{"id":"2b46227c-654f-4048-ba2d-bc3fdfd6effa","arxiv_id":"2412.04439","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"The undercompressive shock surface in three-phase flow has the same boundary structure for identity and capillary-pressure diffusion matrices, and the diffusion choice changes which Riemann solutions are admissible.","lead":"This paper maps the full set of undercompressive shock waves in a standard three-phase flow model, and shows the map has the same topological structure whether diffusion is simple or comes from capillary pressure. It then demonstrates that swapping the diffusion model can remove the undercompressive shock from a concrete Riemann solution.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Topological-similarity claim for B≠I rests on a single-seed continuation with no completeness certificate; missed disconnected components would invalidate the comparison.","rationale":"The reader's weakest assumption identifies exactly the load-bearing gap: the numerical construction for B≠I has no exhaustiveness guarantee. The paper explicitly rests the structural conclusion on extensive numerical experiments (§7.3), and the method seeds from a single saddle-saddle connection near the umbilic point (§7.1), so disconnected components of T could be missed. The analytic construction for B=I is a genuine independent contribution, and the qualitative agreement found numerically is plausible; nothing in the text is internally inconsistent. However, the central claim of topological similarity between the two diffusion matrices is only as strong as the completeness of the numerical search. A global, seed-independent search for the parameter set used in §7.4 would settle the question directly. The reader's CONDITIONAL verdict is therefore appropriate and does not need adjustment.","tokens_in":34382,"tokens_out":3060,"duration_ms":41744,"concrete_test":"Reproduce the type II O case (µw=1.0, µo=2.0, µg=0.75, cow=cog=1) with a search that does not use the umbilic seed: cover the saturation triangle with a fine grid; for each U−, compute the Hugoniot arc H(U−), classify its saddle equilibria for the ODE (16), and detect heteroclinic connections by forward/backward integration of stable and unstable manifolds; record U− if a connection exists. Compare the computed domain DT and boundary types with Fig. 12(a) and §7.3. If the independent search finds any U− outside the displayed regions, or fails to find a displayed one, the claimed topological similarity is not established.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 7.3 concludes that the undercompressive surface for B≠I has 'the same boundaries' as for B=I, explicitly on the basis of 'extensive numerical experiments.' The construction in §7.1 begins from one saddle-saddle connection near the umbilic point obtained from the quadratic approximation [4,16], then continues it over a grid using bisection (§7.2, Algorithms 1–3). This procedure can only discover the connected component of T containing the seed. Nothing in the paper rules out additional components of the undercompressive surface detached from the umbilic neighborhood, or holes bounded by different boundary types. The quadratic approximation guarantees local existence near the umbilic and structural stability under perturbations gives robustness of that local piece, but neither extends to global completeness. The boundary count and type in the four cases listed in §7.3 would change if such an extra component existed; the central claim would then be false. The absence of code or data prevents an independent check. This is not an internal inconsistency—the analysis for B=I is largely self-contained—but the transfer to B≠I is underdetermined by the evidence presented.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies undercompressive shock waves for a 2x2 system of conservation laws modeling three-phase flow in porous media with quadratic Corey relative permeabilities. For the identity diffusion matrix B(U)=I, the authors exploit invariant lines to reduce the system to scalar Buckley-Leverett equations, derive explicit formulas for the undercompressive interval endpoints and for the characteristic boundaries, and construct the undercompressive shock surface T as a ruled surface in the three-dimensional space of saturations and speed. For the capillary-pressure diffusion matrix B(U)≠I, they compute saddle-saddle connections numerically by continuation and bisection, and they claim that the resulting undercompressive surface has the same boundary structure as in the identity case. The paper closes with numerical simulations of a Riemann problem showing that the choice of B changes the admissibility of the nonlocal shock segment and can eliminate the undercompressive shock from the solution.","tokens_in":34647,"tokens_out":4485,"duration_ms":53370,"significance":"If the results hold, the paper provides the