Pith. sign in

REVIEW 5 major objections 6 minor 2 cited by

Kilonova modelling and parameter inference: Understanding uncertainties and evaluating compatibility between observations and models

T0 review · 5 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read A kilonova model's fit to data can be scored by the systematic error margin it needs, and that margin also sets how tight the inferred ejecta properties can be.

desk verdict Useful performance forecast for kilonova follow-up, but the sigma_sys goodness-of-fit metric is only tested under its own Gaussian assumption—handle the AT2017gfo compatibility claim with care. read the letter →

arxiv 2505.21392 v1 pith:45EP3KO7 submitted 2025-05-27 astro-ph.HE astro-ph.IM

classification astro-ph.HEastro-ph.IM
keywords kilonovaparameterinferenceBayesiananalysissystematicuncertaintygoodness-of-fitKullback-LeiblerdivergenceobservationalcadenceAT2017gfo
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper tries to establish that the systematic error margin $\sigma_{\rm sys}$, added in quadrature to observational uncertainties in the likelihood, is not just a nuisance parameter but a useful metric of how well a kilonova model fits the data: the value that maximizes the likelihood (equivalently, makes $\chi^2/{\rm dof}\sim1$) is the recommended goodness-of-fit measure, with lower values meaning better agreement. It then quantifies how much physical information kilonova-only parameter inference can extract, using the Kullback-Leibler divergence between prior and posterior as the measure. For a two-component ejecta model and a follow-up cadence of about one observation per day in \textit{grizyJHK} bands, the expected information gain is about 8 nats, meaning the posterior volume is roughly 3000 times smaller than the prior, and a cadence of one observation every three days is the threshold below which ejecta properties cannot be constrained. Applied to AT2017gfo, the model requires $\sigma_{\rm sys}\sim0.57$ mag to fit the data, and combining the lightcurve with viewing-angle and gravitational-wave distance priors yields ejecta-mass constraints that are compatible with, but tighter than, GW-only estimates.

What carries the argument

The engine of the paper is the likelihood in Eq. (1): each observed magnitude is compared with the model magnitude using a total variance $(\sigma_i^j)^2+\sigma_{\rm sys}^2$, so the single hyperparameter $\sigma_{\rm sys}$ simultaneously broadens or narrows every posterior and, at its likelihood-maximizing value, defines the model's effective goodness of fit. Two supporting metrics carry the performance analysis: the Kullback-Leibler divergence $D_{\rm KL}(P||\pi)$ between posterior and prior, computed from nested-sampling outputs, which turns information gain into a single number in nats, and the ratio of posterior to prior 68% credible intervals, which the paper shows scales roughly as $\sigma_{\rm sys}\sqrt{10/N}$. The model side is a surrogate lightcurve generator built from a grid of radiative-transfer simulations with singular-value-decomposition smoothing and Gaussian-process interpolation, whose 0.1–0.3 mag interpolation residuals are one of the contributions absorbed by $\sigma_{\rm sys}$.

What would settle it

Fit the AT2017gfo dataset with the same model and bin the residuals by time and filter: if the residual scatter grows after the first week, differs between optical and infrared bands, or correlates across filters, a single constant $\sigma_{\rm sys}$ cannot represent the model error, and the proposed goodness-of-fit metric and posterior-scaling predictions would not hold. A simpler calculation is to inject a lightcurve with known time-varying model error and check whether the likelihood-maximizing $\sigma_{\rm sys}$ still yields $\chi^2/{\rm dof}=1$ and the predicted posterior-to-prior ratios.

Watch

Extended reading notes

Core claim

On the paper's own terms, the claim is that the likelihood of Eq. (1), with a single Gaussian systematic term $\sigma_{\rm sys}$ added in quadrature to each photometric error, is the right place to absorb model imperfection, and that the $\sigma_{\rm sys}$ which maximizes the likelihood gives a principled, continuous measure of model-data compatibility. Evidence includes a toy model in which injected noise of 0.5 mag is recovered as $\sigma_{\rm sys}=0.4$ mag (with declared measurement errors of 0.3 mag), and the finding that a fixed $\sigma_{\rm sys}=1$ mag leaves AT2017gfo overfit ($\chi^2/{\rm dof}=0.26$), while freeing $\sigma_{\rm sys}$ gives about 0.57 mag. The paper further shows that posterior widths scale roughly linearly with $\sigma_{\rm sys}$, that a roughly one-per-day \textit{grizyJHK} cadence gives $D_{\rm KL}\sim8$ nats (posterior volume about 3000 times smaller than the prior), and that the information gain falls below 5–6 nats for a one-per-three-day cadence, marking the loss of ejecta constraints. Applied to AT2017gfo, the best-constrained parameter is the wind ejecta mass, and with viewing-angle and GW-distance priors the recovered masses are $\log_{10}(M_{\rm ej}^{\rm dyn}/M_\odot)=-1.92^{+0.09}_{-0.12}$ and $\log_{10}(M_{\rm ej}^{\rm wind}/M_\odot)=-1.24^{+0.09}_{-0.05}$ (95% credible intervals).

