{"id":"0b7b5d16-0a20-42d0-82b9-60250fad3d6a","arxiv_id":"2608.10100","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Long 3D simulations of neutron star merger ejecta show radioactive heating keeps reshaping heavy-element outflows, and 3D light curves are dimmer but no closer to AT2017gfo than 2D ones.","lead":"This paper simulates the debris thrown off by four neutron-star collision models for a full second, long enough to see how radioactive heating changes the outflow's shape and direction. It matters because it tests whether going from 2D to 3D models can close the gap between predicted and observed kilonova light curves, and concludes it cannot.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"2D-vs-3D light-curve comparison conflates dimensionality with ejecta evolution time; the upper-limit conclusion is not cleanly isolated.","rationale":"The reader's weakest assumption concerns missing viscous ejecta limiting the scope of the comparison to AT2017gfo. That is a fair, honestly acknowledged caveat, but it does not directly attack the internal validity of the 2D-vs-3D comparison for the ejecta that are included. My concern is more load-bearing: the comparison itself mixes dimensionality with evolution time, because the 2D profiles are taken from the original THC data at ~100 ms while the 3D profiles come from Athena++ at ~1 s. The paper even states this in Sec. IIE and uses the BLh_150 run to try to bound the evolution-time effect, but BLh_150 is not a same-time 2D control. Therefore the central claim that '2D light curves should be treated as upper limits' and that 'increased dimensionality alone is unlikely to reconcile' is not uniquely supported by the data as presented. This does not invalidate the paper's substantial contributions—the long-term 3D evolution, nucleosynthesis distributions, and explicit identification of missing viscous ejecta are valuable—but it raises the bar for the headline conclusion. Since the reader already issued a CONDITIONAL verdict that requires additional data and controls, my analysis reinforces that condition without moving the verdict; hence UNCHANGED.","tokens_in":35376,"tokens_out":5483,"duration_ms":50825,"concrete_test":"Take the 1 s Athena++ 3D profile of BLh_q1.43 (and DD2_q1.67), perform a mass-weighted azimuthal average over phi exactly as done for the 2D THC-based profiles in [35], and feed this 2D profile to KNEC using the same settings as the existing 2D runs. Compare the resulting light curves with (i) the original 2D-from-THC curves and (ii) the 3D curves. If the new same-time 2D curves lie close to the 3D curves, the reported 2D-vs-3D difference is dominated by evolution time rather than dimensionality, and the '2D as upper limit' conclusion needs revision. If they track the original 2D curves, the confound is minor. A complementary control is to run KNEC on the un-evolved 3D THC data at ~100 ms to see whether a 3D-to-2D difference appears without the Athena++ extension.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim—that 2D ray-by-ray light curves are broadly robust upper limits and that increased dimensionality alone cannot close the AT2017gfo gap—rests on comparing 3D Athena++ profiles at ~1 s (Sec. IIE) with 2D KNEC runs built from azimuthally averaged THC data at r = 300 M_sun that were NOT evolved with Athena++ (Sec. IIE, Fig. 16). The authors explicitly note the 2D results are 'also representative of an earlier homology assumption.' Therefore the 2D-vs-3D difference includes both dimensionality and the ~1 s of additional hydrodynamical evolution, including nuclear heating and the approach to homology that the paper itself finds non-negligible (Fig. 4, and the BLh_150 vs. BLh comparison in Fig. 16). The BLh_150 run is a 3D simulation truncated at 150 ms; it does not provide the necessary control, because it differs from the 2D calculation in both dimensionality and data origin (Athena++ evolved vs. direct THC data). Without a same-time, same-ejecta 2D control—e.g., azimuthally averaging the 1 s Athena++ profiles and feeding those to KNEC—the observed luminosity enhancement of 2D over 3D cannot be uniquely attributed to low dimensionality. Since this comparison is the direct evidence for the abstract's strongest claim, the conclusion is not yet internally clean, independent of the separately acknowledged missing-viscous-ejecta limitation (Secs. IIIA, IIIC).","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents long-term (about 1 s) three-dimensional general-relativistic hydrodynamics simulations of ejecta from four binary neutron star merger simulations, using THC numerical-relativity data injected as boundary conditions into Athena++. The code includes a transition from a nuclear-statistical-equilibrium EOS to the Timmes/Helmholtz EOS, an effective nuclear-heating prescription based on reaction-network fits, and passive scalars for