{"id":"1b0a51ab-1159-4e83-94ad-604300fb6145","arxiv_id":"2506.03325","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A new 3D energy flux model computes averaged underwater sound intensity in wedge-shaped oceans and matches analytic, ray, and parabolic equation results qualitatively.","lead":"This paper derives and tests a fast approximate model for how sound travels in 3D ocean environments like undersea wedges, capturing the way sound bends toward shallow water. It could make 3D underwater acoustic predictions much cheaper, though it currently works only for environments with a constant cross-section.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Favorable wedge comparisons are not quantitatively supported: focusing parameters δ⊤_m/δ⊤_n are undisclosed, no error norms are reported, and the paper admits a ~10 dB offset and an unexplained artifact, so the central claim needs conditional verification.","rationale":"The reader's weakest_assumption focuses on the x-translational invariance and convex-profile restrictions. Those are real scope limitations, but the authors state them clearly and the implemented model is explicitly for constant-cross-section wedges/troughs; they do not by themselves invalidate the demonstrated wedge results. The more immediately load-bearing issue for the central claim is the quality and reproducibility of the validation itself. The paper compares TL fields visually, does not report the focusing parameters δ⊤_m and δ⊤_n that control the convergence-factor lobes, provides no quantitative error norms, and acknowledges both a ~10 dB offset and an unexplained intensity-band artifact in the benchmark environments. Without the parameter values and a sensitivity check, a reader cannot distinguish a model that predicts horizontally refracted intensity from a model whose adjustable focusing parameters have been tuned to mimic the analytic wedge-mode envelopes. This concern does not require rejecting the paper; it is exactly the kind of missing evidence that the reader's conditional verdict should require. Since the reader already reached 'CONDITIONAL' and listed several of these issues, my read does not change the verdict, but it sharpens which condition matters most: report the focusing parameters, run a sensitivity sweep, and provide quantitative TL errors against the analytic and PE reference solutions.","tokens_in":27458,"tokens_out":8846,"duration_ms":100232,"concrete_test":"Re-run the ASA wedge I comparison at 25 Hz while varying the undisclosed focusing parameter δ⊤_n over {1, 2, 5, 10, 20} (with δ⊤_m varied analogously), holding all other settings fixed. Compute the L2 and mean-absolute TL error against the Buckingham analytic solution over the receiver line in Fig. 6, and repeat for wedge III at 25 Hz against the PE solution over the line in Fig. 12. Also run the incoherent-limit case (convergence factor set to unity) to test whether the band artifact in Figs. 11-16 persists. If the error changes by more than ~3 dB across the sweep, or if the artifact is present even without the convergence factor, the favorable agreement is parameter-sensitive and not a robust validation of eq. (127).","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that eq. (127) 'compares favorably' with analytic, ray, and PE solutions for the ASA wedges rests on qualitative visual agreement. Two load-bearing gaps make that support insufficient. First, the convergence factor C in eqs. (89)-(93) depends on the periodic-Gaussian widths F_m and F_n, which are set by the maximum mode-number differences δ⊤_m and δ⊤_n. The paper explicitly states (Section III.C) that δ⊤_n 'affects the size and shape of the focused intensity lobes,' yet the values used in Figs. 5-16 are never reported. If these parameters were tuned to make the wedge-mode envelopes match the analytic solution, the comparison would not be a predictive test of the model's physics. Second, no quantitative error metric is given for any of the TL fields. The paper instead concedes a 'TL offset, approximately 10 dB at 25 km range' in Figs. 6 and 8, and admits in Section III.D that the lossy wedge solutions contain a 'noticeable horizontal intensity band or caustic-like effect' whose 'exact cause is currently unknown.' A 10 dB discrepancy in the benchmark environment and an unexplained artifact directly weaken the 'favorable' comparison. These issues are addressable, but as written the validation evidence does not establish the central claim at the level of confidence the abstract implies.