{"id":"17ed6c8e-2b58-47ec-b0a7-71250bcab3cd","arxiv_id":"2608.06314","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Three-layer stratified water flows are cast in Hamiltonian form using Dirichlet-Neumann operators, yielding a dispersion relation and approximate wave speed formulas.","lead":"This paper derives an energy-based 'Hamiltonian' formulation for the motion of three stacked fluid layers, such as those in the ocean or in lakes, and uses it to derive explicit formulas for the speeds of surface and internal waves. This offers a systematic way to model internal waves in oceans, lakes, and laboratory tanks.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The inversion step (3.31) is not justified: the paper never establishes invertibility of G(η3), G11, Γ12, and D on the stated function spaces, and G(η3) actually has a kernel.","rationale":"The reader's weakest assumption correctly identifies the inversion of D in (3.31) as the most fragile technical premise. I agree that invertibility is asserted without proof. However, the more fundamental obstruction is the use of G^{-1}(η3): the bottom-layer DNO has a nontrivial kernel consisting of constant boundary data, so its inverse is only meaningful on a zero-mean subspace, and the paper does not verify that the arguments entering G^{-1} lie in that subspace. This strengthens the reader's concern and makes it more concrete, but it does not change the verdict: the central Hamiltonian formulation is plausible and the algebra appears internally consistent, yet the paper needs either a proof of invertibility on appropriate spaces or an explicit statement that the construction is formal/local. The leading-order symbol calculation suggests the issue is repairable, which is why I recommend keeping the conditional acceptance rather than rejecting the paper. My agreement with the reader is partial because the reader focused on D, whereas I see the kernel of G(η3) and the missing zero-mean condition as the load-bearing part of the same inversion step.","tokens_in":36667,"tokens_out":33464,"duration_ms":337775,"concrete_test":"On the flat-interface configuration, compute the leading-order symbol of the operator D in (3.31) from (4.5)–(4.7) and verify analytically that it is bounded away from zero for all k ≠ 0 and all admissible densities and depths. Then check whether the quantity Γ21(φ2)_{s2} + Γ22(φ2)_{s3} entering G^{-1} in (3.29) has zero mean for every solution of the linearized system; if it does not, the inversion is not defined on the stated Schwartz space. If the symbol vanishes for some parameter set, or if the zero-mean condition fails, the closed form (3.33) is invalid for those parameters.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The Hamiltonian representation (3.33) rests on solving the operator-valued linear system (3.30) for the five boundary traces in terms of the three momenta ξ_i. This requires inverses of G11, G(η3), Γ12, Γ21, and D, but no hypotheses on the elevations, densities, or function spaces are given under which these inverses exist. The issue is not merely technical: for the flat bottom, the bottom-layer Dirichlet-Neumann operator G(η3) has symbol k tanh(kh_b), which vanishes at k=0, so G(η3)1 = 0 and G^{-1} is not defined on all of S(R). The paper's restriction to zero-mean η_i does not automatically fix this, because the argument of G^{-1} in (3.29), namely Γ21(φ2)_{s2} + Γ22(φ2)_{s3}, is not shown to lie in the range of G(η3). Consequently the explicit closed forms (3.31)–(3.34) are formal, and the claimed reduction from five traces to three canonical variables is not established for configurations where these operators are singular or where the traces fall outside the domains of the inverses. Since the central claim is exactly that the full nonlinear system is Hamiltonian in the six variables (η_i, ξ_i), a failure of this reduction would leave the Hamiltonian defined only on a constrained submanifold, and the equivalence statement would not hold as stated. This is a gap rather than a detected algebraic error; the leading-order symbol of D appears nonzero for all k ≠ 0, so the issue is likely repairable with a suitable function-space setting and zero-mean conditions.