The New Status Qvo? SN 2021qvo is Another 2003fg-like Type Ia Supernova with a Rising Light-Curve Bump

Corresponding author: I. A. Abreu Paniagua

[email protected] keywords: Type Ia supernovae(1728)

Introduction

Type Ia supernovae (SNe Ia) follow a strongly correlated luminosity-width relation (LWR) that relates the light curve shape of a SN Ia to its intrinsic luminosity, allowing for distances to be calculated with high precision [121]. SNe Ia can be used to measure cosmological parameters with high precision using distances derived from this relationship, including the Hubble constant [e.g. 23, 50, 130, 49] and the dark energy equation of state parameter [e.g. 13, 135, 74, 19, 38].

While there is broad consensus that SNe Ia result from an exploding Carbon-Oxygen White Dwarf (CO WD) star [66], there is a large diversity of proposed models explaining SNe Ia, each with their own unique progenitor scenario and explosion mechanism [see 132 for a recent review]. SNe Ia progenitor channels include the single degenerate scenario containing a CO WD and a non-degenerate companion [e.g. 160, 113], the double degenerate scenario containing two degenerate CO WDs, a CO WD and a He WD, or a CO WD and a HeCO WD [e.g. 112, 69, 120], and the core degenerate scenario containing a CO WD and an asymptotic giant branch star [e.g. 79].

Each progenitor scenario has several possible explosion mechanisms that could result in a SN Ia. In the single degenerate scenario, the CO WD accretes material from the non-degenerate companion star (either a red giant star or a main sequence star) until a thermonuclear runaway explosion occurs (generally when the CO WD nears the Chandrasekhar mass; [160, 115]). Accretion-triggered explosions can also occur in the double degenerate scenario [e.g. 113], a detonation of accreting or built-up surface helium may send a converging shock into the CO WD, causing central carbon ignition (and thus a SN Ia, [e.g. 112, 113, 94, 138]). Finally, dynamical or violent mergers may also cause a SN Ia in the double degenerate scenario [e.g. 69, 159, 131, 129, 118, 85, 86, 141].

Most SNe Ia are spectroscopically and photometrically “normal” [e.g. 39] with similar photometric and spectroscopic features [e.g. 109], although at the earliest stages show significant diversity, [61]. However, a growing number of SNe Ia have been discovered with different observational properties, which may arise from differences in the underlying progenitor scenario and/or explosion mechanism. While some SN Ia subtypes follow a luminosity-width relationship (i.e., 1991bg-like and 1991T-like SNe Ia, [46, 87, 123, 46]), other subtypes do not. These include 2002cx-like (Li et al. 2003), 2002es-like [51], and 2003fg like SNe Ia [65]. Because of deviations in parameters like peak absolute magnitude and light-curve decline rate, such SNe Ia are not always standardizable for SN cosmology. Still, these peculiar objects offer an opportunity to understand the extreme ends of the SNe Ia population.

These peculiar SN Ia subtypes are identified through a mix of photometric and spectroscopic characteristics. Several 2003fg-like SNe Ia have abnormally luminous abs mags between −19.5 to −20.4 [56, 134, 42] and decline rates $\Delta m_{15}$ values in the range of 0.7 to 0.9 mag [149] (Ashall et al. 2021), however we now know there is a broad diversity in this subclass with some showing more typical luminosities and decline rates ($\sim$0-1 mag brighter [68] (Lu et al. 2021)). In general, 2003fg-like events are $\sim$1–2 mag brighter than normal SNe Ia [39] and have light curve shapes consistent with or broader than the slowest-declining normal SNe Ia. With these properties in mind, common photometric diagnostic features are a primary $i$-band maximum after the primary $B$-band maximum, weak or missing $i$-band secondary maximum, and a broad light curve (Ashall et al. 2021, although the latter two are seen in only some 2003fg-like SNe Ia).

Spectroscopically, 2003fg-like SNe Ia also exhibit different features than normal SNe Ia. The early-time spectra of 2003fg-like SNe Ia have weaker Ca II features and prominent O I and C II features [149] (Ashall et al. 2021) and have higher UV flux than normal SNe Ia [59, 14]. In general, the C II $\lambda\lambda6580,7234$ features are especially strong for up to $\sim$2 weeks after maximum light. At later times, 2–4 weeks after maximum light, 2003fg-like SN Ia spectra are dominated by Fe II slightly earlier than for normal SNe Ia [142, 150, 28, 119]. Other spectroscopic diagnostics to identify a SNe as 2003fg-like include a lack of an $H$-band break at +10 days in the near-infrared spectra and a low ionization state in nebular-phase spectra. Features common, but not ubiquitous in the 2003fg-like population include strong C ii features after maximum light, low ejecta velocity gradients before maximum light, a lack of Ti ii lines in the peak time spectra, and asymmetric near-infrared nebular lines [e.g. 86, 141, 116].

Recently, three 2003fg-like SNe Ia with sufficiently early detections have been discovered with bump-like features in their early-time emission:² SNe 2020hvf [72], 2021zny [43], and 2022ilv (Srivastav et al. 2023). These three SNe Ia are the only three known 2003fg-like SNe Ia with sufficient observations to ascertain the presence of a rising light curve bump, suggesting these bump features may be common or even ubiquitous in 2003fg-like SNe Ia. Furthermore, these bumps have solely been observed in 2003fg-like SNe Ia and in another SN Ia subtype, 2002es-like SNe Ia, potentially suggesting these objects are linked [59].

Many models attempt to explain the origin of rising light curve bumps in SNe Ia. While the interaction between SN ejecta and a non-degenerate companion in the single degenerate scenario [e.g. 78, 98, 84] has been ruled out for these objects (Dimitriadis et al. 2023; Srivastav et al. 2023; Hoogendam et al. 2024), varying radial distributions of ${}^{56}\mathrm{Ni}$ [e.g. 126, 99], the interaction between SN ejecta and circumstellar material (CSM; e.g., Levanon et al. 2015; Piro & Morozova 2016; Levanon & Soker 2017; Maeda et al. 2023), or a surface He detonation [e.g. 71, 96, 127, 90, 89] may produce an observable “bump” signature. The early colors may be too blue for significant surface ${}^{56}\mathrm{Ni}$ from either mixing or a surface He-detonation, leaving CSM interaction as the preferred explanation [59]. This is consistent with the spectral and light curve evolution, which also suggests a CSM-interacting explosion (Ashall et al. 2021).

In this work, we examine the photometric and spectroscopic properties of SN 2021qvo, a 2003fg-like SN Ia that adds to the class of such objects that exhibit early light-curve bumps. Section 2 presents our observations. Sections 4 and 5 contain our photometric and spectroscopic observations, respectively. In Section 6, we constrain the CSM mass and compare to similar events in the literature. Finally, Section 7 summarizes the main results. Throughout, we use a flat $\Lambda$CDM cosmology with $\Omega_m = 0.3$ and $H_0 = 70\ \mathrm{km\ s^{-1}\ Mpc^{-1}}$.

