ZTF SN Ia DR2: Homogeneous Low-z Supernovae

• The paper presents the largest homogeneous SN Ia dataset to date, featuring 2,629 cosmology-ready events from a well-defined low-z sample.
• It provides forced-photometry in gri bands and 5,138 spectra alongside host galaxy properties, enabling robust re-analysis and light-curve modeling.
• The release underpins studies on standardisation, environmental dependencies, and calibration challenges critical for low-redshift cosmology.

to=arxiv_search 大发时时彩开奖ian 官网群 天天爱彩票提现json {"query":"ZTF SN Ia DR2", "max_results": 10} to=arxiv_search 大发快三是什么json {"query":"ZTF SN Ia DR2", "max_results": 10} ZTF SN Ia DR2 is the second data release of Type Ia supernovae from the Zwicky Transient Facility cosmology science working group. It is the first homogeneous release of several thousand spectroscopically confirmed SNe Ia discovered, followed and classified by ZTF between March 2018 and December 2020, and it contains 3628 nearby objects with $z<0.3$, 3000 with good-to-excellent sampling, and 2667 that pass standard cosmology light-curve quality cuts. Alongside the supernova parameters, the release publishes forced-photometry $gri$ light curves, 5138 spectra, local and global host properties, observing logs, and a Python client. In scope and uniformity, it is presented as the largest SN Ia release to date and as a low-redshift reference sample for studies of standardisation, host dependence, population diversity, and light-curve modeling (Rigault et al., 2024).

1. Release scope and sample definition

The released sample spans roughly $0.001\lesssim z<0.30$, with median $z_{\rm med}\approx0.07$, and includes a volume-limited subset at $z<0.06$ containing nearly a thousand “normal” SNe Ia. In the DR2 processing, SALT2 fits are forced in the rest-frame phase window $\phi\in[-10,+40]$ days. The sample is then reduced by a “good sampling” cut requiring at least 7 distinct phases, at least 2 points before and after peak, and at least 2 filters, yielding $N_{\rm good}=2960\,(82\%)$. Standard SALT2-parameter cuts, $-3\le x_1\le+3$, $-0.2\le c\le+0.8$, $\sigma_{x_1}<1$, $gri$0, and fitprob $gri$1, give $gri$2 cosmology-ready events; after removing recognized peculiar subtypes, about 2629 remain for standard analyses (Rigault et al., 2024).

A parallel simulation program using the skysurvey framework was developed to reproduce the survey selection function and the evolution of SALT2 parameters with redshift. In that analysis, realistic simulations of ZTF cadence, limiting magnitudes, and DR2 selection criteria showed that the post-selection $gri$3 and $gri$4 distributions remain unbiased only below $gri$5, motivating the use of a volume-limited DR2 subset of about 1000 SNe Ia for precision studies of standardisation and environment (Amenouche et al., 2024).

2. Released observables and technical infrastructure

DR2 is not only a catalog of fitted light-curve parameters. It also releases the primary observational products needed for re-analysis. Each supernova has forced-photometry light curves in ZTF $gri$6, with files containing mjd, filter, flux, fluxerr, flag, field_id, and quadrant. Spectra are distributed in ASCII or FITS formats with wavelength, flux, and error arrays, together with header redshifts and classifications. The metadata table includes positions, heliocentric redshifts, SALT2 parameters $gri$7 and covariances, Milky-Way $gri$8, and classification flags. Host-galaxy products include global and local (2 kpc aperture) stellar masses, rest-frame $gri$9 colors, and SN-host directional-light-radius distances. Observing logs are also provided for every P48 exposure in DR2 fields, and the Python toolkit ztfcosmo supports downloading, querying, light-curve and spectrum reading, and SALT2 fitting (Rigault et al., 2024).

This release structure is important because many DR2 analyses depend on re-fitting or re-selecting the sample under different assumptions. The spectral studies, for example, make use of the public rest-frame, Milky-Way-de-reddened spectra, and the photometric analyses exploit the homogeneous forced-photometry products and host-property tables. The design is therefore explicitly oriented toward reproducible, survey-level statistical work rather than a single fiducial cosmology fit (Burgaz et al., 2024).

