{"id":"e4501216-288a-4e51-84c0-19f89fe26ec8","arxiv_id":"2411.10355","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A matrix transport equation derived from replicated field theory yields the full transmission eigenvalue distribution in disordered media, from quasiballistic to diffusive regimes, for waveguides and infinite slabs.","lead":"This paper derives a field-theoretic transport equation, called radiant field theory, for the distribution of transmission eigenvalues in disordered waveguides and slabs, covering both quasiballistic and diffusive regimes. The framework provides a microscopic route to quantities relevant for wavefront shaping and imaging through scattering media, going beyond random matrix theory.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Single-replica reduction is not controlled: the Vandermonde factor in Eq. (47) vanishes at the replica-symmetric saddle point and its R→0 contribution to F(γ) is never estimated, so Eq. (97) is not fully justified in the quasiballistic/grazing regime.","rationale":"The paper is impressive: the Gaussian-integral machinery, the Wigner transform, the Eilenberger-type equation, and the internal check of the semiclassical equation against the full saddle-point equations in Fig. 3 are coherent and nontrivial. The recovery of the bimodal law in the diffusive limit is independent support. However, the central claim (a new microscopic transport equation for transmission eigenvalue statistics) is only as solid as the single-replica approximation, and that approximation contains an unexamined algebraic prefactor whose R→0 limit is not controlled. The authors themselves flag localization as an exclusion, so the grazing-mode T→0 tail should be interpreted with care; but the gap I identify is more specific: even within the claimed nonlocalized regime, the Vandermonde factor in Eq. (47) can affect ⟨ln Z⟩. The deferred wave-equation validation makes this a conditional rather than fatal issue. Since the reader's verdict already reflects this conditionality, I do not propose a change.","tokens_in":37346,"tokens_out":18285,"duration_ms":196625,"concrete_test":"Run the companion-Letter wave-equation simulations for the 2D waveguide of Fig. 3 with W/λ=25.5, L/W=1, periodic transverse boundaries, at L/ℓ=0.2 and L/ℓ=1, accumulating enough realizations to resolve the low-T tail, and overlay ρ(T) on the RFT curves. If the simulated distribution disagrees with the solid curves beyond the sampling error in the small-T region, the single-replica/semiclassical reduction of Secs. II.E and III.A is not validated in the quasiballistic regime; if it agrees, the missing Vandermonde estimate is unlikely to be numerically relevant.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing step is Sec. II.E: Eq. (53) discards the Jacobian |Δ(Q_1,...,Q_R)| of Eq. (47) on the grounds that L(1)∼A≫1. This is not a controlled estimate. The Jacobian is a Vandermonde-type determinant that vanishes at the replica-symmetric point Q_1=...=Q_R selected by the saddle point; its logarithm contributes Σ_{i<j} log|Q_i−Q_j|. Near the saddle this is O(R(R−1) log A), and since ∂_R[R(R−1)] at R=0 is −1, it can survive the replica derivative (33), with a γ-dependence inherited from Q_s(γ). The paper never bounds this contribution, nor explains why it is γ-independent. Appendix C only proves convergence of the iteration in the diffusive regime (ℓ≪L), not in the quasiballistic or grazing-mode regimes where the new predictions live. The infinite-slab T→0 tail is explicitly attributed to grazing modes (μ→0), i.e., the region where this uncontrolled limit is most dangerous. If replica-interaction corrections leak into F(γ), then Eq. (76), Eq. (97), and every derived ρ(T) change. This is the central-claim risk.