Quantitative modelling of type Ia supernovae spectral time series III: Implications for type Ia supernovae standardisation in cosmology

M. R. Magee

arxiv: 2604.22928 · v1 · submitted 2026-04-24 · 🌌 astro-ph.HE · astro-ph.CO

Quantitative modelling of type Ia supernovae spectral time series III: Implications for type Ia supernovae standardisation in cosmology

M. R. Magee This is my paper

Pith reviewed 2026-05-08 10:01 UTC · model grok-4.3

classification 🌌 astro-ph.HE astro-ph.CO

keywords type Ia supernovaeexplosion mechanismscosmological standardizationsub-Chandrasekhar massspectral modellingmachine learninghost galaxy propertiesdistance estimates

0 comments

The pith

Two thirds of type Ia supernovae are best reproduced by sub-Chandrasekhar mass explosions, and standardizing each mechanism separately reduces distance scatter.

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

The paper applies the riddler machine-learning framework to match spectral time series and light curves of the ZTF SN Ia DR2 sample against explosion simulations. It finds that sub-Chandrasekhar mass events provide the best fit for roughly two thirds of the supernovae. Chandrasekhar mass explosions are disfavoured among the fastest-evolving objects, while sub-Chandrasekhar events are preferred for the reddest ones. Standardizing the two classes independently tightens the residuals in distance estimates, implying that some reported trends with host-galaxy mass and light-curve shape may arise from shifts in the underlying population mix rather than additional physics. This distinction matters for cosmology because type Ia supernovae serve as standard candles whose calibration directly affects inferences about the expansion history.

Core claim

Using the riddler framework on the ZTF SN Ia DR2 sample, approximately two thirds of events are best reproduced by sub-Chandrasekhar mass explosions. Chandrasekhar mass explosions are not favoured for the fastest-evolving SNe Ia, while sub-Chandrasekhar mass explosions are favoured for the reddest SNe Ia. Standardising each explosion mechanism independently reduces scatter in distance estimates, and previously claimed environmental and non-linear light curve shape corrections may be due to changes in the relative populations of different explosion mechanisms.

What carries the argument

The riddler machine-learning framework, which matches observed supernova spectra and light curves against libraries of realistic explosion simulations to assign the most likely explosion mechanism.

If this is right

Selecting SNe Ia in massive, passive galaxies produces a more homogeneous sample dominated by violent merger events.
Standardizing Chandrasekhar and sub-Chandrasekhar events separately reduces scatter in cosmological distance estimates.
Apparent environmental and non-linear light-curve corrections may reflect changes in the relative populations of explosion mechanisms rather than intrinsic effects.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Future cosmological analyses could improve precision by first classifying each supernova's likely explosion channel before applying luminosity corrections.
If the fraction of sub-Chandrasekhar events changes with redshift, this population shift could introduce systematic biases in dark-energy measurements that current standardization methods overlook.
Larger and more uniformly observed samples are required to test whether the reported population trends reach statistical significance.

Load-bearing premise

The riddler framework correctly identifies the true explosion mechanism from the spectra and light curves of each supernova.

What would settle it

A set of late-time spectra or nucleosynthesis measurements for events in the sample that systematically contradict the explosion mechanisms assigned by riddler.

Figures

Figures reproduced from arXiv: 2604.22928 by M. R. Magee.

**Figure 1.** Figure 1: Distributions of SN parameters for our primary (red) and secondary (blue) samples, and the parent ZTF SN Ia DR2 cosmology sample (black). Faded lines show histograms of each parameter and solid lines show smoothed kernel density estimates. Note that some SNe Ia have been excluded as they do not have measured host galaxy properties. (𝑡exp), the luminosity, and the inner boundary velocity (𝑣𝑖). To construct… view at source ↗

**Figure 2.** Figure 2: Comparison between spectra of ZTF18aahheaj (black) and emulated spectra produced by riddler for different model types. Emulated spectra are based on the best-fitting model parameters found by riddler for each explosion mechanism. The overall favoured explosion scenario is shown on the left, while all other explosion scenarios are shown on the right. The evidence for each model, determined by ultranest, is … view at source ↗

**Figure 3.** Figure 3: Fits to ZTF18acbxsge as in view at source ↗

**Figure 4.** Figure 4: Fits to ZTF19aabvfwn as in view at source ↗

**Figure 5.** Figure 5: Distributions of log signal-to-noise ratios and redshifts of SNe Ia in our primary and secondary samples. Circles show our primary sample, while diamonds show our secondary sample. Each point shows a spectrum and is coloured based on the overall best-fitting explosion scenario for that SN. Note that SNe Ia in our primary sample have multiple observations and therefore each SN is shown multiple times. Top a… view at source ↗

