Pith. Sign in

REVIEW

Core-Collapse Supernova Rates in the Dark Energy Survey

PaperPDF from arXiv

data files

ReviewThe objections and verdict, as JSON Paper breakdownThe authors' claims, conditions and measurements, as JSON Signed receiptProof this review has not been altered
X Bluesky LinkedIn Reddit Hacker News
K. Boyne, O. Graur, M. Vincenzi, P. Wiseman, M. Grayling, M. Smith, C. Frohmaier, C. Lidman

High Energy Astrophysical Phenomena · Submitted 19 Oct 2026 · 296 references

Pith's verdict: Careful, transparent CC SN rate measurement with genuinely new z<1.3 coverage; the headline slope and secondary claims all carry caveats that need referee attention, especially the unvalidated high-z luminosity function.

T0 · AI review

How this review was made: An AI referee (deepseek/deepseek-v4-flash) read the paper on October 21, 2026 and answered a fixed list of questions about it. T0 is the starting level every paper gets. The levels describe how deeply Pith has checked the paper, never the authors.

  1. T0 AI review An AI read the paper and answered a fixed list of questions.
  2. T1 Audit Automatic checks that the references and files the paper relies on exist and match.
  3. T2 Compute The paper's code or calculations re-run with the same results.
  4. T3 Formal statement The claims written out in a formal math language a computer can read.
  5. T4 Proof Those claims proved, and the proof checked by computer.

How Pith checks papers →

· 3 major objections · 7 minor · Ask about this paper · The compiled record · Read the full review →
Challenge this review Comment (0) Read on arXiv ↗
Cite ▾
arXiv ID2610.27328 (version 1)
Pith IDPEDGQJOW, Pith's permanent record of this paper and its review
CitationC. Frohmaier, C. Lidman, K. Boyne, M. Grayling, M. Smith, M. Vincenzi, O. Graur, P. Wiseman. "Core-Collapse Supernova Rates in the Dark Energy Survey." arXiv:2610.27328 (2026). https://pith.science/paper/PEDGQJOW
Cite this reviewBibTeX for Pith's review of this paper
Claim paper

Why Pith says this

Highlight any text to comment on it.

Referee report

How Pith checks papers → · Full audit of this paper →

The paper

The paper claims

The deaths of massive stars, counted from 2,445 supernovae found by a five-year optical survey, rose as $(1+z)^{2.85}$ from $z=0$ to 1.3 with no sign of the expected turnover.

The reading

Every star more than about eight times the Sun's mass ends its life in a core-collapse supernova, so the rate of these explosions tracks how fast the Universe was making massive stars. This paper aims to measure that rate across nine billion years of lookback time from 2,445 events found by the five-year Dark Energy Survey, giving the most detailed single-survey rate coverage from $z=0$ to 1.3. The central result is that the volumetric rate climbs smoothly as $R_{\rm CC}(z) = (1.58 \pm 0.17)\times10^{-4}\,(1+z)^{2.85 \pm 0.25}\,\mathrm{yr}^{-1}\,\mathrm{Mpc}^{-3}$, with the downturn expected near cosmic noon ($z \approx 2$) not yet visible. A second claim is that stripped-envelope supernovae make up $31 \pm 5$ per cent of all core-collapse events and keep that fraction constant across the full redshift range, which the authors read as evidence that binary companions, not stellar winds alone, strip the exploding stars. A third claim, that the measured rate runs about 70 per cent above recent star-formation-history predictions built on a standard Salpeter initial mass function, bears on whether those histories miss faint galaxies or whether the stellar mass distribution is more top-heavy than assumed.