Load-bearing premise

The load-bearing premise is that all model-data discrepancies can be represented by one constant Gaussian systematic error $\sigma_{\rm sys}$ added to the measurement errors, an assumption the paper does not test for non-Gaussianity, time dependence, or correlation across filters.

Editorial extensions

If this is right

  • The likelihood-maximizing $\sigma_{\rm sys}$ gives a continuous, model-to-model comparable goodness-of-fit score: a model that needs a smaller $\sigma_{\rm sys}$ is the better match to the same event.
  • At a follow-up cadence of about one observation per day per filter in optical plus near-infrared bands, the joint posterior volume shrinks by roughly a factor of 3000 relative to the prior.
  • A cadence of one observation every three days in each band is the practical floor for ejecta-property constraints; below it, the information gain drops below $D_{\rm KL}\sim5$–6 nats and the data mostly bound the distance.
  • Combining kilonova lightcurves with independent viewing-angle and distance priors tightens ejecta-mass estimates beyond what either messenger alone provides, unless the extra priors are nearly redundant with the lightcurve information.
  • Observing strategies can be optimized by filter and time window: early-time points, the $r$ band, and optical plus infrared combinations lift degeneracies most effectively.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If $\sigma_{\rm sys}$ is adopted as a published model-comparison statistic, the natural next step is a formal model-selection test on the difference in optimal $\sigma_{\rm sys}$ between competing models, which the paper does not provide.
  • The predicted cadence thresholds could be inverted into an observing-planning tool: from a target posterior compression, a telescope network could read off the minimal cadence and filter set, though this assumes the Gaussian-error model holds.
  • A direct diagnostic of the constant-Gaussian assumption would be to bin the AT2017gfo fitting residuals by time and filter; correlated or late-time-growing residuals would indicate that a single $\sigma_{\rm sys}$ overstates constraints at some epochs.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 6 minor

Summary. The manuscript presents a methodology-oriented study of kilonova parameter inference. It catalogues sources of uncertainty in kilonova light-curve modelling and inference, proposes the systematic-error hyperparameter sigma_sys in a Gaussian likelihood (Eq. 1) as a goodness-of-fit metric, and uses the Kullback-Leibler divergence between posterior and prior to quantify information gain. Using the Bu2019lm model in the NMMA framework, the authors simulate AT2017gfo-like light curves under different cadences, filter sets, and time windows, and report expected posterior reductions (Tables 3 and 4), a cadence threshold of roughly one observation per three days, and a comparison of simulated versus real inference performance. They then re-analyse AT2017gfo under several prior sets, report an optimal sigma_sys of about 0.57 mag, and claim final ejecta-mass estimates compatible with GW constraints from Dietrich et al. (2020).

Significance. If the proposed metric and performance forecasts were fully validated, the paper would provide practical guidance for kilonova follow-up campaigns and a simple comparative measure of model fit. The paper has clear strengths: a systematic catalogue of uncertainty sources, a substantial simulation campaign (the ManySims recovery tests), careful discussion of cadence/filter effects, and a public data release. However, the central sigma_sys metric is currently validated only under the same Gaussian assumption used to define it, and the AT2017gfo compatibility conclusions rely on priors derived from the comparison data. These issues are load-bearing for the paper's main claims, so they need to be addressed before the results can be taken at face value.

