{"id":"1b2fe384-abf6-4172-a614-d6606ffe0e1b","arxiv_id":"2412.17863","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Using the entropy-based flux limiter in both eccentricity reduction and evolution yields apparent fifth-order convergence in binary neutron star waveform phase.","lead":"This paper applies an entropy-based flux-limiting scheme, previously used only for evolving neutron star mergers, to the step that removes orbital eccentricity from the starting data. In two test binaries, the combined scheme achieves apparent fifth-order convergence in gravitational wave phase, the highest order reported so far.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Fifth-order phase convergence is not established: with fourth-order time integration the asymptotic rate cannot exceed fourth order, so the reported rate is likely pre-asymptotic unless temporal error is proven negligible.","rationale":"The reader's weakest-assumption identification is correct as far as it goes: three resolutions without an independent reference cannot establish an asymptotic order, especially when the middle resolution is used for eccentricity tuning. My concern is a sharper structural version of the same gap. The paper's own previous result [9] reached fourth-order convergence, which is consistent with a fourth-order Runge-Kutta time integrator. Claiming fifth-order requires either a fifth-order time integrator or proof that temporal error is negligible at these resolutions; neither is provided. A CFL-halving run at fixed spatial resolution is a cheap, decisive test: it isolates temporal truncation error from spatial error. If temporal error is significant, the central claim overstates the method's asymptotic order, and the paper becomes a pre-asymptotic convergence study rather than the first demonstration of fifth-order convergence. If temporal error is negligible, the concern is resolved and the fifth-order claim may stand as a spatial-convergence result. This does not change the reader's CONDITIONAL verdict, but it sharpens the required revision: either document the time integrator and add a temporal convergence test, or soften the claim to an apparent, resolution-dependent order. I do not see a reason to move to REJECT, because the physical claims about EFL's improved accuracy and the eccentricity-reduction workflow are plausible and the missing evidence is obtainable.","tokens_in":14171,"tokens_out":7350,"duration_ms":69689,"concrete_test":"Run the pure EFL BAM:95 n=128 configuration twice with CFL = 0.25 and CFL = 0.125, keeping the grid and all other settings fixed, and compute the (2,2)-mode phase difference up to merger. If the phase shift between the two CFL runs is comparable to, or larger than, the MID-HIG phase difference in Fig. 5, temporal truncation error is not subdominant and the apparent fifth-order rate is not an asymptotic spatial convergence order. Also verify the formal order of BAM's time integrator: if it is RK4, the paper should be revised to claim a pre-asymptotic apparent order unless a higher-order integrator is documented.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim (Sec. V) is that using EFL in both eccentricity reduction and evolution yields fifth-order convergent waveform phase at production resolutions. The evidence is the three-resolution self-convergence analysis in Secs. IV A 3 and IV B 3 using Eqs. (16)-(17). The load-bearing assumption is that the measured phase differences are dominated by the spatial discretization of the hydro scheme. This is not checked, and it is in tension with the time integrator. The BAM code, in the BNS implementations cited here ([7,18]), uses a fourth-order Runge-Kutta time integrator; with CFL fixed to 0.25 (Sec. II), dt is proportional to h, so temporal truncation error is O(h^4). For a method-of-lines discretization the global order is bounded by min(spatial, temporal), so the asymptotic phase convergence cannot exceed fourth order unless a fifth-order time integrator is identified, and none is. The observed fifth-order matching over n=64,96,128 therefore likely reflects a pre-asymptotic regime where spatial error still dominates, not true asymptotic fifth-order convergence. Additionally, Eq. (17) as printed gives a negative rescaling factor for p=5 with ni<nj<nk, so the visual dashed-solid matching in Figs. 5, 6, 9, 10 rests on an unclear implementation. No Richardson extrapolation, independent high-resolution reference, or temporal convergence study is provided.