3. Light-curve modeling and the standardisation problem

DR2 is centered on SALT-family empirical modeling. In the standard SALT2 formalism, the rest-frame flux is written as

$0.001\lesssim z<0.30$0

and the standardized distance modulus is typically expressed through the Tripp relation,

$0.001\lesssim z<0.30$1

Within DR2, Rigault et al. found that SALT2 models the ZTF SN Ia optical light curves remarkably well, even though ZTF was not used in training, except for points earlier than $0.001\lesssim z<0.30$2 days from maximum, where the training critically lacks statistics. They further found that the fit is robust against the considered choice of phase range, while the interval $0.001\lesssim z<0.30$3 days is optimal in terms of statistics and accuracy. No significant features were detected in the light-curve fit residuals that could be connected to the host environment, but a small yet significant inconsistency between residuals of blue- and red-SN Ia suggests a phase-dependent color term and motivates SALT2 retraining for precision cosmology (Rigault et al., 2024).

Several DR2 companion studies then used the volume-limited subset to probe the structure of standardisation itself. The stretch distribution is clearly bimodal, with fitted modes at $0.001\lesssim z<0.30$4 and $0.001\lesssim z<0.30$5, and the means of the modes decrease with host stellar mass at a $0.001\lesssim z<0.30$6 significance. The stretch-magnitude relation is non-linear at the $0.001\lesssim z<0.30$7 level. In the broken-$0.001\lesssim z<0.30$8 model, the fitted break is $0.001\lesssim z<0.30$9, with $z_{\rm med}\approx0.07$0 for $z_{\rm med}\approx0.07$1 and $z_{\rm med}\approx0.07$2 for $z_{\rm med}\approx0.07$3. Environmental magnitude offsets remain larger than 0.12 mag for local or global color and mass tracers, and increase to about $z_{\rm med}\approx0.07$4–$z_{\rm med}\approx0.07$5 mag when broken-$z_{\rm med}\approx0.07$6 standardisation is used (Ginolin et al., 2024).

Color standardisation was examined separately with a nearly 1000-object, low-$z_{\rm med}\approx0.07$7 volume-limited sample. The color-residual relation is linear with SN color, the “dustless” subsample has a significantly shorter red tail than the full sample, and the reddening above $z_{\rm med}\approx0.07$8 is inferred to be dominated by host interstellar dust. At the same time, no evolution of the environmental steps with SN color was found, and there are indications that $z_{\rm med}\approx0.07$9 may evolve with stellar host mass, with $z<0.06$0 for low-mass galaxies compared to $z<0.06$1 for the full sample (Ginolin et al., 2024).

DR2 also enabled standardisation tests that do not rely on the Hubble diagram itself. Using 25 sibling pairs in common hosts, a cosmology-independent analysis found $z<0.06$2 and $z<0.06$3 in a single-global-relation fit, with $z<0.06$4 mag at 95% C.L.; allowing a break in stretch gave $z<0.06$5 and $z<0.06$6, suggesting different width-luminosity slopes for slow and fast decliners (Dhawan et al., 2024).

Follow-up model development used DR2 directly in the training set. The SALT3+ model adds a new parameter $z<0.06$7 and finds coherent light-curve variability beyond SALT3, largely as phase-dependent variation in $z<0.06$8 and $z<0.06$9 color curves and in the height of the $\phi\in[-10,+40]$0-band secondary maximum. Neglecting $\phi\in[-10,+40]$1 produces a Hubble-residual trend of $\phi\in[-10,+40]$2 mag with $\phi\in[-10,+40]$3, although no evidence was found for a bias in current cosmological measurements (Kenworthy et al., 13 Feb 2025). A separate Gaussian-process study of 893 DR2 SNe found $\phi\in[-10,+40]$4 correlations between $\phi\in[-10,+40]$5 and the timing and strength of the secondary maximum in the $\phi\in[-10,+40]$6 band, and concluded that the strength of the secondary maximum is a better standardisation parameter than SALT stretch $\phi\in[-10,+40]$7 (Deckers et al., 2024).