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a field-theoretic framework, radiant field theory (RFT), for the statistics of transmission eigenvalues in disordered waveguides and slabs. Starting from a Gaussian disorder model and the Fisher-Lee formula, the authors represent the generating function as a functional integral, apply the replica trick, and approximate the replicated functional integral by a single replica in the nonlocalized regime. The saddle point of the resulting action is a self-consistent Gorkov-type equation, which is then converted by a Wigner transform into a semiclassical matrix transport equation for a 2x2 matrix radiance g(Omega,r), Eq. (97). The central claim is that this equation, together with the boundary conditions Eqs. (116)-(117) and the generating-function relation Eq. (76), yields rho(T) in both quasiballistic and diffusive regimes and for infinite-slab geometries. Numerical solutions are presented for waveguides and slabs; the diffusive limit recovers the bimodal law, while the quasiballistic regime shows shape-dependent, non-universal features.","tokens_in":37621,"tokens_out":5154,"duration_ms":49157,"significance":"The paper is ambitious and potentially important. If the derivation is valid, RFT provides a microscopic transport theory that goes beyond the diffusive DMPK universality class while retaining no fitted parameters; the reported predictions are quantitative and falsifiable by wave-equation simulations, and the companion Letter [31] provides part of that comparison. Strengths include a fully written derivation with multiple internal checks: the reduction to the nonlinear sigma model in the diffusive limit, the recovery of the bimodal law, and the agreement between the modal transport equation and the direct saddle-point solution in Fig. 3. The transport-equation solver Ebsolve is made publicly available. The main risk is the single-replica approximation and the convergence of the iterative solution in the very regimes (quasiballistic, grazing modes) where the new predictions live; these are not controlled in the manuscript.","major_comments":[{"comment":"The single-replica approximation is asserted, not proved. The Jacobian Delta(Q_1,...,Q_R) in Eq. (47) is a Vandermonde-type factor whose logarithm contains a sum over log|Q_i-Q_j|, and it vanishes at the replica-symmetric point Q_1=...=Q_R that the saddle-point argument selects. Through the replica derivative in Eq. (33), this factor can contribute to <ln Z> and hence to F(gamma) in Eq. (76), but the paper never bounds this contribution or shows that it is gamma-independent. Since Eq. (97) and all derived rho(T) rest on Eq. (53), this is a load-bearing gap.","section":"Sec. II.E, Eq. (53)"},{"comment":"The convergence and stability proof for the self-consistent iteration is explicitly limited to the diffusive regime. The text states that the key Green's function expression (C8) holds exactly only for a constant field Q(0) and that the authors assume it also holds for a spatially varying field in the diffusive regime (ell << L). The numerical results in Secs. IV A and IV B include quasiballistic optical thicknesses L/ell = 0.05 and 0.2, so the convergence of the iterative algorithm in the regime of the new predictions is not covered by the proof. A numerical or analytical demonstration of convergence for quasiballistic parameters is needed.","section":"Appendix C"},{"comment":"The T -> 0 region of the infinite-slab distribution is dominated by grazing modes (mu -> 0), and the paper itself warns that these modes may be affected by localization, which is outside the theory. Because the single-replica approximation is least controlled for exactly these near-grazing trajectories, the quantitative T -> 0 tail should be clearly marked as outside the domain of validity, or supported by an additional estimate that addresses the replica-interaction contribution in this limit.","section":"Sec. IV B, Fig. 4"}],"minor_comments":[{"comment":"The sentence 'a+ and b- must be integrated in the forward direction and a+ and b- in the backward direction' contains a typo; the second pair should be a- and b+.","section":"Appendix F, text below Eq. (F9)"},{"comment":"The notation <f>_0^+ is used without reminding the reader of the directional-average definition in Eq. (94); a brief reminder would improve readability.","section":"Appendix D, Eq. (D14)"},{"comment":"The sentence 'It is clear from the above calculation that...' relies on an order-of-magnitude argument; please state more precisely the domain of validity in terms of L, ell, and N_p, and in what sense Eq. (53) is expected to hold.","section":"Sec. II.E"},{"comment":"The caption does not define the color code or line styles for the matrix elements Re Q11, Im Q11, Re Q12, Im Q12, Re Q21, and Im Q21; please add a legend or explicit statement in the caption.","section":"Fig. 6"}],"recommendation":"major_revision","confidential_remarks":"The companion Letter [31] is cited as the source of wave-equation comparison, and the present manuscript is very long; please ask the editor to confirm