We present the discovery of SN 2025mkn, a gravitationally lensed Type II supernova. First detected as a blue transient in Zwicky Transient Facility (ZTF), 0."83 from a z = 0.42 elliptical galaxy, the follow-up SNIFS/UH2.2 m and LRIS/Keck spectra revealed absorption lines at z = 1.371. Later JWST NIRCam imaging shows that the bright transient is a close pair of point sources separated by ~ 0."07, and a 30 times fainter counterimage opposite the lens, for which NIRSpec reveals strong Hα emission also at z = 1.371. The lightcurves and spectra are consistent with the Type II supernova source being magnified ≳100 times, with ∼250 required to reconcile its luminosity with that of nearby events such as SN 2023ixf. Lens models are consistent with such high magnifications, and always show that the faint image arrived first (undetected in earlier ZTF imaging), consistent with the later spectral phase of this fainter image. A fourth image is also predicted and possibly detected in the NIRSpec data. Lightcurve-based time-delay measurements are not possible due to the first image being the faintest; however, the resolved NIRSpec spectra offer a future opportunity for time-delay cosmography through supernova phase measurements.

We present BayeSN-TD, an enhanced implementation of the probabilistic Type Ia supernova (SN Ia) BayeSN spectral energy distribution (SED) model, designed for fitting multiply-imaged gravitationally lensed Type Ia supernovae (glSNe Ia). BayeSN-TD fits for magnifications and time delays across multiple images while marginalizing over an achromatic Gaussian process-based treatment of microlensing, to allow for time-dependent deviations from a typical SN Ia SED caused by gravitational lensing by stars in the lensing system. BayeSN-TD is able to robustly infer time delays and produce well-calibrated uncertainties, even when applied to simulations based on a different SED model and incorporating chromatic microlensing, strongly validating its suitability for time-delay cosmography. We then apply BayeSN-TD to publicly available photometry of the glSN Ia SN H0pe, inferring time delays between images BA and BC of d and d along with absolute magnifications β for each image, ⁠, ⁠, and ⁠. Combining our constraints on time delays and magnifications with existing lens models of this system, we infer km s⁻¹ Mpc⁻¹⁠, consistent with previous analysis of this system; incorporating additional constraints based on spectroscopy yields  km s Mpc⁠. While this is not yet precise enough to draw a meaningful conclusion with regard to the ‘Hubble tension’, upcoming analysis of SN H0pe with more accurate photometry enabled by template images, and other glSNe, will provide stronger constraints on ⁠H₀; BayeSN-TD will be a valuable tool for these analyses.

Multiply imaged supernovae (SNe) provide a novel means of constraining the Hubble constant (H0). Such measurements require a combination of precise models of the lensing mass distribution and an accurate estimate of the relative time delays between arrival of the multiple images. Only two multiply imaged SNe, Refsdal and H0pe, have enabled measurements of H₀ thus far. Here we detail the third such measurement for SN Encore, a z = 1.95 Type Ia SN discovered in JWST/NIRCam imaging. We measure the time delay, perform simulations of additional microlensing and millilensing systematics, and combine with the mass models of Suyu et al. in a double-blind analysis to obtain our H0 constraint. Our final time-delay measurement is Delta t1b,1a=-39.8-3.3+3.9 days, which is combined with seven lens models weighted by the likelihood of the observed multiple image positions for a result of H0=66.9-8.1+11.2kms-1Mpc-1 . The uncertainty on this measurement could be improved significantly if template imaging is obtained. Remarkably, a sibling to SN Encore (SN "Requiem") was discovered in the same host galaxy, making the MACS J0138.0-2155 cluster the first system known to produce more than one observed multiply imaged SN. SN Requiem has a fourth image that is expected to appear within a few years, providing an unprecedented decade-long baseline for time-delay cosmography and an opportunity for a high-precision joint estimate of H₀.

A significant number of Lyman-break galaxies (LBGs) with redshifts are expected to be observed by the upcoming Vera C. Rubin Observatory Legacy Survey of Space and Time (LSST). This will enable us to probe the Universe at higher redshifts than is currently possible with cosmological galaxy clustering and weak lensing surveys. However, accurate inference of cosmological parameters requires precise knowledge of the redshift distributions of selected galaxies, where the number of faint objects expected from LSST alone will make spectroscopic based methods of determining these distributions extremely challenging. To overcome this difficulty, it may be possible to leverage the information in the large volume of photometric data alone to precisely infer these distributions. This could be facilitated using forward models, where in this paper we use stellar population synthesis (SPS) to estimate uncertainties on LBG redshift distributions for a 10 yr LSST (LSSTY10) survey. We characterize some of the modelling uncertainties inherent to SPS by introducing a flexible parametrization of the galaxy population prior, informed by observations of the galaxy stellar mass function (GSMF) and cosmic star formation rate density (CSFRD). These uncertainties are subsequently marginalised over and propagated to cosmological constraints in a Fisher forecast, leveraging galaxy clustering and lensing of the cosmic microwave background (CMB). Assuming a known dust attenuation model for LBGs, we forecast constraints on the σ₈ parameter comparable to Planck CMB constraints.