4. Host environment, large-scale structure, and dust

One of the central uses of DR2 is the isolation of environmental effects from pure light-curve phenomenology. In the volume-limited sample, host-dependent magnitude offsets are detected with local or global stellar mass and rest-frame $\phi\in[-10,+40]$8 color. In parallel, image-decomposition work on 728 DR2 hosts found that stretch correlates linearly with model-derived galaxy color in both ellipticals and disk galaxies, with significances of $\phi\in[-10,+40]$9 and $N_{\rm good}=2960\,(82\%)$0, respectively. For SNe Ia in disk-containing galaxies, a further $N_{\rm good}=2960\,(82\%)$1 linear trend was found between stretch and model-derived local $N_{\rm good}=2960\,(82\%)$2-band surface brightness. SN color, by contrast, shows little correlation with host environment, apart from a possible dust effect in surface brightness at $N_{\rm good}=2960\,(82\%)$3 for SNe Ia in disk galaxies (Senzel et al., 2024).

Environmental trends are not confined to host-integrated properties. In the vicinity of galaxy clusters, the stretch distribution is better fit by a model that includes cluster-centric distance, with $N_{\rm good}=2960\,(82\%)$4 relative to a distance-independent model. Inside $N_{\rm good}=2960\,(82\%)$5, the mean stretch shifts by $N_{\rm good}=2960\,(82\%)$6, while the color shows a smaller shift of $N_{\rm good}=2960\,(82\%)$7. This was interpreted as further support for stellar-population age as the driver of the bimodal stretch distribution (Ruppin et al., 2024).

At larger scales, DR2 was cross-matched to SDSS void catalogs and Voronoi-volume density estimates. In that analysis, the authors found no statistically significant trend of either $N_{\rm good}=2960\,(82\%)$8 or $N_{\rm good}=2960\,(82\%)$9 with void-centric distance $-3\le x_1\le+3$0, but they did find that local Voronoi volumes affect mostly the fraction of low- and high-stretch SNe. The study concluded that, with current statistics, under-dense environments should not cause biases in supernova analyses and can largely be treated as a proxy for host-galaxy color (Aubert et al., 2024).

A distinct, combined-survey analysis anchored by DR2 addressed dust directly. Using ZTF together with SDSS, PS1, and DES to form a volume-limited sample over $-3\le x_1\le+3$1, the fitted dust column-density parameter $-3\le x_1\le+3$2 was found to increase steadily with redshift at $-3\le x_1\le+3$3 significance in the fiducial sample and to vary strongly with host stellar mass, peaking near $-3\le x_1\le+3$4. The same work found a strong correlation between host mass and the color-luminosity coefficient $-3\le x_1\le+3$5, at $-3\le x_1\le+3$6, even when accounting for the quantity of dust in a host galaxy (Popovic et al., 2024). This suggests that DR2 host-dependence results cannot be reduced to a single dust-correction narrative.

5. Spectroscopic diversity and subtype taxonomy

The release is unusually rich spectroscopically. DR2 contains 5138 spectra, and dedicated classification work used a custom SNID template library of 370 templates, wrapped by pysnid, together with an in-house typingapp that collected more than 14,000 human classifications from 32 users. Only 0.8% of spectra failed to converge. After arbiter logic and vetting, 34 objects were rejected as non-Ia, and the full DR2 sample was partitioned into normal, 91T-like, 91bg-like, 03fg-like, Iax, Ia-CSM, 02es-like, 18byg-like, plus an un-subtyped remainder. The corresponding full-sample counts are 2511 normal, 292 91T-like, 92 91bg-like, 29 03fg-like, 23 Iax, 14 Ia-CSM, 8 02es-like, 4 18byg-like, and 655 not sub-typed. In the volume-limited regime, efficiency-corrected subclass fractions are 75.4% normal, 12.2% 91T-like, 6.1% 91bg-like, 4.5% Iax, 0.8% 03fg-like, 0.3% Ia-CSM, 0.5% 02es-like, and 0.2% 18byg-like (Dimitriadis et al., 2024).

A separate near-maximum spectroscopic analysis of 482 volume-limited DR2 SNe measured Si II $-3\le x_1\le+3$7 and $-3\le x_1\le+3$8 velocities and pseudo-equivalent widths, quantified host contamination, and introduced a new “04gs-like” subclass. Simulations showed that host contamination can depress pEW by 20–50% for at least 50% host-flux contamination, while velocities shift by at most 500 km s$-3\le x_1\le+3$9. After applying a $-0.2\le c\le+0.8$0 cut, the clean sample contained 18 “04gs-like” events, about 5% of the sample, spectroscopically bridging normal and 86G-like SNe Ia. TARDIS modeling showed that cooler temperatures alone are unable to explain these spectra and that changes in elemental abundances are also required (Burgaz et al., 2024).