major comments (5)
  1. [Section 2.3, Eq. (1), Fig. 6] The validating toy model injects Gaussian noise with variance (0.5 mag)^2 and declares sigma_meas = 0.3 mag, so the recovered optimum sigma_sys = 0.4 mag is guaranteed by construction. The paper does not test the proposed goodness-of-fit metric against structured residuals, such as per-filter offsets, time-correlated errors, or outliers, which are precisely the systematic effects that sigma_sys is meant to absorb. Without a posterior predictive check or a per-filter/per-epoch residual analysis, the claim that the maximum-likelihood sigma_sys is a reasonable goodness-of-fit metric for real kilonova data is unsupported.
  2. [Section 3.2, Table 5] The statement that Bu2019lm 'needs' sigma_sys ~ 0.57 mag and therefore cannot fit AT2017gfo within the observational uncertainties is circular in a technical sense: sigma_sys is estimated from the same dataset to which it is then applied as a compatibility criterion. Because sigma_sys is chosen to make chi2/dof ~ 1 by construction, the quoted chi2/dof values (0.26, 0.90, 0.96, 0.97) do not constitute independent evidence of good or bad fit. A meaningful test would condition on part of the data and evaluate predictions for the remaining data, or report the full residual covariance rather than a single scalar variance.
  3. [Section 3.2, Table 6, prior sets C and D] The conclusion that KN-only inference with Bu2019lm gives ejecta masses compatible with GW constraints is partly built into the priors. Prior set D restricts log10 Mdyn_ej and log10 Mwind_ej to ranges derived from the Dietrich et al. (2020) GW posterior, and the resulting KN posterior is then compared with that same GW posterior. This demonstrates that the KN likelihood is not strongly inconsistent with the GW-informed region, but it cannot by itself establish compatibility or complementarity between independent messengers. The authors should either use a prior derived from data not used in the comparison, or quantify the relative evidence for prior sets C versus D in a way that does not reuse the comparison data.
  4. [Section 3.1, Tables 3-4, Fig. 8] The performance forecasts (DKL ~ 8 nats, posterior volume about 3000 times smaller than the prior, and the 'one observation per three days' threshold) are conditional on a fixed Gaussian sigma_sys = 1 mag and on simulated surrogate light curves with declared 0.05 mag errors. Table 3 itself shows that posterior sizes scale by roughly a factor of two when sigma_sys changes from 1 to 0.5 mag, so the numerical forecasts cannot be quoted without an accompanying uncertainty on sigma_sys. The paper should present DKL as a function of sigma_sys, or as a distribution over it, and should soften the 'cannot constrain ejecta properties' threshold claim, which is not tied to an explicit criterion on any specific parameter.
  5. [Section 2.3] When sigma_sys is treated as a free parameter and sampled over, no prior on sigma_sys is specified. The reported 'maximum likelihood' value of 0.57 mag for AT2017gfo depends on whether one maximizes the profile likelihood, the posterior, or the joint likelihood; this should be stated explicitly, since the value is used throughout Section 3.2 as the basis for fixing sigma_sys = 0.6 mag.
minor comments (6)
  1. [Abstract and Section 3.2] There are several typos: 'sometime' should be 'sometimes' in the abstract, 'Bu2029lm' in Section 3.2 should be 'Bu2019lm', and 'toymoydel' in Appendix A.1 should be 'toymodel'.
  2. [Table 5] The table caption states 'Posterior 95% credible intervals' while the text in Section 3.2 sometimes reports 68% or 95% intervals without consistent notation; please clarify which intervals are shown in each part of the table and in the text.
  3. [Eq. (1)] The notation in Eq. (1) uses the same symbol m_j^i for both observed and model magnitudes; using e.g. m^{obs} and m^{model} would improve clarity.
  4. [Section 3.2, Fig. 15] The comparison of the two posterior peaks via chi2/dof values of 0.97 versus 0.96 is not a likelihood comparison; the authors should report delta log-likelihood or log-evidence if they wish to compare the peaks.
  5. [Figure 7] The iteration number on the y-axis is plotted as a continuous variable, which can be misleading; discrete markers or a rasterized rendering would better represent the 50 independent realizations.
  6. [Section 2.2.2] There is a punctuation error in 'to optimize computing cost;.'; the sentence should end with a period rather than a semicolon.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the sigma_sys metric is an explicit reparameterization of the reduced chi-square, and the performance forecasts are simulation-based with an external AT2017gfo consistency check.

full rationale

The derivation chain is self-contained and does not reduce to its own inputs. Eq. (1) defines the likelihood with sigma_sys added in quadrature to the observational errors, and Section 2.3 explicitly shows that, varying only sigma_sys, the likelihood is maximal when the reduced chi-square is of order unity. The proposed goodness-of-fit metric is therefore a transparent, acknowledged reparameterization of the minimized chi-square: a lower sigma_sys means smaller residuals relative to the error budget. This is not a hidden circularity because the paper presents it as a definition and recommendation, not as an independent prediction. The AT2017gfo value sigma_sys ~ 0.57 mag is obtained by maximizing the likelihood on that same dataset, and the paper does not claim to predict it from first principles; it is a fitted characterization of model-data discrepancy. The performance forecasts (DKL ~ 8 nats for 1-per-day grizyJHK, ~3.5 nats for 1-per-5-days, and the ~3000x posterior-volume reduction) are produced by simulating Bu2019lm lightcurves with fixed sigma_sys = 1 mag and a stated prior set, and the subsequent analysis of AT2017gfo with the same fixed sigma_sys is an external consistency check against real data rather than a re-use of the fitted sigma_sys. Citations to NMMA (Pang et al. 2023) and to GW-only constraints (Dietrich et al. 2020) include coauthors of the present paper, but these are public code and independent analyses used as tools or external inputs, not as authority to forbid alternatives. The unverified Gaussian and constant-sigma_sys assumption is a genuine robustness and correctness risk for the metric and for the numerical DKL values, but it is a stated modeling assumption, not a circular step.