What carries the argument

The argument is carried by the rate inversion $R_{\rm CC}(z_i) = N_{\rm CC}(z_i)/[\epsilon_{\rm class}(z_i)\,\epsilon_{\rm detect}(z_i)\,V(z_i)\,T^{\rm rest}_{\rm control}(z_i)]$, which turns corrected event counts per redshift bin into a volumetric rate per year per cubic megaparsec. Its three decisive components are the two-stage random forest classifier that separates supernovae from active galactic nuclei and then core-collapse from Type Ia events; the detection efficiency measured by injecting simulated supernova light curves into the survey, which drops steeply with redshift; and the missing-fraction dust-extinction correction applied uniformly across all redshifts. A physical forward model, a single-Gaussian luminosity function with mean $M_0=-16.40$ mag and width $\sigma=1.52$ mag folded through the survey's limiting magnitude of 23.96 mag, reproduces the simulation-derived correction factors and anchors the high-redshift end, where the efficiency is only about $6\times10^{-4}$ and the completeness correction exceeds a thousand. The power-law fit to the corrected rates and the stripped-envelope fraction are the outputs that carry the scientific conclusions.

Extended reading notes

Core claim

The paper's discovery is the most finely binned volumetric core-collapse supernova rate yet measured by a single survey, spanning $0<z<1.3$ in eleven redshift bins. After correcting the 2,445-event sample for classification incompleteness, redshift uncertainty, dust extinction, and a survey detection efficiency that falls to 0.0006 in the top bin, the rates follow the power law $R_{\rm CC}(z)=R_0(1+z)^\alpha$ with $R_0=(1.58 \pm 0.17)\times10^{-4}\,\mathrm{yr}^{-1}\,\mathrm{Mpc}^{-3}$ and $\alpha=2.85 \pm 0.25$. The expected turnover near cosmic noon is not yet apparent, so massive-star formation was still climbing at $z=1.3$. Scaled with a Salpeter initial mass function, four recent cosmic star-formation histories sit about 70 per cent below these rates; the gap closes with a Baldry--Glazebrook IMF and 8--40 $M_\odot$ progenitors, or if the star-formation histories are missing a population of low-luminosity galaxies. The subtype analysis finds the stripped-envelope share of all core-collapse events is $31 \pm 5$ per cent and constant with redshift, which the paper uses to argue against stripped-envelope supernovae originating solely in massive single stars.

Load-bearing premise

Everything above $z\approx0.6$ is recovered by dividing observed counts by detection efficiencies that fall to 0.0006, correction factors above a thousand, under the paper's own stated assumption that distant core-collapse supernovae were no brighter than the local templates used in the simulations, with only a mild luminosity-evolution test applied.

Editorial extensions

If this is right

  • If the measured power law is right, the core-collapse supernova rate, and with it the rate of massive-star formation, rose monotonically across $0<z<1.3$, and the turnover expected near cosmic noon must lie at higher redshift.
  • The roughly 70 per cent excess over Salpeter-scaled star-formation histories implies that either current histories miss a substantial population of faint galaxies or the high-mass slope of the stellar mass distribution is shallower than Salpeter; both options can be tested against galaxy census data.
  • The redshift-invariant stripped-envelope fraction of $31\pm5$ per cent implies that whatever strips the exploding stars' envelopes did not change over nine billion years, disfavouring metallicity-driven winds of single massive stars as the sole mechanism.
  • The paper's methodological recommendation is that star-formation histories should henceforth be compared only with dust-extinction-corrected supernova rates, since mixing corrected and uncorrected measurements creates the appearance of agreement.

Reading between the lines

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

  • A natural extension the paper does not carry out is pushing the same counting method into $z\approx1.5$--2 with deeper near-infrared imaging; a flattening there would locate the cosmic-noon turnover, while continued growth would strengthen the case that the turnover is genuinely late.
  • If the 70 per cent offset is real, dust-unbiased probes of recent star formation, such as the Galactic $^{26}$Al gamma-ray line or radio searches that see through dust, should independently show higher-than-canonical supernova rates; the paper cites local evidence of this kind, and the prediction extends to other star-forming galaxies.
  • The roughly 12,000 survey light curves whose photometry failed to converge in the cosmology pipeline are noted by the authors as an under-represented population; classifying those transients would test whether the sample is missing a long-lived or unusual supernova population that could shift the intermediate-redshift rates.
Original abstract