Data

SN 2021qvo was discovered in a small, faint galaxy (see Section 3) by the Asteroid Terrestrial-impact Last Alert System [154] on June 21, 2021 with an $o$-band magnitude of $\sim$19.5 mag [153]. The location of SN 2021qvo is (\alpha,\delta) = (22^\mathrm{h}08^\mathrm{m}12\fs08,+07^\circ07^\prime00\farcs78).

The Young Supernova Experiment [75] classified SN 2021qvo two days after its discovery on June 23, 2021 [4] with a classification spectrum from the Alhambra Faint Object Spectrograph and Camera on the Nordic Optical Telescope. They used SuperFit [64] and SNID [16] to classify the event as a SN Ia, finding that the transient matched the template of a SN Ia around $-10$ days.

Photometric Data

Photometric observations of SN 2021qvo primarily include the ATLAS “cyan” and “orange” filters (approximately equivalent to $g+r$ and $r+i$ bandpasses, respectively) and the Pan-STARRS $griz$ filters, spanning from MJD $\sim$59346 to $\sim$59880. We obtained the ATLAS photometry from the online ATLAS forced photometry service [144, 139]. The ATLAS forced photometry has a detection of SN 2021qvo before discovery on 17 June at $4.7\sigma$ significance.

Before the ATLAS discovery, serendipitous pre-discovery observations were taken by YSE with the Pan-STARRS Giga Pixel Camera 1 (GPC1; Kaiser & Pan-STARRS Team [77]; Pier & Chambers [125]). YSE is a time-domain survey using the Pan-STARRS telescopes designed to obtain $griz$ light curves of transients up to $z = 0.2$. YSE can detect events as faint as $\sim$20.5 mag in $z$ and $\sim$21.5 mag in $gri$, observing 1500 deg$^2$ every 3 days (Jones et al. [75]). The Pan-STARRS photometry used in this manuscript has already been published in YSE DR1 (Aleo et al. [3]). As described in that data release, the Pan-STARRS photometric data collected by YSE is processed by the University of Hawai‘i Institute for Astronomy’s Image Processing Pipeline (Magnier et al. [101]). The first detection by YSE was on 17 June 2021 in the $g$ and $i$ bands, approximately 17 days before maximum light and $\sim$4 days before the ATLAS discovery. After discovery, we used the YSE-PZ interface (Coulter et al. [37]) to triage the data and organize follow-up observations of this event. See Table 1 and Table 2 for the ATLAS and Pan-STARRS photometric data respectively.

MJD	Filter	Magnitude	Mag err
58402.41	c	21.24	0.60
58406.37	o	21.01	0.53
58410.36	o	18.09	1.09
⋮	⋮	⋮	⋮
59592.21	o	19.86	0.59

MJD	Filter	Magnitude	$\sigma_m$
59382.57	$g$	20.04	0.07
59382.57	$i$	20.25	0.08
59383.57	$g$	20.50	0.08
$\vdots$	$\vdots$	$\vdots$	$\vdots$
59527.21	$i$	21.24	0.23

Beginning near maximum light, imaging of SN 2021qvo was also obtained in the $BVri$ bands with the 1 m Nickel telescope at Lick Observatory. The images were calibrated using bias and sky flat-field frames following standard procedures. PSF photometry was performed, and photometry was calibrated using Pan-STARRS photometric standards (Flewelling et al. [47]).

Hubble Space Telescope Data

The location of SN 2021qvo was also imaged by the Hubble Space Telescope (HST; GO 16691) on MJD 59841, approximately 440 days after $g$-band maximum, in the F555W and F850LP filters. To see if these data could place meaningful constraints on the presence of CSM interacting with the SN ejecta at late times (i.e., Graham et al. [53]), we aligned the HST imaging to Pan-STARRS $r$ using 13 stars visible in all images. Our astrometric alignment procedures followed those of Angus et al. [5], including star detection via Source Extractor (Bertin & Arnouts [12]) to identify common stars in all images. Specifically, we removed sources within 20 pixels of the edges of the images, required sources to have ellipticity $< 0.5$ and FWHM less than five times the mode of the FWHM distribution of detected sources, in order to exclude extended or irregular objects. After applying these criteria, we visually confirmed 13 high-confidence point sources common to both the HST and Pan-STARRS images and used these for the transformation. We performed iterative fitting to determine the best astrometric solution with the IRAF tool GEOMAP, and used IRAF’s GEOTRAN tool to apply the resulting transformation (Tody [152]). The uncertainty on the position was approximately 2 HST pixels. We selected a region of the environment and calculated the probability that a chance alignment occurs within a 2-pixel radius aperture to be 0.01%, indicating that the source is unlikely to be a chance coincidence.

Unfortunately, there was no unambiguous detection of SN 2021qvo, although there was a possible detection in F555W within the uncertainty interval of the best-fit position. Because SN 2021qvo is within the isophotal radius of its host galaxy, we measured the photometry from the cosmic ray-corrected images within a 5-pixel radius using zero points and aperture corrections provided by the WFC3 team.⁴ For sky subtraction, we used the photutils package from AstroPy to find pixels along the same elliptical isophote relative to the host-galaxy center of SN 2021qvo; we constructed an annulus around the host galaxy with a 10-pixel width to match the diameter of the photometric aperture.

Due to the lack of a galaxy template, we cannot be confident that we have detected the SN rather than galaxy flux; the measured magnitude of F555W = 26.25 $\pm$ 0.28 mag (AB) is insufficient to provide meaningful late-time CSM interaction constraints.

Spectroscopic Data and Redshift Determination

SN 2021qvo has seven epochs of spectroscopic data spanning $\sim$−13 days to $\sim$+123 days relative to maximum light (Table 3). Spectra were taken with ALFOSC, the Kast Double Spectrograph (Miller & Stone [106]), the Goodman High Throughput Spectrograph {^a}Phase relative to rest frame $B$-band maximum on MJD 59401.2.

{^b}Host-galaxy light only.

(Clemens et al. [36]), the Gemini Multi-Object Spectrograph (Hook et al. [63]), and the Low-Resolution Imaging Spectrometer (Oke et al. [117]). An additional host-galaxy spectrum was taken with LRIS on 2025-07-27. Table 3 logs the spectroscopic data for SN 2021qvo. The Kast and LRIS data were reduced using the UCSC Spectral Pipeline.⁵ All spectra were corrected for bias, overscan, and flat-field frames. Wavelength calibrations used arc lamps acquired during the same night and the same instrument settings. When possible, the telluric bands were removed using standard star observations.

UT Date	MJD	Epoch{^a}	Telescope	Instrument
	[days]	[days]
2021-06-23	59388.00	−13.2	NOT	ALFOSC
2021-07-09	59404.40	3.2	Lick	KAST
2021-07-15	59410.16	9.0	Lick	KAST
2021-07-19	59414.82	13.6	Goodman	HST
2021-10-14	59501.01	99.8	Gemini-S	GMOS
2021-10-14	59501.06	99.9	Gemini-S	GMOS
2021-11-06	59524.00	122.8	Keck I	LRIS
2025-07-27{^b}	60883.40	1482.2	Keck I	LRIS

To determine the redshift from the final host-galaxy spectrum, we use the XCSAO package (Kurtz et al. [83]). XCSAO compares spectra to a set of galaxy spectral templates and measures the best-fit cross-correlation redshift; from the host-galaxy spectrum of SN 2021qvo, we find a precise redshift of 0.04205.

