{"id":"6ae58b56-725c-4a0c-9c16-91e25bc32436","arxiv_id":"2608.02223","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"A DG discretization of the IGR-regularized Euler equations stabilizes shocks and preserves fine-scale flow features in 1D/2D benchmarks without limiters or artificial viscosity.","lead":"A new discontinuous Galerkin discretization is built for the information geometric regularized (IGR) compressible Euler equations, where an auxiliary elliptic equation produces an 'entropic pressure' that smooths shocks. Benchmark tests in 1D and 2D indicate the scheme stabilizes shocks without limiters or artificial viscosity, though accuracy in smooth regions is degraded.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Fig. 2 contradicts 'minimal activation': IGR-DG converges at 2nd order vs 4th order for p=3 in a smooth pre-shock simple wave, so the regularization measurably perturbs smooth flow, undercutting the central claim.","rationale":"The reader's weakest-assumption analysis already identifies the same load-bearing issue: the 'minimal Σ activation in smooth regions' premise is contradicted by the paper's own Figure 2 pre-shock convergence study (2nd order vs 4th order for p=3). My independent reading confirms this is the strongest single concern. The method may still be a useful contribution as a shock-stabilization mechanism, and the availability of both Julia and MFEM implementations is real supporting evidence. But the advertised property of selective, non-intrusive PDE-level regularization — the basis for claiming an advantage over limiter/viscosity methods — is not supported by the data as presented. The concern is not that IGR disagrees with conventional shock capturing; it is that the paper's own convergence results show the regularization perturbs smooth flow. This warrants keeping the CONDITIONAL verdict rather than accepting the central claim as stated. I do not see a reason to move to REJECT: the numerical method appears stable across many benchmarks, and the issue is empirically addressable with better reference solutions and an explicit α-selection/scaling rule.","tokens_in":29993,"tokens_out":5577,"duration_ms":54610,"concrete_test":"Reproduce Example 1 on the same grids with p=3 through t=0.2 (pre-shock), using three variants: (i) IGR-DG with fixed α0=0.3 as in Fig. 2; (ii) IGR-DG with α=α0 h^2 scaling as in §5.1.3; (iii) the same DG discretization with Σ≡0. Compare L1 density errors to the exact smooth simple-wave solution, valid until shock formation. If (i) shows ~2nd order and (iii) ~4th order while (ii) recovers high order, the degradation is caused by the entropic pressure and is removable only by making α mesh-dependent — a material qualification to the claim of a fixed PDE-level regularization. Additionally, tabulate max|Σ| vs h in a smooth run; under fixed α it should not vanish as h→0, directly falsifying 'minimal activation.'","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central value proposition is that the entropic pressure is 'minimal in smooth regions' (abstract; §2.1 'Selective Σ activation'), so IGR-DG preserves high-order DG accuracy away from shocks. The load-bearing condition is that the right-hand side of Eq. (5), (tr Du)^2 + tr((Du)^2), is negligible for smooth flows. It is not: any non-uniform smooth velocity field gives a nonzero right-hand side, and Eq. (5) then produces a nonzero Σ at O(α) for fixed α in exactly the regions the abstract calls 'minimal.' The paper's own Figure 2 and §5.1.1 quantify the consequence: before shock formation in the Example 1 simple wave, IGR-DG at p=3 converges at only second order while DG+limiter converges at fourth order. Section 6 concedes this directly: 'although IGR-DG reduces the order of accuracy before shock formation.' Since P+Σ enters the momentum and energy fluxes globally, the smooth flow is being modified by the regularization, not preserved. This is an internal inconsistency between the abstract's 'minimal activation' claim and the paper's own convergence data, not merely a disagreement with an external consensus. The claim that IGR-DG 'retains fine-scale flow features as polynomial order is increased' is most valuable in precisely the smooth/under-resolved regions where this accuracy loss occurs.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a discontinuous Galerkin (DG) discretization of the information-geometric regularization (IGR) of the compressible Euler equations. The entropic pressure Sigma is obtained from an auxiliary elliptic equation (Eq. 5), discretized with SIPG in the same DG space as the hyperbolic conservation law, and enters the Euler fluxes only through the augmented pressure P+Sigma (Eq. 7). The authors test the resulting IGR-DG method on one- and two-dimensional benchmarks, including shock formation, smoothed Riemann problems, Shu-Osher-type fine-scale problems, shock-bubble interaction, and a supersonic cylinder, comparing against a TVB-limited DG baseline. They claim shock stabilization without shock-capturing limiters or artificial viscosity, preservation of fine-scale features, and selective activation of Sigma concentrated in strongly compressive regions.","tokens_in":30318,"tokens_out":6264,"duration_ms":56394,"significance":"If the central claims hold, the paper would provide a useful PDE-level alternative to limiter- and viscosity-based shock capturing for high-order DG, with the