Over its five-year campaign (2013-2017), the Dark Energy Survey (DES) supernova (SN) survey recorded 31,636 light curves. Using a custom-built, two-stage random forest classifier, we identified 2,445 likely core-collapse SNe (CC SNe) within the survey rate analysis footprint. With this sample, we measure the volumetric CC SN rate from redshift z = 0 to 1.3 in 11 redshift bins, representing the most detailed single-survey coverage to date across nine billion years of lookback time. Over this interval, the CC SN rates are well fit by a power law R_CC(z) = R_0 (1 + z)^\alpha, with R_0 = 1.58 \pm 0.17 \times 10^{-4} \text{ yr}^{-1} \text{ Mpc}^{-3} and \alpha = 2.85 \pm 0.25. The rate turnover, expected at cosmic noon, is not yet apparent. Because CC SNe trace recent star formation, we compare our measurements with recent cosmic star-formation history (CSFH) models. Our measurements are ~70 per cent higher than CSFH models using a Salpeter initial mass function (IMF) to predict CC SN rates. Replacing the Salpeter IMF with a Baldry and Glazebrook IMF and a CC SN progenitor mass range of 8-40 M_\odot bridges the gap. Alternatively, CSFHs may miss a hitherto unidentified population of low-luminosity galaxies. We further find that the local ratio of stripped-envelope (SE) SNe to all CC SNe is 31 \pm 5 per cent, remaining constant from z = 0 to 1.3. Combined with an IMF analysis accounting for the "missing progenitor problem" of hydrogen-rich SNe, our results favour a mixture of single massive stars and binaries as SE SN progenitors, with the redshift-invariant ratio favouring binary channels.

The compiled record

The paper's own sentences, arranged, kept beside the review and separate from it. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.

Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-10-21: 1 request, 8,607 input tokens.

What the abstract itself states, as probabilities: claims first 0.34 · states limitation 0.77 · human participants 0.02 · controlled comparison 0.35 · randomized assignment 0.02 · states quantitative result 0.99 · compares with prior results 0.96 · mentions code or data availability 0.03.

The abstract, sentence by sentence

arXiv listing abstract v1. Results and the main claim are underlined; .

Over its five-year campaign (2013-2017), the Dark Energy Survey (DES) supernova (SN) survey recorded 31,636 light curves. Using a custom-built, two-stage random forest classifier, we identified 2,445 likely core-collapse SNe (CC SNe) within the survey rate analysis footprint. With this sample, we measure the volumetric CC SN rate from redshift z = 0 to 1.3 in 11 redshift bins, representing the most detailed single-survey coverage to date across nine billion years of lookback time. Over this interval, the CC SN rates are well fit by a power law R_CC(z) = R_0 (1 + z)^\alpha, with R_0 = 1.58 \pm 0.17 \times 10^{-4} \text{ yr}^{-1} \text{ Mpc}^{-3} and \alpha = 2.85 \pm 0.25. The rate turnover, expected at cosmic noon, is not yet apparent. Because CC SNe trace recent star formation, we compare our measurements with recent cosmic star-formation history (CSFH) models. Our measurements are ~70 per cent higher than CSFH models using a Salpeter initial mass function (IMF) to predict CC SN rates. Replacing the Salpeter IMF with a Baldry and Glazebrook IMF and a CC SN progenitor mass range of 8-40 M_\odot bridges the gap. Alternatively, CSFHs may miss a hitherto unidentified population of low-luminosity galaxies. We further find that the local ratio of stripped-envelope (SE) SNe to all CC SNe is 31 \pm 5 per cent, remaining constant from z = 0 to 1.3. Combined with an IMF analysis accounting for the "missing progenitor problem" of hydrogen-rich SNe, our results favour a mixture of single massive stars and binaries as SE SN progenitors, with the redshift-invariant ratio favouring binary channels.