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"Doulis, Bernuzzi, and Tichy extend the entropy-based flux-limiting (EFL) hydrodynamics scheme from dynamical evolution to the eccentricity-reduction stage of binary neutron star (BNS) initial-data construction with the SGRID code. Two reduction algorithms are exercised, one driven by radial velocity and eccentricity and one driven by radial velocity and orbital angular velocity, and they are applied to the equal-mass nonspinning configuration BAM:95 and the unequal-mass spinning configuration MPA1q1.6. For BAM:95 the paper compares three scheme combinations (pure HO-LLF, hybrid HO-LLF reduction with EFL evolution, and pure EFL) at three resolutions (n=64, 96, 128 points on the finest level); for MPA1q1.6 the pure EFL data is evolved at the same three resolutions. Gravitational-wave phase differences between resolutions are analyzed by self-convergence with the rescaling factor of Eq. (17) for the (2,2), (3,2), and (4,4) modes. The paper reports second-order convergence for pure HO-LLF, third-order for the hybrid, and fifth-order for the pure EFL cases, and concludes that using EFL in both the eccentricity reduction and the evolution yields fifth-order convergent waveforms at current production resolutions, claims the first demonstration of fifth-order convergence in BNS waveform production, and estimates the pure EFL phase error at merger to be a factor of about 2 smaller than HO-LLF at matched resolution.","tokens_in":14426,"tokens_out":23668,"duration_ms":199553,"significance":"The methodological contribution is original and clearly described: embedding the EFL scheme in the SGRID eccentricity-reduction loop, for two reduction variants, with iteration tables (Tables II and IV) and grid specifications (Table VI) that make the procedure reproducible. A practical strength is that the entropy limiter is used without tuning its scale parameter c_E (Sec. II, c_E = 1 by default). The demonstration on a spinning unequal-mass system (MPA1q1.6) in addition to the equal-mass case strengthens the generality of the method. If the convergence claims hold, the result is significant for numerical relativity waveform production, moving BNS phase convergence from the second-order standard of most HRSC schemes (and the fourth-order result of the authors' previous EFL paper) to fifth order at production resolutions, with a measured phase-error improvement over the HO-LLF scheme in the same code.","major_comments":[{"comment":"The central claim of fifth-order convergence is not supported as a statement about asymptotic convergence. The manuscript never specifies the temporal integration scheme; the cited BAM implementations [7,18] evolve the GRHD system with a fourth-order Runge-Kutta method of lines, and with the Courant factor fixed to 0.25 (Sec. II) the time step is proportional to the spatial step h. The temporal truncation error is therefore O(h^4), which bounds the global asymptotic order of the method-of-lines discretization at four. The observed fifth-order rescaling over (64,96,128) must consequently be a pre-asymptotic effect unless temporal error is shown to be subdominant at these resolutions, and the paper provides no temporal convergence study, no Richardson extrapolation, and no independent high-resolution reference. The Sec. V statement that these results are the first demonstration of fifth-order convergence in BNS waveform production therefore overstates the evidence; either supply an error-decomposition study (for example a time-step refinement test at fixed grid, or Richardson extrapolation that includes the temporal order) or re-scope the claim to pre-asymptotic, spatial-error-dominated convergence at production resolutions.","section":"IV A 3; IV B 3; V"},{"comment":"Equation (17), which defines the rescaling factor used to test the convergence order, is inconsistent with its stated purpose. For (ni,nj,nk)=(64,96,128) and p=5 it evaluates to s = 1 - (2/3)^5 / ((2/3)^5 - (1/2)^5) = -0.31, whereas the factor that maps the MID-HIG phase difference onto the LOW-MID difference for a p-th order error is (nk/nj)^p ((nj/ni)^p - 1) / ((nk/nj)^p - 1) = 8.6 for this resolution triple. As printed, the dashed curves in Figs. 5, 6, 9, and 10 cannot coincide with the solid curves in the way the text describes. Please correct Eq. (17) (or state the sign and ordering convention under which it