Assumptions & free parameters 1 free parameters · 5 assumptions · 0 invented entities

The analysis rests on one kilonova model and its surrogate, and on the assumption that all unmodeled discrepancies are absorbed by a single fitted Gaussian sigma_sys. No new physical entities are introduced. The central free parameter is sigma_sys, which is fitted to data and then used as a metric.

free parameters (1)
  • sigma_sys (systematic error margin) = 0.57 mag (optimal for AT2017gfo); also fixed at 1.0, 0.6, 0.5, 0.2, 0 mag in analyses
    A free hyperparameter added to observational uncertainties in Eq. 1. Its optimal value is determined by maximizing the likelihood on the analyzed dataset, and it is then used as a goodness-of-fit metric. This is a fitted quantity, not a measured constant.
assumptions (5)
  • domain assumption The Bu2019lm kilonova model (Bulla 2019) with its four parameters (M_dyn, M_wind, phi, cos(iota), D_L) provides a sufficiently accurate description of kilonova emission for the performance predictions.
    The entire simulation study (sim17, ManySims) and the AT2017gfo analysis rely on this model and its POSSIS grid. Model-to-model variations of 1-3 mag are acknowledged in Section 2.2.1.
  • domain assumption The systematic error can be modeled as a Gaussian with constant variance sigma_sys^2 added to the observational variance.
    Stated in Section 2.3: the 'true' modeled magnitude is assumed to follow a Gaussian distribution of width sigma_sys around the model prediction. If the model-data discrepancy is non-Gaussian or time/filter-dependent, the sigma_sys metric and the predicted posterior sizes are miscalibrated.
  • domain assumption The surrogate model reconstructs the POSSIS grid lightcurves with RMSE < 0.3 mag for most filters (0.25 mag median).
    Section 2.2.2. The interpolation error is a component of sigma_sys, but the surrogate accuracy varies across filters (up to ~0.6 mag for 2massh/ks), which could bias performance predictions.
  • standard math DKL computed with Eq. 3 (mean log-likelihood minus log evidence) is an unbiased estimator of the information gain.
    Used throughout Section 2.4 and 3.1 to compare inference performance. This is a standard identity in nested sampling.
  • standard math The likelihood (Eq. 1) factorizes over independent observations.
    Assumes independent photometric data points, which is standard but may be violated by correlated calibration or extinction uncertainties.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Kilonova modelling and parameter inference: Understanding uncertainties and evaluating compatibility between observations and models." pith.science (2026). https://pith.science/paper/45EP3KO7

@misc{pith2026250521392,
  author       = {Pith},
  title        = {Pith review of: Kilonova modelling and parameter inference: Understanding uncertainties and evaluating compatibility between observations and models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/45EP3KO7}},
  note         = {Machine review of arXiv:2505.21392}
}
abstract

In the study of optical transients, parameter inference is the process of extracting physical information, i.e. constraints on the source's characteristics, by comparing the observed lightcurves to the predictions of different models and finding the model and parameter combination that make the closest match. In the developing field of the study of kilonovae (KNe), systematic uncertainties in modelling are still very large, and many models struggle to fit satisfactorily the whole multi-wavelength dataset of the AT2017gfo kilonova, associated to the Binary Neutron Star (BNS) merger GW170817. In a multi-messenger context, we sometime observe tensions between KN-only inference results and constraints from other messengers. In order to discuss the compatibility of KN models with observations and with the information derived from other messengers, we detail the process of Bayesian parameter inference, identifying the many sources of uncertainty embedded in KN analyses. We highlight the systematic error margin hyperparameter $\sigma_{\rm sys}$, which can be exploited as a metric for a model's goodness-of-fit. We then discuss how to assess the performance of parameter inference analyses by quantifying the information gain using the Kullback-Leibler divergence between prior and posterior. Using the example of the Bu2019lm model with the NMMA Bayesian inference framework, we showcase the expected performance that dedicated KN follow-ups with telescope networks could reasonably reach, highlighting the different factors (observational cadence, error margins) that influence such inference performances. We finally apply our KN analysis to the dataset of AT2017gfo to validate our performance predictions and discuss the complementarity of multi-messenger approaches.

Figures

Figures reproduced from arXiv: 2505.21392 by the authors.