background contribution method result interpretation main claim

What the classifier reports for each sentence. The number is the probability it gave the label.
#sentencerolethe authors' own workstates the main conclusionstrength
s0 Over its five-year campaign (2013-2017), the Dark Energy Survey (DES) supernova (SN) survey recorded 31,636 light curves. method 0.82 0.88 0.02 asserted 1.00
s1 Using a custom-built, two-stage random forest classifier, we identified 2,445 likely core-collapse SNe (CC SNe) within the survey rate analysis footprint. method 0.86 0.97 0.05 asserted 0.94
s2 With this sample, we measure the volumetric CC SN rate from redshift z = 0 to 1.3 in 11 redshift bins, representing the most detailed single-survey coverage to date across nine billion years of lookback time. contribution 0.74 0.96 0.15 asserted 1.00
s3 Over this interval, the CC SN rates are well fit by a power law R_CC(z) = R_0 (1 + z)^\alpha, with R_0 = 1.58 \pm 0.17 \times 10^{-4} \text{ yr}^{-1} \text{ Mpc}^{-3} and \alpha = 2.85 \pm 0.25. result 1.00 0.96 0.34 asserted 1.00
s4 The rate turnover, expected at cosmic noon, is not yet apparent. result 0.93 0.81 0.06 asserted 0.99
s5 Because CC SNe trace recent star formation, we compare our measurements with recent cosmic star-formation history (CSFH) models. method 0.41 0.67 0.03 asserted 0.92
s6 Our measurements are ~70 per cent higher than CSFH models using a Salpeter initial mass function (IMF) to predict CC SN rates. result 0.97 0.87 0.18 asserted 1.00
s7 Replacing the Salpeter IMF with a Baldry and Glazebrook IMF and a CC SN progenitor mass range of 8-40 M_\odot bridges the gap. interpretation 0.82 0.63 0.14 asserted 0.89
s8 Alternatively, CSFHs may miss a hitherto unidentified population of low-luminosity galaxies. interpretation 0.99 0.11 0.05 hedged possibility 1.00
s9 We further find that the local ratio of stripped-envelope (SE) SNe to all CC SNe is 31 \pm 5 per cent, remaining constant from z = 0 to 1.3. result 1.00 0.97 0.09 asserted 1.00
s10 Combined with an IMF analysis accounting for the "missing progenitor problem" of hydrogen-rich SNe, our results favour a mixture of single massive stars and binaries as SE SN progenitors, with the redshift-invariant ratio favouring binary channels. interpretation 0.99 0.89 0.38 hedged possibility 0.67

What kind of paper the classifier reports this is: empirical study 1.00 · evidence from analysis of existing data 0.51.

What was examined, and what was not

sources read
  • arXiv listing abstract v1
not examined
  • full text
  • figures and tables
  • references and citing sentences
  • relations to other papers
levels run
  • P abstract
not run
  • S sections
  • T body sentences
  • C citations
  • R cross-paper relations

Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. Every number in the measurements table is the authors', checked by code. Record as JSON