Host-galaxy Properties

We estimated the host-galaxy properties of SN 2021qvo using the Blast web application (Jones et al. 2024),⁶ Blast estimates physical parameters of transient host galaxies by measuring photometry from GALEX (Martin et al. [103]), SDSS (York et al. [164]; Blanton et al. [15]; Ahumada et al. [1]), Pan-STARRS (Chambers et al. [29]), DECam (Dey et al. [40]), 2MASS (Skrutskie et al. [143]), and WISE (Wright et al. [161]). Blast uses global host-galaxy apertures that are adjusted for the point-spread function (PSF) of each individual filter/instrument to measure the photometry and fits to stellar population synthesis models via the Prospector-$\alpha$ model (Leja et al. [88]; Johnson et al. [73]), which is implemented via the SBI++ neural posterior estimator for increased speed (Wang et al. [157]).

The host galaxy of SN 2021qvo is faint, with PS1 $g \simeq22.1$ mag, and is nearly coincident with the SN location, at $(\alpha, \delta)$ of (22:08:12.11, +07:07:00.84). SN 2021qvo occurred in an unusually faint, low-mass host galaxy with a stellar mass $\log(M_\star/M_\odot) = 7.83^{+0.17}_{-0.24}$ dex. Although this is less massive even than most reported 2003fg-like hosts, which prefer low-mass galaxies (Ashall et al. 2021; Lu et al. 2021), the host galaxy of at least one — SN 2022ilv — is even fainter (Srivastav et al. 2023). Blast also reports host-galaxy masses $\log(M_\star/M_\odot) < 9$ dex for 2003fg-like SNe 2007if and iPTF13asv (see also [134, 27]). While not highly star-forming, there is an indication of recent star formation in the host of SN 2021qvo, with a global star formation rate of $\log_{10}(\mathrm{SFR}\ M_\odot/\mathrm{yr}) = -1.82^{+0.59}_{-0.78}$ and a global specific star formation rate $\log(\mathrm{sSFR}\ \mathrm{yr}^{-1}) = -9.70^{+0.82}_{-0.79}$. The host-galaxy dust is not well constrained by Blast, with best-fit $A_V$ consistent with $\sim$1-4 magnitudes of extinction. The stellar metallicity is poorly constrained by the broadband photometry alone, but Prospector finds modest evidence that the host has a low metallicity, $\log(Z_\star/Z_\odot) = -1.20^{+0.64}_{-0.47}$, which is typical of 2003fg-like SNe Ia and may imply low-metallicity progenitors (e.g., [35, 68] (Lu et al. 2021; Ashall et al. 2021)).

Compared to the larger population of SNe, SN 2021qvo is relatively unique. It has a mass lower than 98.7% of SNe Ia hosts measured by Blast, and 99.5% of transients overall; the comparison set includes 5408 spectroscopically classified SNe, of which 3943 are SNe Ia and eight have been classified as 2003fg-like. Its sSFR is somewhat less extreme; it is higher than 84% of SNe Ia and 81% of all transients. Lastly, SN 2021qvo’s host-galaxy metallicity is lower than 87% of transient hosts measured by Blast, and 89% of SNe Ia.⁷

Photometry

In this section, we discuss the photometric properties of SN 2021qvo, including the light curve, the colorstretch parameter $s_{BV}$, the color curves, and the peak magnitudes, and compare them to other SN Ia subtypes. This analysis will help characterize the properties of SN 2021qvo and solidify it as a 2003fg-like SN event. The photometric data for SN 2021qvo are summarized in Table 1 and Table 2.

Light Curves

Figure 1 shows the light curve of SN 2021qvo, with the inset showing the rising light curve bump around 18.5 days before peak light. The bump is most clearly visible in the $g$ band, with a $\sim4.4\sigma$ detection of the decrease in flux. The $i$ band also shows a signature attributable to a bump: a nearly flat light-curve evolution over a span of $\sim3$ days nearly 2 weeks before maximum light is highly atypical of all known SNe Ia. A natural explanation for the flat $i$-band light curve is that SN 2021qvo rises through or peaks at the first $i$-band detection, declines, and again rises through the second $i$-band detection. Alternatively, the light curve could be declining through the second point, but the later evolution requires it to rise again. Finally, the $o$-band light curve also exhibits a $4.7\sigma$ flux excess. Given the multi-band signatures, especially in the $g$ band, which is well above the YSE detection threshold, we conclude there is a highly significant rising light curve bump in SN 2021qvo.

We next use the SuperNova in object-oriented Python (SNooPy) software [21] (Burns et al. 2014) to determine the light-curve parameters of SN 2021qvo. The SNooPy package is designed to fit SN Ia photometry to an empirical light-curve model. Specifically, we use SNooPy’s EBV_model2, which fits the supernova magnitudes, time of $B$-band maximum light, host galaxy extinction, and a light-curve shape parameter that can be either the “color–stretch” parameter $s_{BV}$, the time difference between the $B$-band maximum and the reddest point in the $(B - V)$ color curve normalized by 30 days (Burns et al. 2014), or $\Delta m_{15}(B)$, the difference in $B$-band magnitude between peak and 15 days after peak [121]. Here, we use $s_{BV}$, which is a more recent and improved stretch parameter (Burns et al. 2014); it also allows easy comparison to measurements from previous 2003fg-like samples (Ashall et al. 2021). The SNooPy fit to SN 2021qvo is shown in Figure 2; note that the SNooPy model assumes SN 2021qvo is a normal SN Ia and therefore predicts a secondary maximum in the $i$ and $z$ bands; however, secondary $i$- and $z$-band maxima are not present in 2003fg-like SNe Ia ([149] (Ashall et al. 2021)).

To minimize errors resulting from the lack of 2003fg-like SN Ia fitting templates in SNooPy, we first used the built-in Gaussian process function to fit all our available bands with the early time bump masked out. We use these fits to set priors on the time of maximum light in the subsequent EBV_model2 fits. SN 2021qvo peaks in the $B$-band⁸ on $MJD = 59401.2 \pm1.00$ days at $m_{g,\max} = 17.42 \pm0.03$ mag and has a color stretch parameter $s_{gr} = 1.210 \pm0.116$ (using the linear correlation found by Ashall et al. [7] we calculate an $s_{BV} = 1.184 \pm0.121$). Correcting only for the redshift-derived distance, we estimate a $g$-band peak absolute magnitude of $M_{g,\max} = -19.3 \pm0.249$ mag. This ranks SN 2021qvo among the lower-luminosity 2003fg-like SNe Ia such as SN 2012dn (Chakradhari et al. [28]) and SN 2022pul (Siebert et al. [141]).

