{"id":"555619d1-4dda-489f-8d2e-aeb3dfe5b85e","arxiv_id":"2411.18813","paper_version":3,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":9,"one_line_summary":"Fast ejecta in neutron star mergers come from two mechanisms, an equatorial spray at first contact and a quasi-spherical bounce of the compressed remnant, and they appear even in prompt-collapse cases.","lead":"This paper uses full general-relativistic simulations of neutron star mergers to trace the origin of fast-moving ejected matter. It finds two launch mechanisms and shows that even mergers that quickly collapse to a black hole still fling out fast material.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 30/70 fast-ejecta split and its bounce dominance are not convergence-tested; the paper's own caveats say the fast tail is resolution-limited, and the split is quoted without per-run scatter.","rationale":"The paper's genuine strength is the Lagrangian particle tracing in full general relativity, which gives a direct, physically interpretable view of where fast ejecta originate; the two-mechanism picture is plausible and consistent with earlier hints. However, the strongest quantitative claim—the 30/70 split and the bounce component's dominance—is exactly the part that rests on the numerically least secure population. The authors themselves warn that the fast tail is resolution-limited and that the highest-velocity distribution is not fully trusted. A controlled resolution sequence is absent; the only comparison with higher particle number is confounded by code changes. Independent grid-based simulations find smaller fast-ejecta masses, and the one published convergence study in this area (Dean et al. 2021) requires much finer grids than used here, although its Newtonian/axisymmetric approximations prevent a direct transfer. The prompt-collapse conclusion, presented as a notable finding, is based on a single run. These are correctness risks, not internal contradictions; they do not falsify the mechanism, but they do mean the quantitative statements should be treated as provisional. Since the reader already reached CONDITIONAL with high confidence, this stress-test supports that verdict without changing it.","tokens_in":21650,"tokens_out":4440,"duration_ms":41616,"concrete_test":"Run a controlled resolution sequence with SPHINCS_BSSN for APR3_13_13 or MPA1_13_13 at 2M, 4M, and 8M particles (finest AMR spacing 369 m, roughly 260 m, and 185 m), using identical initial data, EOS, and particle-tracing criteria. Recompute Table 2 entries and the spray/bounce mass split for v>0.4c. If the split shifts by more than 10 percentage points, or if the v>0.6c ejecta mass changes by more than a factor of two across resolutions, the headline numbers are not converged. Also repeat the branch-identification exercise at two different snapshot times to test the subjectivity of the velocity-branch assignment. For the prompt-collapse claim, run at least two additional prompt-collapse binaries and check whether the fast-ejecta mass and velocity distribution remain similar to SLy_14_14.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central quantitative claim—a 30/70 spray/bounce split of ejecta with v>0.4c, with the bounce component reaching larger velocities—is not supported by a convergence test of the fast tail. Section 5 (Case I) admits that peak velocities are resolution-limited, that faster ejecta exist but are too low-mass to resolve, and that the highest-velocity distribution in Fig. 15 is not fully trusted. The only comparison with earlier runs in Sec. 5 changes both the code version and the particle-to-mesh mapping, so it cannot distinguish physical convergence from numerical changes. Independent grid-based simulations (Radice et al. 2018; Combi & Siegel 2023) find systematically lower fast-ejecta masses, and Dean et al. (2021) report convergence only at roughly 20 m resolution, nearly twenty times finer than the 369 m finest grid used here. If the fast tail is under-resolved, the relative weights of the two mechanisms can shift: since the bounce component is claimed to be faster than the spray, an unresolved high-velocity tail would preferentially remove bounce mass, biasing the 30/70 ratio. The paper never reports the split per simulation or with uncertainties, so the headline ratio cannot be evaluated. In addition, the prompt-collapse statement that 'even prompt collapse ejects fast matter with similar properties' rests on a single simulation (SLy_14_14), with only a vague reference to unpublished runs. Thus the mechanism identification is plausible, but the quantitative split and the prompt-collapse generalization are not yet secured.