Figures
Figure 1. Timeline and mind-map schematic of the analysis pipeline. The analysis proceeds from left to right along the main timeline, with each stage branching vertically. Both upper and lower branches follow a top-to-bottom reading order. The branches summarize data selection, classification and redshift estimation, efficiency and extinction corrections, uncertainty analysis, rate inputs, and final science outputs. High-precision photometry was obtained with the Scene Mod￾elling Photometry (SMP) pipeline (Brout et al. 2019; Sanchez et al. ´ 2024), applied to all 31,636 DIFFIMG candidates.…
Figure 2. Feature distributions for the SN–AGN classifier in the 𝑧 = 0.7–0.8 to 1.0–1.1 transient classification interval. SN distributions are shown in green and AGN distributions in orange. AGN (Pred) SN (Pred) Predicted Label AGN (True) SN (True) True Label 197 2 0 199 0.0 0.1 0.2 0.3 Feature importance Bands showing a fall then a rise Sig. 1/2 season slopes (all seasons) Mean flux contrast between seasons Scatter in peak timing across 4 bands Largest slope (flux change rate - event) Colour change from peak to +40 days Variance (non-event seasons) Presence of a clear peak flux Secondary…
Figure 3. Confusion matrix (left) and Random Forest feature importance rankings (right) for the SN–AGN classifier across the 𝑧 = 0.7–0.8 to 1.0–1.1 transient classification interval. MNRAS 000, 1–38 (2026)
Figure 4. Second stage random forest classifier: feature distributions for SNe classified as SNe Ia and CC SNe for the 𝑧 = 0.7–0.8 to 1.0–1.1 transient classification interval. SNe Ia distributions are shown in blue, CC SNe distributions in green. CC SNe (Pred) SNe Ia (Pred) Predicted Label CC SNe (True) SNe Ia (True) True Label 68 6 5 69 0.00 0.05 0.10 0.15 0.20 0.25 0.30 Importance Bands showing a fall then rise Presence of clear peak flux Colour change peak to +40 days (event) Largest slope (flux change rate - event) Light curve width at half max (event) Peak abs. magnitude (event season)
Figure 5. Confusion matrix (left) and feature importance (right) for the CC SN – SN Ia random forest classifier in the 𝑧 = 0.8–0.9 bin. MNRAS 000, 1–38 (2026)
Figure 6. Multi-band (𝑔, 𝑟, 𝑖, 𝑧) DIFFIMG light curves for a representative transient classified as an H-rich CC SN (a) and an SE SN (b). Data points are shown in wavelength-space flux density units (10−20 erg cm−2 s −1˚A −1 ), scaled directly from the raw pipeline frequency flux measurements (1 𝜇Jy ≈ 3.39 × 10−20 erg cm−2 s −1˚A −1 at the DECam pivot wavelength). Observed data points are overlaid with automated GP regression models (dashed black curves, shaded 1𝜎 envelopes) and filter peak epochs (vertical dashed lines). MNRAS 000, 1–38 (2026)
Figure 7. The fractional breakdown of the CC SN sample as a function of redshift (𝑧), divided into bins of Δ𝑧 = 0.1. The lines track the percentage share of spectroscopic redshifts (spec-𝑧; orange squares), host-galaxy photometric redshifts (photo-𝑧; blue circles), and template-marginalized PSNID redshifts (PSNID-𝑧; purple triangles). Total sample sizes (𝑛) for each subset across the entire redshift range are listed in the legend. simulations account for these issues as robustly as possible given the observational constraints of the data set. For the purposes of this analysis, spec-𝑧…
Figure 8. Contour density maps comparing spectroscopic redshifts (spec-𝑧) with alternative redshift estimates for CC SNe. The upper panel shows photo-𝑧 versus spec-𝑧, while the lower panel shows PSNID-𝑧 versus spec-𝑧. Contours represent the relative density of events, and green markers indicate the mean and 1𝜎 uncertainty range within each redshift bin. Minimum error floors have been applied to the lowest (𝑧 ≤ 0.1) and highest (𝑧 ≥ 0.8) bins to mitigate the impact of sparse sampling in the paired spec-𝑧 reference sample, preventing unphysically narrow or non-monotonic uncertainty bounds…
Figure 9. Observer-frame control time 𝑇obs, derived from the simulated ex￾plosion period (which mirrors the actual overall DES-SN survey period, including non-observing periods/seasons). 5.2 Detection efficiency and classification correction factor The detection efficiency, 𝜖detect, quantifies the fraction of intrinsic SNe successfully detected by the survey. We estimate this efficiency by conducting a simulation that follows the methodology described in Vincenzi et al. (2021, Section 4), as implemented to evaluate CC SN contamination for the 5-yr DES cosmology analysis (DES Collaboration…