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies two-dimensional, inviscid, incompressible, irrotational water waves in a three-layer fluid with a free surface, two interfaces, and a flat bottom. The main claim is that the full nonlinear governing equations admit a Hamiltonian formulation in six variables: the three surface/interface elevations and three momenta obtained from traces of the velocity potentials, with the Hamiltonian expressed through Dirichlet-Neumann operators for each layer. From this formulation the paper derives the linear dispersion relation, analyzes the long-wave limit as a bi-cubic equation for the squared wave speeds, gives bounds and approximate formulas for the six real speeds, and then treats the rigid-lid model, including its linear dispersion relation, the long- and short-wave limits, and a Boussinesq approximation. Limits to two-layer models are recovered in several places and compared with the literature.","tokens_in":37007,"tokens_out":5308,"duration_ms":53796,"significance":"If the technical gaps discussed below are closed, this paper would be a useful and fairly comprehensive reference: it extends the two-layer Hamiltonian/DN-operator framework of Craig, Guyenne and Kalisch to the three-layer free-surface case, provides explicit dispersion relations and root bounds, and derives a Boussinesq system for the rigid-lid model. The algebraic derivations are largely explicit, the two-layer limits are checked against known formulas, and the approximate speed formulas are tested against numerical roots in Table 1. The paper is formal-analytical rather than numerical or rigorous in its functional-analytic aspects, but the scope and the systematic use of DN operators are genuine strengths.","major_comments":[{"comment":"The reduction from the five boundary traces to the three momenta ξ_i is the load-bearing step of the Hamiltonian formulation, but the derivation uses the inverses G^{-1}(η3), G11^{-1}, Γ12^{-1}, Γ21^{-1}, and D^{-1} without stating hypotheses on the elevations, densities, or function spaces. The bottom-layer DN operator has leading symbol k tanh(k h_b), so G(η3) annihilates constants and is not invertible on all of S(R); the zero-mean condition on the η_i does not by itself put the argument Γ21(φ2)_{s2}+Γ22(φ2)_{s3} into the range of G(η3). Thus the closed forms (3.31)–(3.34) are formal, and the claimed equivalence between the Euler system and the six-variable Hamiltonian system is not established for configurations near k=0 or when the relevant traces fall outside the domains of the inverses. This issue appears repairable with a suitable zero-mean Sobolev setting and a verification of the range conditions, but as written it is a gap in the central claim.","section":"§3.4, Eqs. (3.28)–(3.32)"},{"comment":"The long-wave analysis treats the cubic P(X) as having three real positive roots: the bounds in (5.20) and the approximate speeds (5.23), (5.26), and (5.27) all rely on that assumption. The paper cites [2,4,70] for the reality of the roots, and Proposition 5.1 only proves that any real roots lie in (0, gH), not that three real roots exist. Since the dispersion relation is central and Table 1 compares approximate formulas with the numerical roots of (5.1), the proof should be supplied (for example via a discriminant computation) or the precise hypotheses under which the cited results apply should be stated.","section":"§5.1, Eq. (5.1) and Proposition 5.1"}],"minor_comments":[{"comment":"The unit normal n1 is said to be attached to the surface y=-h1+η1(x,t), but the free surface is defined in (2.1) as y=h1+η1(x,t); this is a typo in the definition of n1.","section":"§3.4, after Eq. (3.18)"},{"comment":"The variation of the trace Φ3 is written without its left-hand side; it should read δΦ3 = (φ3,y)_{s3} δη3 + (δφ3)_{s3}.","section":"Eq. (3.9)"},{"comment":"The operator D introduced in (3.31) for the inverse of the 2×2 system is distinct from the Fourier multiplier D=-i∂x defined in (4.4); using different symbols for these two objects would avoid a confusing collision of notation.","section":"§3.4 and Remark 4.1"},{"comment":"There are several typographical slips, including “Bousinessq” in Section 1 and some nonstandard hyphenation; these should be corrected during editorial processing.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThe headline: this is a competent, systematic extension of the Craig et al. Hamiltonian formalism to three-layer flows with a free surface, and the formal machinery works. The main thing to know before refereeing it: the paper never justifies the operator inversion that reduces five boundary traces to three momenta, and one of the operators involved (G(η3)) actually has a kernel. That is a repairable gap, not a dead end, but it needs to be fixed.