**Figure 7.** Figure 7: As in view at source ↗

**Figure 8.** Figure 8: Distributions of SALT2 𝑥1 and 𝑐 parameters for our primary and secondary samples. Circles show our primary sample, while diamonds and faded lines show our secondary sample. Each point is coloured based on the overall best-fitting explosion scenario. estingly however, for our secondary sample, we find that the DOD scenario appears to be marginally favoured for lower resolutions compared to other scenarios. … view at source ↗

**Figure 10.** Figure 10: Comparison between SALT2 𝑐 and host extinction estimated from our best-fitting riddler models. Uncertainties on 𝐴 ℎ𝑜𝑠𝑡 𝑉 are given as 1𝜎 errors. The grey shaded region approximately shows intrinsically blue SNe Ia reddened by dust. We find a number of outliers with high 𝑐 and low 𝐴 ℎ𝑜𝑠𝑡 𝑉 that may be a result of our prior assumptions. burning within the helium shell results in the production of significa… view at source ↗

**Figure 11.** Figure 11: gives the distributions of host galaxy properties for each model type. Our results show some indications of a link between the global host galaxy mass and the mass of the exploding white dwarf, with sub-Chandrasekhar mass progenitors exhibiting a slight bias towards higher mass galaxies. The VM model is not favoured for any SNe Ia hosted by galaxies with masses below ∼ 109.3 𝑀⊙ in our primary sample. The … view at source ↗

**Figure 12.** Figure 12: Cumulative probability distributions of Hubble residuals assuming no environmental corrections for our cosmology sample (dashed line) and different explosion mechanisms or progenitor scenarios. Panels a & b show Hubble residuals for our primary and secondary samples, respectively, calculated based on parameters determined from fits to the full cosmology sample. Panel c shows our combined primary plus seco… view at source ↗

**Figure 13.** Figure 13: Hubble residuals with no 𝑥1 correction as a function of 𝑥1. The dashed and solid lines show the best-fitting linear and broken 𝛼 corrections to the Tripp equation. Adapted from fig. 8 of Ginolin et al. (2025b). and quality cuts. Furthermore, <1 per cent have multiple spectra. Our primary sample is therefore limited by the inclusion of only a small number of SNe Ia, while fits to our secondary sample may n… view at source ↗

read the original abstract

The physics driving type Ia supernovae (SNe~Ia) standardisation in cosmology remains poorly-understood. Recent advances however mean that it is now possible to systematically analyse the explosion properties of large numbers of cosmological SNe~Ia. To that end we use riddler, a machine learning based framework for rapidly modelling SNe~Ia based on realistic explosion simulations, to perform quantitative spectral modelling of the Zwicky Transient Facility SN~Ia DR2 sample and determine their best-fitting explosion mechanism(s). We find that approximately two thirds of our sample is best reproduced by sub-Chandrasekhar mass explosions. Analysing their light curve and host galaxy properties, we find that Chandrasekhar mass explosions are not favoured for the fastest-evolving SNe~Ia, while sub-Chandrasekhar mass explosions are favoured for the reddest SNe~Ia. Due to the differences in their environments, selecting SNe~Ia in massive, passive galaxies could produce a homogeneous sample of violent merger SNe~Ia. We show that standardising each explosion mechanism independently reduces scatter in distance estimates and previously claimed environmental and non-linear light curve shape corrections may be due to changes in the relative populations of different explosion mechanisms. Although a step forward towards understanding SNe~Ia physics in cosmology, we highlight a number of limitations affecting our conclusions, including sample biases and small numbers. We therefore cannot assess the statistical significance of our results and they should be treated with caution. Larger and more uniformly observed samples will be key to determining the significance of any trends hinted at here.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper applies riddler to ZTF DR2 and reports ~2/3 sub-Chandrasekhar events with possible standardization gains, but the result rests on unvalidated ML assignments and a sample the authors themselves flag as biased and small.

read the letter

The core new piece is the quantitative split from running the existing riddler framework on the public ZTF SN Ia DR2 sample, plus the follow-up claim that mechanism-specific standardization reduces distance scatter and that some environmental or light-curve trends may just reflect changing population fractions. They also note that violent-merger events might be isolated by selecting massive passive hosts. That application to recent data is the concrete step forward; the discussion of how explosion channels could explain existing corrections is a reasonable way to connect the modeling to cosmology practice. The work is honest about its limits, which helps. The main weakness is that the two-thirds fraction and the standardization improvement both depend on riddler correctly mapping spectra and light curves to the true mechanism. The abstract gives no held-out accuracy, no cross-checks against independent modeling, and no robustness tests on the assignments, even though the training simulations are known to have incomplete physics. The authors correctly say sample biases and small numbers prevent any significance assessment, but that leaves the population claim and the practical gain as exploratory hints rather than firm results. No error bars or sensitivity checks are described for the classification step itself. This is worth a reading group for people who follow SN Ia standardization or ML applications in transients; the trends are concrete enough to discuss even if they need follow-up. I would not cite the fractions yet, but the paper deserves peer review because the data application is new and the idea of mechanism-dependent standardization is worth referee scrutiny on the validation side.

