Pith. sign in

REVIEW 4 major objections 6 minor 27 references

Core-collapse supernova parameter estimation with the upcoming Vera C. Rubin Observatory

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read This paper shows, from 6,730 simulated LSST light curves of core-collapse supernovae, that photometry alone cannot precisely recover explosion energy and nickel mass, because the optical bands do not constrain bolometric luminosity…

desk verdict A useful, honestly-caveated forecast that LSST photometry alone will not pin down CCSN explosion parameters, but the quantitative FoMs are pipeline-specific and lack a baseline comparison. read the letter →

arxiv 2506.04184 v1 pith:EDKQH3GC submitted 2025-06-04 astro-ph.HE

classification astro-ph.HE
keywords core-collapsesupernovaeLSSTVeraC.RubinObservatorylightcurveparameterestimationbolometricluminosityCASTORprogenitormassnickel
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

Before the Vera C. Rubin Observatory begins its Legacy Survey of Space and Time, this paper asks what the survey's optical light curves alone will tell us about how core-collapse supernovae explode. The authors run 22,663 simulated supernova explosions through realistic LSST observing conditions, including cadence, depth, saturation, redshift, and extinction, leaving 6,730 usable light curves, and feed those to CASTOR, a pipeline that reconstructs synthetic spectra and estimates distance, extinction, explosion energy, nickel mass, and progenitor mass. The paper's central result is a cautionary one: LSST photometry alone will not give a comprehensive and precise characterization. The bottleneck is spectral coverage, since with only optical bands the bolometric luminosity is poorly constrained and errors flow into explosion energy and nickel yield, while the redshift-extinction degeneracy remains hard to break. The authors conclude that for the scientifically most interesting supernovae, spectroscopic and especially infrared follow-up will be needed to obtain accurate parameters.

What carries the argument

The central machinery is CASTOR, a machine-learning pipeline that estimates supernova parameters from light curves alone. It first matches each target's Gaussian-process-interpolated light curves against a catalog of 106 spectroscopically observed core-collapse supernovae, and then builds synthetic spectra by two-dimensional Gaussian-process interpolation that combines the target's photometric flux densities with the matched reference supernova's observed spectra. Physical assumptions then convert these spectra into parameters, including spherical symmetry, Arnett's virial theorem for ejecta and progenitor mass, the Lusk and Baron model for nickel mass from the linear decay of bolometric luminosity, and Cardelli-McCall extinction laws. The load-bearing result is that bolometric luminosity is estimated from absolute-magnitude light curves in the six LSST bands, whose limited spectral range the paper argues is the reason energy and nickel mass come out wrong.

What would settle it

Apply CASTOR to a spectroscopically complete sample of nearby core-collapse supernovae observed in LSST-like bands, comparing photometric-only bolometric luminosities, explosion energies, and nickel masses with values measured from full spectral energy distributions; if the photometric-only estimates agree with the spectral values within the mean relative deviations quoted in the paper across a sample that includes reddened and higher-redshift events, the central negative claim would be overturned.

Watch

Extended reading notes

Core claim

On its own terms, the paper's discovery is a quantitative mismatch between true and recovered parameters. Across the 6,730-event sample, the recovered and true distributions differ markedly: distance has a Kullback-Leibler divergence of 1.20 with a mean relative deviation of 2.75; extinction 0.53 and 1.69; explosion energy 1.37 and 2.61; and nickel mass 2.15 and 15.9. Progenitor mass is the exception, recovered reasonably well (KL 0.18 to 0.74, relative deviation 0.52 to 0.84) through a virial-theorem relation between energy and ejecta velocity. The paper traces the failures to LSST's optical-only spectral coverage: the bolometric luminosity cannot be measured accurately from the limited wavelength range, and the error propagates to the derived energy and nickel mass. It further shows that fixing extinction substantially improves distance, energy, and progenitor-mass estimates, and that saturation gaps near maximum light are a major source of outliers. The conclusion on the paper's own terms is that LSST photometry alone will not suffice for comprehensive and precise core-collapse supernova parameter estimation.

Load-bearing premise

The load-bearing premise is that two supernovae with similar light curves also have similar spectra, so a synthetic spectrum built by grafting the target's photometry onto a photometrically matched reference's observed spectra is trustworthy; if photometric twins are not spectral twins, the bolometric luminosity, energy, and nickel mass are systematically biased.