applies) so that the rescaling procedure is reproducible.","section":"IV A 3, Eq. (17)"},{"comment":"The convergence order is inferred entirely by visual matching of rescaled phase differences against an assumed integer p; the text reports second-, third-, and fifth-order matches without a quantitative measure of the mismatch, without error bars on the phase differences, and without stating the retarded-time interval over which each match is evaluated. Because the fifth-order claim is the paper's central result, please add a quantitative diagnostic, for example the ratio dphi(LOW,MID)/dphi(MID,HIG) as a function of retarded time compared with the values predicted for integer orders, or a locally fitted convergence order p(u) with uncertainty estimates.","section":"IV A 3; IV B 3"},{"comment":"The eccentricity-reduction loop measures the residual eccentricity and applies the stopping criterion using n=96 evolutions (Sec. III, steps ii-iii; the resolution is fixed in TABLE VI), so the middle member of the convergence triple is also the resolution used to tune the initial data. For the pure EFL case the final residual eccentricity is the largest of the three BAM:95 cases (TABLE V: e=0.7e-3 versus 0.4e-3 for pure HO-LLF), and the contribution of the residual eccentricity to the phase differences entering Eqs. (16)-(17) is not quantified. Please assess the sensitivity of the reported convergence order to this choice, for example by measuring e from the proper distance at all three resolutions for the final initial data or by re-running the reduction at n=128, and state how residual eccentricity affects the phase-difference comparison near merger.","section":"III; IV A 3; TABLE VI"}],"minor_comments":[{"comment":"The sentence that the expected rate of convergence is 5th-order for all cases is asserted without derivation; given that the temporal integration and the Z4c metric evolution are lower-order than the fifth-order reconstruction, the expected global rate requires justification, presumably as the expected spatial-error-dominated pre-asymptotic rate rather than an asymptotic one.","section":"IV A 3"},{"comment":"The conclusions state that fifth-order convergence is observed in the (3,2) and (4,4) modes without the qualifications present in the body: for BAM:95 the higher modes (and the (2,2) mode in the early post-merger phase) show third-order convergence through merger and in the early post-merger regime (Figs. 5 and 6), while for MPA1q1.6 the fifth-order trend is claimed to persist through the early post-merger (Figs. 9 and 10). Please state the convergence claim per mode and per phase interval so that the conclusions match the figures.","section":"V; IV A 3; IV B 3"},{"comment":"The comparison that the pure EFL phase error is a factor of about 2 smaller than pure HO-LLF compares the MID-HIG phase difference at merger, which is a resolution-specific self-convergence estimate, and the two cases have different residual eccentricities (TABLE V); please state explicitly that the factor refers to that estimate at the quoted resolutions and comment on its sensitivity to the eccentricity difference.","section":"IV A 3"},{"comment":"The figure captions do not define the gray shaded regions (described in the text as differences in merger times between runs), nor the sign and ordering convention of the plotted phase differences (which curve is dphi(LOW,MID) and which is the rescaled dphi(MID,HIG)); please add these definitions to the captions.","section":"Figs. 5, 6, 9, 10"},{"comment":"The numerical setup does not report the gravitational-wave extraction radius (or radii) used for Psi4 and the multipolar strain, the order of the spatial finite differencing and the Kreiss-Oliger dissipation used in the Z4c metric evolution, or the order of the time integrator actually used in these runs; reporting these is necessary to assess whether phase convergence could be limited by extraction or gauge effects.","section":"IV A 1; IV B 1"}],"recommendation":"major_revision","confidential_remarks":"The methodological core of the paper (EFL in the eccentricity reduction loop) is sound and appropriate for this journal, and I would not recommend rejection: the eccentricity-reduction demonstrations and the three-scheme comparison are useful regardless of how the fifth-order result is ultimately interpreted. The main editorial risk is the novelty claim of a first demonstration of fifth-order convergence