practical advantage that the regularization is independent of mesh structure and polynomial order details. The availability of both a Julia implementation and an MFEM implementation is a strength and supports reproducibility. However, the significance is conditional on two load-bearing points: (i) the entropic pressure really is 'minimal' in smooth regions so that high-order accuracy is preserved, and (ii) the reported error comparisons are against trustworthy reference solutions. Both points are currently problematic, as detailed below.","major_comments":[{"comment":"The abstract and §2.1 state that Sigma has 'minimal activation in smooth regions' and that the regularization preserves the underlying solution elsewhere. This is contradicted by the paper's own data. For any non-uniform velocity field, the right-hand side of Eq. (5) is nonzero, so at fixed alpha > 0 the entropic pressure is generically O(alpha) in smooth regions and enters the momentum and energy fluxes through Eq. (7). Figure 2 shows that, before shock formation in Example 1, IGR-DG at p=3 converges at only second order in density, momentum, and energy, while the DG+limiter baseline converges at fourth order. Section 6 explicitly concedes that 'IGR-DG reduces the order of accuracy before shock formation.' This is not a localized, negligible effect: it is a systematic modification of smooth flow. The claim of 'minimal activation' must be substantially revised or quantified, because it i","section":"Abstract, §2.1, Eq. (5), Fig. 2, §6"},{"comment":"For all smoothed Riemann problems, the reference solutions are obtained from fine-grid DG+limiter simulations. Since the initial data are smoothed, the classical self-similar solutions are not exact, so the reported L1 errors measure the difference from the baseline method's output rather than from a physical/analytic solution. This is especially problematic in the modified Shu-Osher example, where the baseline is claimed to damp fine-scale oscillations: if the reference itself is smoothed, IGR-DG's lower 'error' may partly reflect a better match to the reference's smoothing, or a different kind of smoothing. Similarly, the p-refinement study in §5.1.3 uses an IGR-DG solution with m_ref=1024, p_ref=3, and alpha_0,ref=1.0 as the reference, which favors the method being compared. The quantitative convergence and error comparisons in Figs. 2, 4, 6, 8, and 9 need a more neutral reference, su","section":"§5.1, paragraph after Eq. (32); §5.1.3, Fig. 10"},{"comment":"The regularization strength alpha_0 is selected per problem to ensure stable evolution (e.g., 0.3 for the simple wave, 0.5 or 0.7 for Sod, 4.0 in the fixed-DOF Shu-Osher study, 32.0 for the shock-bubble case). Figure 12 provides only a p=1 threshold curve for a sinusoidal compressive wave and gives no predictive formula or criterion for choosing alpha_0 for a new problem. The claim in §5.1.4 that this provides a 'physically interpretable basis for parameter selection rather than ad hoc tuning' is not supported by the experiments. Furthermore, the p-refinement scaling alpha = alpha_0 h_eff^2 in §5.1.3, with h_eff = 2h/(p+1), is introduced without derivation or sensitivity study; the conclusion that IGR-DG has 'nearly constant error' across p may depend on this specific scaling and on the choice alpha_0=4.0. A method advertised as removing limiter-parameter tuning should come with a more p","section":"§5.1.4, §5.1.3, parameter selection"},{"comment":"The abstract says the method 'stabilizes shocks without shock-capturing limiters or artificial viscosity,' with a caveat that positivity-preserving methods may still be required. In the supersonic-cylinder example (Example 11), a Zhang-Shu positivity-preserving limiter is enabled. This is not necessarily a logical contradiction, since the limiter is not a shock-capturing limiter, but the central comparison in the conclusion says 'without shock-capturing limiters' rather than 'without any limiter.' The role of the positivity limiter in the benchmark suite should be stated clearly in the summary and conclusion so that readers do not infer limiters are absent in all tested configurations.","section":"§5.2.2, Example 11, abstract/conclusion"}],"minor_comments":[{"comment":"The local Lax-Friedrichs coefficient is written as lambda = (|u_n^-|+c^-, |u_n^+|+c^+), which looks like a pair rather than the expected maximum of the two local speeds. Please clarify the notation.","section":"§4.2, Eq. (23)"},{"comment":"For the same numerical example, the snapshot and convergence figures report different alpha_0 values: the Sod snapshots use alpha_0=0.7 while the convergence study uses alpha_0=0.5, and the high-frequency perturbation problem uses 0.5 in Fig. 7 but 0.7 in Fig. 8. If the parameter is intentionally changed between the snapshot and the convergence run, this should be stated; otherwise the captions are inconsistent.","section":"Fig. 3/4 and Fig. 7/8"},{"comment":"The figure caption repeats the same 'E, omega, Sigma, t=10.0' panel identifier four times, which appears to be an editing artifact. Please clean up the caption.","section":"Fig. 18 caption"},{"comment":"Eq. (24) defines the SIPG numerical trace with the penalty parameter eta, but the implementation section does not state the value or formula used for eta, despite noting the scaling eta ~ p^2/h. Please report the actual choice used in the experiments.","section":"§4.2, SIPG