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper derives a three-dimensional energy flux propagation model for underwater acoustics. Starting from a 3D Helmholtz Green's function, the authors perform a double vertical/transverse mode-sum cross-product, apply adiabatic WKB modefunctions, map the modal continuum to solid-angle integrals via a differential-topology Jacobian, and add integration kernels for convergence focusing, transparent-boundary mode stripping, bottom attenuation along transverse cycles, and Lloyd's-mirror directivity. The final transmission-loss expression is Eq. (127). The model is implemented in the open-source Matlab package 'Tethys' and is demonstrated on the ASA wedge benchmarks I and III, with comparisons to analytic, Bellhop3D, and parabolic-equation solutions, plus a demonstration of 3D adiabatic cycle trajectories.","tokens_in":27740,"tokens_out":3551,"duration_ms":48913,"significance":"If the central claim holds, this would be the first practical 3D energy flux model that captures horizontal refraction of incoherent acoustic intensity, and its algorithmic structure (globally precomputed cycle integrals, receiver-independent field evaluation) could offer a substantial efficiency advantage over marching methods. The paper is also commendable for presenting the derivation in enough detail to be checked, for shipping an open-source implementation, and for demonstrating qualitative wedge-mode lobe positions and 3D adiabatic cycle trajectories. However, the current manuscript does not quantitatively support the 'compares favorably' claim, leaves key convergence-factor parameters undisclosed, and adopts environmental restrictions that are narrower than the abstract's '3D ocean acoustic environments' wording. These gaps are addressable, but they are load-bearing for the paper's central claims.","major_comments":[{"comment":"The validation is entirely qualitative: no error norms, mean/median absolute TL errors, or correlation metrics are reported for any of the comparisons against analytic, PE, or ray solutions. The paper itself concedes a 'TL offset, approximately 10 dB at 25 km range' in Figs. 6 and 8. Since the abstract and Section III claim that fields 'compare favorably,' the absence of quantitative metrics makes this the central supporting evidence insufficient. Please add range/depth-averaged error metrics for the line-array comparisons and, where appropriate, spatial error maps for the 2D slices.","section":"Section III.C-D, Figs. 5-16"},{"comment":"The convergence factor depends on the maximum mode-number differences δ⊤_m and δ⊤_n through F_m = 1 + 2δ⊤_m and F_n = 1 + 2δ⊤_n, and Section III.C states that δ⊤_n 'affects the size and shape of the focused intensity lobes.' Yet the values of δ⊤_m and δ⊤_n used for Figs. 5-16 are never reported. Without these values and a sensitivity study, it is impossible to determine whether the lobe agreement with the analytic wedge solution is a predictive test of the model's physics or a consequence of tuned focusing parameters. Please report the parameter values and demonstrate robustness across a reasonable range.","section":"Section II.G, eqs. (89)-(93), and Section III.C"},{"comment":"The lossy penetrable wedge results contain a 'noticeable horizontal intensity band or caustic-like effect' that the text describes as 'likely an artifact' whose 'exact cause is currently unknown.' This artifact appears in the benchmark environment used to support the model's central claim, and it is not confined to a peripheral plot. The manuscript needs either to explain and remove the artifact, or to quantify its spatial extent and show that it does not compromise the claimed agreement with the PE reference solution in the regions used for validation.","section":"Section III.D, Figs. 11-16"},{"comment":"The model's derivation assumes the environment is translationally symmetric along the forward x-axis, which eliminates back-propagation and makes the forward wavenumber k_xmn constant. This is a strong restriction: it excludes along-axis geological or oceanographic variability, so the model is not a general 3D solver. The abstract's phrase '3D ocean acoustic environments' and Section V's 'generalized 3D waveguides' overstate the scope. Please either qualify the claims explicitly to 'transversely range-dependent, longitudinally invariant' environments, or implement and validate the forward adiabatic x-range-dependence that Section IV only discusses as future work.","section":"Section II.A and Section V"},{"comment":"The abstract claims that the required computational effort is 'predominantly independent of range and frequency,' but