Editorial extensions

If this is right

  • For the roughly ten million supernovae LSST will discover, photometric-only estimates of explosion energy will carry typical relative deviations above 100 percent, so population-level studies of explosion energies cannot rest on LSST light curves alone.
  • Nickel mass is the hardest parameter, with a mean relative deviation of 15.9, and about 17 percent of events lack any observable linear decay phase, so most LSST nickel-mass estimates will be usable only as order-of-magnitude values.
  • Progenitor mass is the most robust photometric-only product, with relative deviations of 0.52 to 0.84, so LSST alone can still map the red-supergiant progenitor mass distribution, though with a wide neutron-star-to-black-hole remnant interval.
  • Because fixing extinction markedly improves distance, energy, and progenitor mass, a modest program of spectroscopy aimed at measuring extinction would leverage the scientific value of the entire LSST sample.
  • Saturated peaks, affecting about 19 percent of the filtered sample, are disproportionately responsible for outliers, so follow-up observations in less-saturated bands, particularly in the infrared, would pay large dividends.

Reading between the lines

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

  • If the paper's negative result holds, LSST's scientific yield for core-collapse supernova physics will depend on a tiered observing strategy: the survey itself can deliver progenitor masses and coarse rates, while a small fraction of events receives spectroscopic follow-up for extinction and bolometric calibration.
  • The same CASTOR pipeline could be tested on existing dense spectrophotometric samples of nearby core-collapse supernovae before LSST begins, directly checking whether photometric twins are also spectral twins; if they are not, the reference-spectrum method, rather than LSST itself, would be the limiting factor.
  • The simulation grid used here covers only five progenitor masses from 10 to 18 solar masses and two nickel masses, so the conclusion is most secure for typical Type II-P supernovae and should not be extrapolated to stripped-envelope, superluminous, or interacting supernovae without new simulations.
  • The paper's finding that cadence has little effect on accuracy, because Gaussian-process interpolation smooths cadence differences, suggests that sparse but regular LSST sampling will not be the main limitation; filter coverage and saturation will be.
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

4 major / 6 minor

Summary. The paper assesses how well upcoming LSST photometry alone will constrain core-collapse supernova (CCSNe) parameters. Using 22,663 simulated LSST light curves generated from a grid of STELLA red-supergiant explosions (Moriya et al. 2023), the authors propagate the simulations through magnitude-limit and saturation cuts, leaving 6,730 events, and then interpolate the light curves with Gaussian processes. The CASTOR pipeline matches each target light curve to a catalog of 106 observed CCSNe, constructs synthetic spectra by grafting the reference SN's spectra onto the target light curves, and estimates distance, extinction, explosion energy, nickel mass, and progenitor mass. Two figures of merit, the Kullback-Leibler divergence and a mean relative deviation, are used to compare true and estimated parameter distributions. The paper finds large reconstruction errors for nickel mass and energy, moderate errors for distance and extinction, and relatively good progenitor-mass estimates; it concludes that LSST photometry alone will not suffice for precise characterization and that complementary spectral and infrared follow-up will be essential.

Significance. If the central conclusion holds, this is a useful quantitative expectation for Rubin follow-up planning: it suggests that precise explosion parameters for the bulk of CCSNe will require spectroscopy or additional bandpasses. The study's strengths include the use of a large public simulation grid, clearly defined figures of merit, and two additional tests that isolate the effects of extinction degeneracy and saturation. The paper is transparent about several of its own limitations, such as the redshift incompleteness of the reference catalog and the use of the same Hubble law for true and estimated distances. However, the central negative claim rests on an untested assumption about the correspondence between photometric similarity and spectral similarity, and the quantitative figures of merit are computed on a heavily filtered subsample. These issues make the conclusion plausible but not yet fully established.