trace"}],"recommendation":"major_revision","confidential_remarks":"The paper contains reproducible code and a clear numerical methodology, and the qualitative stability of the proposed IGR-DG discretization on the tested benchmarks appears plausible. However, the abstract's 'minimal activation' claim is directly contradicted by the paper's own convergence data (Fig. 2 and §6), and the error-reference methodology is not sufficiently neutral to support the quantitative claims. These issues are load-bearing for the paper's central value proposition and need to be resolved or substantially reframed before publication. If the authors can supply a principled alpha selection strategy and stronger reference solutions, the revised manuscript could be suitable for the journal; in its current form, I do not think the central claims are established as stated."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The real contribution here is concrete: this is the first DG semidiscretization of the IGR-regularized Euler equations, coupling a DG Euler discretization with an SIPG solve for the entropic pressure in the same approximation space. The authors ship both Julia and MFEM code, and the benchmark suite is broad. The headline empirical result — stable shocks without limiters or artificial viscosity across 1D and 2D problems — is supported by the plots. The fixed-DOF p-refinement study is also genuinely informative: IGR-DG keeps error roughly flat as p increases, while the TVB limiter baseline gets worse. That is a real and useful observation.\n\nThe soft spots are real, though none are fatal. The abstract's \"minimal activation in smooth regions\" is contradicted by the paper's own Figure 2: pre-shock, p=3 IGR-DG converges at second order while DG+limiter converges at fourth order. Section 6 concedes this directly (\"reduces the order of accuracy before shock formation\"), so the paper is not hiding it — but the abstract overclaims, and the tension matters because the value proposition is precisely that the regularization preserves smooth flow. The per-problem tuning of alpha0 is a second weakness: the stability results show that a stabilizing parameter exists, not how to choose it without hand-fitting. The reference-solution methodology also biases the comparison, since errors are measured against fine-grid DG+limiter output; and the 2D experiments lack the baseline and error quantification that the abstract promises. None of these undercut the central stability result, but they do mean the paper is not yet the clean case for IGR-DG that the abstract suggests.\n\nWhat the paper does well: it is honest about its own limitation in Section 6, it provides reproducible artifacts, and the MFEM independent implementation is a nice check that the behavior is not tied to one discretization framework. The math is coherent, and the method does what it claims on the tested benchmarks.\n\nThis paper deserves a serious referee. The right outcome is probably conditional acceptance after the abstract is toned down, the pre-shock accuracy loss is treated as a known tradeoff rather than an incidental detail, and the alpha-selection question is addressed head-on (even a sensitivity study would help). I would bring it to a reading group and cite it; the method is new enough and well-enough executed to be worth engaging with.","headline":"A genuinely new DG realization of IGR with stable shocks on benchmarks, but the abstract's 'minimal activation' claim does not survive the paper's own convergence study.","tokens_in":30870,"tokens_out":1619,"would_cite":true,"duration_ms":16828,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["65M60","76N15","35L65"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims that a discontinuous Galerkin discretization of information-geometric regularization stabilizes shocks in the compressible Euler equations without shock-capturing limiters or artificial viscosity, while preserving fine-sca","keywords":["discontinuous Galerkin","shock capturing","information geometric regularization","compressible Euler equations","hyperbolic-elliptic systems","entropic pressure","symmetric interior penalty Galerkin","high-order methods"],"falsifier":"Run a smooth, non-shocking Euler flow (e.g., the nonlinear periodic simple wave of Example 1) on the same grid with the same polynomial order, comparing IGR-DG against unregularized DG with no limiter. If, before shock formation, the L1 error of IGR-DG does not approach that of the unregularized method under mesh refinement, then Σ is not minimal in smooth regions and the claim of selective activation fails directly.","tokens_in":29820,"feed_emoji":"💥","tokens_out":7329,"duration_ms":60438,"temperature":0.7,"pith_summary":"The paper develops a discontinuous Galerkin (DG) realization of information-geometric regularization (IGR) for the compressible Euler equations. The key idea is to regularize shocks at the PDE level by adding an entropic pressure Σ, obtained from an auxiliary elliptic equation, so that the Euler fluxes carry P+Σ instead of P alone. In numerical experiments, the resulting IGR-DG method stabilizes shocks without limiters or artificial viscosity, keeps Σ localized near compression, and retains fine-scale features better than a characteristic TVB-limited DG baseline, especially at higher polynomial order. If correct, this gives a genuinely inviscid, PDE-level alternative to