the manuscript provides no runtime measurements, scaling experiments, or complexity analysis to support this. The implementation section describes vectorization and parallelization but not how the cost varies with range, frequency, or receiver count. Please include at least a scaling benchmark that varies frequency and range, with wall-clock times, to substantiate the efficiency claim.","section":"Abstract and Section III.A"}],"minor_comments":[{"comment":"There is a typo in 'the averge acoustic energy density'; it should read 'average.'","section":"Section II.F"},{"comment":"The WKB limiter in Eq. (71) and the elevation-angle floor in Eq. (74) introduce heuristic or user-set elements. Please clarify how the constants in the limiter are chosen and report the exact |θ|^⊥, |θ|^⊤, and ε_ϕ values used for each figure, since these affect the computed TL fields.","section":"Equations (71) and (74)"},{"comment":"The side-by-side 2D TL panels would be much more informative if they used a shared color scale and colorbars; without colorbars the reader cannot compare intensity levels across models.","section":"Figures 5, 7, 9, 11, 13, 15"},{"comment":"The implementation relies on Matlab's contour-search routines for the adiabatic invariance mapping; a brief statement about the grid resolution and convergence of that search would help reproducibility.","section":"Section III.A"}],"recommendation":"major_revision","confidential_remarks":"The derivation is ambitious and the open-source implementation is a genuine strength. The main barrier to acceptance is quantitative validation and disclosure of the convergence-factor parameters; the current evidence does not yet support the strength of the abstract's claims. I would encourage the editor to request a revision rather than reject, because the identified issues appear fixable within the manuscript's scope."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read this if you care about energy flux methods or cheap 3D-ish propagation solvers. The genuinely new thing is a working 3D energy-flux model: solid-angle integration of a double-mode-sum cross-product, a bivariate convergence factor, transparent-boundary mode-stripping, and adiabatic cycle-tracking. That combination is not in Weston or Harrison, and the open-source Matlab code makes the contribution reproducible. The paper is also honest about several limitations, which counts in its favor.\n\nWhat it does well: the wedge-mode envelope agreement in Fig. 7 is a real success. The model puts horizontal-refraction intensity lobes in roughly the right places, and the comparisons are against independent analytic, PE, and ray solutions, so the central claim is not fitted to the benchmarks. The cycle-tracking demo in Figs. 17-18 is a nice piece of machinery and looks physically sensible.\n\nThe soft spots are in the validation, not the derivation's core idea. First, the focusing widths δ⊤_m and δ⊤_n are never reported, and the text itself says δ⊤_n affects the size and shape of the intensity lobes. Without those values and a sensitivity study, the favorable wedge comparisons cannot rule out tuning, and the stress-test is right to flag that. Second, there are no quantitative error norms anywhere. The paper concedes a ~10 dB TL offset at 25 km in the ideal wedge and an unexplained horizontal intensity band in the lossy wedge. Those are load-bearing: the abstract says \"compares favorably,\" and a 10 dB offset plus an unknown artifact weakens that claim. Third, the implemented environment is x-translation-invariant (Section II.A), so the model is 2.5D, not fully 3D, despite the abstract's wording. Fourth, the efficiency claim has no runtime scaling data; it is plausible but unmeasured.\n\nNone of these are fatal. The derivation is detailed enough to check, the code is public, and the qualitative agreement supports the core mechanism. I would send this to a serious referee but expect major revision: report the parameter values, add error metrics and sensitivity, clarify the scope, and give one runtime scaling plot.