major comments (4)
  1. [Sec. 3.2, Table 1] The assumption that 'two photometrically similar SNe have spectra that behave in the same way' is load-bearing and untested. CASTOR builds the synthetic spectra of each target by grafting the observed spectra of a chi-square-selected reference SN onto the target light curves, and the bolometric luminosity, explosion energy, nickel mass, and progenitor mass all propagate from those synthetic spectra. The reference catalog is limited to z < 0.1, has no y-band coverage, and has sparse u/z coverage, while the target simulations extend to z ~ 0.6. A light-curve match does not guarantee spectral similarity, and a spectral mismatch would bias bolometric corrections in a way that could produce exactly the large FoM values reported. I recommend a concrete validation: use the STELLA spectra themselves to build mock targets, run the full CASTOR pipeline, and check whether the matched reference spectra reproduce the true bolometric luminosity and derived parameters. Without such a test, the conclusion that LSST alone is insufficient may reflect an artifact of the template assumption rather than a fundamental limit of LSST photometry.
  2. [Sec. 3.2, Table 1] The distance figure of merit is partly tautological. The text states that both the true distances used in the simulations and the estimated distances are obtained with the same Hubble law and H0 = 70 km/s/Mpc, so the distance FoM mainly tests how well the redshift is recovered from the synthetic spectral lines, not how well LSST measures distances as an independent probe. This is acknowledged in Sec. 3.2, but the paper later interprets distance as one of the best-reconstructed parameters. Because energy and nickel mass propagate from the distance estimate, their FoM values inherit this limitation. Please reframe the distance results as 'redshift-based distance under an assumed cosmology' and, ideally, add a test with a different H0 or with a cosmology-free distance indicator to assess the robustness of the distance and downstream parameter conclusions.
  3. [Sec. 2, Secs. 6.3-6.4] The FoM statistics describe only a heavily filtered subset, not the full simulated sample. Of the 22,663 initial simulations, only 6,730 survive the L2 magnitude, saturation, and point-count cuts. Further cuts remove 14% of events as energy outliers, roughly 20% of events lack a measurable linear decay and are excluded from the nickel-mass analysis, 31.8% of nickel-mass estimates are flagged as outliers, and 8.5% of progenitor-mass estimates are excluded. The abstract and conclusions should explicitly state that the insufficiency claim applies to the detectable, well-sampled subset, and the paper should report event counts at each selection step and a sensitivity analysis to the outlier cuts so that Table 1 is not interpreted as representative of all LSST CCSNe.
  4. [Sec. 4.2, Fig. 4] The conclusion that observation cadence has little impact on parameter estimation is an artifact of the Gaussian-process interpolation performed before the analysis. The text itself says this is 'a direct consequence of using pre-interpolated datasets,' because the GP step smooths out differences between cadence strategies before the FoM binning is applied. The binning therefore cannot test the effect of cadence on the raw LSST sampling. This is not presented as a limitation but as a finding. Please compute the cadence binning on the original L2 sampling (with realistic gaps) or explicitly state that the flat trend in Fig. 4 reflects the interpolation procedure, not an intrinsic insensitivity to cadence.
minor comments (6)
  1. [Sec. 1] The introduction states that LSST will identify 'approximately 10 million changes in the sky each night,' which conflicts with the later statement that 'around 10 million supernovae will be observed over the course of a decade.' Please clarify the intended numbers.
  2. [Sec. 3.1] Please provide a versioned DOI or software repository reference for CASTOR v2.0 and a more complete description of the 106 reference SNe (subtypes, redshift range, number of spectra per object) so that the template-matching experiment is reproducible.
  3. [Fig. 3] The figure caption states that the distributions are binned logarithmically and normalized differently per panel, but the axes are not fully labeled and the binning scheme is not specified. Because the KL divergence is binning-dependent, please show the bin edges and state whether the KL values in Table 1 are computed on counts or densities.
  4. [Eq. (2)] In the discrete KL divergence formula, P and Q are described as the true and estimated distributions, but it is not specified whether they are histograms of counts or normalized densities, nor how the number of bins is chosen. This matters for the interpretation of the DKL values in Tables 1 and 2.
  5. [Sec. 6.4] The text reports a median estimated nickel mass of 10^-3 solar masses, but the simulated grid only contains nickel masses of 0.001 and 0.01 solar masses. Please clarify which subsample this median refers to and why the median is not one of the two grid values.
  6. [Table 2] Columns (a) and (b) are only identified in the table notes. Please add column headers directly to the table for readability.

Circularity Check

2 steps flagged · score 4.0 of 10

Distance validation is partly self-referential (true and estimated distances share the same Hubble law and H0); the central LSST-insufficiency claim otherwise rests on independent simulations and is not circular.

  1. self definitional [Section 3.2, Parameters (page 3)]
    "Moreover, the distance of each object was estimated via a Hubble law, with H0 = 70 km s−1 Mpc−1. We emphasize that this may not be the best approximation of distance of local galaxies, due to peculiar velocity effects with the consequence of higher errors of reconstruction at short distances. However, since the true distance values were also obtained under the same assumption, this error only propagates to the other parameters without a ffecting the distance comparison itself."