electron fraction, entropy, expansion timescale, and injection flag. The authors analyze ejecta dynamics and geometry, report a homology parameter and heating diagnostics, post-process tracers with the WinNet nuclear network to obtain element distributions and nucleosynthesis yields, and compute multi-angle kilonova light curves with KNEC using both 3D and azimuthally-averaged 2D profiles. The main claims are that nuclear heating significantly affects the ejecta dynamics and redistributes heavy elements on the sky, that the 56Ni-56Co-56Fe chain dominates heating around 100 days, that 2D ray-by-ray light curves are broadly robust but should be treated as upper limits, and that increased dimensionality alone cannot bridge the gap between ab-initio simulated light curves and AT2017gfo observations.","tokens_in":35695,"tokens_out":4055,"duration_ms":41106,"significance":"If the central claims hold, this is a valuable step forward: it provides one of the first consistent extensions of NR merger ejecta to second-long timescales in 3D with non-NSE thermodynamics and coupled nuclear heating, and it quantifies how early truncation of the hydrodynamical evolution affects nucleosynthesis and element sky maps. The paper includes useful internal diagnostics, such as the mass-flux comparison in Appendix A, the homology parameter in Fig. 4, and the E_theta^Kinetic diagnostic in Fig. 6, which support the dynamical conclusions. The authors also release tracer data on Zenodo upon acceptance, which is commendable. The heating-rate input is taken from external SkyNet-based fits rather than fitted to observations, and the 56Ni dominance at 100 days is a post-processing result from WinNet, so the light-curve comparison is not circular. However, the headline conclusion about 2D-versus-3D light curves and the AT2017gfo gap is weakened by a methodological confound in the comparison, as detailed below.","major_comments":[{"comment":"The central 2D-versus-3D light-curve comparison conflates dimensionality with the additional ~1 s of hydrodynamical evolution. As stated in Sec. IIE, the 2D KNEC runs are built from azimuthally averaged THC data at r = 300 M_sun, without the Athena++ 3D extension, and the authors explicitly note that the 2D results are 'also representative of an earlier homology assumption.' The BLh_150 run in Fig. 16 is not a control for this comparison: it is a 3D Athena++ run truncated at 150 ms, differing from the 2D calculation in both dimensionality and data origin. Therefore the observed luminosity enhancement of 2D over 3D cannot be uniquely attributed to the axisymmetry assumption. The conclusion in Sec. IV that 'the 2D ray-by-ray approach remains a robust but upper-limit estimate' and that 'increased dimensionality alone is not sufficient' is not yet cleanly supported. A conclusive test would be to azimuthally average the 1 s Athena++ profiles and feed those into KNEC, or otherwise perform a same-time, same-ejecta 2D control.","section":"Sec. IIE and Sec. IIIC, Fig. 16"},{"comment":"The ejecta inventory used in the comparison is incomplete in a way that limits the AT2017gfo conclusion. The authors state in Sec. IIIA that 'mass ejection from long-lived remnants would still continue past the NR simulated time,' and in Sec. IIIC they attribute a large part of the AT2017gfo gap to missing viscous ejecta. The 3D-vs-2D light-curve comparison, however, is based only on the ejecta that crossed r = 300 M_sun during the ~100 ms covered by the THC simulations. If the later disk outflows are massive and lanthanide-rich, the relative importance of dimensionality versus missing ejecta components could be different for the total ejecta. The paper should either restrict the dimensionality claim to the modeled ejecta component or provide a quantitative estimate of how the missing component would affect the upper-limit interpretation.","section":"Sec. IIIA and Sec. IIIC"},{"comment":"The absence of a resolution study weakens the quantitative claims about light curves and element distributions. The Athena++ grid uses N_theta = 16 and N_phi = 32, which is coarse for capturing the small-scale features of tidal arms and the lanthanide curtain, and the KNEC mapping uses 512 angular sections derived from this grid. No convergence test is presented for the 3D profiles, the opening angles quoted in the abstract (e.g., 15 to 30 degrees), or the light-curve differences in Figs. 16-18. Since several conclusions are phrased in terms of angular structure and luminosity differences of order tens of percent, a resolution study (at least for one model) is needed to establish that the reported 3D features are converged.","section":"Sec. IIA and Sec. IIIC"},{"comment":"The treatment of Ye as a fixed passive scalar after injection is a potentially load-bearing simplification for the light-curve comparison. The authors note in Sec. IIIA that beta-decay and