in BNS waveform production: given that the cited BAM implementation uses fourth-order time integration, an asymptotic fifth-order claim cannot hold unless an error-source decomposition is supplied, so acceptance should be conditional on re-scoping that claim or on new temporal-error evidence. The sign issue in Eq. (17) and the absence of quantitative convergence diagnostics further weaken the evidence as it stands. I do not see a citation-pattern problem; references to the authors' own prior work are appropriate given the continuity of the EFL development, although the increment from the fourth-order claim of [9] to fifth order here is exactly what needs the strongest support."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: the new thing is EFL applied to SGRID eccentricity reduction, and the paper claims a first-time fifth-order phase convergence for BNS waveforms. I read it as a useful method advance with an over-stated headline claim.\n\nWhat works: the setup is clean. The authors compare pure HO-LLF, hybrid EFL (HO-LLF for eccentricity reduction, EFL for evolution), and pure EFL on the same BAM:95 configuration, and then demonstrate pure EFL on a spinning unequal-mass system. The eccentricity reduction procedures are described in enough detail to reproduce, and the convergence plots for (2,2), (3,2), and (4,4) are a genuine step beyond the usual two-resolution checks. Using EFL in the initial-data construction is a legitimate extension of their own prior work, and the result is not in the earlier literature. Citation practice looks fine; the prior EFL work is clearly credited.\n\nThe soft spots, in order of size. First, the fifth-order claim cannot be asymptotic if the time integration is RK4 with dt proportional to h, which is what BAM uses per the cited implementations. A method-of-lines scheme's global order is capped by the time integrator, so the observed fifth-order phase differences over n=64,96,128 are most likely pre-asymptotic spatial-error dominance. Without a temporal convergence study or an independent high-resolution reference, calling it fifth-order convergence is too strong. That is load-bearing, because the paper's abstract and conclusions rest on it. Second, Eq. (17) as printed gives a negative rescaling factor for p=5 with the stated resolution ordering; the dashed curves in the convergence figures cannot be what the formula says. That may be a typo, but it undermines the visual evidence as presented. Third, the eccentricity reduction is tuned at n=96, which is also the middle resolution, and pure EFL retains a slightly larger residual eccentricity than HO-LLF; so the attribution of the improvement to EFL-in-eccentricity-reduction is not completely isolated.\n\nThese are all addressable in a revision. I would send it to peer review, and I would ask for a temporal convergence check and a corrected convergence analysis before accepting the headline claim.","headline":"Useful method advance, but the fifth-order convergence claim is over-strong given the time integrator; deserves a serious referee.","tokens_in":15002,"tokens_out":5948,"would_cite":true,"duration_ms":51201,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Entropy-based flux limiter gives 5th-order convergent BNS waveforms","keywords":["neutron star mergers","gravitational waves","numerical relativity","entropy-based flux limiting","eccentricity reduction","convergence order","high-resolution shock-capturing","initial data"],"falsifier":"Evolve the pure EFL BAM:95 configuration at a fourth resolution (for example n=160 or n=192 on the finest level) and compare the phase difference from the n=128 run with the fifth-order rescaling factor from Eq. (17); if the measured rate drops below fifth order, or if the n=96-based eccentricity-reduction tuning makes the phase differences depend on resolution in a non-power-law way, the fifth-order claim is falsified. Alternatively, compute a Richardson extrapolation of the three existing runs; an extrapolated order clearly below 5 would also refute the stated convergence rate.","tokens_in":13922,"feed_emoji":"🌌","tokens_out":8327,"duration_ms":67510,"temperature":0.7,"pith_summary":"This paper claims that using the entropy-based flux limiting (EFL) scheme not just for the time evolution of binary neutron star mergers but also inside the eccentricity-reduction loop that constructs the initial data raises the phase convergence of the emitted gravitational