Color Curves

Figure 3 shows the $g-r$ and $r-i$ color curves for SN 2021qvo and a comparison sample of other 2003fg-like SNe Ia from the Carnegie Supernova Project corrected for Milky-Way extinction but without host-galaxy extinction correction. The comparison sample includes SNe 2007if (Akerlof et al. [2]), 2009dc (Puckett et al. [128]), LSQ12gqw (Baltay et al. [11]), 2013ao (Ashall et al. (2021)), LSQ14fmg (Baltay et al. [11]), CSP14abk (Ashall et al. (2021)), ASASSN-15hy (Holoien et al. [58]), and KISS15n (Ashall et al. (2021)).

SN 2021qvo shows a color evolution consistent with other 2003fg-like SNe Ia. As detailed by Ashall et al. (2021), the $r-i$ curve demonstrates the greatest differences between individual 2003fg-like events in terms of their bluest point and color curve shape. Although the data are somewhat limited, SN 2021qvo is bluer in $r-i$ near peak and appears to have larger color changes as a function of phase than other 2003fg-like SNe Ia, with the largest color difference being $\Delta(r-i) = 0.49$ mag between the bluest and reddest point and a $\Delta(r-i) = 0.11$ mag within 15 days at later epochs. This leads to a bluest point around 8 days post $B$-band peak at $r-i = -0.33$ mag, similar to the bluest range of most other 2003fg-like SNe Ia and redder than normals (see Figure 9 in Ashall et al. (2021)). We then see a flattening after the reddest point around $\sim40$ days until $\sim80$ days, consistent with other 2003fg-like events. After this, the large uncertainties make inferring the color evolution of SN 2021qvo tenuous.

The $g-r$ curve shows a pre-peak evolution consistent with the early colors of KISS15n, starting around $g-r=0.01$ mag before reaching its bluest point $\sim2$ days before peak. The curve then becomes redder like the other 2003fg-like SNe Ia in an almost monotonic nature. After reaching its reddest point at around 40 days, SN 2021qvo gradually becomes bluer at later times. This continues until 80 days, when again, the large uncertainties prevent robust conclusions about the color evolution at later phases.

Luminosity Width Relation

The LWR (Phillips [121]) relates the light curve shape of a SN Ia to its intrinsic luminosity. Normal SNe Ia have a tightly correlated LWR, with the more luminous SNe Ia having slower decline rates, and the less luminous SNe Ia having faster decline rates. 2003fg-like SNe Ia are often located on the luminous and broad area of the LWR parameter space. However, 2003fg-like SNe Ia span a variety of luminosities and light curve shapes, including some extremely luminous events (e.g., SN 2007if, Scalzo et al. [134]; SN 2009dc, Taubenberger et al. [150]; Silverman et al. [142]; and LSQ14fmg, Hsiao et al. [68]) and some underluminous events (relative to their light curve shape, e.g., ASASSN-15hy Lu et al. 2021).

Figure 4 shows SN 2021qvo on the $B$-band vs $s_{BV}$ LWR diagram. We compare SN 2021qvo to normal SNe Ia from CSP-I (Krisciunas et al. [81]; Burns et al. [23]) and confirmed 2003fg-like SNe Ia from (Ashall et al. 2021). The 2003fg-like sample includes SNe 2006gz (Hicken et al. [56]), 2007if (Akerlof et al. [2]), 2009dc (Puckett et al. [128]), 2012dn (Chakradhari et al. [28]), 2013ao (Ashall et al. 2021), LSQ14fmg (Baltay et al. [11]), ASASSN-15hy (Holoien et al. [58]), and ASASSN-15pz (Chen et al. [32]). Although it is the least luminous object in the sample, the location of SN 2021qvo on the LWR is coincident with the known sample of 2003fg-like SNe. SN 2021qvo is a notably underluminous 2003fg-like, instead blending into the normal SN Ia population rather than some of the more over-luminous 2003fg-like SNe Ia. We note, however, a couple of reasons for this underluminosity. Due to the sparsity of our $B$-band data, SnooPy was unable to determine a $B$-band peak absolute magnitude using a Gaussian Process function, so we instead use the nearest data point as our estimate. This happens to be 4 days post-peak, so we expect a peak $B$-band magnitude similar to SN 2012dn if accounting for this. Additionally, correcting for the relatively red color of SN 2021qvo (SNooPy finds a value of $E(B - V)_{\mathrm{host}} = 0.151 \pm0.037$ mag) would increase its absolute magnitude by $\sim0.5$–$1$ mag in Figure 4.

Photometric Sub-Type Identification

Ashall et al. [7] presented a method to photometrically differentiate between subtypes using the color stretch parameters, $s_{BV}$ or $s_{gr}$,⁹ and the difference between the time of $i$-band maximum versus either the $B$-band or $g$-band maximum. The time at which the $i$-band peaks relative to other bands is the physical driver behind why this diagnostic works. It traces the speed of the recombination front of the Fe-group elements in the ejecta and, thus, the temperature evolution and adiabatic cooling rate. In particular, 1991bg-like, 2002cx-like, and 2003fg-like SNe Ia are easily identified in this parameter space because their $i$-band maximum occurs after their $B$-band or $g$-band maximum times. For more extensive discussion of this diagnostic, see Ashall et al. [7].

Figure 5 shows $s_{gr}$ versus the $g$- and $i$-band times of maximum ($t^{i}_{\max} - t^{g}_{\max}$), with SN 2021qvo located in the 2003fg-like SN Ia region. Of the 2003fg-like SNe Ia from this sample, SN 2021qvo has the smallest time difference between the $g$ and $i$ bands yet is still consistent with all other 2003fg-likes in the diagnostic, with $i$-band maximum coming $\sim0.7$ days after the $g$-band maximum.

Characterizing the Early-Time Bump

As summarized in Hoogendam et al. [59], the rising light curves of SNe Ia can be categorized as one of “single power-law”, “double power-law”, and “bump.” Most SNe Ia light curves fall under the “single” category, where the light curve behavior is well fit by a single power law. In the case of “double” light curves, their best fits can be characterized by broken or two-component power laws. “Bump” cases have a non-monotonic rising light curve, with a declining light curve within the timescales of $\sim1$–$4$ days before rising again to maximum light. These bumps have been observed in the UV and optical for 2003fg-like SNe Ia ([72, 43] (Srivastav et al. 2023)) and 2002es-like SNe Ia (e.g., [25] (Srivastav et al. 2023)).

A bump can be observed for SN 2021qvo in both the Pan-STARRS and ATLAS data. Plotting the data from both telescopes onto Figure 6 shows nonmonotonic behavior in the rise of the light curve (the $g$ band has the strongest bump, but lack of data during the exponential rise makes it impossible to fit a single-band Gaussian+$t^2$ model similar to previous works (e.g., [41, 136, 104, 61 b] (Wang et al. 2024; Ye et al. 2024)). Beginning with an ATLAS $o$-band observation around 16 days from $i$-band peak, a decrease of nearly 0.8 mag occurs in $g$ band within a day before the quick rise indicative of SN light curves.