Figure 1
Figure 1. Breakdown of the sources of uncertainties and the sections in which they are treated in this work. focusing on the Bu2019lm model) and discuss the dif￾ferent sources of uncertainties involved. Then, in Sec￾tion 3 we study the typical performances of such frame￾works to see what scientific output can be expected from future KN observations, and compare these predictions with a reanalysis of AT2017gfo. 2. EXPLORATION … view at source ↗
Figure 2
Figure 2. Distribution histogram of the observational mag￾nitudes uncertainties for the AT2017gfo dataset of Cough￾lin et al. (2018). The median value is 0.09 mag. generally by the order of few tenths of mag. For exam￾ple, the magnitude uncertainties of AT2017gfo span a [0.01-0.4] mag range, with most images with 1σ errors around ∼ 0.1 mag, as shown in [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Top: Example of a representative Bu2019lm lightcurve in 2massh-band simulated by POSSIS (correspond￾ing to ejecta parameters Mdyn ej = 0.005M⊙, Mwind ej = 0.09M⊙, ϕ = 45 degrees and cos(θ) = 0.6) in absolute magni￾tude, in blue, and its corresponding reconstruction from the first 10 SVD coefficients, in orange. Bottom: Residual dif￾ference between the two curves. This SVD reduction seems to smooth out a Poisson nois… view at source ↗
Figures from the paper (12 more)
Figure 4
Figure 4. Figure 4: Quartiles of the distribution of the RMS error (in mag) between the simulated model and its reconstruction using a Gaussian Process Regression interpolator, for all the Bu2019lm grid lightcurvves, and across different filters. to this are the 2massh and 2massks filters…
Figure 6
Figure 6. Figure 6: Posterior distribution of the distance, the best￾fit log-likelihood log L( ⃗θbest) and χ 2 /dof for different fixed values of σsys; resulting from NMMA Bu2019lm parameter in￾ferences (default prior set A from [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: Posterior distributions (distance and ejecta masses) and K-L divergences for different lighturve realizations sampled from the sim17 seed with a 1perday cadence (grizyJHK filters); inference with prior set A1 (from [PITH_FULL_IMAGE:figures/full_fig_p010_7.png]
Figure 8
Figure 8. Figure 8: Posterior distributions (Left: Luminosity distance. Right: Dynamical ejecta mass) and K-L divergences when analysing different sim17 lightcurves (grizyJHK) with prior set A1 (from [PITH_FULL_IMAGE:figures/full_fig_p010_8.png]
Figure 9
Figure 9. Figure 9: Posterior wind ejecta distributions and K-L di￾vergences when trying to recover the sim17 lightcurves with 170817like cadence, analysed with prior set A1 (from Ta￾ble 6), using different filters on their own. 2.4 1.6 lo g 1 0 M dy n ej grizy JHK grizyJHK 1.5 1.0 0.5 lo…
Figure 10
Figure 10. Figure 10: Posterior contours of KN-only (Bu2019lm) analysis of AT2017gfo, with prior set B1 (from [PITH_FULL_IMAGE:figures/full_fig_p011_10.png]
Figure 11
Figure 11. Figure 11: Posterior distance distributions and K-L diver￾gences when trying to recover the sim17 lightcurves with 2perday cadence (grizyJHK) truncated to different time windows, analysed with prior set A1 (from [PITH_FULL_IMAGE:figures/full_fig_p012_11.png]
Figure 12
Figure 12. Figure 12: Median K-L divergence obtained with parame￾ter inference (prior set A1 from [PITH_FULL_IMAGE:figures/full_fig_p013_12.png]
Figure 13
Figure 13. Figure 13: , the median DKL(P||π) is about 1 nat greater on the high end of the wind mass range compared to the low end. Properties of expected performance —Given the relative homogeneity of parameter inference constraints, we now explore different properties of posterior distri…
Figure 14
Figure 14. Figure 14: Size of lightcurve variations (Top: Individual lightcurves. Bottom: Relative difference between filters) when moving along axes of the parameter space in steps of 10% of the prior interval, averaged over time and across the whole parameter space. messengers or alterna…
Figure 15
Figure 15. Figure 15: Parameter inference of AT2017gfo using Bu2019lm with different constraints enforced in the priors: Prior set A0.6 in blue, B0.6 in green, C0.6 in red. For each analysis, the overlayed point and solid lines show the position of the best-fit (i.e highest-likelihood) par…
Figure 16
Figure 16. Figure 16: Size of variation across 1perday cadence realizations, plotted against the median posterior distribution interval, in the inclination angle estimates of mock22 lightcurves analysed with prior set A, with varying σsys choices. Coughlin, M. W., Dietrich, T., Margalit, B…

Discussion (0). Sign in to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Binary neutron stars in the next-generation era: Multi-messenger detection prospects and constraints on the equation of state, mass distribution, and cosmology

    astro-ph.HE 2026-07 conditional novelty 6.0 of 10

    With ET (and ET+CE), mock multi-messenger BNS catalogues yield ~40–500 EM counterparts per year and, under ideal recovery, constrain R1.4 to ~0.2 km and H0 to ~1 km s−1 Mpc−1.