\n\nWho it is for: ocean acousticians who want a fast incoherent solver for wedge/trough parameter studies and are comfortable with the energy-flux approximations. Not for people who need coherent fields or fully general 3D environments. I'd bring it to a reading group as a 'maybe' - the derivation is worth discussion even before the validation is tightened.","headline":"A genuinely new 3D energy-flux derivation with public code and real qualitative wedge agreement, but the favorable-validation claim needs parameter disclosure and error metrics.","tokens_in":28278,"tokens_out":2759,"would_cite":false,"duration_ms":33656,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["76Q05"],"pacs":[],"model":"deepseek-v4-flash","headline":"A three-dimensional acoustic propagation model built entirely from energy flux methods computes horizontally refracted incoherent intensity as one integral over source solid angles, with cost that does not grow with range or frequency.","keywords":["three-dimensional ocean acoustics","energy flux model","horizontal refraction","adiabatic modes","ray-mode analogy","WKB approximation","transmission loss","wedge benchmark"],"falsifier":"Run the model in a waveguide with measurable along-axis variation — for instance a wedge whose slope or seabed attenuation changes with $x$ — and compare its transmission loss with a 3D parabolic-equation solution or scaled tank data; if the horizontally refracted pattern diverges beyond the expected incoherent averaging difference, the translational-invariance assumption is doing the essential work. A second decisive test targets the convex-profile assumption: pick an environment whose transverse effective wavenumber profile is non-convex (a trough with a ridge, where tunneling can carry energy past a barrier) and check whether a reference solution shows transmitted intensity where the model's unimodal angular distribution predicts a hard shadow.","tokens_in":27247,"feed_emoji":"🌊","tokens_out":16157,"duration_ms":155229,"temperature":0.7,"pith_summary":"This paper claims that energy flux methods — the ray-invariant approach to computing average acoustic intensity in bounded waveguides — can be extended from two dimensions to three. The central assertion is that equation (127), a single integral over source solid angles, reproduces the horizontally refracted, incoherent acoustic intensity in a waveguide whose properties vary across the transverse direction, with computational effort that is predominantly independent of range and frequency. The paper demonstrates this on the two standard wedge benchmark environments, where its transmission-loss fields compare favorably with an analytic ideal-wedge solution, with ray-tracing solutions, and with a 3D parabolic-equation solution. A sympathetic reader would care because horizontal refraction is precisely the effect that separates true three-dimensional propagation from repeated 2D slices, and it has so far demanded full-wave solvers; a 3D energy flux model would make incoherent 3D field estimates cheap enough for repeated queries.","feed_headline":"First practical 3D energy flux model captures horizontal refraction","feed_subtitle":"Wedge transmission loss from one solid-angle integral, at any range and frequency, matching standard reference solutions.","key_machinery":"The load-bearing object is the unfolded double mode-sum cross-product of the 3D Helmholtz Green's function, converted into an integral over source solid angles $(\\theta_0, \\phi_0)$ by a Jacobian built from 'differential chains' — derivative identities that link the vertical and transverse cycle phase integrals back to the source take-off angles through adiabatic invariance of the mode numbers. Inside the integral, the mechanism that produces horizontal refraction is the bivariate convergence factor $C$, a periodic-Gaussian approximation of neighboring-mode interference that acts as an eigenray filter, concentrating incoherent intensity onto the adiabatic cycle trajectories whose accumulated phase matches the source-to-receiver path; without it the field would be a smooth gradient with no convergence or shadow structure. Cycle-tracking complements the filter by numerically inverting open partial-cycle-distance integrals to recover the wave's transverse and vertical position and momentum, $(y(x), z(x), k_y(x), k_z(x))$, along the forward axis, and those trajectories drive the transparent-boundary mode-stripping kernel $R_y$ and the bottom-attenuation product integral $R_z$.","core_discovery":"The central claim is that the 3D solid-angle energy flux model of equation (127), derived purely from energy flux principles, is the first implemented and validated three-dimensional energy flux propagation model: it captures horizontally refracted incoherent acoustic intensity in waveguides with transverse variability (wedges and troughs) and resolves effects, such as wedge-mode envelopes and horizontal shadow zones, that are invisible to 2D treatments. The model begins as a double mode-sum cross-product of vertical and transverse WKB modes, is transformed by a Jacobian into an integral over