Spectroscopy

The spectroscopic data available for SN 2021qvo include seven optical spectra taken between $-13$ and $+123$ days from maximum light. These spectra are listed in Table 3 and their normalized evolution shown in Figure 7.

Spectral Line Measurements

To quantify the spectral evolution of SN 2021qvo, we measure velocities and pseudo-equivalent widths (pEWs) for the most prominent spectral features.

Both of these quantities were extracted with the semiautomated line fitting Python program, Measure Intricate Spectral Features In Transient Spectra (misfits, [57]¹⁰) following the procedure in [60].

To derive the velocity of a feature, we used the velocity.gaussian function, which determines a pseudo-continuum and then fits a Gaussian profile to the absorption feature, the minimum of which is taken as the velocity. We then use the width.shallowpew function to measure the pEW of a feature with the same pseudo-continuum, which integrates the absorption feature to extract a pEW ([52]). Both of these functions estimate the respective uncertainties by computing 10000 realizations using a Monte Carlo approach. The $1\sigma$ standard deviation of the posterior samples serves as the measurement’s uncertainty. For both of these measurements, we employed the lowpass and rawsmooth arguments to remove high-frequency noise by passing the spectra through a low-pass filter to smooth them. Figure 8 shows SN 2021qvo’s Si II spectral velocity and pEW evolutions derived from misfits in comparison to other 2003fg-like SNe and the median values of a normal SNe Ia catalog ([140]). The comparative 2003fg-like sample is the same as in Figure 4 with the addition of SN LSQ12gpw ([11]).

The Branch Diagram

Branch et al. [18] found that the maximum light pEWs of the Si II $\lambda5972$ versus Si II $\lambda6355$ features differentiate SN Ia into four groups based on their spectral features: core normal, shallow Si, broad line, and cool. The “broad line” SNe Ia have a broader Si II $\lambda6355$ feature than the “core normal” SNe Ia, perhaps indicating a high-velocity component. “Shallow silicon” SNe Ia have shallower Si II features, especially Si II $\lambda5972$. Lastly, “cool” SNe Ia have stronger Si II $\lambda5972$ features due to lower temperatures than core-normal SNe Ia.

Like most 2003fg-like SNe Ia (Ashall et al. 2021), SN 2021qvo is located in the shallow Si region of the Branch diagram (Figure 9). The Si II $\lambda5972$ pEW feature for SN 2021qvo is near the average of all 2003fg-like SNe. In contrast, the pEW of the Si II $\lambda6355$ feature is unusually low compared to other 2003fg-like SNe Ia events. This could indicate a higher explosion energy for SN 2021qvo, which may doubly ionize Si and produce Si III features instead of Si II. Only the extremely overluminous SN 2007if ($M_V=-20.4$ mag; [134]) has a lower Si II $\lambda6355$ pEW than SN 2021qvo.

Spectral Comparison

We compare SN 2021qvo to pre- and post-peak spectra from other notable SNe Ia in Figure 10 and Figure 11. Figure 10 compares the $\sim-13$ day spectrum of SN 2021qvo to four other SNe Ia with similarly early spectra. Around $-11$ days, the Si II $\lambda6355$ feature in the 2003fg-like SNe Ia (SNe 2009dc and SN 2021qvo) is present, in contrast to other overluminous subtypes such as 1991T and 1999aa-like SNe where weak Si II features appear after $-7$ days (e.g., [124]). Compared to SN 2009dc, SN 2021qvo has a stronger Si II $\lambda6355$ feature but lacks the prominent C II $\lambda6580$ feature seen in SN 2009dc. In contrast to SN 2011fe, a canonically normal SN Ia with strong Si II $\lambda5972$, $\lambda6355$ features, the rest of the sample has considerably shallower Si II double line features, a trait consistent with the pre-peak and near-peak spectral behavior of these overluminous subtypes (See Figure 9).

Figure 11 compares SN 2021qvo to the same SNe Ia as Figure 10 at $\sim13$ days post peak. While these SNe Ia have spectra that significantly differ at $-13$ to $-10$ days, by $+13$ to $+15$ days they are mostly similar (Ashall et al. 2021). This is due to the dispersion of the outer layers in these later days, leading to the recession of the photosphere well into the ${}^{56}\mathrm{Ni}$ region. ${}^{56}\mathrm{Ni}$ and its products then influence much of the spectra regardless of the subtype, resulting in spectral similarities at these later epochs.

One difference in the $\sim15$ days post-peak spectra may be in the line profile in the 5200 Å to 5400 Å region. gion. SNe 2009dc and 2021qvo have two distinct peaks, whereas SNe 1991T and 1999aa have one peak and a blended decline, and SN 2011fe has a protruding knee feature. A detailed abundance tomography is beyond the scope of this work, but a population-level comparison of 2003fg-like SNe Ia and other normal and overluminous SNe Ia to understand the 5200-5400 region may reveal new insights into the explosion symmetry (perhaps these profiles arise from ejecta asymmetries that can be more easily seen in the NIR; O’Hora et al. [116]), physics (differences in burning may produce different elemental distributions), or temperature (resulting in different ionization fractions).

Constraints and Comparison of CSM in 2003fg-like SNe Ia

In this section, we constrain the explosion parameters of SN 2021qvo, including CSM interaction. We then examine our results alongside the CSM parameter constraints of other comparable 2003fg-like SNe in the literature.

Constraining CSM Interaction

We use the Modular Open Source Fitter for Transients (MOSFiT) to model SN 2021qvo (Guillochon et al. [54]). MOSFiT is a Python-based package that fits transient data to determine parameter distributions using Monte Carlo methods. Within MOSFiT, we use the built-in csmni model, which uses the CSM-ejecta (Chatzopoulos et al. [31]) and $^{56}\mathrm{Ni}$-$^{56}\mathrm{Co}$ radioactive decay (Nadyozhin [110]) models to fit the light curve of SN 2021qvo. In our modeling, we infer parameters by assuming that the early time bump is due to CSM interaction, although see § 6.3 for additional discussion.

The csmni model assumes that both ejecta-CSM interaction and $^{56}\mathrm{Ni}$ decay contribute to the luminosity of the SN Ia. A dense CSM may form through mass loss of the progenitor WD (Hachisu et al. [55]), leftover material from the WD’s main sequence stage (Tsebrenko & Soker 2015), or donor material from a companion star (Moriya et al. [107]). Once the progenitor WD undergoes a SN Ia explosion, the SN ejecta interacts with the CSM, resulting in two shock waves, one moving back into the ejecta and the other carrying energy and the CSM outwards. These shocks then ionize the expanding material, resulting in detectable emission and a contribution to the SN luminosity. Such a model may explain overluminous SNe Ia by attributing the additional luminosity to CSM interaction (Chevalier [33]; Chevalier & Fransson [34]; Chatzopoulos et al. [31]).