\n\nWhat's genuinely new: the free-surface three-layer Hamiltonian written entirely in DN operators, with the explicit closed form (3.33)-(3.34). The rigid-lid dispersion relation is already in de Melo Viríssimo & Milewski and Ratliff, and the long-wave multi-layer dispersion relation is classical (Stokes/Greenhill). The paper's real contributions are the DN-based derivation of the Boussinesq system in Section 7.5, the two-layer limits, and the practical bound/approximation analysis of the bi-cubic in Section 5. The Table 1 comparison is useful and honest.\n\nThe soft spots, in order of importance. (1) The inversion step around (3.28)-(3.31) is purely formal. G(η3) is the bottom-layer DN operator with symbol k tanh(k h_b), which vanishes at k=0; its inverse is not defined on all of Schwartz space, and the argument Γ21(φ2)_s2+Γ22(φ2)_s3 is not shown to lie in its range. Similar domain issues apply to D. The Hamiltonian is only a legitimate functional on the full phase space if these inverses exist on the relevant spaces. A suitable zero-mean Sobolev setting probably fixes it, but the paper needs to say so. (2) There's no comparison with Camassa et al. [9], whose title promises Hamiltonian aspects of three-layer stratified fluids; the authors should state explicitly what is new relative to that work. (3) Minor typos: the free surface is written as y=-h1+η1 in the definition of n1, and a few notational slips. These are cosmetic.\n\nThe rest is sound: the variation computations follow the standard Legendre-transform route, the two-layer limits check out, and the cited results are real. The paper is not a paradigm shift but it is a useful reference for anyone deriving reduced models for lakes, ocean thermoclines, or lab experiments.\n\nMy recommendation: yes, send to peer review. A serious referee should push for a rigorous function-space treatment of the inversion step and a sharper statement of novelty versus [9], but the paper's core is worth refereeing.","headline":"A solid formal extension of the Craig et al. Hamiltonian machinery to three-layer free-surface flows, with a repairable gap in the operator inversion that needs a function-space fix.","tokens_in":37533,"tokens_out":3542,"would_cite":true,"duration_ms":37891,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35Q31","35Q35"],"pacs":[],"model":"deepseek-v4-flash","headline":"The full nonlinear three-layer water-wave equations admit a Hamiltonian form in six surface variables.","keywords":["three-layer water waves","Hamiltonian formulation","Dirichlet-Neumann operators","internal waves","dispersion relation","Boussinesq approximation","rigid lid model","stratified flows"],"falsifier":"Compute the symbol of $D$ for a monochromatic perturbation of wavenumber $k$ and search over admissible densities and depths for a zero eigenvalue; if one exists, the closed Hamiltonian (3.33) fails there. Alternatively, numerically solve the full linear system (4.11)-(4.12) across a random sample of ocean- and lake-like parameters and compare with the approximate speeds (5.23)-(5.27); a parameter region with relative error exceeding a few percent would delimit the approximations' validity.","tokens_in":36472,"feed_emoji":"🌊","tokens_out":7238,"duration_ms":73278,"temperature":0.7,"pith_summary":"Three-layer water flows — a free surface, two internal interfaces, a flat bottom, and constant density in each layer — obey strongly coupled nonlinear equations. The paper claims that, for irrotational flow, these equations are exactly equivalent to a Hamiltonian system in six one-dimensional variables: the three elevations $\\eta_1,\\eta_2,\\eta_3$ and three momenta $\\xi_1,\\xi_2,\\xi_3$ built from density-weighted traces of the velocity potentials. The Hamiltonian is written explicitly using Dirichlet-Neumann operators for each layer, so all information about the fluid interior is condensed into boundary quantities. From this formulation the paper derives the linear dispersion relation, analyzes its long-wave limit where the squared wave speed $c^2$ satisfies a bi-cubic equation, and provides bounds and approximate formulas for the six real propagation speeds. A sympathetic reader would