    The distance figure of merit compares estimated distances to 'true' distances that were generated with the same Hubble law and the same H0=70 value used by the estimator. The KL divergence and FoM for distance therefore cannot validate absolute distance calibration; they measure only internal consistency of the redshift/distance conversion. Because bolometric luminosity, energy, and nickel mass all propagate from distance, the reported errors in those parameters inherit this self-referential distance scale. The paper explicitly acknowledges the shared assumption but still presents the distance FoM as a reconstruction result.

  2. self definitional [Section 4.2, Relative deviation from true value (page 4)]
    "Notably, the observation cadence has little impact on parameter estimation. This is a direct consequence of using pre-interpolated datasets, which e ffectively eliminate differences between various cadence strategies."

    The paper reports as a finding that cadence has little impact, but this is guaranteed by the analysis design: light curves are GP-interpolated before parameter estimation, which removes cadence information by construction. The 'result' is therefore not an empirical property of LSST observing strategies but a consequence of the interpolation step, as the quoted sentence itself concedes. This is a minor, non-central example of a conclusion reducing to its input.

full rationale

The paper is a simulation-based self-test rather than an external prediction. Its main conclusion—that LSST photometry alone yields poor bolometric luminosity, energy, and nickel estimates—is derived from comparing CASTOR estimates against STELLA-simulated truths, with spectra taken from an observed CCSN catalog; that chain is not circular. The clearest circular element is the distance figure of merit: Section 3.2 states that true distances were obtained under the same Hubble-law/H0 assumption used by the estimator, so distance FoM/KL values measure internal consistency rather than absolute distance accuracy. This also propagates into luminosity-dependent parameters, though it does not fully determine the central negative result, since the dominant reported errors are driven by filter coverage, saturation, extinction, and template mismatch. A second, minor by-construction result is the claim that cadence has little impact, which the paper itself attributes to GP pre-interpolation. Self-citations to CASTOR are backed by the simulation tests in this paper and are not load-bearing in a circular way. The untested photometric-similarity assumption is a correctness/validity risk, not a circularity.

Assumptions & free parameters 4 free parameters · 8 assumptions · 0 invented entities

The central claim rests on a chain of physical assumptions (spherical symmetry, adiabaticity, mass conservation, Hubble law, spectral similarity) and on heuristic GP hyperparameters. No new entities are introduced. The Hubble-law assumption is notable because the benchmark truth is generated under the same law, and the spectral-similarity assumption is the most fragile link for the derived bolometric quantities.

free parameters (4)
  • GP kernel amplitude A = set to mean magnitude of the light curve
    In Eq. (1), the amplitude A of the Matern kernel is chosen heuristically as the mean magnitude value, affecting the interpolation and hence all derived parameters.
  • GP kernel lengthscale sigma = mixture of mean, max, and min cadence
    The lengthscale in Eq. (1) is set to a mixture of the mean, maximum, and minimum cadence of each light curve, a hand-chosen hyperparameter.
  • Wavelength lengthscale in spectral GP = 70 Angstrom
    In Sec. 3.1, the wavelength lengthscale for the 2D GP building synthetic spectra is fixed at 70 Angstrom, a non-data-driven choice.
  • Time lengthscales in spectral GP = minimum and maximum sampling step
    In Sec. 3.1, the time lengthscales are fixed as the minimum and maximum sampling step of the reference SN, chosen by hand.
assumptions (8)
  • domain assumption Spherical symmetry of the explosion
    Stated in Sec. 3.2 among the physical assumptions used for parameter estimation.
  • domain assumption Perfect adiabaticity at peak luminosity and complete conversion of explosion energy into kinetic energy with 99.9% neutrino and 0.1% photon partition
    Stated in Sec. 3.2 as a canonical assumption for energy estimation.
  • domain assumption Perfect conservation of mass and canonical nucleosynthesis processes
    Stated in Sec. 3.2 as a basis for progenitor mass estimation.
  • domain assumption Hubble law with H0 = 70 km/s/Mpc for distance estimation
    Stated in Sec. 3.2; the true distance values are generated under the same assumption, making the distance comparison partly tautological.
  • domain assumption Photometrically similar SNe have similar spectra
    Stated in Sec. 3.2 as the key assumption for building synthetic spectra from a reference SN.
  • domain assumption Cardelli extinction law with RV = 3.1
    Used in Sec. 3.2 when a direct color-based extinction estimate is not possible.
  • domain assumption Lusk & Baron (2017) model for nickel mass from linear decay of bolometric luminosity
    Used in Sec. 3.2; the nickel mass estimate depends on this assumed relation.
  • standard math Arnett (1982) virial theorem relating ejecta mass, energy, and velocity
    Used in Sec. 3.2 to convert energy and velocity into ejecta mass and hence progenitor mass.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Core-collapse supernova parameter estimation with the upcoming Vera C. Rubin Observatory." pith.science (2026). https://pith.science/paper/EDKQH3GC