dissipation-based shock capturing for high-order DG methods.","feed_headline":"Entropic pressure stabilizes DG shock flows without limiters","feed_subtitle":"A PDE-level extra pressure, solved from an elliptic equation, replaces limiter-based shock capturing in high-order Euler simulations.","key_machinery":"The central object is the entropic pressure Σ, defined by the elliptic equation ρ⁻¹Σ − α∇·(ρ⁻¹∇Σ) = α[(tr Du)² + tr((Du)²)]. It is the mechanism that redistributes compression: it replaces the developing discontinuity with a smooth profile of width proportional to √α, without any viscous flux term. In the DG setting, Σ is computed by SIPG from the same polynomial space as the conserved variables, and enters the Euler equations only through the augmented pressure P+Σ in the momentum and energy fluxes; this preserves conservative structure and makes the elliptic solve the only global coupling beyond the elementwise numerical flux.","core_discovery":"The central claim is that coupling a standard DG transport discretization (with a local Lax–Friedrichs flux) to a symmetric interior penalty Galerkin (SIPG) solve for the entropic pressure Σ yields stable shock resolution in Euler flows without any limiter or artificial viscosity. The regularization is defined by the elliptic equation ρ⁻¹Σ − α∇·(ρ⁻¹∇Σ) = α[(tr Du)² + tr((Du)²)], where α controls strength and the length scale over which compression is redistributed; the same polynomial space is used for both hyperbolic and elliptic variables, and Σ enters the Euler fluxes only through the augmented pressure P+Σ. The paper's benchmarks — shock formation, propagation, shock–shock and shock–vort","pith_inferences":["The paper's own Figure 2 shows that before a shock forms, IGR-DG converges at second order for p=3 versus fourth order for the limiter baseline; if this contamination is generic, the method trades smooth-flow accuracy for shock stability, and a scale- or sensor-dependent α would be needed to recover the full order.","The reported accuracy comparisons use fine-grid DG+limiter solutions as references for the smoothed Riemann problems, so the L1 errors measure agreement with the competing method rather than with an exact solution.","A predictive formula linking α to local compressive intensity would be needed to make the method parameter-free in practice; across the paper's tests α ranges from 0.3 to 32.","If the smooth-region contamination is confirmed, a natural testable extension is to apply Σ only to the high-wavenumber part of the velocity gradient, aiming to preserve formal accuracy while keeping shock stabilization."],"forward_implications":["If correct, high-order DG methods can resolve post-shock oscillations and fine-scale structures that limiter-based methods typically damp.","The method removes the need for explicit shock sensors or limiting procedures, replacing them with one elliptic solve per time step.","The formulation is conservative and uses a common approximation space, so it may transfer to other hyperbolic systems sharing the same conservative structure, such as magnetohydrodynamics.","The fixed-total-DOF p-refinement study suggests IGR-DG avoids the error growth with polynomial order seen in TVB-limited DG for shock-dominated flows.","Multidimensional benchmarks show the regularization stays localized to the shock, preserving symmetry and avoiding grid-aligned artifacts."],"fun_headline_variants":["Entropic pressure stabilizes DG Euler shocks without limiters","Elliptic entropic pressure lets DG skip limiters for shocks","PDE-level entropy pressure replaces shock limiters in DG","DG shock flows stabilized by elliptic entropic pressure","High-order DG captures shocks via entropy pressure, no limiter"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The load-bearing premise is that the entropic pressure Σ remains negligible in smooth regions of the flow, so the regularization does not degrade the underlying high-order DG accuracy away from shocks; the paper's own pre-shock convergence data show the regularization measurably slows convergence in smooth flow, putting this premise at risk.","fun_headline_variants_meta":{"raw":{"variants":["Entropic pressure stabilizes DG Euler shocks without limiters","Elliptic entropic pressure lets DG skip limiters for shocks","PDE-level entropy pressure replaces shock limiters in DG","DG shock flows stabilized by elliptic entropic pressure","High-order DG captures shocks via entropy pressure, no limiter"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001049,"raw_usage":{"total_tokens":4277,"prompt_tokens":807,"completion_tokens":3470,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":551,"completion_tokens_details":{"reasoning_tokens":3399}},"tokens_in":551,"tokens_out":3470,"duration_ms":21606,"temperature":1.0,"reasoning_tokens":3399,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T11:11:33.774203+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run a smooth, non-shocking Euler flow (e.g., the nonlinear periodic simple wave of Example 1) on the same grid with the same polynomial order, comparing IGR-DG against unregularized DG with no limiter. If, before shock formation, the L1 error of IGR-DG does not approach that of the unregularized method under mesh refinement, then Σ is not minimal in smooth regions and the claim of selective activation fails directly.","supporting_citations":[],"review_version":1}