Referee Report

2 major / 2 minor

Summary. The manuscript applies the riddler machine-learning framework (trained on explosion simulations) to perform quantitative spectral time-series modeling of the ZTF SN Ia DR2 sample. It reports that approximately two thirds of the objects are best reproduced by sub-Chandrasekhar-mass explosions, identifies trends linking explosion mechanism to light-curve evolution and host-galaxy properties, and shows that standardizing each mechanism separately reduces scatter in distance estimates. The authors interpret previously claimed environmental and non-linear light-curve corrections as possible population effects and explicitly caution that sample biases, small numbers, and the inability to assess statistical significance limit the robustness of the conclusions.

Significance. If the riddler assignments can be shown to be reliable, the work supplies a physical interpretation for part of the observed diversity in SN Ia standardization, potentially allowing mechanism-specific corrections that improve cosmological distance precision. The suggestion that host-galaxy and light-curve trends reflect shifts in explosion-mechanism populations rather than intrinsic corrections is a concrete, testable hypothesis. The manuscript's own caveats, however, keep the immediate cosmological impact modest until validation and larger samples are available.

major comments (2)

[Methods (riddler application and sample modeling)] The central population fraction (~2/3 sub-Chandrasekhar) and all downstream claims rest on the riddler framework's ability to map observed spectral time series to the correct explosion mechanism. No quantitative validation metrics (accuracy on held-out simulations, confusion matrices, or cross-checks against independent radiative-transfer modeling) are reported for the classification step itself. This is load-bearing for the standardization and population inferences.
[Results (standardization and scatter analysis)] The claimed reduction in distance-estimate scatter when standardizing each mechanism independently is presented without error bars, bootstrap uncertainties, or any statistical test. Given the explicit statement that statistical significance cannot be assessed, the quantitative improvement remains unquantified and cannot yet support the stronger claim that prior corrections are due to mechanism population changes.

minor comments (2)

[Discussion] The abstract and main text repeatedly note sample biases and small numbers; a dedicated limitations subsection with a quantitative breakdown of selection effects would improve clarity.
[Throughout] Notation for the three mechanism classes (Chandrasekhar, sub-Chandrasekhar, violent merger) should be defined once in the methods and used consistently in all figures and tables.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and detailed review of our manuscript. Their comments highlight important aspects of validation and statistical rigor that we address below. We have revised the manuscript to incorporate additional metrics and to moderate certain claims while preserving the cautious tone already present in the original text.

read point-by-point responses

Referee: [Methods (riddler application and sample modeling)] The central population fraction (~2/3 sub-Chandrasekhar) and all downstream claims rest on the riddler framework's ability to map observed spectral time series to the correct explosion mechanism. No quantitative validation metrics (accuracy on held-out simulations, confusion matrices, or cross-checks against independent radiative-transfer modeling) are reported for the classification step itself. This is load-bearing for the standardization and population inferences.

Authors: We agree that demonstrating the reliability of the riddler classifications is essential. The framework was previously validated on explosion simulations in the papers introducing it, but we acknowledge that specific performance metrics for the current ZTF DR2 application were not reported. In the revised manuscript we will add a dedicated subsection presenting accuracy on held-out simulations and a confusion matrix for the mechanism assignments. Cross-checks against independent radiative-transfer codes are computationally prohibitive at the scale of our sample; we will explicitly discuss this limitation and note that the assigned mechanisms are consistent with qualitative expectations from the literature. revision: yes
Referee: [Results (standardization and scatter analysis)] The claimed reduction in distance-estimate scatter when standardizing each mechanism independently is presented without error bars, bootstrap uncertainties, or any statistical test. Given the explicit statement that statistical significance cannot be assessed, the quantitative improvement remains unquantified and cannot yet support the stronger claim that prior corrections are due to mechanism population changes.