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"Using full-GR Lagrangian SPH simulations (SPHINCS_BSSN) with two million particles, this paper studies fast dynamic ejecta (v > 0.4c) in nine neutron star merger configurations covering four equations of state and several mass ratios. By identifying velocity-radius branches in the ejecta and tracing particles backward, the authors distinguish two ejection mechanisms: a 'spray' component from the shear interface at first contact, confined mostly to the orbital plane, and a 'bounce' component from the strongly compressed central remnant bouncing back, ejected more isotropically. They report that roughly 30% of fast ejecta are spray and 70% bounce, with bounce material reaching about 0.1c higher velocities, and they find a similar fast-ejecta component even in the one prompt-collapse case (SLy_14_14). They fit mass-momentum profiles M(>γβ) with a broken power law, use these to compute kilonova afterglow light curves, and apply shock-breakout theory to predict gamma-ray/X-ray flares. The paper's central claim is that these two mechanisms account for the mildly relativistic ejecta and have EOS-dependent observational signatures.","tokens_in":21999,"tokens_out":5310,"duration_ms":43721,"significance":"If the mechanism identification and the 30/70 split hold up, this is an important step: prior work identified fast ejecta but did not cleanly separate a two-component origin. The Lagrangian particle tracing here gives a direct and visually convincing kinematic identification of the spray and bounce episodes, and the prompt-collapse simulation is suggestive. The paper also produces concrete, falsifiable predictions for kilonova afterglow detectability (MeerKAT/DSA-2000) and for shock-breakout gamma-ray flares, with the explicit caveat that these depend on extrapolating a barely resolved fast tail. Strengths include the full-GR treatment, the particle-history approach, the survey of four equations of state, and the honest discussion of resolution limitations. However, the quantitative central claim—the 30/70 split—is not yet backed by a convergence test or per-run uncertainty quantification, and the prompt-collapse generalization rests on a single published run. These points prevent acceptance as a quantitative account, though the mechanism identification itself is plausible and well illustrated.","major_comments":[{"comment":"The 30%/70% spray/bounce split is reported as a single global number, without per-run values, uncertainties, or a statement of how the particles were partitioned into the two components. Table 2 shows that the fast ejecta mass m_ej,0.4 varies by a factor of about seven across runs (from 1.3e-4 to 9.0e-4 M_sun), so a single ratio without scatter is not evaluable. Please report the split per simulation, state the exact selection criteria (the three unbound criteria of Sec. 5 plus the v>0.4c threshold), and provide an uncertainty estimate, e.g., by bootstrap resampling over particles.","section":"Abstract and Sec. 5 (first paragraph after Tab. 2)"},{"comment":"The paper explicitly states that peak velocities are resolution-limited, that faster ejecta exist but are too low-mass to be resolved, and that 'we do not fully trust the velocity distribution shown in Fig. 15 at the highest velocities.' Because the bounce component is claimed to be the faster one, an unresolved high-velocity tail would preferentially remove bounce mass and could bias the 30/70 ratio toward the spray component. The comparison with Rosswog et al. 2022 in Sec. 5 changes both the code version and the particle-to-mesh mapping, so it cannot serve as a clean convergence test. Please provide a resolution study (even a single higher-resolution run) or a quantitative estimate of the unresolved mass fraction and its expected partitioning between spray and bounce.","section":"Sec. 5, Case I"},{"comment":"The claim that prompt collapse ejects fast matter with 'similar properties' to the non-collapsing case is based on a single simulation included in the paper (SLy_14_14), with only a vague reference to 'other simulations' that are not listed in Tab. 1. Since the prompt-collapse case drives a distinctive kilonova afterglow prediction in Sec. 5.1 (the dashed curve in Fig. 16), either provide details of those additional runs (masses, EOS, collapse time, fast-ejecta properties) or explicitly state that the prompt-collapse result is a single-case demonstration that requires further study.","section":"Sec. 4.5"},{"comment":"The fast-ejecta masses reported here are larger by an order of magnitude than those in Radice et al. (2018) and Combi & Siegel (2023). The paper proposes an explanation (their higher vacuum density) but does not test it. Since the 30/70 split is a mass ratio, agreement with independent codes on absolute fast-ejecta mass is relevant to its robustness. Please either quantify the sensitivity of the fast-ejecta mass and the spray/bounce split to the background density and resolution, or present a side-by-side comparison that isolates the cause of the discrepancy.","section":"Sec. 4.6"}],"minor_comments":[{"comment":"There is a typo in the sentence following Eq. (9): 'the the color temperature' should read 'the color temperature'.","section":"Sec. 5.2"},{"comment":"In the caption, the symbol 't_bo' is used for both the pulse duration and the typical photon energy; please use 'T_bo' for the