Early-time spectra provide another axis of diversity. In a DR2 study of pre-maximum Si II $-0.2\le c\le+0.8$1, about 250 spectra out of a high-quality sample of about 750 showed robust high-velocity components. After correcting for detection efficiency, $-0.2\le c\le+0.8$2 of spectra before $-0.2\le c\le+0.8$3 days show HVFs, compared to $-0.2\le c\le+0.8$4 between $-0.2\le c\le+0.8$5 and $-0.2\le c\le+0.8$6 days and $-0.2\le c\le+0.8$7 between $-0.2\le c\le+0.8$8 and $-0.2\le c\le+0.8$9 days. No significant difference was found between HVF and non-HVF subsamples in SALT2 $\sigma_{x_1}<1$0, peak magnitude, decline rate, host-galaxy stellar mass, or local color, supporting the idea that Si II HVFs are ubiquitous across the SN Ia population (Harvey et al., 6 Feb 2025).

6. Cosmological use, calibration limits, and later follow-up

DR2 demonstrates that a multi-thousand-object low-$\sigma_{x_1}<1$1 Hubble diagram can be built from a single survey. The reported natural Hubble-diagram scatter is about 0.33 mag, the standardized scatter for the full sample is about 0.165 mag, and a plateau scatter of about 0.150 mag is obtained in the interval $\sigma_{x_1}<1$2, where peculiar-velocity and SN-redshift errors are minimized (Rigault et al., 2024).

A common misconception is that sample size alone makes DR2 immediately suitable for cosmological parameter inference. The release documentation states the opposite: the photometric accuracy of DR2 is not yet suited for cosmological parameter inference, and a later DR2.5 release was planned to address this. The principal limitations are that the forced photometry is calibrated to about 1% relative but about 5% absolute, and that a read-out nonlinearity, the “pocket effect,” since late 2019 induces a percent-level bias in $\sigma_{x_1}<1$3 and $\sigma_{x_1}<1$4 (Rigault et al., 2024).

That limitation was sharpened by scene-modeling photometry work. Scene-modeling photometry reached repeatability on stars better than 1%, but identified the “pocket effect” as a flux-dependent PSF distortion that introduces photometric and astrometric biases of order 1–7%, well above the $\sigma_{x_1}<1$5 target for dark-energy work. Comparing scene-modeling photometry to the released forced photometry, the fitted stretch and color are consistent aside from a 10 mmag shift in color, whereas the absolute calibration shifts by 90 mmag. The paper therefore concluded that DR2 cannot yet be used for accurate cosmological inference, even though many population-level DR2 results are robust against the current processing differences (LaCroix et al., 4 Sep 2025).

Low-redshift cosmology in DR2 is also limited by peculiar velocities. Using realistic ZTF-like simulations and then applying the method to a 904-object DR2 sample in $\sigma_{x_1}<1$6, Carreres et al. showed that one must include the full peculiar-velocity covariance matrix derived from the velocity power spectrum. Neglecting peculiar velocities and their correlations shifts the intercept $\sigma_{x_1}<1$7 and corresponds to a shift of the inferred $\sigma_{x_1}<1$8 of about $\sigma_{x_1}<1$9 km s$gri$00 Mpc$gri$01, while slightly underestimating the error bar (Carreres et al., 2024).

Later follow-up within the DR2 program extended these issues to subtype-dependent systematics. One study of 277 near-peak spectra found that the canonical host-mass step arises almost entirely from normal-velocity SNe Ia, with $gri$02 mag and $gri$03 mag, implying that a universal mass-step correction could mis-correct high-velocity events (Burgaz et al., 2 Sep 2025). Taken together, these analyses position ZTF SN Ia DR2 as both a release and an experimental platform: it already supports precise work on SN Ia population structure and standardisation, while simultaneously documenting the calibration, modeling, and environmental effects that must be controlled before a fully cosmology-grade low-redshift anchor can be claimed.