first systematic construction of the undercompressive shock surface for a physically motivated capillary diffusion matrix, alongside an explicit analytical construction for the identity matrix. The analytic part of the paper is a genuine strength: Sections 4--6 contain closed-form expressions for the endpoints S, F, B0, B1, B2, YΓ, and YΓ, and the reduction along invariant lines is derived from the governing equations rather than assumed. The paper also makes a concrete, falsifiable prediction: that the shape of the undercompressive surface is insensitive to the choice of diffusion matrix, while the solution of the Riemann problem can change qualitatively. However, the central structural claim for B≠I rests on numerical continuation from a single seed, with no completeness certificate, and one key analytic claim is justified by a figure rather than a proof. The absence of code or data also prevents independent verification of the numerical experiments.","major_comments":[{"comment":"The central claim that 'when calculating the undercompressive boundaries with B(U) ≠ I, we obtain the same boundaries as those discussed for B(U) = I' is supported only by 'extensive numerical experiments.' The construction in §7.1 starts from one saddle-saddle connection near the umbilic point, obtained from the quadratic approximation in [4,16], and then continues it over a grid using the bisection procedures of Algorithms 1--3. Such a continuation can only discover the connected component of the undercompressive surface T that contains the seed; nothing in the paper rules out additional disconnected components, or holes bounded by different boundary types, away from the umbilic neighborhood. If such a component existed, the boundary count and the claimed topological similarity in the four cases enumerated in §7.3 would fail. This is not an internal inconsistency, but the evidence presented does not determine the global structure of T. The authors should either provide a completeness argument for the continuation, demonstrate coverage by multiple independent seeds and sweeping strategies, or soften the conclusion to a conjecture. In addition, providing the code/data used for Figures 12--14 would allow independent checking.","section":"§7.3"},{"comment":"Claim 6.1, which states that for 0 < νG ≤ 1 there is no left state L on the invariant line [G,D) forming an undercompressive shock with a right state R in (U, D0], is justified by inspection of the speed diagram in Figure 9: the proof says 'According to Fig. 9' and identifies the Bethe-Wendroff point S and the states F and F1 graphically. This claim is load-bearing because it establishes the gap between U and D0 and thereby determines the boundary structure in Lemma 6.3 and in the analogous numerical case 2 of §7.3. A figure-based argument is not a proof, especially since the same figure is used to assert that no orbit connects M to R along the invariant line. The authors should supply an analytic proof, or at least a machine-checkable computation, of the ordering and the absence of an admissible connection.","section":"§6, Claim 6.1"},{"comment":"The paper states that 'numerical solid evidence shows that the mixed double contact locus exists only when the umbilic point lies in one of the regions of type II' and then uses the position of XG to decide whether a compatibility-loss boundary appears (Remark 6.3, Figure 8(b)). This assertion is not proven, and it affects the boundary classification for the identity case as well as the interpretation of the non-identity case. If the 'only when' statement is not known analytically, it should be marked as a numerical observation, and its role in the boundary classification should be made explicit.","section":"§4.3 and Remark 6.3"}],"minor_comments":[{"comment":"The sentence 'the Rankine-Hugoniot condition constitutes two equations for the five variables U+, U+, and σ' contains a typo: it should read U-, U+, and σ.","section":"§3"},{"comment":"Definition 6.1 says 'the set of pairs of states pS and xM in Ω × R+', but pS and xM are triples in Ω × R+, not pairs. Rephrase to 'the set of points (U-, σ) and (U+, σ) in Ω × R+' or the equivalent.","section":"§6, Definition 6.1"},{"comment":"The phrase 'For νΓ > 8 por (νΓ−)2/νΓ > 8' contains a typo: 'por' should be 'or'.","section":"§5, Corollary 5.1"},{"comment":"The algorithms rely on a 'Sotomayor line s', which is described only as 'a line conveniently placed in the phase space'. A precise definition or a reference for the Sotomayor line would make the numerical procedure reproducible.","section":"Appendix A"},{"comment":"The caption of Figure 12(b) mentions the z axis but does not state its meaning; the text identifies it as shock speed, so the caption should be updated accordingly.","section":"Figure 12"},{"comment":"The notation νΓ− is introduced in Definition 4.2 and used later as (νΓ−)2/νΓ, but the connection between