temperature to avoid ambiguity.","section":"Fig. 18 caption"},{"comment":"Gutiérrez et al. (2024a) and (2024b) are both listed with the same arXiv identifier (2408.15973); if these are distinct papers please provide the correct identifiers, and if they are the same, merge the citations.","section":"References"},{"comment":"The fitted broken power-law parameters (γ0β0, s_ft, s_KN) that feed the afterglow calculations are described qualitatively but not tabulated. Please provide a table of these fitted parameters for each simulation so the afterglow predictions can be reproduced.","section":"Sec. 5.1 and Eq. (1)"}],"recommendation":"major_revision","confidential_remarks":"The paper is within scope for MNRAS and makes a plausible, well-illustrated contribution to the fast-ejecta mechanism discussion. The main risk is that the headline quantitative split is not yet supported by a resolution study or uncertainty quantification; I found no evidence of circularity or non-disclosure of prior work. A major revision addressing the per-run split, the resolution sensitivity, and the prompt-collapse evidence would make the paper acceptable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. First, this is the cleanest Lagrangian tracing of fast ejecta in full GR to date, and it makes a credible case for two distinct launch mechanisms: an equatorial spray at first contact and a more spherical bounce of the compressed remnant. Second, the headline 30/70 split and the prompt-collapse generalization are both weaker than the abstract implies; the split is not convergence-tested and the prompt-collapse claim rests on one run.\n\nWhat's actually new: previous simulations saw fast ejecta and both mechanisms had been suggested separately, but here they are separated with particle histories in full GR, and the prompt-collapse case (SLy_14_14) showing similar fast ejecta is something I don't recall seeing elsewhere. The exponential mass-momentum profile, with a shallow segment at gamma-beta 0.1-0.2, is a useful update to broken power-law assumptions in afterglow models. The observational discussion is careful, especially the explicit admission that the highest-velocity tail is unresolved and the two bracketing cases used for breakout.\n\nSoft spots. The 30/70 ratio is quoted without per-run scatter or error bars, and by the authors' own admission peak velocities are resolution-limited and the fastest ejecta are too low-mass to resolve. The comparison to their earlier runs changes both the code version and the particle-to-mesh mapping, so it cannot serve as a convergence test. Independent grid-based simulations find systematically lower fast-ejecta masses, and Dean et al.'s convergence study suggests roughly 20 m resolution is needed, far finer than the 369 m finest grid here. If the fast tail is under-resolved, the spray/bounce weighting can shift because the bounce component is claimed to be faster. The prompt-collapse result is based on one simulated case plus a vague reference to unpublished runs—fine as a report, not as a settled claim. The afterglow and breakout predictions depend on fitted and assumed microphysical parameters; the authors say so, but the abstract leans harder on the split than the body's caveats.\n\nThe citation pattern is honest: earlier suggestions by Metzger et al. and Radice et al. are credited, and conflicting results are discussed openly. The mechanism identification may well be robust, but the quantitative claims need convergence work before they are taken as settled.\n\nWho it's for: anyone working on kilonova afterglows, shock breakout, or ejecta production in neutron star mergers. It deserves a serious referee: the core question is important, the method is state of the art, and the weaknesses are fixable with more simulations and honest error bars. I'd accept it for review despite my skepticism about the specific ratio.","headline":"Plausible two-mechanism picture for fast NSM ejecta, with a genuinely new prompt-collapse result, but the 30/70 split is a resolution-sensitive number the paper itself doesn't fully back.","tokens_in":22556,"tokens_out":1559,"would_cite":true,"duration_ms":14373,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper claims that fast ejecta (above $0.4c$) from neutron star mergers come from two mechanisms — a spray from the contact interface and a dominant bounce of the compressed remnant — and that prompt collapse to a black hole still…","keywords":["neutron star mergers","dynamic ejecta","fast ejecta","prompt collapse","kilonova afterglow","shock breakout","general relativistic hydrodynamics","equation of state"],"falsifier":"Run one equal-mass configuration (for example APR3 with $2\\times1.3\\,M_\\odot$) at substantially higher resolution, or with a different method that resolves the shear interface to roughly $20$ m, and re-measure the mass with $v_\\infty>0.4c$ and its division into spray versus bounce branches. A bounce fraction that drops below the spray fraction, or fast-ejecta masses that change by more than