  2. nmma: An extended Bayesian framework for Nuclear Multimessenger Astronomy in the Era of Next-Generation Detectors

    astro-ph.IM 2026-07 accept novelty 6.0 of 10

    nmma now jointly samples nuclear EoS parameters with GW and EM data via TOV emulators and Fiesta surrogates, delivering 20–60× speedups and future H0–nuclear constraints.

Reference graph

Works this paper leans on

70 extracted references · 5 canonical work pages · cited by 2 Pith papers

  1. [1]

    P., et al

    Abbott, B. P., et al. 2017 a , Phys. Rev. Lett., 119, 161101, 10.1103/PhysRevLett.119.161101

  2. [2]

    2017 b , Astrophys

    ---. 2017 b , Astrophys. J. Lett., 848, L13, 10.3847/2041-8213/aa920c

  3. [3]

    2017 c , Astrophys

    ---. 2017 c , Astrophys. J. Lett., 848, L12, 10.3847/2041-8213/aa91c9

  4. [4]

    Anand, S., Pang, P. T. H., Bulla, M., et al. 2023, Chemical Distribution of the Dynamical Ejecta in the Neutron Star Merger GW170817. 2307.11080

  5. [5]

    S., et al

    Andrade, C., Alserkal, R., Manzano, L. S., et al. 2025, Publ. Astron. Soc. Pac., 137, 034102, 10.1088/1538-3873/adbfbc

  6. [6]

    2013, Astrophys

    Berger, E., Fong, W., & Chornock, R. 2013, Astrophys. J. Lett., 774, L23, 10.1088/2041-8205/774/2/L23

  7. [7]

    K., et al

    Blanchard, P. K., et al. 2017, Astrophys. J. Lett., 848, L22, 10.3847/2041-8213/aa9055

  8. [8]

    2024, Astron

    Breschi, M., Gamba, R., Carullo, G., et al. 2024, Astron. Astrophys., 689, A51, 10.1051/0004-6361/202449173

Show all 70 references
  1. [9]

    2021, Mon

    Breschi, M., Perego, A., Bernuzzi, S., et al. 2021, Mon. Not. Roy. Astron. Soc., 505, 1661, 10.1093/mnras/stab1287

  2. [10]

    2024, Astrophys

    Brethauer, D., Kasen, D., Margutti, R., & Chornock, R. 2024, Astrophys. J., 975, 213, 10.3847/1538-4357/ad7d83

  3. [11]

    2014, Astron

    Buchner, J., Georgakakis, A., Nandra, K., et al. 2014, Astron. Astrophys., 564, A125, 10.1051/0004-6361/201322971

  4. [12]

    2019, Mon

    Bulla, M. 2019, Mon. Not. Roy. Astron. Soc., 489, 5037, 10.1093/mnras/stz2495

  5. [13]

    2023, Mon

    ---. 2023, Mon. Not. Roy. Astron. Soc., 520, 2558, 10.1093/mnras/stad232

  6. [14]

    2018, Astrophys

    Cantiello, M., et al. 2018, Astrophys. J. Lett., 854, L31, 10.3847/2041-8213/aaad64

  7. [15]

    W., Dietrich, T., Margalit, B., & Metzger, B

    Coughlin, M. W., Dietrich, T., Margalit, B., & Metzger, B. D. 2019, Mon. Not. Roy. Astron. Soc., 489, L91, 10.1093/mnrasl/slz133

  8. [16]

    W., et al

    Coughlin, M. W., et al. 2018, Mon. Not. Roy. Astron. Soc., 480, 3871, 10.1093/mnras/sty2174

  9. [17]

    W., Antier, S., Dietrich, T., et al

    Coughlin, M. W., Antier, S., Dietrich, T., et al. 2020, Nature Communications, 11, 10.1038/s41467-020-17998-5

  10. [18]

    A., et al

    Coulter, D. A., et al. 2017, Science, 358, 1556, 10.1126/science.aap9811

  11. [19]

    J., Sneden, C., Lawler, J

    Cowan, J. J., Sneden, C., Lawler, J. E., et al. 2021, Reviews of Modern Physics, 93, 10.1103/revmodphys.93.015002

  12. [20]

    2022, Mon

    Curtis, S., M\"osta, P., Wu, Z., et al. 2022, Mon. Not. Roy. Astron. Soc., 518, 5313, 10.1093/mnras/stac3128

  13. [21]

    W., Pang, P

    Dietrich, T., Coughlin, M. W., Pang, P. T. H., et al. 2020, Science, 370, 1450, 10.1126/science.abb4317

  14. [22]

    2017, Class

    Dietrich, T., & Ujevic, M. 2017, Class. Quant. Grav., 34, 105014, 10.1088/1361-6382/aa6bb0

  15. [23]