the superscript '−' and the difference μα−μβ could be made more prominent, since it is easy to confuse with a negative exponent.","section":"§4.3"}],"recommendation":"major_revision","confidential_remarks":"The paper is within the scope of the journal and the analytic construction for B=I is a solid contribution. My main reservation is that the headline claim for B≠I is a numerical observation without a completeness certificate, and one of the key analytic claims is justified by a figure. I would be willing to reconsider after the authors either supply a proof or explicitly reframe the B≠I structural claim as a conjecture supported by experiments, and after they make the numerical evidence reproducible. The heavy reliance on the authors' own prior work, especially [30], is not inappropriate but should be supplemented by making the algorithms and data available."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Worth your time if you care about nonclassical shocks in three-phase flow. The genuinely new thing is Sections 4-6: an explicit ruled-surface construction of the full undercompressive shock surface for the identity viscosity matrix, using the invariant lines and closed-form formulas for the boundary states S and F. That part is clean, mostly self-contained, and the boundary lemmas (6.1-6.3) follow properly from the analytic Theorem 5.1/5.2. The comparison with the capillary matrix (Section 7) is also new, and Section 7.4's example - where switching diffusion makes the undercompressive shock disappear from the Riemann solution - is a nice, concrete illustration that the diffusion matrix matters.\n\nThe soft spots are real but localized. First, Claim 6.1 is justified by a figure (Fig. 9), not by an analytic argument. The claim matters for the nu_Gamma <= 1 case, where the undercompressive surface has a gap, and a referee should ask for a proof or a more rigorous numerical verification. Second, and more important, the central structural claim for B != I - that the surface has exactly the same boundary types as for B = I - rests on 'extensive numerical experiments' with no code, no data, and no completeness or convergence certificate. The stress-test note worries about single-seed continuation, but Section 7.3 also describes a grid-based contour search, so the actual procedure is broader than the note suggests. That said, the broader procedure still cannot rule out disconnected components or holes smaller than the grid resolution. So the right summary is: the B != I comparison is plausible and well-illustrated, but it is not established to the same standard as the identity case. That should be stated more cautiously in the text, and the code/data should be archived.\n\nThe self-citation pattern is heavy but not abusive: the wave-manifold framework and several lemmas come from prior work, and the paper builds on them honestly. I would send this to a serious referee, mainly for the analytic construction and the numerical comparison, with the expectation that the B != I claim be reframed as a numerical conjecture or supported by a completeness argument. I would not cite the B != I topological claim in my own work until that is settled.","headline":"Solid analytic construction of the undercompressive surface for the identity matrix plus a plausible numerical analogue for capillary diffusion, but the topological-similarity claim for B != I is a numerical finding without a completeness certificate.","tokens_in":689,"tokens_out":1347,"would_cite":false,"duration_ms":91012,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35C06","35D30","76S05","76T30"],"pacs":[],"model":"deepseek-v4-flash","headline":"The undercompressive shock surface for capillary-pressure diffusion has the same four boundary types as the identity-diffusion surface, yet the diffusion matrix can remove the undercompressive shock from a Riemann solution.","keywords":["undercompressive shock waves","three-phase flow in porous media","viscous profile criterion","capillary pressure","Riemann problem","wave manifold","saddle-saddle connection","non-strictly hyperbolic conservation laws"],"falsifier":"Run the same continuation for $B(U)\\neq I$ seeded from a saddle-saddle connection near the opposite edge of the saturation triangle, away from the umbilic point, and compare the resulting boundary curves with the umbilic-seeded surface; the paper's claim predicts that no new boundary component appears. Alternatively, repeat the construction for an anisotropic capillary matrix with $c_{ow}\\neq c_{og}$ and check whether the four boundary types persist.","tokens_in":34219,"feed_emoji":"🌊","tokens_out":9242,"duration_ms":80119,"temperature":0.7,"pith_summary":"This paper constructs the complete surface of undercompressive shock waves for three-phase flow in porous media, under two diffusion matrices: the identity matrix and the physical capillary-pressure matrix. For the identity