care because this gives one energy-based framework from which surface and internal wave models — including rigid-lid, Boussinesq, and two-layer limits — follow systematically.","feed_headline":"Three-layer wave equations become one Hamiltonian system","feed_subtitle":"Six surface variables carry the full nonlinear dynamics; long-wave speeds follow from a single cubic equation.","key_machinery":"The carrying object is the Dirichlet-Neumann operator for each layer: the boundary map that sends a harmonic function's trace on a layer's boundary to its normal derivative, with the surface geometry encoded in the elevations. The top and middle layers each get a $2\\times2$ operator matrix and the bottom layer a single operator; their leading Fourier-multiplier symbols are $k\\coth(kh_i)$, $k\\,\\mathrm{csch}(kh_i)$ and $k\\tanh(kh_b)$. The algebraically central object is the operator $D=(\\rho_2 G+\\rho_3\\Gamma_{22})\\Gamma_{12}^{-1}(\\rho_2G_{11}+\\rho_1\\Gamma_{11})-\\rho_1\\rho_3\\Gamma_{21}$, whose invertibility lets the five potential traces be written as functions of $(\\eta_i,\\xi_i)$; this converts the total energy into the explicit Hamiltonian (3.33). The paper then uses systematic Taylor expansions of the DN operators in the elevations to move from the nonlinear Hamiltonian equations to the linear dispersion relation and, in the rigid-lid case, to a Boussinesq system.","core_discovery":"The central claim is a Hamiltonian reformulation of the full nonlinear three-layer irrotational water-wave problem. With elevations as coordinates and momenta defined by $\\xi_1=\\rho_1\\Phi_1^1$, $\\xi_2=\\rho_2\\Phi_2^2-\\rho_1\\Phi_1^2$, $\\xi_3=\\rho_3\\Phi^3-\\rho_2\\Phi_2^3$, the authors show that the Euler equations and all boundary conditions are equivalent to $\\delta H/\\delta \\eta_i=-\\xi_{i,t}$ and $\\delta H/\\delta \\xi_i=\\eta_{i,t}$ for $i=1,2,3$. The kinetic energy is expressed through three-layer Dirichlet-Neumann operators; inverting an operator $D$ expresses the five velocity-potential traces in terms of the three momenta, yielding a closed Hamiltonian $H(\\eta_i,\\xi_i)$ with coefficients $A_{ij}$ built from the DN operators. Linearization about the rest state produces a $6\\times 6$ system whose characteristic equation is the dispersion relation; in the long-wave limit it becomes a cubic in $c^2$, whose six real roots are the three right-moving and three left-moving wave speeds. The same DN machinery, with the upper surface fixed flat, yields the rigid-lid system, its bi-quadratic dispersion relation, a coupled Boussinesq system, and the two-layer free-surface limits.","pith_inferences":["The six-variable reduction likely extends to $N$ layers by continuing the density-weighted momentum definition, since equations (3.38)-(3.51) already display a recursive structure; this is an extrapolation, not proven here.","Because the Boussinesq and rigid-lid systems are obtained by expanding DN operators inside a Hamiltonian framework, one would expect them to inherit exact energy conservation; a numerical check of conserved quantities would be a direct test.","The approximate formulas (5.23)-(5.27) could be used to predict the three mode speeds in a laboratory three-layer tank, and the predicted polarity relation between interface displacements is a concrete observable signature.","A spectral check of $D$ for the extreme-density laboratory cases in Table 1 would show whether the closed-form Hamiltonian (3.33) remains valid outside the oceanographic parameter range."],"forward_implications":["A single energy functional now generates the full nonlinear motion of the free surface and both interfaces, so no separate interface-by-interface derivation is needed.","Linearizing the Hamiltonian equations gives the dispersion relation at arbitrary wavelength, and its long-wave limit is a bi-cubic equation in $c^2$ with six real roots: three right-moving and three left-moving modes with speeds of different orders.","The bounds and approximations in Section 5 give practical estimates of the three positive speeds directly from densities and layer thicknesses, and they match numerical roots for ocean, lake, and laboratory parameter sets.","In the rigid-lid case the same DN formalism yields a bi-quadratic dispersion relation whose two positive speeds have opposite interface polarities, and a Boussinesq system follows by expanding the operators.","Setting $\\rho_3=\\rho_2$ or $\\rho_1=0$ recovers the known two-layer free-surface dispersion relation and Boussinesq system, so the three-layer model contains earlier two-layer models as limits."],"supporting_citations":[{"why":"Supplies the Hamiltonian long-wave expansion method and the two-layer free-surface dispersion relation and Boussinesq system that this paper extends and recovers.","marker":"[26]"},{"why":"Provides the Legendre-transform ansatz that defines the momenta from weighted potential traces on surfaces and interfaces.","marker":"[27]"},{"why":"Provides the linear-system/exponential solution framework and equatorial internal-wave parameter values used in the dispersion analysis.","marker":"[19]"},{"why":"Establishes the Hamiltonian formulation for stratified rotational flows whose density-weighting structure motivates the momenta here.","marker":"[20]"},{"why":"Introduces the variable-reduction ansatz for interfacial problems that reduces the number of dynamical variables.","marker":"[6,7]"},{"why":"Supplies the recursive DN-operator expansion technique used to obtain model equations in various regimes.","marker":"[42]"},{"why":"Provides the three-layer shallow-water long-wave dispersion relation and amplitude-polarity relations used for comparison.","marker":"[53]"},{"why":"Gives the root bounds and classification for the cubic equation used to locate the six wave speeds.","marker":"[58,59]"},{"why":"Supplies the classic two-layer superimposed-liquids dispersion relation recovered as the two-layer limit.","marker":"[47]"}],"fun_headline_variants":["Three-layer waves fold into a single Hamiltonian system","Six wave speeds emerge from three-layer Hamiltonian","Dirichlet-Neumann operators unify three-layer water waves","Three-layer wave dynamics becomes one Hamiltonian math","Hamiltonian structure tames three-layer water flows"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The derivation assumes that the operator $D$ used to solve for the velocity-potential traces can be inverted on the spaces of allowed surface elevations and densities; the paper does not prove this, and if $D$ degenerates for some configuration the explicit closed-form Hamiltonian is not valid.","fun_headline_variants_meta":{"raw":{"variants":["Three-layer waves fold into a single Hamiltonian system","Six wave speeds emerge from three-layer Hamiltonian","Dirichlet-Neumann operators unify three-layer water waves","Three-layer wave dynamics becomes one Hamiltonian math","Hamiltonian structure tames three-layer water flows"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000348,"raw_usage":{"total_tokens":1998,"prompt_tokens":1132,"completion_tokens":866,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":748,"completion_tokens_details":{"reasoning_tokens":794}},"tokens_in":748,"tokens_out":866,"duration_ms":8725,"temperature":1.0,"reasoning_tokens":794,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T05:37:26.687625+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the symbol of $D$ for a monochromatic perturbation of wavenumber $k$ and search over admissible densities and depths for a zero eigenvalue; if one exists, the closed Hamiltonian (3.33) fails there. Alternatively, numerically solve the full linear system (4.11)-(4.12) across a random sample of ocean- and lake-like parameters and compare with the approximate speeds (5.23)-(5.27); a parameter region with relative error exceeding a few percent would delimit the approximations' validity.","supporting_citations":[{"cited_title":"Craig, P","cited_arxiv_id":null,"evidence_quote":"Supplies the Hamiltonian long-wave expansion method and the two-layer free-surface dispersion relation and Boussinesq system that this paper extends and recovers."},{"cited_title":"Craig, P","cited_arxiv_id":null,"evidence_quote":"Provides the Legendre-transform ansatz that defines the momenta from weighted potential traces on surfaces and interfaces."},{"cited_title":"de Melo Vir ´ ıssimo, P.A","cited_arxiv_id":null,"evidence_quote":"Provides the three-layer shallow-water long-wave dispersion relation and amplitude-polarity relations used for comparison."},{"cited_title":"Lamb,Hydrodynamics, 6 ed., Cambridge University Press, Cambridge, 1932","cited_arxiv_id":null,"evidence_quote":"Supplies the classic two-layer superimposed-liquids dispersion relation recovered as the two-layer limit."}],"review_version":1}