Figure 10. Detection efficiency of DES CC SNe as a function of redshift 𝑧, showing the fraction of simulated CC SNe detected as a fraction of all intrinsic simulated explosions. assumptions, and (ii) the method directly measures the fraction of the comoving volume in which SNe are detectable given the survey’s cadence, depth and selection criteria. The primary limitation is that the efficiency is averaged over all magnitudes within each redshift bin. However, this effect is reduced by adopting narrow bins (Δ𝑧 = 0.1). Following detection, the classification algorithm is run on the combined…
Figure 11. Physical forward model fit to the binned CC SN selection correc￾tion factors (log10 (1/𝜖¯)) as a function of redshift 𝑧. Navy square markers represent empirical correction factors derived from snana simulations, with inner and outer error bars depicting 1𝜎 and 2𝜎 Clopper–Pearson uncertain￾ties, respectively. The solid red curve displays the best-fit continuous forward model (𝜖0 = 0.5, 𝑚𝑐 = 23.96 mag) optimized via 𝜒 2 minimization. the highest-redshift regime is anchored by a combined empirical bin spanning 𝑧 = 1.0–1.3. This high-𝑧 bin captures 50 valid detec￾tions out of…
Figure 12. Comparison of DES CC SN rates with previous observations (hor￾izontal bars denote the redshift interval). The 𝑥-axis is scaled by log(1 + 𝑧), providing a linear mapping to the logarithmic cosmic scale factor, log(𝑎 −1 ). There are two main reasons for the use of this scale: (i) it linearizes the power-law evolution of the rate, making the trend in our data easier to inter￾pret as a straight line, and (ii) it stretches the low-redshift regime to prevent data crowding, while allowing the x-axis to extend to 𝑧 = 5.2, to show the high-redshift results from Strolger et al. (2015) and…
Figure 13. Comparison of CC SN volumetric rates (𝑅CC) as a function of redshift (𝑧). The blue circles show the original deterministic rates derived via standard quadrature error propagation. The orange squares present the me￾dian rates from the hybrid MC simulation, which self-consistently convolves Poisson counting statistics with asymmetric systematic uncertainty profiles (classification, efficiency, and dust extinction). For visual clarity, markers are horizontally offset by Δ𝑧 = ±0.006. 1.0 < 𝑧 < 1.3, which encompasses the full DES redshift range. We tested alternative binning…
Figure 14. DES volumetric rates of SE and H-rich CC SNe, together with the SE fraction as a proportion of all CC SN rates, in six redshift bins spanning 𝑧 = 0 to 1.3. The widths of the horizontal bars indicate the bin sizes. et al. (2020) and Schulze et al. (2021) using PTF data. However, our result is in excellent agreement with two independent sets of inter￾mediate literature values: the 30–35 per cent fractions reported by Smartt (2009) and Perley et al. (2020) and the 30–34 per cent frac￾tions obtained by Graur et al. (2017b) and Shivvers et al. (2017) from re-analyses of the LOSS sample.…
Figure 15. Comparison of DES CC SN rates with previous observations and CSFH models using each model’s IMF assumptions and assuming a progenitor range of 8–40 𝑀⊙ and an IMF stellar mass range of 0.1–100 𝑀⊙ (horizontal bars denote the redshift interval). The 𝑥-axis is scaled by log(1 + 𝑧), providing a linear mapping to the logarithmic cosmic scale factor, log(𝑎 −1 ). There are two main reasons for the use of this scale: (i) it linearizes the power-law evolution of the rate, making the trend in our data easier to interpret as a straight line, and (ii) it stretches the low-redshift regime to…
Figure 16. Subplots show different SE SN progenitor scenarios: (a) single, massive stars; (b) binary systems; (c) a combination of single stars and binaries. The area of the IMF that contain H-rich SN progenitors is coloured light blue; single-star SE SN progenitors in light red; and binary SE SN progenitors in light pink. Under this framework, the inferred progenitor mass range for H-rich SNe spans 8–25.2 M⊙ (Figure 16c). This composite sce￾nario provides a flexible interpretation that reconciles the observed subtype fractions with expected stellar population demographics. By accounting for…

Figures reproduced from arXiv:2610.27328 by the authors.

References

Discussion

Disagree with the review, or know something it missed? Leave a comment, highlight any passage above to comment on it, or challenge this review formally.