matrix the construction is analytic, reducing along invariant lines to a scalar two-phase conservation law and producing explicit formulas for the boundary states. For the capillary matrix the construction is numerical, tracking saddle-to-saddle connections of the traveling-wave ODE. The central claim is that the two surfaces have the same topological structure, with the same four boundary types appearing in the same configurations for each umbilic-point type. The paper then shows, in a concrete Riemann problem, that the capillary matrix can eliminate the undercompressive shock from the solution, so the diffusion model changes which wave sequences are admissible.","feed_headline":"Shock surface survives switch to capillary diffusion","feed_subtitle":"Two diffusion matrices share four boundary types, but capillary diffusion can erase the undercompressive shock.","key_machinery":"The load-bearing object is the undercompressive surface $T$ inside the wave manifold $W$, the three-dimensional set of shock triples $(U^-, U^+, \\sigma)$ satisfying the shock jump condition after a blow-up removes the trivial solutions $U^-=U^+$. For $B(U)=I$, the construction uses the invariant lines along which the three-phase system reduces to a scalar two-phase conservation law; the paper derives the effective flux and explicit boundary-state formulas, making $T$ a ruled surface over the segments $[G,D]$, $[W,E]$, and $[O,B]$. For $B(U)\\neq I$, the machinery is the traveling-wave ODE $dU/d\\eta = \\mathrm{Adj}(B(U))[F(U)-F(U^-)-\\sigma(U-U^-)]$ together with a bisection algorithm that measures the signed distance between the stable and unstable manifolds of the two saddle equilibria along a transversal line, locating the saddle-to-saddle and saddle-to-saddle-node connections that form the boundary of $T$.","core_discovery":"An undercompressive shock is a discontinuity whose viscous profile connects two saddle equilibria of the traveling-wave system, so it satisfies the viscous profile criterion rather than the classical shock inequalities. The paper establishes that the undercompressive surface $T$ inside the wave manifold has the same boundary structure for $B(U)=I$ and for the capillary-pressure matrix (10)-(11): slow characteristic, fast characteristic, undercompressive characteristic, and genuine undercompressive boundaries, arranged in the same patterns for umbilic points of type I and type II and for each viscosity-ratio regime. For $B(U)=I$, $T$ is a ruled surface lying over the invariant lines $[G,D]$, $[W,E]$, and $[O,B]$, with explicit formulas for the boundary states $D_0$, $D_1$, $D_2$, $Y_\\Gamma$, and $Y^\\Gamma$; for $B(U)\\neq I$, the surface is found numerically and projects to two-dimensional regions of the saturation triangle. The same Riemann problem with states $L=(0.475708, 0.02608)^T$ and $R=(0.235578, 0.670876)^T$ produces different solutions: with $B(U)=I$ it contains an undercompressive shock, while with the capillary matrix the undercompressive shock is absent and a different fast shock segment becomes admissible.","pith_inferences":["Because the boundary types are carried by saddle-to-saddle and saddle-to-saddle-node connections, which are structurally stable, the same four-boundary taxonomy is likely to persist for other diffusion matrices in the same physical class; this is an inference beyond the paper's two matrices.","A direct test of exhaustiveness would seed the numerical continuation from each of the 31 umbilic-region configurations and from boundary states away from the umbilic point; the paper's claim predicts the same boundary-type counts in every region.","If the capillary matrix can remove the undercompressive shock in a single Riemann problem, then in heterogeneous reservoir simulations models with $B(U)=I$ may over-predict non-classical waves; computing the Riemann solution for the same $L$, $R$ with the two matrices at finite grid resolution would make this quantitative."],"forward_implications":["Riemann solvers can reuse the same four boundary-type logic for both diffusion models, only replacing the analytic surface by the numerically computed one.","For the capillary matrix, undercompressive shocks occupy a two-dimensional region of the saturation triangle rather than one-dimensional invariant-line segments, so capillary diffusion broadens the set of left and right states that can be joined by an undercompressive shock.","The admissibility of nonlocal shock segments changes with the diffusion matrix: a segment that is only partly admissible for $B=I$ can become fully admissible for the capillary matrix.","There exist Riemann problems whose solution contains an undercompressive shock for $B=I$ but not for the capillary matrix, so the viscous profile criterion ties the wave sequence to the physical diffusion model.","For viscosity ratio $\\nu_\\Gamma\\leq 1$ the undercompressive surface has a gap between $U$ and $B_0$, so undercompressive shock amplitude is bounded away from zero and the gap is filled by transitional rarefaction waves."],"supporting_citations":[{"why":"introduced transitional (undercompressive) waves and the viscous profile construction that defines the shocks studied here.","marker":"[4]"},{"why":"showed that for $B=I$ the admissible non-classical shocks lie on invariant lines, the basis of the analytic construction.","marker":"[9]"},{"why":"classifies the umbilic point as type I or II for the quadratic-permeability model, fixing the boundary-type regimes.","marker":"[11]"},{"why":"exhibits undercompressive shocks for the quadratic approximation near the umbilic point, the seed for the numerical search.","marker":"[16]"},{"why":"proved structural stability of undercompressive shocks under viscosity matrix perturbations, supporting the similarity claim.","marker":"[20]"},{"why":"derives the capillary-pressure diffusion matrix used for $B(U)\\neq I$ and the instability region governing it.","marker":"[21]"},{"why":"provides the wave-manifold global formalism in which the undercompressive surface $T$ is embedded.","marker":"[23]"},{"why":"gives the analytic shock-locus branches and uniqueness results for the identity case used in the explicit formulas.","marker":"[29]"},{"why":"develops the global wave-curve continuation and diffusive Riemann solution framework used for the numeric construction.","marker":"[30]"}],"fun_headline_variants":["Capillary diffusion can erase undercompressive shocks","Same shock surface, different Riemann solutions","Undercompressive shock structure survives diffusion swap","Capillary matrix alters shock outcomes in porous flow","Diffusion choice decides undercompressive shock presence"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The numerical claim that the capillary-pressure surface has the same structure as the identity surface rests on the search finding the entire undercompressive surface, but the search is seeded from one saddle-saddle connection near the umbilic point and continued over a grid, with no proof that a disconnected component of the surface cannot be missed.","fun_headline_variants_meta":{"raw":{"variants":["Capillary diffusion can erase undercompressive shocks","Same shock surface, different Riemann solutions","Undercompressive shock structure survives diffusion swap","Capillary matrix alters shock outcomes in porous flow","Diffusion choice decides undercompressive shock presence"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00068,"raw_usage":{"total_tokens":3078,"prompt_tokens":925,"completion_tokens":2153,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":541,"completion_tokens_details":{"reasoning_tokens":2085}},"tokens_in":541,"tokens_out":2153,"duration_ms":19772,"temperature":1.0,"reasoning_tokens":2085,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T21:24:15.164723+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same continuation for $B(U)\\neq I$ seeded from a saddle-saddle connection near the opposite edge of the saturation triangle, away from the umbilic point, and compare the resulting boundary curves with the umbilic-seeded surface; the paper's claim predicts that no new boundary component appears. Alternatively, repeat the construction for an anisotropic capillary matrix with $c_{ow}\\neq c_{og}$ and check whether the four boundary types persist.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"introduced transitional (undercompressive) waves and the viscous profile construction that defines the shocks studied here."},{"cited_title":"Isaacson, D","cited_arxiv_id":null,"evidence_quote":"showed that for $B=I$ the admissible non-classical shocks lie on invariant lines, the basis of the analytic construction."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"exhibits undercompressive shocks for the quadratic approximation near the umbilic point, the seed for the numerical search."},{"cited_title":"Marchesin, A","cited_arxiv_id":null,"evidence_quote":"proved structural stability of undercompressive shocks under viscosity matrix perturbations, supporting the similarity claim."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"derives the capillary-pressure diffusion matrix used for $B(U)\\neq I$ and the instability region governing it."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"provides the wave-manifold global formalism in which the undercompressive surface $T$ is embedded."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"gives the analytic shock-locus branches and uniqueness results for the identity case used in the explicit formulas."},{"cited_title":"Lozano, Diffusive effects in Riemann solutions for the three phase flow in porous media, Ph.D","cited_arxiv_id":null,"evidence_quote":"develops the global wave-curve continuation and diffusive Riemann solution framework used for the numeric construction."}],"review_version":1}