that the overlap with the Letter is handled appropriately and that the journal's length and scope expectations are met. In addition, the saddle-point code is currently available only upon request, whereas Ebsolve is deposited; for full reproducibility, depositing the saddle-point solver as well would be preferable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know two things about this paper: it delivers a genuinely new matrix transport equation for transmission eigenvalue statistics, and it ships actual code and numerical benchmarks. The framework goes beyond DMPK by dropping the isotropy hypothesis and the diffusive limit, and it handles open geometries like the infinite slab that wave-equation solvers can't touch. The diffusive limit correctly recovers the bimodal law, and the transport equation matches the direct saddle-point solution in Fig. 3. That's real evidence, not hand-waving.\n\nThe main soft spot is exactly what the stress-test flags: the single-replica approximation in Sec. II.E. The paper assumes the Vandermonde Jacobian from Eq. (47) is negligible because the single-replica action scales as A >> 1. That's plausible, but it's asserted, not proved. The Jacobian vanishes at the replica-symmetric saddle point, so its logarithm is not obviously small, and the replica derivative at R=0 can pick up contributions that are not controlled. The paper never bounds the γ-dependence of such contributions, and Appendix C's convergence proof only covers the diffusive regime. This matters because the new quasiballistic and grazing-mode predictions live exactly where the approximation is least tested. I'm not saying the theory is wrong; I'm saying the central-claim risk is real and should be addressed head-on.\n\nOther soft spots are more minor. The infinite-slab T→0 tail is explicitly attributed to grazing modes, which the authors concede may be localization-sensitive. Evanescent modes are neglected. And the comparison to wave-equation simulations is deferred to the companion Letter rather than shown here. That's reasonable for a long paper, but it makes the in-paper validation thinner than the abstract suggests.\n\nWho is this for? Mesoscopic physicists working on wavefront shaping, disordered transport, and the statistics of transmission eigenchannels. They'll find the framework useful and the numerical method reproducible. The paper deserves a serious referee, but I'd want the authors to either control the Jacobian contribution or move the wave-equation comparison into this paper before accepting. Recommend: accept conditional on those revisions.","headline":"A serious, novel field theory for transmission eigenvalue statistics with real numerical benchmarks, but the single-replica reduction is a genuine gap that needs tightening before I'd fully trust the quasiballistic predictions.","tokens_in":38113,"tokens_out":3532,"would_cite":true,"duration_ms":38431,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A matrix transport equation derived from a replicated field theory yields the transmission eigenvalue distribution in disordered media, unifying the quasiballistic and diffusive regimes and reproducing the bimodal law in the diffusive…","keywords":["transmission eigenvalues","disordered media","field theory","matrix radiance","radiative transfer","quasiballistic transport","replica method","wavefront shaping"],"falsifier":"A practical falsifier: in a wide but finite disordered waveguide with absorbing lateral boundaries, solve the stationary wave equation (2) for many disorder realizations at optical thickness $L/\\ell = 0.2$ and compare the measured transmission eigenvalue density to the RFT curves of Fig. 5. The theory predicts a lobe structure near $T=0$ from grazing modes; if these lobes are absent or their locations scale differently with width, the semiclassical treatment of grazing modes is wrong. A more direct check: evaluate the exact replica integral retaining the Jacobian $\\Delta(Q_1,\\dots,Q_R)$ for a system with dimensionless conductance $A$ of order 1; the discrepancy with the single-replica saddle point marks where the theory breaks down.","tokens_in":37127,"feed_emoji":"🌊","tokens_out":9984,"duration_ms":89603,"temperature":0.7,"pith_summary":"The paper sets out to compute the full distribution of transmission eigenvalues—the weights of the open channels of coherent transmission through a disordered medium—starting only from the statistics