the stated few $10^{-5}\\,M_\\odot$, would show that the 30/70 partition is resolution-dominated rather than a physical signature.","tokens_in":21445,"feed_emoji":"💥","tokens_out":9799,"duration_ms":88087,"temperature":0.7,"pith_summary":"Fast ejecta from neutron star mergers — the small fraction of matter leaving the merger at velocities above $0.4c$ — are often treated as a single population. This paper argues, using full general-relativistic Lagrangian simulations, that they form through two distinct mechanisms: roughly 30% are sprayed sideways from the shear interface where the two stars first touch, and the remaining ~70% are thrown out when the strongly compressed central remnant bounces back. The bounce component is more isotropic and faster by about $0.1c$, so its leading parts can catch up with and shock the spray component. Even a merger that promptly collapses to a black hole still emits fast ejecta with similar properties, because the outer layers bounce even as the inner core is swallowed. If correct, these mechanisms tie the fast ejecta — and their observable shock-breakout gamma-ray flares and radio kilonova afterglows — to the equation of state and binary parameters.","feed_headline":"Fast ejecta from neutron-star mergers are 30% spray, 70% bounce","feed_subtitle":"The two mechanisms set where the ejecta go, and so what radio and gamma-ray signals to expect.","key_machinery":"The load-bearing tool is the Lagrangian particle hydrodynamics code SPHINCS_BSSN, which solves the full Einstein equations on an adaptive mesh while evolving matter with freely moving SPH particles; each run uses two million particles with a finest grid spacing of 369 m. Because particles carry their identities, the authors can identify ejecta 'branches' in a velocity-versus-radius plot at a given snapshot and then trace each branch backward to its launch site at first contact. This backward tracing is what separates the equatorial 'spray' component from the nearly isotropic 'bounce' component and reveals that the bounce follows strong compression of the remnant. The mass-momentum profile $M(>\\gamma\\beta)$ is the quantitative diagnostic connecting the simulations to observations.","core_discovery":"On the paper's own terms, the central discovery is that the mildly relativistic dynamic ejecta seen in merger simulations are not a single outflow. By color-coding SPH particles according to the velocity branch they lie on and tracing them backward, the authors find a first 'spray' pulse emitted from the contact interface along the orbital plane (about 30% of the mass with $v_\\infty>0.4c$) and later 'bounce' pulses (about 70%) launched when the remnant rebounds from deep general-relativistic compression. The bounce ejecta are more spherical and faster by $\\sim0.1c$, so they interact with and brake against earlier spray ejecta, shaping the final angular distribution. The discovery extends to prompt collapse: even when a black hole forms within a free-fall time, fast ejecta appear, with slow material swallowed and fast material escaping, so the observable fast component resembles that of non-collapsing cases.","pith_inferences":["Editorial extension: if the bounce component is the dominant fast-ejecta channel and its strength tracks the softness of the equation of state, then early radio observations of mergers could serve as an equation-of-state diagnostic that is independent of gravitational-wave tidal measurements.","Editorial extension: the authors' resolution caveat implies the true fast tail extends beyond the simulated $\\sim0.8c$, so the cutoff-based shock-breakout estimates are lower limits; power-law-extrapolated estimates may be closer to reality, and the 30/70 split could shift once finer resolutions resolve more of the spray component.","Editorial extension: the same two-mechanism picture may apply to neutron star-black hole mergers and to mergers with stronger magnetic fields or neutrino losses, none of which are included here; those processes could alter the relative importance of spray versus bounce.","Editorial extension: a direct cross-check would be to compare tracer-particle histories from an Eulerian grid code on the same binaries; if that code also finds a spray-first, bounce-second sequence with a 30/70 split, the mechanism is method-independent, and if not, the split is likely tied to the SPH treatment of the contact interface."],"forward_implications":["Because the spray component stays near the orbital plane while the bounce component expands almost isotropically, the angular distribution of fast ejecta is set by the collision between consecutive pulses, not by a single launch event.","Prompt collapse does not suppress fast ejecta; instead it removes the slow material, making the ejecta velocity distribution peak above $0.2c$ and giving the kilonova afterglow a distinct early-time rise that could act as a merger-outcome diagnostic.","Softer equations of state produce larger peak velocities and more detectable kilonova afterglow; stiffer equations of state