e± capture can drive Ye to ~0.4 on the second timescale, and the KNEC opacity model in Sec. IIE uses a constant opacity set by the initial Ye. Since the lanthanide fraction and hence the opacity depend on Ye, the 2D-versus-3D luminosity comparison could be affected by the frozen-Ye assumption. The a posteriori pressure-error check in Sec. IIC (12% maximum for 90% of the mass) bounds the thermodynamic impact but does not directly bound the opacity or light-curve impact. A test with a simple Ye-evolution prescription, or at least an estimate of the opacity uncertainty, would strengthen the light-curve conclusions.","section":"Sec. IIC and Sec. IIIA"}],"minor_comments":[{"comment":"The last label in the legend reads 'Kinetic Energy' twice; one of the entries should presumably be 'E_theta^Kinetic' as defined in Eq. (19).","section":"Fig. 6 caption"},{"comment":"There is a typo: 'accelatation in the Lagrangian grame' should read 'acceleration in the Lagrangian frame.'","section":"Sec. IIIA surrounding Eq. (16)"},{"comment":"The sentence 'the AB magnitudes are increased when the dimensionality was increased' is ambiguous; since larger AB magnitudes correspond to dimmer objects, the wording should be clarified (e.g., 'the AB magnitudes become larger, i.e., the model becomes dimmer').","section":"Sec. IIIC, paragraph near Fig. 18"},{"comment":"The phrase 'The right captions are insets of the left plots' appears to mean 'the right panels are insets of the left plots' and should be reworded.","section":"Fig. 15 caption"},{"comment":"The projection weights p_k are defined with a vector solid-angle integral; the notation is somewhat terse. A brief explanation of how the bins are constructed and how the weights are normalized would improve reproducibility.","section":"Sec. IIE, Eq. (14)"}],"recommendation":"major_revision","confidential_remarks":"The paper is built on the authors' own recent pipeline (Athena++ implementation, KNEC, and the [35] 2D comparison), which is natural in a fast-moving field but means the 2D baseline is not independent. The central comparative claim would be substantially more convincing with the same-time 2D control suggested in my first major comment. In addition, the paper's scope is somewhat broader than its strongest result; the dynamical and nucleosynthesis parts are solid, but the abstract's AT2017gfo sentence overreaches relative to the caveats in Sec. IIIC. These are fixable with additional analysis or a more careful framing, so I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The real news here is the data: four BNS merger ejecta profiles evolved to about one second in full 3D with nuclear heating coupled to the hydrodynamics, plus element-distribution sky maps and a systematic 2D-vs-3D ray-by-ray light-curve comparison. That combination is new and useful, even though previous work reached similar timescales with different methods. The paper also does a nice job with internal diagnostics: the homology parameter, the E_theta^Kinetic energy, the 150-ms versus 1-s element redistribution, and the Appendix A mass-flux comparisons all hang together and support the qualitative picture. The nucleosynthesis findings (56Ni/56Co dominance at about 100 days, the absence of 44Ti, the 88Sr overproduction) are concrete and testable, and the authors are honest about the missing viscous ejecta and the simplified Ye treatment. The soft spot is the central 3D-vs-2D comparison. The authors claim that increased dimensionality alone is unlikely to close the gap with AT2017gfo, but the 2D runs are not evolved in Athena++: they use azimuthally averaged THC data at r=300 M_sun, so they differ from the 3D runs in both dimensionality and in about one second of additional hydrodynamical evolution, including heating and the approach to homology. The authors explicitly say the 2D results are 'also representative of an earlier homology assumption,' which is precisely the confound. The BLh_150 run helps but does not fully isolate dimensionality, because it is still a 3D run truncated at 150 ms, not a 2D version of the 1-s profiles. So the abstract's strongest claim is not cleanly supported by its own comparison. The conclusion may still be true, and the paper's other arguments (mass overestimation, analytic scaling) point in the same direction, but the specific causality is muddier than the abstract suggests. Minor issues: no resolution convergence study, no error bars on light curves, and the opacity model is deliberately simplified. These are not fatal for a work of this scope, but they should be acknowledged more prominently. The data availability statement promises tracer release after peer review, which is fine. This paper deserves a serious referee. The simulation data are valuable, the analysis is mostly careful, and the conclusions are interesting even if the headline claim about dimensionality needs softening or a cleaner