For $^{56}\mathrm{Ni}$ and the subsequent $^{56}\mathrm{Co}$ decay emission, the primary power source of the SN light curve, the model assumes centrally concentrated $^{56}\mathrm{Ni}$. As the WD explodes, the $^{56}\mathrm{Ni}$ near the core expands outward with the SN ejecta. The $^{56}\mathrm{Ni}$ then decays to $^{56}\mathrm{Co}$, releasing high-energy photons absorbed by the ejecta and re-emitted at lower energies. This primarily powers the observed light curve’s peak luminosity and quick rise. The remaining $^{56}\mathrm{Co}$ then further decays to stable $^{56}\mathrm{Fe}$, releasing more photons that power the later phases of the supernova (Chatzopoulos et al. [30]). This $^{56}\mathrm{Ni}$ scenario models most normal SNe Ia well and leads to the Arnett [6] relation that can be used to estimate $^{56}\mathrm{Ni}$ mass (e.g., Stritzinger et al. [148]).

At early times, the CSM interaction may produce a significant fraction of the flux; thus, MOSFiT’s csmni combination model can account for early time excess fluxes (see also SN 2022ywc; Srivastav et al. 2023). Its constituent CSM interaction and $^{56}\mathrm{Ni}$ and $^{56}\mathrm{Co}$ decay models are built from Chatzopoulos et al. [31] and Nadyozhin [110], respectively. MOSFiT’s csmni model uses 10 free parameters: total ejecta mass ($M_{\mathrm{ej}}$), the fraction of $^{56}\mathrm{Ni}$ mass with respects to total ejecta mass ($f_{\mathrm{Ni}} = \frac{M_{\mathrm{Ni}}}{M_{\mathrm{ej}}}$), kinetic energy ($E_{\mathrm{k}}$), mass of the CSM shell ($M_{\mathrm{CSM}}$), inner radius of CSM shell ($R_0$), CSM density at inner radius $R_0$ ($\rho_0$), time of explosion relative to first observation ($t_{\mathrm{exp}}$), minimum temperature ($T_{\mathrm{min}}$), host galaxy extinction ($A_V^{\mathrm{host}}$), and white noise variance ($\sigma$): a variable added to measurement uncertainties to allow additional unmodeled errors (Guillochon et al. [54]).

We apply a conservative upper bound on $\rho_0$, as it is not well constrained by the data. We compute the maximum value of $\rho_0$ assuming electron opacity from Thomson scattering and MOSFiT’s quadratic power-law density profile. For a maximum optical depth of $\tau= 5$ — above which essentially no photons could escape without interaction — and a Hydrogen CSM, this gives a maximum density of $\rho= 6 \times10^{-8}$ g cm$^{-3}$ when assuming a minimum CSM radius of 2000 km (approximately equal to the surface of a high-mass white dwarf; Caiazzo et al. [24]).

Our other priors follow Srivastav et al. (2023, their Table 1) with only a few minor differences. Primarily, $R_0$ and $M_{\mathrm{CSM}}$ are slightly more conservative in this analysis; we use a lower limit on $R_0$ of 2000 km and a lower limit on $M_{\mathrm{CSM}}$ of $10^{-7}$ $M_\odot$. Additionally, the lower bound on $T_{\min}$ is 2000 K instead of 100 K, $E_k$ spans from $0.01 \times10^{51}$ erg to $4 \times10^{51}$ erg (slightly more conservative), and $M_{\mathrm{ej}}$ has a higher maximum of $10M_\odot$.

We also use the host-galaxy redshift of $z = 0.04205$ to derive the distance; the luminosity distance assumes a flat $\Lambda$CDM model with $\Omega_m = 0.3$, $\Omega_\Lambda= 0.7$, and $H_0 = 70$ km s Mpc$^{-1}$. Although not all parameters are well constrained by our limited data, our conservative choice in priors ensures that uncertainties are propagated accurately to the parameters of interest.

Lastly, because our initial fitting resulted in poor fits to the light curves after peak, we use an integrated light curve to fit the later-time data. We found that MOSFiT’s best fit to the full data set initially used CSM interaction to fit the near-peak light curve, in an attempt to rectify discrepancies of up to $\sim1$ mag in the colors near maximum light, while leaving the early time bump unmodeled. For this reason, we fit the full light curve up to $-5$ days relative to peak and, for the later-time data, we use SNooPy to integrate fluxes with Gaussian process interpolation across the optical bandpasses to produce an integrated optical light curve across (approximately) the griz bands and spanning $\sim$-$5$ to $+45$ days relative to peak.¹¹ The early-time data therefore constrains MOSFiT’s CSM model, while the later-time light curve constrains MOSFiT’s default model, which assumes the average light curve is powered by $^{56}\mathrm{Ni}$ and $^{56}\mathrm{Co}$ decay.

Our results from MOSFiT using 20 000 MCMC steps—to ensure full sampling of the posterior—are shown in Figure 12. Although we have significant degeneracies, particularly between the density, the CSM mass, and the radius at which the CSM begins (e.g., a closer-in CSM and higher density produce similar light curves to farther-out CSM and lower density), we find that non-zero CSM mass best explains the observed light curves (Figure 13).¹² MOSFiT prefers CSM beginning at $\sim0.05$–$4$ AU, with a best-fit mass of $\log_{10}(M_{\mathrm{CSM}}/M_{\odot})=-2.33^{+0.26}_{-0.15}$. There is also a long, low-probability tail to CSM masses of up to $\sim0.03M_{\odot}$ that is not fully ruled out by our data. We note that using a less-conservative density prior, such as $\rho=10^{-11}\ \mathrm{g\ cm^{-3}}$ following Srivastav et al. (2023), changes the mean CSM mass by just 0.09 dex, well within the errors.

We omit the extinction, minimum temperature, and white-noise variance from Figure 12 for visual clarity. The host-galaxy extinction is constrained to $\lesssim0.4$ mag at the $1\sigma$ level, using priors that range from 0 to 5.6 mag following Srivastav et al. (2023). The minimum temperature is $\log_{10}(T_{\min}\ \mathrm{K})=3.85^{+0.05}_{-0.16}$; however, it has a bimodal distribution with peaks at $\sim3.7$ and $\sim3.9$ dex. Lastly, the variance suggests an added model (or data) uncertainty of $\sim0.08$ mag is needed for a reduced $\chi^2\simeq1$.

Fitting Variants

While the MOSFiT fitting results are largely consistent with previous work (see § 6.3 below), we derive an unphysical $^{56}\mathrm{Ni}$ fraction near 1.0; this is likely because the $^{56}\mathrm{Ni}$ model is not always a good approximation to the colors of 2003fg-like SNe Ia. This high $^{56}\mathrm{Ni}$ fraction is driven by the near-peak and later-time light curve; excluding the early data for SN 2021qvo does not significantly affect our measurement. Although fitting the later-time light curves is not the primary goal of this analysis, we perform two additional tests to ensure the robustness of our CSM mass estimates. First, we fit only the pre-peak light curve, finding that the csmi model can match the early colors well and returns a consistent ($0.9\sigma$) best-fit of $\log_{10}(M_{\mathrm{CSM}}/M_{\odot})=-1.85^{+1.022}_{-0.46}$.