    2021, Astrophys

    Fong, W., et al. 2021, Astrophys. J., 906, 127, 10.3847/1538-4357/abc74a

  16. [24]

    J., Fryer, C

    Fontes, C. J., Fryer, C. L., Hungerford, A. L., Wollaeger, R. T., & Korobkin, O. 2020, Mon. Not. Roy. Astron. Soc., 493, 4143, 10.1093/mnras/staa485

  17. [25]

    2017, Astrophys

    Gao, H., Zhang, B., L\"u, H.-J., & Li, Y. 2017, Astrophys. J., 837, 50, 10.3847/1538-4357/aa5be3

  18. [26]

    2017, Astrophys

    Goldstein, A., et al. 2017, Astrophys. J. Lett., 848, L14, 10.3847/2041-8213/aa8f41

  19. [27]

    W., Dietrich, T., et al

    Heinzel, J., Coughlin, M. W., Dietrich, T., et al. 2021, Mon. Not. Roy. Astron. Soc., 502, 3057, 10.1093/mnras/stab221

  20. [28]

    2013, Phys

    Hotokezaka, K., Kiuchi, K., Kyutoku, K., et al. 2013, Phys. Rev. D, 87, 024001, 10.1103/PhysRevD.87.024001

  21. [29]

    2019, Nature Astron., 3, 940, 10.1038/s41550-019-0820-1

    Hotokezaka, K., Nakar, E., Gottlieb, O., et al. 2019, Nature Astron., 3, 940, 10.1038/s41550-019-0820-1

  22. [30]

    Jhawar, S., Wouters, T., Pang, P. T. H., et al. 2025, Phys. Rev. D, 111, 043046, 10.1103/PhysRevD.111.043046

  23. [31]

    2020, Nature Astron., 4, 77

    Jin, Z.-P., Covino, S., Liao, N.-H., et al. 2020, Nature Astron., 4, 77. 1901.06269

  24. [32]

    2016, Nature Commun., 7, 12898, 10.1038/ncomms12898

    Jin, Z.-P., Hotokezaka, K., Li, X., et al. 2016, Nature Commun., 7, 12898, 10.1038/ncomms12898

  25. [33]

    2017, Nature, 551, 80, 10.1038/nature24453

    Kasen, D., Metzger, B., Barnes, J., Quataert, E., & Ramirez-Ruiz, E. 2017, Nature, 551, 80, 10.1038/nature24453

  26. [34]

    2021, Astrophys

    Korobkin, O., et al. 2021, Astrophys. J., 910, 116, 10.3847/1538-4357/abe1b5

  27. [35]

    Kullback, S., & Leibler, R. A. 1951, The Annals of Mathematical Statistics, 22, 79, 10.1214/aoms/1177729694

  28. [36]

    2024, Mon

    Kunert, N., et al. 2024, Mon. Not. Roy. Astron. Soc., 527, 3900, 10.1093/mnras/stad3463

  29. [37]

    P., et al

    Lamb, G. P., et al. 2019, 10.3847/1538-4357/ab38bb

  30. [38]

    M., & Schramm, D

    Lattimer, J. M., & Schramm, D. N. 1974, Astrophys. J. Lett., 192, L145, 10.1086/181612

  31. [39]

    J., et al

    Levan, A. J., et al. 2024, Nature, 626, 737, 10.1038/s41586-023-06759-1

  32. [40]

    1998, Astrophys

    Li, L.-X., & Paczynski, B. 1998, Astrophys. J. Lett., 507, L59, 10.1086/311680

  33. [41]

    Metzger, B. D. 2017, Living Rev. Rel., 20, 3, 10.1007/s41114-017-0006-z

  34. [42]

    D., Martinez-Pinedo, G., Darbha, S., et al

    Metzger, B. D., Martinez-Pinedo, G., Darbha, S., et al. 2010, Mon. Not. Roy. Astron. Soc., 406, 2650, 10.1111/j.1365-2966.2010.16864.x

  35. [43]

    2021, Astron

    Mochkovitch, R., Daigne, F., Duque, R., & Zitouni, H. 2021, Astron. Astrophys., 651, A83, 10.1051/0004-6361/202140689

  36. [44]

    P., Anderson, J., & Lu, W

    Mooley, K. P., Anderson, J., & Lu, W. 2022, Nature, 610, 273, 10.1038/s41586-022-05145-7

  37. [45]

    P., Deller, A

    Mooley, K. P., Deller, A. T., Gottlieb, O., et al. 2018, Nature, 561, 355, 10.1038/s41586-018-0486-3

  38. [46]

    2021, Astrophys

    Nedora, V., Bernuzzi, S., Radice, D., et al. 2021, Astrophys. J., 906, 98, 10.3847/1538-4357/abc9be

  39. [47]