of the random potential, without the isotropy or slice-stacking assumptions of standard random-matrix transport theory. The proposed framework, radiant field theory, expresses the disorder-averaged generating function through the solution of a self-consistent transport equation for a $2\\times 2$ matrix radiance $g(\\Omega,\\mathbf{r})$. Solving this equation yields the density $\\rho(T)$ for quasiballistic waveguides, where random-matrix theory fails, and for infinite slabs, an open geometry inaccessible to direct wave-equation solvers; in the diffusive limit it returns the universal bimodal law. A sympathetic reader would care because this is a first microscopic route to transmission statistics and, potentially, to wavefront-shaping observables in realistic disordered samples.","feed_headline":"Equation predicts transmission eigenvalues beyond the diffusive limit","feed_subtitle":"A transport equation yields full eigenvalue statistics from microscopic disorder, from quasiballistic to slab geometries.","key_machinery":"The central object is the matrix radiance $g(\\Omega,\\mathbf{r})$, a $2\\times 2$ matrix defined as a direction-resolved integral of the Wigner transform of the retarded Green's function over the momentum shell at $k$. It generalizes the scalar specific intensity of radiative transfer: like a radiance it depends on position and propagation direction $\\Omega$, but its off-diagonal entries carry the phase coherence needed to describe transmission eigenchannels. The load-bearing mechanism is the matrix transport equation (97), a quasiclassical transport equation that combines ballistic streaming with a self-consistent scattering term $\\frac{1}{2\\ell}[\\tilde{Q},g]$ and contact terms at the two coupling surfaces; its self-consistency condition $\\tilde{Q}(\\mathbf{r}) = \\oint d\\Omega\\, g(\\Omega,\\mathbf{r})/S_d$ closes the problem. The equation is solved numerically through a parametrization that enforces $g^2=1$ automatically and converts the evolution into linear equations for two-component vectors, integrated with an exponential integrator. This machinery turns the microscopic Gaussian disorder into a tractable transport description valid beyond the diffusion limit.","core_discovery":"The paper's central claim is that the disorder-averaged transmission eigenvalue density can be obtained from the solution of the semiclassical matrix transport equation (97) for the matrix radiance $g(\\Omega,\\mathbf{r})$. The derivation starts from a replicated Gaussian field theory for the generating function $F(\\gamma)$; in the nonlocalized regime (large dimensionless conductance $A\\gg 1$) replica interactions are negligible, the saddle point reduces to a single replica, and the saddle-point equations become a self-consistent equation for the $2\\times 2$ matrix field $\\tilde{Q}(\\mathbf{r})$. Wigner transforming this equation in the semiclassical limit gives the transport equation, whose solution $g$ satisfies $g^2=1$ and $\\operatorname{tr} g=0$, with boundary conditions that fix only three of four matrix components per direction. Inserting $g$ into the generating function gives $\\rho(T) = \\pi^{-1} T^{-2} \\operatorname{Im} F(1/T+i0^+)$. Numerical solutions reproduce the bimodal law at $L/\\ell = 5$, produce shape-dependent distributions in quasiballistic waveguides, and access the infinite-slab limit with a lobe structure near $T=0$ that the paper attributes to grazing modes.","pith_inferences":["The paper leaves implicit that the same matrix radiance $g(\\Omega,\\mathbf{r})$ could be evaluated at interior positions to compute the spatial intensity profile of an optimized wavefront—an observable that the current theory, unlike random-matrix approaches, has the microscopic machinery to address directly.","A quantitative test that the paper does not run: because the quasiballistic solution depends on the direction weights $\\mu_n$ only through the directional means (E9), two waveguides with different cross-section shapes but the same set of weights should exhibit identical quasiballistic transmission eigenvalue densities.","The grazing-mode lobes near $T=0$ are a prediction that could be probed in wide finite waveguides with absorbing lateral boundaries; if localization suppresses or shifts these lobes, the single-replica approximation would be the likely culprit, exactly as the paper warns."],"forward_implications":["In the quasiballistic regime ($L \\lesssim \\ell$), the distribution $\\rho(T)$ is no longer universal: it depends on the waveguide shape through the direction weights $\\mu_n$, and the accompanying analytical approximation (D14)–(D16) is accurate near the peak at $T=1$.","The same matrix transport equation describes an infinite slab, a geometry inaccessible to direct simulation of the wave equation; the resulting density extends down to $T=0$ and acquires a lobe structure from grazing modes as the waveguide width grows.","In the diffusive limit the framework recovers the bimodal law $\\rho(T) = \\bar{T}/(2T\\sqrt{1-T})$, showing that the microscopic theory contains the universal random-matrix result as a special case.","The derivation supplies a route to observables beyond the eigenvalue distribution, such as interior eigenchannel profiles and energy-deposition statistics, and to physical effects including absorption and incomplete channel control."],"supporting_citations":[{"why":"Supplies the random-matrix baseline, including the bimodal law and the DMPK equation, that the new theory must reproduce and goes beyond.","marker":"[14]"},{"why":"Provides the duplication technique and Green's-function route to transmission statistics from which the field-theoretic derivation grows.","marker":"[53]"},{"why":"Gives the Green's-function formula for transmission matrix elements, used to lift the eigenvalue problem into the volume.","marker":"[60]"},{"why":"Provides the quasiclassical transport equation whose structure the central matrix transport equation (97) mirrors.","marker":"[50]"},{"why":"Supplies the radiative transfer equation context and the discrete-ordinates method used for the directional integrals.","marker":"[30]"},{"why":"The numerical solver that produces the solid curves in Figs. 3–5.","marker":"[68]"},{"why":"The companion Letter, which benchmarks the waveguide predictions against wave-equation simulations and extends the framework to absorption and incomplete channel control.","marker":"[31]"}],"fun_headline_variants":["Radiant field theory yields all transmission eigenvalue statistics","Beyond diffusive limit: transport equation predicts eigenvalue distribution","From quasiballistic to slab: new theory captures transmission eigenvalues","Microscopic transport equation goes beyond DMPK for eigenvalue statistics"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole derivation assumes that the disorder average is governed by a single replica of the replicated field theory, i.e. that interactions between replicas are negligible; if that fails—as it does for strongly localized systems—the matrix transport equation and all derived distributions are invalid.","fun_headline_variants_meta":{"raw":{"variants":["Radiant field theory yields all transmission eigenvalue statistics","Beyond diffusive limit: transport equation predicts eigenvalue distribution","From quasiballistic to slab: new theory captures transmission eigenvalues","Microscopic transport equation goes beyond DMPK for eigenvalue statistics"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001834,"raw_usage":{"total_tokens":7224,"prompt_tokens":974,"completion_tokens":6250,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":590,"completion_tokens_details":{"reasoning_tokens":6181}},"tokens_in":590,"tokens_out":6250,"duration_ms":39440,"temperature":1.0,"reasoning_tokens":6181,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T19:42:01.056784+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A practical falsifier: in a wide but finite disordered waveguide with absorbing lateral boundaries, solve the stationary wave equation (2) for many disorder realizations at optical thickness $L/\\ell = 0.2$ and compare the measured transmission eigenvalue density to the RFT curves of Fig. 5. The theory predicts a lobe structure near $T=0$ from grazing modes; if these lobes are absent or their locations scale differently with width, the semiclassical treatment of grazing modes is wrong. A more direct check: evaluate the exact replica integral retaining the Jacobian $\\Delta(Q_1,\\dots,Q_R)$ for a system with dimensionless conductance $A$ of order 1; the discrepancy with the single-replica saddle point marks where the theory breaks down.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the random-matrix baseline, including the bimodal law and the DMPK equation, that the new theory must reproduce and goes beyond."}],"review_version":1}