Second, we use SnooPy’s integrated light curve computed from $-5$ to $+45$ days with MOSFiT’s default model, which assumes the light curve is powered by $^{56}$Ni and $^{56}$Co decay, to derive the $^{56}$Ni fraction. This yields a constraint on the $^{56}$Ni fraction between 0.53 and 0.65, which we then use as a flat prior in the pre-peak, multi-color light-curve fit. We note that the bolometric fit alone yields a slightly higher ejecta mass of $\sim0.87^{+0.15}_{-0.09}\ M_{\odot}$, but it is consistent with the baseline results at $\sim1\sigma$ significance. From this analysis, we find $\log_{10}(M_{\mathrm{CSM}}/M_{\odot})=-1.66^{+1.04}_{-0.72}$, again consistent with the unconstrained fit and demonstrating that the $^{56}$Ni fraction and CSM mass are not strongly correlated and can be constrained independently (Figure 12). We treat the MOSFiT parameter estimation discussed in § 6.1 as our baseline result, but note that the early-time results alone may favor a slightly higher CSM mass.

As a third test, we fix $M_{\mathrm{ej}}$ to the Chandrasekhar mass to control for potential model degeneracies in ejecta mass when using the early time light curve. We find only a minimal change to the CSM mass, with a derived value of $\log_{10}(M_{\mathrm{CSM}}/M_{\odot})=-2.26^{+0.42}_{-0.20}$ that differs from our baseline results by only 0.07 dex. This is further evidence that the derived CSM mass is not highly sensitive to uncertainties in the ejecta mass.

Lastly, including possible HST detections at late times does not significantly change the fit, with CSM masses on order $10^{-2}$ remaining the best-fit solution. The late-time epochs are therefore consistent with a normally declining $^{56}$Ni-powered light curve.

As an independent check on the high $^{56}$Ni fraction, we also measure the $^{56}$Ni mass and total ejecta mass from the Khatami & Kasen [80] model following the procedure in Dimitriadis et al. [43]. We adopt the same parameters as Dimitriadis et al. [43], including a constant opacity of $\kappa=0.1\ \mathrm{cm}^{2}\ \mathrm{g}^{-1}$ and a value of $\beta=1.6$ (a dimensionless constant) as suggested by Table 6 of Khatami & Kasen [80]. We estimate the peak luminosity of SN 2021qvo with extrabol as $2.706\times10^{43}\ \mathrm{erg}\ \mathrm{s}^{-1}$ (Thornton et al. [151]) and adopt the best-fit time from explosion to maximum light determined by MOSFiT of $-19.6$ days.

From this model, we derive a $^{56}$Ni mass of $1.78\ M_{\odot}$ and an ejecta mass of $1.89\ M_{\odot}$ (higher than the best-fit MOSFiT ejecta mass of $0.75\ M_{\odot}$), for a $^{56}$Ni fraction of 0.94. We note that we derive $^{56}$Ni fractions of greater than or nearly one for other 2003fg-like SNe; for SN 2021zny we measure a $^{56}$Ni fraction of 1.8,¹³ for SN 2009dc, we find a $^{56}$Ni fraction of 1.6, and for SN 2012dn we find a $^{56}$Ni fraction of 0.97. In practice, this may mean that bright and slow-evolving transients like 2003fg-like SNe Ia require an additional power source beyond $^{56}$Ni to explain their luminosities. We leave a more in-depth analysis of this topic for future work.

Comparison to Other Rising Light Curve Bump Type Ia Supernovae

SN 2021qvo is another SN Ia with a non-monotonic, rising light-curve bump and is one of fewer than 10 SNe Ia to have such a feature. It is just the fourth rising light curve bump SN Ia with CSM mass estimates. Previous SNe Ia with rising light-curve bumps include iPTF14atg (Cao et al. [26]; Kromer et al. [82]), SN 2019yvq (Miller et al. [105]; Tucker et al. [156]; Burke et al. [20]), SN 2020hvf (Jiang et al. [72]), SN 2021zny (Dimitriadis et al. [43], although this claim has been disputed by Fausnaugh et al. [45]), SN 2022ilv (Srivastav et al. 2023), and the 2002es-like SN 2022ywc (Srivastav et al. 2023). We also note that SN 2022pul, although lacking early-time coverage, has nebular-phase data that are consistent with expectations from a CSM-ejecta interaction progenitor model (Siebert et al. [141]).

The existing sample of early-time bump SNe Ia all have CSM mass estimates of order $10^{-2}$ to $10^{-3}\ M_{\odot}$. SN 2022ilv (Srivastav et al. 2023) had an estimated $M_{\mathrm{CSM}}\simeq10^{-2}$ to $10^{-3}\ M_{\odot}$ of material, while SN 2022ywc had $0.050\pm0.006M_{\odot}$ (which had a much stronger and longer-duration bump; Srivastav et al. 2023). SN 2021zny had a best-fit $0.04\ M_{\odot}$ of material (Dimitriadis et al. [43]). Our best-fit CSM mass is fully consistent with the mass range from these previous analyses (Figure 14).

Of these early-time bump SNe, only SN 2022ywc and 2022vqz (Xi et al. 2024), both 2002es-like SNe, have been fit with the MOSFiT CSM+Ni model. Overall, while SN 2022ywc has a much stronger shock-cooling signature than 2021qvo, reaching an absolute magnitude of $\sim-19$ in the ATLAS $o$ band, we find that the physical parameters of SN 2021qvo are broadly consistent within the uncertainties. Specifically, our ejecta mass, $\sim0.74M_{\odot}$, is $\sim2.7\sigma$ lower than SN 2022ywc ($\sim1.24M_{\odot}$), while our explosion energy is consistent within $1\sigma$. The CSM density is somewhat higher but just at $\sim1.2\sigma$ significance; we emphasize that our measurement from SN 2021qvo is extremely uncertain. The best-fit CSM radius in SN 2021qvo is also consistent with SN 2022ywc and it is degenerate with CSM mass and density. SN 2022vqz, which has a relative bump strength that is much more consistent with SN 2021qvo, also has CSM masses that are consistent with SN 2021qvo within the uncertainties ($((M_{\mathrm{CSM}}/M_{\odot})=5.6^{+4.2}_{-1.5}\times10^{-3})$) and a slightly higher ejecta mass of $\sim0.87^{+0.11}_{-0.06}\ M_{\odot}$.

Previous studies have examined alternative explanations for early-time excesses, including bumps, in SNe Ia. These models include surface He-detonations (e.g., Maeda et al. [96]; Polin et al. [127]), surface ${}^{56}\mathrm{Ni}$ mixing (e.g., Piro & Morozova [126]; Magee et al. [100]), and non-degenerate companion interaction (e.g., Kasen [78]; Maeda et al. [98]). However, each of these models has challenges to explain the observations, in particular, the blue UV colors and occurrence rates (see Hoogendam et al. [59], for further discussion). To date, the existing 2003fg-like events with observed light-curve bumps paint a consistent picture of CSM around their progenitor systems, with future statistical samples and fitting methods (e.g., Sarin et al. [133]) likely to enable additional tests of this model.