@misc{pith2026250604184,
  author       = {Pith},
  title        = {Pith review of: Core-collapse supernova parameter estimation with the upcoming Vera C. Rubin Observatory},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EDKQH3GC}},
  note         = {Machine review of arXiv:2506.04184}
}
read the original abstract

The Vera Rubin Observatory's Legacy Survey of Space and Time (LSST) is expected to revolutionize time-domain optical astronomy as we know it. With its unprecedented depth, the LSST will survey the southern hemisphere sky, generating nearly 32 trillion observations over its nominal 10-year operation. Among these, approximately 10 million will be supernovae (SNe). These observations will uniquely characterize the SN population, enabling studies of known and rare SN types, detailed parameterization of their light curves, deep searches for new SN progenitor populations, the discovery of strongly lensed SNe, and the compilation of a large, well-characterized sample of superluminous SNe. We analyzed a sample of 22663 simulations of LSST light curves for core collapse SNe (CCSNe), modeled using the radiative transfer code STELLA. We analyzed this dataset with the software CASTOR, which enables the reconstruction of synthetic light curves and spectra via a machine learning technique that allows one to retrieve the complete parameter map of a SN. For each parameter we compared the observed and the true values, determining how LSST light curves alone will contribute to characterize the progenitor and the explosion. Our results indicate that LSST alone will not suffice for a comprehensive and precise characterization of progenitor properties and explosion parameters. The limited spectral coverage of LSST light curves (in most cases) does not allow for the accurate estimation of bolometric luminosity, and consequently, of the explosion energy and nickel yield. Additionally, the redshift-absorption degeneracy is difficult to resolve without supplementary information. These findings suggest that for the most interesting SNe, complementary follow-up observations using spectrographs and optical facilities (particularly in the infrared bands) will be essential for accurate parameter determination.

Figures

Figures reproduced from arXiv: 2506.04184 by the authors.

Figure 1
Figure 1. Distribution of redshift (top left panel), extinction (top right), cadence (bottom left), and number of filters (bottom right) of the 6730 SNe in the L2 level of data. comparable to the LSST ones in terms of throughput (Gunn et al. 1998). Therefore, we considered only the SNe that had observa￾tions in these filters, reducing the catalog of reference SNe to a total of 106 [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 3
Figure 3. Comparison between the true (dark blue) and the observed (gold) distribution for each estimated parameter. Except for the extinction one, every plot is log scaled and the distributions are binned logarithmically. The normalization (y axes) changes for each distribution. Note that, because the bins are logarithmically distributed, the density is estimated differently for each bin. 5.1. Breaking degeneracy As was ment… view at source ↗
Figure 4
Figure 4. Relative deviation of distance (first column from the left), extinction (second column), energy (third column), mass of nickel (fourth column), and mass of progenitor (fifth column) from the true values in bins of redshift (first row from the top), extinction (second row), cadence (third row), number of filters (fourth row), and luminosity at peak (fifth row). Each point represents the relative deviation (FoM) for e… view at source ↗
Figures from the paper (1 more)
Figure 5
Figure 5. Figure 5: Mean deviation (FoM) of the observed mass of nickel from its true value binned in terms of the number of points available in the linear decay phase. 6.4. Mass of nickel The mass of nickel is the most critical parameter, due to instru￾mental limitations, dependencies, a…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

27 extracted references · 22 canonical work pages

  1. [1]

    Arnett, W. D. 1982, ApJ, 253, 785

  2. [2]

    V ., Blinnikov, S

    Baklanov, P. V ., Blinnikov, S. I., & Pavlyuk, N. N. 2005, Astron- omy Letters, 31, 429

  3. [3]

    2021, extinction: Dust extinction laws, Astrophysics Source Code Library, record ascl:2102.026