We study star formation over (Formula presented) 12 Gyr using pop-cosmos, a generative model trained on 26-band photometry of (Formula presented) 420 000 COSMOS2020 galaxies (Spitzer IRAC (Formula presented)). The model learns distributions over 16 stellar population synthesis parameters via score-based diffusion, matching observed colours and magnitudes. We use pop-cosmos to compute the cosmic star formation rate density (SFRD) to (Formula presented) by directly integrating individual galaxy SFRs. The SFRD peaks at (Formula presented), (Formula presented) later than previous canonical estimates, with peak value (Formula presented) (Formula presented). We classify star-forming (SF) and quiescent (Q) galaxies using specific SFR (sSFR) (Formula presented) yr(Formula presented), comparing with (Formula presented) colour selection. The sSFR criterion yields up to 20 per cent smaller Q fractions across (Formula presented), with (Formula presented) -selected samples contaminated by galaxies with sSFR up to (Formula presented) yr(Formula presented). Our sSFR-selected stellar mass function shows a negligible number density of low-mass ((Formula presented)) Q galaxies at (Formula presented), where colour-selection shows a prominent increase. Non-parametric star formation histories around the SFRD peak reveal distinct patterns: SF galaxies show gradually weakening correlations between their recent and earlier SFRs, implying increasingly stochastic star formation towards early epochs. Q galaxies exhibit full correlation ((Formula presented)) during the most recent (Formula presented) 300 Myr, then sharp decorrelation with earlier SF epochs, marking clear quenching transitions. Massive ((Formula presented)) galaxies quench on a time-scale of (Formula presented) Gyr, with mass assembly concentrated in their first 3.5 Gyr. Finally, active galactic nucleus (AGN) activity (infrared torus luminosity fraction) peaks as massive ((Formula presented)) galaxies approach the transition between SF and Q states, declining sharply once quiescence is established. This provides evidence that AGN feedback operates in a critical regime during the (Formula presented) Gyr quenching transition.

The delay in arrival time of the multiple images of gravitationally lensed supernovae (glSNe) can be related to the present-day expansion rate of the universe, ⁠H₀. Despite their rarity, Rubin Observatory’s Legacy Survey of Space and Time (Rubin-LSST) is expected to discover tens of galaxy-scale glSNe per year, many of which will not be resolved due to their compact nature. Follow-up from ground- and space-based telescopes will be necessary to estimate time delays to sufficient precision for meaningful H₀ constraints. We present the glimpse model (GausSN Light curve Inference of Magnifications and Phase Shifts, Extended) that estimates time delays with resolved and unresolved observations together for the first time, while simultaneously accounting for dust and microlensing effects. With this method, we explore best follow-up strategies for glSNe observed by Rubin-LSST. For unresolved systems on the dimmest end of detectability by Rubin-LSST, having peak i-band magnitudes of 22–24 mag, the time delays are measured to as low as 0.7 d uncertainty with 6–8 epochs of resolved space-based observations in each of 4–6 optical and NIR (near-infrared) filters. For systems of similar brightness that are resolved by ground-based facilities, time delays are consistently constrained to 0.5–0.8 d precision with six epochs in four optical and NIR filters of space-based observations or eight epochs in four optical filters of deep ground-based observations. This work improves on previous time-delay estimation methods and demonstrates that glSNe time delays of ~ 10 − 20 d can be measured to sufficient precision for competitive H₀ estimates in the Rubin-LSST era.