waveforms to fifth order. The evidence comes from full numerical-relativity runs of two configurations: a non-spinning equal-mass pair and a spinning, unequal-mass pair, each evolved at three resolutions. For the pure EFL case, phase differences between resolutions follow the fifth-order rescaling factor for the dominant (2,2) mode and for the (3,2) and (4,4) subdominant modes, through inspiral and into the early post-merger regime. At matched resolution, the pure EFL waveform phase error is about a factor of two smaller than the same code's standard high-order scheme (HO-LLF), which converges only at second order. The point of the claim: waveform production for gravitational-wave astronomy needs controlled numerical error, and reaching optimal convergence at production resolutions would make templates more accurate at a fixed computational cost.","feed_headline":"Entropy-based flux limiter gives 5th-order convergent BNS waveforms","feed_subtitle":"At matched resolution, its waveform phase error is about half that of the standard high-order scheme.","key_machinery":"The load-bearing object is the entropy production function $\\nu = \\min(c_E |R|, 1)$, built from the entropy residual $R = \\partial_t s + v^i \\partial_i s$ of the relativistic hydrodynamics equations. It is a shock detector: at each cell interface the numerical flux is the convex combination $\\theta \\hat{f}^{\\mathrm{HO}} + (1-\\theta)\\hat{f}^{\\mathrm{LO}}$ with $\\theta_{i\\pm1/2} = 1 - \\frac{1}{2}(\\nu_i + \\nu_{i\\pm 1})$, so the scheme continuously drops from a high-order unfiltered flux to a stable WENOZ-based flux exactly where entropy production flags non-smooth flow. Around that limiter sits the eccentricity-reduction procedure: short EFL evolutions fit the binary's proper distance to a Keplerian ansatz, and the fit parameters correct either (radial velocity, eccentricity) or (radial velocity, orbital angular velocity), repeated until residual eccentricity stops decreasing. The mechanism concentrates numerical dissipation at stars' surfaces and merger shocks while leaving smooth inspiral regions to the high-order flux, which is why the global waveform phase error can decrease at fifth order.","core_discovery":"On the paper's own terms, the discovery is that the entropy-based flux limiter—used in both the iterative eccentricity reduction of SGRID initial data and the subsequent BAM evolution—yields gravitational waveforms whose phase converges at fifth order in the grid spacing at the currently used production resolutions. This is presented as the first demonstration of fifth-order convergence in binary neutron star waveform production. The convergence holds for the dominant (2,2) multipole and for subdominant (3,2) and (4,4) modes; a hybrid run that uses EFL only in evolution, with the standard HO-LLF scheme for eccentricity reduction, converges at third order, and the pure HO-LLF run converges at second order. The authors attribute the improvement to the combination of eccentricity-reduced initial data and the EFL switch, which confines numerical dissipation to the non-smooth features detected by entropy production.","pith_inferences":["If the fifth-order rate persists at higher resolutions, the three-resolution runs shown here would already sit in the asymptotic regime, meaning error bars on waveform phase could be assigned from the measured scaling rather than from conservatively assuming second order.","The same entropy-switch construction could plausibly be transplanted to other finite-difference or conservation-law codes; the two test cases here are not a proof of generality, especially for high mass ratio, high spin, or exotic equations of state.","A stronger test would be a fourth resolution or a Richardson extrapolation against a high-resolution reference; the paper's visual matching of the rescaling factor does not by itself prove the asymptotic order.","If fifth-order convergence survives such a test, waveform template banks for parameter estimation could in principle quote per-mode phase-error budgets from the measured convergence rate, reducing the need for ad hoc safety factors."],"forward_implications":["The pure EFL pipeline produces waveform phase errors roughly a factor of two smaller than HO-LLF at the same resolution, so existing production grids can deliver more accurate waveforms without adding resolution.","Fifth-order convergence extends beyond the dominant (2,2) mode to the (3,2) and (4,4) multipoles, improving the fidelity of subdominant-mode physics near and after merger.","Using EFL in the eccentricity-reduction step is essential for the full gain: replacing only the evolution flux with EFL gives third order, while replacing both gives fifth order.","The two-parameter eccentricity-reduction variants cover both standard and more extreme (spinning, unequal-mass) BNS configurations, widening the class of initial data that can be prepared with this scheme."],"supporting_citations":[{"why":"Established the EFL scheme for BNS evolution and reported up to fourth-order convergence; the present paper extends it to eccentricity reduction.","marker":"[9]"},{"why":"Supplies the BAM:95 configuration and the pure HO-LLF reference run that the paper compares against.","marker":"[30]"},{"why":"Describes the original SGRID eccentricity-reduction algorithm based on radial velocity and eccentricity corrections.","marker":"[23]"},{"why":"Describes the modified eccentricity-reduction algorithm based on radial velocity and orbital angular velocity, used for the spinning unequal-mass case.","marker":"[24]"},{"why":"Provides the SGRID constant-rotational-velocity initial-data construction for spinning BNS configurations.","marker":"[14]"},{"why":"Developed high-order WENOZ schemes in numerical relativity, the reconstruction basis for the fluxes here.","marker":"[7]"},{"why":"Provides the rescaling factor used to read off the experimental convergence order from phase differences.","marker":"[33]"},{"why":"Origin of the entropy-limited hydrodynamics approach from which the entropy residual and limiter concept derive.","marker":"[12]"}],"fun_headline_variants":["Entropy-based flux limiter achieves 5th-order BNS convergence","BNS waveforms hit 5th-order convergence with EFL scheme","EFL scheme yields 5th-order BNS phase convergence","Eccentricity-reduced BNS data with EFL: 5th-order convergence","Entropy limiter boosts BNS initial data to 5th-order phase"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The claim assumes the three chosen resolutions are already in the regime where phase differences shrink at a fixed order of the grid spacing, and that using the middle resolution to tune the eccentricity reduction does not bias those phase differences.","fun_headline_variants_meta":{"raw":{"variants":["Entropy-based flux limiter achieves 5th-order BNS convergence","BNS waveforms hit 5th-order convergence with EFL scheme","EFL scheme yields 5th-order BNS phase convergence","Eccentricity-reduced BNS data with EFL: 5th-order convergence","Entropy limiter boosts BNS initial data to 5th-order phase"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000743,"raw_usage":{"total_tokens":3257,"prompt_tokens":832,"completion_tokens":2425,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":448,"completion_tokens_details":{"reasoning_tokens":2330}},"tokens_in":448,"tokens_out":2425,"duration_ms":14726,"temperature":1.0,"reasoning_tokens":2330,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T10:44:01.797645+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Evolve the pure EFL BAM:95 configuration at a fourth resolution (for example n=160 or n=192 on the finest level) and compare the phase difference from the n=128 run with the fifth-order rescaling factor from Eq. (17); if the measured rate drops below fifth order, or if the n=96-based eccentricity-reduction tuning makes the phase differences depend on resolution in a non-power-law way, the fifth-order claim is falsified. Alternatively, compute a Richardson extrapolation of the three existing runs; an extrapolated order clearly below 5 would also refute the stated convergence rate.","supporting_citations":[{"cited_title":"Entropy based flux limiting scheme for conservation laws","cited_arxiv_id":"2401.04770","evidence_quote":"Describes the original SGRID eccentricity-reduction algorithm based on radial velocity and eccentricity corrections."},{"cited_title":"original","cited_arxiv_id":null,"evidence_quote":"Describes the modified eccentricity-reduction algorithm based on radial velocity and orbital angular velocity, used for the spinning unequal-mass case."},{"cited_title":"The gravitational waveform de- rived from the MPA1q1.6 simulation is shown in FIG","cited_arxiv_id":null,"evidence_quote":"Developed high-order WENOZ schemes in numerical relativity, the reconstruction basis for the fluxes here."}],"review_version":1}