control. I would engage with it in a reading group and likely cite it. Recommend sending to peer review with requests for a clearer comparison and more explicit caveats.","headline":"A valuable but slightly over-claimed 3D ejecta evolution paper: the new data and the 2D-vs-3D light-curve comparison are worth engaging, but the central dimensionality conclusion is weakened by a confounded control.","tokens_in":741,"tokens_out":2350,"would_cite":true,"duration_ms":34290,"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":"Full 3D simulations of binary neutron star merger ejecta show that 2D ray-by-ray kilonova light curves are robust upper limits, and that increased dimensionality alone cannot reconcile theoretical models with the AT2017gfo observations.","keywords":["kilonova","binary neutron star mergers","ejecta evolution","nuclear heating","r-process nucleosynthesis","ray-by-ray light curves","AT2017gfo","3D hydrodynamics"],"falsifier":"Re-run the same 3D pipeline after adding the missing viscous disk outflows (or otherwise extending the simulation until the remnant disk has finished ejecting): if the 3D bolometric light curve then matches AT2017gfo while the azimuthally averaged 2D curve still falls short, the paper's central claim that dimensionality is not the missing ingredient is falsified.","tokens_in":35191,"feed_emoji":"💥","tokens_out":11196,"duration_ms":93909,"temperature":0.7,"pith_summary":"The paper asks whether the standard shortcut of averaging a neutron-star merger's ejecta into an axisymmetric 2D profile before computing the kilonova light curve is reliable. It runs the same ejecta in full 3D out to about one second after merger, with nuclear heating included, and finds that the 2D light curves are broadly correct in shape but systematically too bright: they are upper limits. The negative side of the result is that going to 3D makes the gap between simulated and observed light curves (specifically the kilonova AT2017gfo) slightly wider, not narrower. The paper concludes that missing ejecta components, most likely neutron-rich outflows from the remnant disk, are the more plausible explanation for the discrepancy than the assumed geometry. This matters because it changes where the community should look for the missing physics.","feed_headline":"3D modeling can't bridge the kilonova gap to AT2017gfo","feed_subtitle":"3D ejecta dim the kilonova; 2D curves are upper limits, so missing mass, not geometry, likely explains the gap.","key_machinery":"The argument is carried by a pipeline that uses the ejecta crossing a radius of 300 solar masses during the numerical-relativity window as a time-dependent inner boundary for a 3D general-relativistic hydrodynamics evolution on a Schwarzschild background, with a hybrid equation of state that switches from a nuclear-statistical-equilibrium table to a low-density Helmholtz-type EOS, and with nuclear heating injected through fitted heating-rate tables. The identical ejecta are then mapped onto a Lagrangian ray-by-ray radiation-hydrodynamics light-curve code, once as fully 3D multi-angle profiles and once after a mass-weighted azimuthal average, so that the only difference between the two light-curve sets is the dimensionality of the input. A Newtonian homology parameter quantifies how far the ejecta still are from free homologous expansion at one second, and tracer trajectories post-processed through a nuclear reaction network produce the element sky-maps and isotopic yields.","core_discovery":"Using four numerical-relativity merger simulations as boundary data, the authors evolve the ejected matter in full 3D for about one second with a transition equation of state and fitted nuclear heating rates. They find that nuclear heating delays the onset of homologous expansion, keeps non-radial motions alive in the most asymmetric binary, and reshuffles heavy elements on the sky: extending the evolution from ~150 ms to ~1 s widens the polar region containing 90% of the heavy-element mass from within 15 degrees of the equator to within 30 degrees. The central comparative discovery is that azimuthally averaged (2D) ray-by-ray light curves are broadly robust but should be treated as upper limits; the 3D light curves peak later, are dimmer, and depend more strongly on viewing angle for unequal-mass binaries, with the 3D emission matching its 2D counterpart only for observers who look straight into the densest part of a lanthanide curtain. From this the authors conclude that dimensionality alone is unlikely to close the gap between ab-initio models and the observed AT2017gfo kilonova, and that the gap points instead to missing ejecta, most plausibly viscous disk outflows.","pith_inferences":["If the 2D upper-limit interpretation is right, many previously published kilonova light-curve calculations that azimuthally averaged 3D merger data may have systematically overestimated