source departure angles $(\\theta_0, \\phi_0)$, and is then modulated by four kernels: a bivariate convergence factor that focuses intensity onto converging adiabatic ray cycles, a transverse mode-stripping kernel for transparent boundaries, a product-integral bottom-attenuation factor accumulated along transverse cycle trajectories, and a Lloyd's-mirror directivity kernel for boundary interference. The paper reports favorable comparisons of transmission-loss fields with the analytic ideal-wedge solution, ray tracing, and a 3D split-step Fourier parabolic-equation model for the lossy penetrable wedge, and it demonstrates 3D adiabatic cycle trajectories obtained by inverting partial cycle integrals rather than by marching. The validation runs activate the transverse-convergence portion of the factor; the paper states that full implementation of the vertical-convergence part is still under investigation.","pith_inferences":["The validation runs shown in the paper activate only the transverse part of the convergence factor, and the full vertical-convergence implementation is reported as still under investigation; the testable next step is to exercise the complete bivariate factor in an environment where vertical and transverse focusing are coupled, such as a wedge with a depth-dependent sound-speed profile, and compare","The derivation implies a production architecture the paper leaves implicit: every cycle integral and wavenumber profile depends only on the environment, so they could be precomputed once and reused across many source-receiver queries, turning the solver into a lookup-table field generator for repeated use in the same region.","Because the integrand is smooth and slowly varying, the solid-angle integral is a natural candidate for GPU-batched evaluation; the paper does not explore this, but it would make the model attractive for inversion loops and real-time sonar performance prediction that need many transmission-loss fields.","The horizontal intensity-band artifact of unknown cause that the authors flag in the lossy wedge fields could be isolated numerically: they suspect the splitting of the angular integration domain about $(\\theta_0, \\phi_0) = (0, 0)$, so varying that splitting convention is a concrete experiment that would confirm or rule out the mechanism."],"forward_implications":["A receiver's field is an independent quadrature over source angles, so no range-marching and no mode-eigenvalue search is needed; the cost is set by the angular grid and the cycle-integral interpolations, which is what makes the effort 'predominantly independent' of range and frequency.","Horizontally refracted structure — wedge-mode envelopes, convergence zones, and horizontal shadow zones — appears directly in the incoherent transmission-loss field, meaning 3D refractive effects can be predicted without computing fine-scale coherent interference.","3D adiabatic ray cycles can be produced by inverting partial cycle integrals rather than by marching, yielding positions $(y(x), z(x))$ and wavenumber components $(k_y(x), k_z(x))$ along the forward axis.","Bottom attenuation is accumulated along oriented transverse cycles with range-dependent grazing angles, so even an environment with uniform sediment properties produces mode-dependent loss that varies along each horizontal trajectory.","Environments of wedge or trough type with transparent outer boundaries become tractable in incoherent 3D, extending the range of problems energy flux methods can address beyond the strictly 2D slice."],"supporting_citations":[{"why":"Supplies the 2D convergence-factor formulation, including the periodic-Gaussian approximation of neighboring-mode interference, which the bivariate convergence factor generalizes.","marker":"[15]"},{"why":"Contains the original energy flux / ray-invariant derivation that this paper extends from two to three dimensions.","marker":"[2]"},{"why":"Defines the standard wedge benchmark environments and parameters used for all inter-model transmission-loss comparisons.","marker":"[31]"},{"why":"Provides the analytic integral-transform solution for the ideal wedge that serves as the lossless reference field.","marker":"[32]"},{"why":"Provides the 3D split-step Fourier parabolic-equation reference solutions for the lossy penetrable wedge.","marker":"[35]"},{"why":"The author's prior work that introduced the mode-stripping kernel, asymmetric energy-flux integration, and inversion of partial cycle distances, all incorporated here.","marker":"[1]"},{"why":"Establishes the ray invariant under the adiabatic approximation, which