Summary

In this manuscript, we present SN 2021qvo, one of a growing class of 2003fg-like SNe Ia with bumps in their early-time light curve flux. We analyze photometric data from YSE and ATLAS beginning nearly 17 days before peak, plus an optical spectroscopic time series spanning $-13$ to $+123$ days relative to maximum light.

SN 2021qvo occurred in a low-mass dwarf galaxy with $\log(M_\ast/M_\odot) = 7.83^{+0.17}_{-0.24}$ dex and has evidence for low stellar metallicity, characteristic of other known 2003fg-like host galaxies. Our analysis of the light-curve shape, luminosity, and Si II pEW and velocities indicate that SN 2021qvo is fully consistent with the 2003fg-like subtype. After classifying SN 2021qvo, we test the leading progenitor theory for overluminous SNe Ia with an early-time bump, CSM-ejecta interaction. We use the MOSFiT csmni model to estimate the interaction of ejecta with CSM in the early time light curve, finding a best-fit CSM mass of $\log_{10}(M_{\mathrm{CSM}}/M_\odot) = -2.33^{+0.26}_{-0.15}$, which is in broad agreement with constraints from other 2003fg-like SNe Ia in the literature. Despite being unable to constrain degeneracies between density and radius fully, we find significant evidence for a non-zero CSM to explain the early-time observed bump.

SN 2021qvo is the latest example of a 2003fg-like SN Ia with a rising light curve bump. The origin of these non-monotonic bumps, which have only been seen in 2002es-like and 2003fg-like SNe Ia (Hoogendam et al. [59]), remains intriguing. Recent observational analyses (e.g., Ashall et al. (2021); Hoogendam et al. [59]) have found the most likely explanation to be SN ejecta interacting with CSM, but the origin of this CSM is not yet well constrained. Possible explanations include an AGB stellar envelope that has lost most if not all of its H and He (Kashi & Soker [79]; Ilkov & Soker [70]; Soker & Bear [145]), or from ejected CO during the merger event of two white dwarfs (Shen et al. [137]; Levanon et al. [92]). As the number of SNe Ia with rising light curve bumps increases due to continued early-time discovery, classification, and follow-up efforts, our understanding of the properties and origin of this plausible enshrouding material will improve.

Acknowledgments

We thank Nikhil Sarin for helpful discussions.

I.A.A.P. acknowledges support from the Research Experience for Undergraduate program at the Institute for Astronomy, University of Hawaiʻi, funded through NSF grant no. 2050710, and would like to thank the Institute for Astronomy for their hospitality during the course of this project. W.B.H. acknowledges support from the National Science Foundation Graduate Research Fellowship Program under Grant Nos. 1842402 and 2236415. Any opinions, findings, conclusions, or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation. D.O.J. acknowledges support from NSF grants AST-2407632 and AST-2429450, NASA grant 80NSSC24M0023, and HST/JWST grants HST-GO-17128.028, HST-GO-16269.012, and JWST-GO-05324.031, awarded by the Space Telescope Science Institute (STScI), which is operated by the Association of Universities for Research in Astronomy, Inc., for NASA, under contract NAS5-26555. The UCSC team is supported in part by NASA grants 80NSSC23K0301 and 80NSSC24K1411; and a fellowship from the David and Lucile Packard Foundation to R.J.F. C.G. is supported by a VILLUM FONDEN Young Investigator Grant (VIL25501) and a Villum Experiment grant (VIL69896). M.R.S. is supported by a STScI postdoctoral fellowship.

The Young Supernova Experiment (YSE) and its research infrastructure is supported by the European Research Council under the European Union’s Horizon 2020 research and innovation programme (ERC Grant Agreement 101002652, PI K. Mandel), the Heising-Simons Foundation (2018-0913, PI R. Foley; 2018-0911, PI R. Margutti), NASA (NNG17PX03C, PI R. Foley), NSF (AST–1720756, AST–1815935, PI R. Foley; AST–1909796, AST-1944985, PI R. Margutti), the David & Lucille Packard Foundation (PI R. Foley), VILLUM FONDEN (project 16599, PI J. Hjorth), and the Center for AstroPhysical Surveys (CAPS) at the National Center for Supercomputing Applications (NCSA) and the University of Illinois Urbana-Champaign.

Pan-STARRS is a project of the Institute for Astronomy of the University of Hawai‘i, and is supported by the NASA SSO Near Earth Observation Program under grants 80NSSC18K0971, NNX14AM74G, NNX12AR65G, NNX13AQ47G, NNX08AR22G, 80NSSC21K1572, and by the State of Hawai‘i. The Pan-STARRS1 Surveys (PS1) and the PS1 public science archive have been made possible through contributions by the Institute for Astronomy, the University of Hawai‘i, the Pan-STARRS Project Office, the Max-Planck Society and its participating institutes, the Max Planck Institute for Astronomy, Heidelberg and the Max Planck Institute for Extraterrestrial Physics, Garching, The Johns Hopkins University, Durham University, the University of Edinburgh, the Queen’s University Belfast, the Harvard-Smithsonian Center for Astrophysics, the Las Cumbres Observatory Global Telescope Network Incorporated, the National Central University of Taiwan, STScI, NASA under grant NNX08AR22G issued through the Planetary Science Division of the NASA Science Mission Directorate, NSF grant AST-1238877, the University of Maryland, Eotvos Lorand University (ELTE), the Los Alamos National Laboratory, and the Gordon and Betty Moore Foundation.

Some of the data presented herein were obtained at the W. M. Keck Observatory, which is a private 501(c)3 non-profit organization operated as a scientific partnership among the California Institute of Technology, the University of California, and the National Aeronautics and Space Administration. The Observatory was made possible by the generous financial support of the W. M. Keck Foundation.

A major upgrade of the Kast spectrograph on the Shane 3 m telescope at Lick Observatory was made possible through generous gifts from the Heising-Simons Foundation as well as William and Marina Kast. Research at Lick Observatory is partially supported by a generous gift from Google.

The data presented here were obtained in part with ALFOSC, which is provided by the Instituto de Astrofísica de Andalucía (IAA) under a joint agreement with the University of Copenhagen and NOT.

This work has made use of data from the Asteroid Terrestrial-impact Last Alert System (ATLAS) project. The Asteroid Terrestrial-impact Last Alert System (ATLAS) project is primarily funded to search for near earth asteroids through NASA grants NN12AR55G, 80NSSC18K0284, and 80NSSC18K1575; byproducts of the NEO search include images and catalogs from the survey area. This work was partially funded by Kepler/K2 grant J1944/80NSSC19K0112 and HST GO-15889, and STFC grants ST/T000191/1 and ST/S006109/1. The ATLAS science products have been made possible through the contributions of the University of Hawai‘i Institute for Astronomy, the Queen’s University Belfast, the Space Telescope Science Institute, the South African Astronomical Observatory, and The Millennium Institute of Astrophysics (MAS), Chile.

Facilities: Pan-STARRS. The HST data used in this study can be found at MAST: 10.17909/6bs7-7x11.

Software: astropy ([9, 10]), SNooPy ([21]).