_tokens":4539}},"tokens_in":596,"tokens_out":4600,"duration_ms":26794,"temperature":1.0,"reasoning_tokens":4539,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T14:00:48.851385+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Fix $\\sigma$, $\\rho_o$, and $\\alpha$, and vary the medium correlation length $\\ell_c$: the theory predicts that the maximum of the scintillation index is unchanged, so a statistically significant change in the peak value would falsify the $X_o$-only dependence. A second check is the predicted relation $S_z^{(c)}=2S_z^{(pc)}+1$ between coherent-speckle and partially coherent settings with identical parameters.","supporting_citations":[{"cited_title":"Metzger, R","cited_arxiv_id":null,"evidence_quote":"Reports the incoherent branched-flow experiments this paper explains; supplies the rotating and static diffuser setup and the qualitative coherence effects."},{"cited_title":"Shapiro and R.W","cited_arxiv_id":null,"evidence_quote":"Derives the white-noise paraxial scintillation equation for plane waves that the present theory generalizes to speckled initial fields."},{"cited_title":"Foley and M.S","cited_arxiv_id":null,"evidence_quote":"Provides the diffusion-approximation theorems for random ordinary differential equations used to pass from the ray equations to the Markov generator."},{"cited_title":"Garnier and K","cited_arxiv_id":null,"evidence_quote":"The split-step Fourier scheme used to solve the closed PDE for $\\tilde\\Pi$ and generate the theoretical curves."},{"cited_title":"Pradas, A","cited_arxiv_id":null,"evidence_quote":"Argues that ray and caustic theory cannot predict intensity values because interference matters; the gap the present paper fills."},{"cited_title":"Nye, Natural Focusing and Fine Structure of Light: Caustics and Wave Dislocations, CRC Press, 1999","cited_arxiv_id":null,"evidence_quote":"Establishes universal branched-flow statistics and identifies the $\\ell_c/\\sigma^{2/3}$ scale behind the definition of $z_c$."}],"review_version":1}