grounds the invariant mode-number mapping used throughout the derivation.","marker":"[4]"},{"why":"Analyzes horizontal shadow zones in the horizontal plane, the phenomenon the convergence factor and finite-frequency angle cutoffs reproduce.","marker":"[14]"}],"fun_headline_variants":["First 3D energy flux model captures horizontal refraction","3D flux model solves wedge acoustics at any range and frequency","Energy flux model goes 3D, matches benchmark wedge solutions","Horizontal refraction in 3D from an energy flux integral"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The construction assumes the environment is unchanged along the forward $x$-axis, so a wave's forward wavenumber stays constant and no wave ever turns back; wherever depth, sediment, or water-column properties change along the propagation direction, the adiabatic mode mapping and cycle tracking on which everything else rests no longer hold.","fun_headline_variants_meta":{"raw":{"variants":["First 3D energy flux model captures horizontal refraction","3D flux model solves wedge acoustics at any range and frequency","Energy flux model goes 3D, matches benchmark wedge solutions","Horizontal refraction in 3D from an energy flux integral"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000312,"raw_usage":{"total_tokens":1816,"prompt_tokens":1030,"completion_tokens":786,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":646,"completion_tokens_details":{"reasoning_tokens":717}},"tokens_in":646,"tokens_out":786,"duration_ms":7879,"temperature":1.0,"reasoning_tokens":717,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T11:06:00.566959+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the model in a waveguide with measurable along-axis variation — for instance a wedge whose slope or seabed attenuation changes with $x$ — and compare its transmission loss with a 3D parabolic-equation solution or scaled tank data; if the horizontally refracted pattern diverges beyond the expected incoherent averaging difference, the translational-invariance assumption is doing the essential work. A second decisive test targets the convex-profile assumption: pick an environment whose transverse effective wavenumber profile is non-convex (a trough with a ridge, where tunneling can carry energy past a barrier) and check whether a reference solution shows transmitted intensity where the model's unimodal angular distribution predicts a hard shadow.","supporting_citations":[{"cited_title":"Ray convergence in a flux-like propagation formulation","cited_arxiv_id":null,"evidence_quote":"Supplies the 2D convergence-factor formulation, including the periodic-Gaussian approximation of neighboring-mode interference, which the bivariate convergence factor generalizes."},{"cited_title":"Guided propagation in a slowly varying medium","cited_arxiv_id":null,"evidence_quote":"Contains the original energy flux / ray-invariant derivation that this paper extends from two to three dimensions."},{"cited_title":"Numerical solutions of range-dependent benchmark problems in ocean acoustics","cited_arxiv_id":null,"evidence_quote":"Defines the standard wedge benchmark environments and parameters used for all inter-model transmission-loss comparisons."},{"cited_title":"Acoustic propagation in a wedge-shaped ocean with perfectly reflecting bound- aries","cited_arxiv_id":null,"evidence_quote":"Provides the analytic integral-transform solution for the ideal wedge that serves as the lossless reference field."},{"cited_title":"A higher-order tangent linear parabolic-equation solution of three-dimensional sound propagation","cited_arxiv_id":null,"evidence_quote":"Provides the 3D split-step Fourier parabolic-equation reference solutions for the lossy penetrable wedge."},{"cited_title":"Foundations for Development of Three-Dimensional Energy Flux Acoustic Propa- gation Models","cited_arxiv_id":null,"evidence_quote":"The author's prior work that introduced the mode-stripping kernel, asymmetric energy-flux integration, and inversion of partial cycle distances, all incorporated here."},{"cited_title":"Ray and wave invariants for SOFAR channel propagation","cited_arxiv_id":null,"evidence_quote":"Establishes the ray invariant under the adiabatic approximation, which grounds the invariant mode-number mapping used throughout the derivation."},{"cited_title":"Acoustic shadow zones in the horizontal plane","cited_arxiv_id":null,"evidence_quote":"Analyzes horizontal shadow zones in the horizontal plane, the phenomenon the convergence factor and finite-frequency angle cutoffs reproduce."}],"review_version":1}