    2021, Mon

    Nicholl, M., Margalit, B., Schmidt, P., et al. 2021, Mon. Not. Roy. Astron. Soc., 505, 3016, 10.1093/mnras/stab1523

  40. [48]

    2021, Mon

    O'Connor, B., et al. 2021, Mon. Not. Roy. Astron. Soc., 502, 1279, 10.1093/mnras/stab132

  41. [49]

    2017, Astrophys

    Palmese, A., et al. 2017, Astrophys. J. Lett., 849, L34, 10.3847/2041-8213/aa9660

  42. [50]

    Pang, P. T. H., et al. 2023, Nature Commun., 14, 8352, 10.1038/s41467-023-43932-6

  43. [51]

    2018 a , Astrophys

    Radice, D., Perego, A., Hotokezaka, K., et al. 2018 a , Astrophys. J., 869, 130, 10.3847/1538-4357/aaf054

  44. [52]

    2018 b , Astrophys

    Radice, D., Perego, A., Zappa, F., & Bernuzzi, S. 2018 b , Astrophys. J. Lett., 852, L29, 10.3847/2041-8213/aaa402

  45. [53]

    2024, Astrophys

    Ragosta, F., Ahumada, T., Piranomonte, S., et al. 2024, Astrophys. J., 966, 214, 10.3847/1538-4357/ad35c1

  46. [54]

    C., Fong, W., Kilpatrick, C

    Rastinejad, J. C., Fong, W., Kilpatrick, C. D., Nicholl, M., & Metzger, B. D. 2025, Astrophys. J., 979, 190, 10.3847/1538-4357/ad9c77

  47. [55]

    C., et al

    Rastinejad, J. C., et al. 2022, Nature, 612, 223, 10.1038/s41586-022-05390-w

  48. [56]

    2024, Astrophys

    Sarin, N., & Rosswog, S. 2024, Astrophys. J. Lett., 973, L24, 10.3847/2041-8213/ad739d

  49. [57]

    F., & Finkbeiner , D

    Schlafly , E. F., & Finkbeiner , D. P. 2011, , 737, 103, 10.1088/0004-637X/737/2/103

  50. [58]

    2006, Bayesian Analysis, 1, 833, 10.1214/06-BA127

    Skilling, J. 2006, Bayesian Analysis, 1, 833, 10.1214/06-BA127

  51. [59]

    J., et al

    Smartt, S. J., et al. 2017, Nature, 551, 75, 10.1038/nature24303

  52. [60]

    2025, Astrophys

    Stratta, G., et al. 2025, Astrophys. J., 979, 159, 10.3847/1538-4357/ad9b7b

  53. [61]

    2020, Mon

    Tanaka, M., Kato, D., Gaigalas, G., & Kawaguchi, K. 2020, Mon. Not. Roy. Astron. Soc., 496, 1369, 10.1093/mnras/staa1576

  54. [62]

    R., Levan, A

    Tanvir, N. R., Levan, A. J., Fruchter, A. S., et al. 2013, Nature, 500, 547, 10.1038/nature12505

  55. [63]

    K., Arcones , A., K \"a ppeli , R., et al

    Thielemann , F. K., Arcones , A., K \"a ppeli , R., et al. 2011, Progress in Particle and Nuclear Physics, 66, 346, 10.1016/j.ppnp.2011.01.032

  56. [64]

    2018, Nature Commun., 9, 4089, 10.1038/s41467-018-06558-7

    Troja, E., et al. 2018, Nature Commun., 9, 4089, 10.1038/s41467-018-06558-7

  57. [65]

    2019, Mon

    ---. 2019, Mon. Not. Roy. Astron. Soc., 489, 2104, 10.1093/mnras/stz2255

  58. [66]

    2022, Nature, 612, 228, 10.1038/s41586-022-05327-3

    ---. 2022, Nature, 612, 228, 10.1038/s41586-022-05327-3

  59. [67]

    A., et al

    Villar, V. A., et al. 2017, Astrophys. J. Lett., 851, L21, 10.3847/2041-8213/aa9c84

  60. [68]

    T., Fryer, C

    Wollaeger, R. T., Fryer, C. L., Chase, E. A., et al. 2021, Astrophys. J., 918, 10, 10.3847/1538-4357/ac0d03

  61. [69]

    2015, Nature Commun., 6, 7323, 10.1038/ncomms8323

    Yang, B., Jin, Z.-P., Li, X., et al. 2015, Nature Commun., 6, 7323, 10.1038/ncomms8323

  62. [70]

    2021, The Astrophysical Journal, 917, 24, 10.3847/1538-4357/abfe5e

    Zhu, J.-P., Wu, S., Yang, Y.-P., et al. 2021, The Astrophysical Journal, 917, 24, 10.3847/1538-4357/abfe5e

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.