    Barbary, K. 2021, extinction: Dust extinction laws, Astrophysics Source Code Library, record ascl:2102.026

  4. [4]

    & Sorokina, E

    Blinnikov, S. & Sorokina, E. 2004, Ap&SS, 290, 13

  5. [5]

    Woosley, S. E. 1998, ApJ, 496, 454

  6. [6]

    I., Röpke, F

    Blinnikov, S. I., Röpke, F. K., Sorokina, E. I., et al. 2006, A&A, 453, 229

  7. [7]

    2018, ApJ, 853, 57

    Bose, S., Dong, S., Pastorello, A., et al. 2018, ApJ, 853, 57

  8. [8]

    & Wheeler, J

    Branch, D. & Wheeler, J. C. 2017, Supernova Explosions (Springer)

Show all 27 references
  1. [9]

    A., Clayton, G

    Cardelli, J. A., Clayton, G. C., & Mathis, J. S. 1989, ApJ, 345, 245 De Jaeger, T., González-Gaitán, S., Anderson, J., et al. 2015, ApJ, 815, 121

  2. [10]

    Filippenko, A. V . 1997, ARA&A, 35, 309

  3. [11]

    2019, LSST Alerts: Key Numbers

    Graham, M., Bellm, E., Guy, L., Slater, C., & Dubois-Felsmann, G. 2019, LSST Alerts: Key Numbers

  4. [12]

    E., Carr, M., Rockosi, C., et al

    Gunn, J. E., Carr, M., Rockosi, C., et al. 1998, AJ, 116, 3040

  5. [13]

    M., Bianco, F

    Hambleton, K. M., Bianco, F. B., Street, R., et al. 2023, Publica- tions of the Astronomical Society of the Pacific, 135, 105002

  6. [14]

    & Pinto, P

    Hamuy, M. & Pinto, P. A. 2002, ApJ, 566, L63 Ivezi´c, Ž., Kahn, S. M., Tyson, J. A., et al. 2019, ApJ, 873, 111

  7. [15]

    & Leibler, R

    Kullback, S. & Leibler, R. A. 1951, The Annals of Mathematical Statistics, 22, 79 LSST Collaboration, Abell, P. A., Allison, J., et al. 2009, arXiv e-prints, arXiv:0912.0201

  8. [16]

    Lusk, J. A. & Baron, E. 2017, PASP, 129, 044202

  9. [17]

    McCall, M. L. 2004, AJ, 128, 2144

  10. [18]

    J., Sorokina, E

    Moriya, T. J., Sorokina, E. I., & Chevalier, R. A. 2018, Space Science Reviews, 214, 1

  11. [19]

    J., Subrayan, B

    Moriya, T. J., Subrayan, B. M., Milisavljevic, D., & Blinnikov, S. I. 2023, PASJ, 75, 634

  12. [20]

    X., & Bloom, J

    Poznanski, D., Prochaska, J. X., & Bloom, J. S. 2012, MNRAS, 426, 1465

  13. [21]

    Rasmussen, C., Bousquet, O., Luxburg, U., & Rätsch, G. 2004, Advanced Lectures on Machine Learning: ML Sum- mer Schools 2003, Canberra, Australia, February 2 - 14, 2003, Tübingen, Germany, August 4 - 16, 2003, Revised Lectures, 63-71 (2004), 3176

  14. [22]

    R., Law, N

    Rau, A., Kulkarni, S. R., Law, N. M., et al. 2009, PASP, 121, 1334

  15. [23]

    P., Kirshner, R

    Schmidt, B. P., Kirshner, R. P., Eastman, R. G., et al. 1994, AJ, 107, 1444

  16. [24]

    2024, MNRAS, 533, 3053

    Simongini, A., Ragosta, F., Piranomonte, S., & Di Palma, I. 2024, MNRAS, 533, 3053

  17. [25]

    2025, CASTOR: v2.0 - 2025-03-05

    Simongini, A., Ragosta, F., Piranomonte, S., & DI PALMA, I. 2025, CASTOR: v2.0 - 2025-03-05

  18. [26]

    M., & Janka, H.-T

    Sukhbold, T., Ertl, T., Woosley, S., Brown, J. M., & Janka, H.-T. 2016, ApJ, 821, 38

  19. [27]

    A., Zimmerman, G

    Weaver, T. A., Zimmerman, G. B., & Woosley, S. 1978, ApJ, 225, 1021 Article number, page 9 of 9

Pith tools

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