the emitting mass; re-reading those results as upper limits could revise inferred r-process yields from GW170817/AT2017gfo.","The predicted one-sided lanthanide bullet in asymmetric mergers implies a measurable late-time spectropolarimetric signature; archival or future polarimetric observations of kilonovae could test this geometry directly.","The paper's four models span only two equations of state and two mass-ratio ranges; whether the 2D-versus-3D upper-limit behavior is universal across the neutron-star population remains an open question that a broader parameter survey could settle.","A natural next step the authors do not take is to map the 1 s 3D profiles into a full 3D radiative-transfer solver rather than ray-by-ray sections; the ray-by-ray approximation itself could be tested by comparing its 3D output against a true multi-dimensional transport calculation."],"forward_implications":["Published kilonova light curves computed from azimuthally averaged profiles should be interpreted as upper limits on the bolometric luminosity, not as best estimates.","For unequal-mass mergers, the kilonova's brightness and color depend strongly on viewing angle, with a one-sided 'lanthanide bullet' producing a red, slow-declining direction and a bluer, faster direction opposite to it.","The 56Ni to 56Co to 56Fe decay chain dominates the heating around 100 days, and the associated gamma-ray lines at 846.77 and 1238.288 keV should be strongest when the binary is viewed close to face-on.","The residual mismatch with AT2017gfo is more likely explained by missing neutron-rich viscous disk outflows than by the difference between 2D and 3D geometry, redirecting modeling effort toward including late-time disk ejecta.","Since homologous expansion is not yet reached at one second in the asymmetric models, radiative-transfer light curves that assume homology at earlier times may carry a systematic offset; evolving the ejecta closer to the ~2.5 s homology time would produce more reliable inputs."],"supporting_citations":[{"why":"Supplies the 2D ray-by-ray light curves and 56Ni yields that form the baseline for the dimensionality comparison.","marker":"[35]"},{"why":"Provides the ray-by-ray radiation-hydrodynamics code used to compute both the 3D and 2D light curves.","marker":"[45]"},{"why":"Provides the effective nuclear heating-rate fits and the thermalization/opacity model used in the hydrodynamics and light-curve computation.","marker":"[46]"},{"why":"Defines the underlying analytic heating-rate fit that [46] extends, setting the nuclear energy source term in the ejecta evolution.","marker":"[70]"},{"why":"Provide the numerical-relativity merger simulations whose ejecta at 300 solar masses are injected as time-dependent boundary conditions.","marker":"[66–68]"},{"why":"The nuclear reaction network used to post-process tracer trajectories for element abundances and 56Ni masses.","marker":"[44]"},{"why":"Recent finding that homologous expansion sets in only around 2.5 s, used to support the paper's conclusion that 1 s evolutions are still dynamically active.","marker":"[57]"}],"fun_headline_variants":["3D dims kilonovas, but gap to AT2017gfo persists","Nuclear heating widens heavy-element lobes in mergers","Missing ejecta, not dimensionality, explains kilonova gap","3D light curves are dimmer, but still can't fit AT2017gfo","2D kilonova light curves are upper limits, 3D confirms"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The comparison assumes that the ejecta recorded in the first ~100 ms of the numerical-relativity simulations are the whole inventory powering the light curve, even though later viscous disk outflows are missing.","fun_headline_variants_meta":{"raw":{"variants":["3D dims kilonovas, but gap to AT2017gfo persists","Nuclear heating widens heavy-element lobes in mergers","Missing ejecta, not dimensionality, explains kilonova gap","3D light curves are dimmer, but still can't fit AT2017gfo","2D kilonova light curves are upper limits, 3D confirms"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000855,"raw_usage":{"total_tokens":3824,"prompt_tokens":1166,"completion_tokens":2658,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":782,"completion_tokens_details":{"reasoning_tokens":2561}},"tokens_in":782,"tokens_out":2658,"duration_ms":18562,"temperature":1.0,"reasoning_tokens":2561,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T04:14:50.083529+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-run the same 3D pipeline after adding the missing viscous disk outflows (or otherwise extending the simulation until the remnant disk has finished ejecting): if the 3D bolometric light curve then matches AT2017gfo while the azimuthally averaged 2D curve still falls short, the paper's central claim that dimensionality is not the missing ingredient is falsified.","supporting_citations":[],"review_version":1}