Type Ia supernovae (SNe Ia) are standardizable candles: their peak magnitudes can be corrected for correlations between light-curve properties and their luminosities to precisely estimate distances. Understanding SN Ia standardization across wavelength improves methods for correcting SN Ia magnitudes. Using 150 SNe Ia from the Foundation Supernova Survey and Young Supernova Experiment, we present the first study focusing on SN Ia standardization properties in the z band. Straddling the optical and near-infrared, SN Ia light in the z band is less sensitive to dust extinction and can be collected alongside the optical on CCDs. Pre-standardization, SNe Ia exhibit less residual scatter in z-band peak magnitudes than in the g and r bands. SNe Ia peak z-band magnitudes still exhibit a significant dependence on light-curve shape. Post-standardization, the z-band Hubble diagram has a total scatter of root mean square =0.195 mag. We infer a z-band mass step of  -0.105±0.031 mag, which is consistent within 1σ of that estimated from gri data, assuming R_v=2.61⁠. When assuming different R_v values for high and low mass host galaxies, the z band and optical mass steps remain consistent within 1⁠σ. Based on current statistical precision, these results suggest dust reddening cannot fully explain the mass step. SNe Ia in the z band exhibit complementary standardizability properties to the optical that can improve distance estimates. Understanding these properties is important for the upcoming Vera Rubin Observatory and Nancy G. Roman Space Telescope, which will probe the rest-frame z band to redshifts 0.1 and 1.8.

We present a simple method for assessing the predictive performance of high-dimensional models directly in data space when only samples are available. Our approach is to compare the quantiles of observables predicted by a model to those of the observables themselves. In cases where the dimensionality of the observables is large (e.g., multiband galaxy photometry), we advocate that the comparison is made after projection onto a set of principal axes to reduce the dimensionality. We demonstrate our method on a series of two-dimensional examples. We then apply it to results from a state-of-the-art generative model for galaxy photometry () that generates predictions of colors and magnitudes by forward simulating from a 16-dimensional distribution of physical parameters represented by a score-based diffusion model. We validate the predictive performance of this model directly in a space of nine broadband colors. Although motivated by this specific example, we expect that the techniques we present will be broadly useful for evaluating the performance of flexible, nonparametric population models of this kind, and other settings where two sets of samples are to be compared.

We present an extension of the pop-cosmos model for the evolving galaxy population up to redshift z ∼ 6. The model is trained on distributions of observed colors and magnitudes, from 26-band photometry of ∼420,000 galaxies in the COSMOS2020 catalog with Spitzer IRAC Channel 1 < 26 mag. The generative model includes a flexible distribution over 16 stellar population synthesis (SPS) parameters, and a depth-dependent photometric uncertainty model, both represented using score-based diffusion models. We use the trained model to predict scaling relationships for the galaxy population, such as the stellar mass function, star-forming main sequence, and gas phase and stellar metallicity versus mass relations, demonstrating reasonable to excellent agreement with previously published results. We explore the connection between mid-infrared emission from active galactic nuclei (AGN) and star formation rate, finding high AGN activity for galaxies above the star-forming main sequence at 1 ≲ z ≲ 2. Using the trained population model as a prior distribution, we perform inference of the redshifts and SPS parameters for 429,669 COSMOS2020 galaxies, including 39,588 with publicly available spectroscopic redshifts. The resulting redshift estimates exhibit minimal bias (median[Δz] = −8 × 10−4), scatter (σMAD = 0.0132), and outlier fraction (6.19%) for the full 0 < z < 6 spectroscopic compilation. These results establish that pop-cosmos can achieve the accuracy and realism needed to forward model modern wide, deep surveys for Stage IV cosmology. We publicly release pop-cosmos software, mock galaxy catalogs, and COSMOS2020 redshift and SPS parameter posteriors.

Context. Type Ia supernovae (SNe Ia) are a key probe in modern cosmology, as they can be used to measure luminosity distances at gigaparsec scales. Models of their light curves are used to project heterogeneous observed data onto a common basis for analysis. Aims. The SALT model currently used for SN Ia cosmology describes SNe as having two sources of variability, accounted for by a color parameter c, and a “stretch” parameter x1. We extend the model to include an additional parameter we label x2, to investigate the cosmological impact of currently unaddressed light-curve variability. Methods. We constructed a new SALT model, that we dub “SALT3+”. This model was trained by an improved version of the SALTshaker code, using training data combining a selection of the second data release of cosmological SNe Ia from the Zwicky Transient Facility and the existing SALT3 training compilation. Results. We find additional, coherent variability in supernova light curves beyond SALT3. Most of this variation can be described as phase-dependent variation in g − r and r − i color curves, correlated with a boost in the height of the secondary maximum in i-band. These behaviors correlate with spectral differences, particularly in line velocity. We find that fits with the existing SALT3 model tend to address this excess variation with the color parameter, leading to less informative measurements of supernova color. We find that neglecting the new parameter in light-curve fits leads to a trend in Hubble residuals with x2 of 0.039 ± 0.005 mag, representing a potential systematic uncertainty. However, we find no evidence of a bias in current cosmological measurements. Conclusions. We conclude that extended SN Ia light-curve models promise mild improvement in the accuracy of color measurements, and corresponding cosmological precision. However, models with more parameters are unlikely to substantially affect current cosmological results.