Authors: We concur that the reported scatter reduction lacks formal uncertainties and statistical tests, consistent with the manuscript's own statement that statistical significance cannot be assessed given sample biases and limited numbers. In the revision we will include bootstrap uncertainties on the distance-scatter values. We will also revise the text to describe the reduction as an observed trend and to present the interpretation that prior environmental and light-curve corrections may reflect mechanism population shifts as a testable hypothesis rather than a definitive conclusion. revision: partial

Circularity Check

0 steps flagged

No significant circularity; results driven by external ML model applied to independent data

full rationale

The derivation applies the pre-trained riddler ML framework (external simulations) to ZTF DR2 observations to assign explosion mechanisms, then reports population fractions and standardization gains from those assignments. No paper equation or step reduces the ~2/3 sub-Chandrasekhar fraction, mechanism-specific standardization improvements, or environmental trends to a quantity defined by the present fit. Self-citations to the riddler framework exist but are not load-bearing; the framework is independently trained and the paper explicitly flags sample biases and lack of statistical significance. The chain remains open to external validation.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claims rest on the untested accuracy of the riddler ML classifier for distinguishing explosion mechanisms and on the assumption that the observed sample reflects the underlying population without major selection bias. No free parameters or invented entities are explicitly introduced in the abstract.

axioms (1)

domain assumption The riddler framework, trained on explosion simulations, can reliably assign the dominant explosion mechanism to observed spectra and light curves.
This is the load-bearing premise that converts spectral fits into statements about Chandrasekhar versus sub-Chandrasekhar populations.

pith-pipeline@v0.9.0 · 5590 in / 1416 out tokens · 57816 ms · 2026-05-08T10:01:13.170500+00:00 · methodology

discussion (0)

Forward citations

Cited by 2 Pith papers

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

On the origin of the environmental step: A BayeSN view of the ZTF SN Ia DR2
astro-ph.CO 2026-05 unverdicted novelty 6.0

BayeSN analysis of ZTF SN Ia DR2 data shows a persistent ~0.1 mag environmental step that is intrinsic to the supernovae, not explained by differing dust properties.
On the origin of the environmental step: A BayeSN view of the ZTF SN Ia DR2
astro-ph.CO 2026-05 unverdicted novelty 5.0

BayeSN analysis of ZTF Type Ia supernovae confirms a ~0.1 mag intrinsic environmental step in standardized brightness that is not explained by differences in dust extinction properties.

Reference graph

Works this paper leans on

2 extracted references · 1 canonical work pages · cited by 1 Pith paper

[1]

write newline

" write newline "" before.all 'output.state := FUNCTION fin.entry write newline FUNCTION new.block output.state before.all = 'skip after.block 'output.state := if FUNCTION new.sentence output.state after.block = 'skip output.state before.all = 'skip after.sentence 'output.state := if if FUNCTION not #0 #1 if FUNCTION and 'skip pop #0 if FUNCTION or pop #1...
[2]

A., Leloudas G., Nomoto K., Motohara K., 2010, @doi [ ] 10.1088/0004-637X/708/2/1703 , https://ui.adsabs.harvard.edu/abs/2010ApJ...708.1703M 708, 1703

Maas, Andrew L, Hannun, Awni Y, and Ng, Andrew Y. Rectifier nonlinearities improve neural network acoustic models. In ICML, volume 30, 2013. @ARTICLE maeda--2010b, author = Maeda , K. and Taubenberger , S. and Sollerman , J. and Mazzali , P. A. and Leloudas , G. and Nomoto , K. and Motohara , K. , title = " Nebular Spectra and Explosion Asymmetry of Type ...

work page doi:10.1088/0004-637x/708/2/1703 2013

[1] [1]

write newline

" write newline "" before.all 'output.state := FUNCTION fin.entry write newline FUNCTION new.block output.state before.all = 'skip after.block 'output.state := if FUNCTION new.sentence output.state after.block = 'skip output.state before.all = 'skip after.sentence 'output.state := if if FUNCTION not #0 #1 if FUNCTION and 'skip pop #0 if FUNCTION or pop #1...

[2] [2]

A., Leloudas G., Nomoto K., Motohara K., 2010, @doi [ ] 10.1088/0004-637X/708/2/1703 , https://ui.adsabs.harvard.edu/abs/2010ApJ...708.1703M 708, 1703

Maas, Andrew L, Hannun, Awni Y, and Ng, Andrew Y. Rectifier nonlinearities improve neural network acoustic models. In ICML, volume 30, 2013. @ARTICLE maeda--2010b, author = Maeda , K. and Taubenberger , S. and Sollerman , J. and Mazzali , P. A. and Leloudas , G. and Nomoto , K. and Motohara , K. , title = " Nebular Spectra and Explosion Asymmetry of Type ...

work page doi:10.1088/0004-637x/708/2/1703 2013