produce steeper mass-momentum profiles, so radio observations before peak can discriminate between equations of state.","The ejecta mass-momentum profile is better described by an exponential or a broken power law with an extra shallow segment at $0.1\\lesssim\\gamma\\beta\\lesssim0.2$, which changes predicted afterglow light curves compared with standard broken-power-law assumptions.","A cocoon-driven shock breaking out of these fast ejecta yields a short gamma-ray flare of $10^{45}$-$10^{47}$ erg lasting $0.001$-$1$ s, with parameters consistent with the GRB seen from GW170817 for a jet-launch delay near 1 s."],"supporting_citations":[{"why":"The Eulerian study that previously attributed most high-velocity ejecta to the remnant's first bounce; this paper's bounce component is the direct comparison and extension.","marker":"Radice et al. (2018)"},{"why":"The analysis that located fast ejecta at the shock-heated interface between the neutron stars; supplies the 'spray' hypothesis this paper quantifies.","marker":"Metzger et al. (2015)"},{"why":"Early Lagrangian simulations that found fast ejecta and highlighted thermal effects; provides the baseline for the present particle-based method.","marker":"Bauswein et al. (2013)"},{"why":"The resolution study of the shear interface that argued for convergence of fast-ejecta mass in grid codes; this paper argues its Newtonian and axisymmetric setup cannot capture the GR bounce.","marker":"Dean et al. (2021)"},{"why":"Established the kilonova-afterglow observable from mildly relativistic ejecta that this paper re-predicts from its measured mass-momentum profiles.","marker":"Hotokezaka et al. (2018)"},{"why":"The shock-breakout theory used to translate the simulated fast polar ejecta into gamma-ray flare predictions.","marker":"Nakar (2020)"},{"why":"The kilonova-afterglow model used by the paper to compute radio light curves from the simulated ejecta profiles.","marker":"Sadeh et al. (2023)"},{"why":"Independent simulations that found two ejection pulses in some cases, supporting the spray/bounce distinction.","marker":"Combi & Siegel (2023)"},{"why":"Earlier SPHINCS_BSSN simulations with the same code family whose fast-ejecta masses provide the comparison for the current improved runs.","marker":"Rosswog et al. (2022)"}],"fun_headline_variants":["Mergers: 70% bounce back, 30% spray out","Fast ejecta split: 70% bounce, 30% spray","Spray then bounce: how mergers make fast ejecta","Neutron star mergers: fast ejecta from spray and bounce","Bounce dominates: 70% of fast ejecta, spray 30%"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The conclusion that fast ejecta split roughly 30/70 into spray and bounce assumes that two-million-particle runs with a finest grid spacing of 369 m resolve the small fast-moving ejecta population. The authors state that peak velocities are likely resolution-limited, that matter at even higher velocities is too low in mass to be resolved, and that they do not fully trust the highest-velocity part of the distribution; if that fast tail is under-resolved, the measured split and the dominance of the bounce could be numerical artifacts.","fun_headline_variants_meta":{"raw":{"variants":["Mergers: 70% bounce back, 30% spray out","Fast ejecta split: 70% bounce, 30% spray","Spray then bounce: how mergers make fast ejecta","Neutron star mergers: fast ejecta from spray and bounce","Bounce dominates: 70% of fast ejecta, spray 30%"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001318,"raw_usage":{"total_tokens":5426,"prompt_tokens":1059,"completion_tokens":4367,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":675,"completion_tokens_details":{"reasoning_tokens":4272}},"tokens_in":675,"tokens_out":4367,"duration_ms":26658,"temperature":1.0,"reasoning_tokens":4272,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T10:51:14.456744+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run one equal-mass configuration (for example APR3 with $2\\times1.3\\,M_\\odot$) at substantially higher resolution, or with a different method that resolves the shear interface to roughly $20$ m, and re-measure the mass with $v_\\infty>0.4c$ and its division into spray versus bounce branches. A bounce fraction that drops below the spray fraction, or fast-ejecta masses that change by more than the stated few $10^{-5}\\,M_\\odot$, would show that the 30/70 partition is resolution-dominated rather than a physical signature.","supporting_citations":[{"cited_title":"D., 2021, @doi [ApJ] 10.3847/1538-4357/ac1f20 , https://ui.adsabs.harvard.edu/abs/2021ApJ...921..161D 921, 161","cited_arxiv_id":null,"evidence_quote":"The resolution study of the shear interface that argued for convergence of fast-ejecta mass in grid codes; this paper argues its Newtonian and axisymmetric setup cannot capture the GR bounce."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Earlier SPHINCS_BSSN simulations with the same code family whose fast-ejecta masses provide the comparison for the current improved runs."}],"review_version":1}