SN 2019yvq Does Not Conform to SN Ia Explosion Models

New ASAS-SN g-band observations were reduced using a fully automated pipeline (Shappee et al. 2014; Kochanek et al. 2017) based on the ISIS image subtraction package (Alard & Lupton 1998; Alard 2000). Each photometric epoch combines three dithered 90 s image exposures subtracted from a reference image. In addition to the standard pipeline, we rebuilt the reference image, excluding any images with $\mathrm{JD}\geqslant 2\,458\,812$ to prevent any flux contamination from the SN.

We then used the IRAF package apphot to perform aperture photometry with a 2 pixel, or approximately 160, radius aperture on each subtracted image, generating a differential light curve. The photometry was calibrated using the AAVSO Photometric All-Sky Survey (Henden et al. 2015). All subtractions were inspected manually, and images with clouds or other systematic issues are excluded from the final light curve.

SN 2019yvq was observed by TESS (Ricker et al. 2015) during the mission's Sector 21 and 22 operations, from January 21.94 to 2020 March 17.96 UTC. These observations narrowly miss the peak of the light curve, but provide excellent high-cadence monitoring of its subsequent decline. TESS observes in a single ∼6000–10000 Å broadband filter with an effective wavelength of ∼8000 Å that is comparable to that of the Johnson–Cousins I band.

We reduced the TESS data using the image subtraction procedure of Vallely et al. (2019), which implements a version of the ASAS-SN pipeline optimized for use with the TESS Full-Frame Images (FFIs). As in Vallely et al. (2019) and Holoien et al. (2019), due to the large pixel scale of the TESS CCDs we chose to construct independent reference images for each sector rather than try to rotate a single reference image for use across multiple sectors. For each sector, reference images were built using the first 100 FFIs of good quality, excluding those with sky background levels or PSF widths above average for the sector as well as those associated with mission-provided data quality flags. The measured fluxes were converted into physical TESS-band fluxes using an instrumental zero point of 20.44 electrons per second in the FFIs, based on the values provided in the TESS Instrument Handbook (Vanderspek et al. 2018). The flux offset between the two sector light curves was determined by using a linear extrapolation of the last 1.5 days of Sector 21 and the first 1.5 days of Sector 22. Because the supernova had already attained maximum light prior to the first TESS observations, the reference images necessarily include flux from the transient. Precise calibration of the absolute flux is unimportant for our analysis, and for display purposes we simply normalize the TESS photometry to the Zwicky Transient Facility (ZTF; Bellm et al. 2019) i-band photometry presented in Miller et al. (2020).

The data reduction process for the optical spectra, observed with GMOS on the Gemini-North telescope, generally follows the Gemini Data Reduction Cookbook. ⁸ Each night has observations at two separate grating angles to remove gaps from the final spectra. Raw frames are bias- and overscan-subtracted, flat-fielded, and mosaicked to reconstruct the monolithic detector. lacosmic (van Dokkum 2001) is used to detect and reject cosmic rays. Wavelength solutions are derived from arc-lamp exposures, and the spectral response curves are generated from spectrophotometric standard stars for each grating tilt. After extracting each spectrum, the individual spectra from each night are combined into a single spectrum and the standard deviation of the individual spectra is used to estimate the uncertainty in the combined spectrum.

We acquired r-band imaging on both nights to reliably flux-calibrate the GMOS spectra. These images are reduced using the same basic process as the spectroscopic observations including bias- and overscan-subtraction, flat-fielding, and cosmic-ray rejection. The World Coordinate System solution provided by the Gemini iraf package is optimized with astrometry.net (Lang et al. 2010). After the astrometric calibration procedure, we use r-band photometry from the Pan-STARRS survey (Chambers et al. 2016; Flewelling et al. 2020) to calibrate each image. The first epoch of our GMOS observations overlaps with the end of the ZTF light curve (Miller et al. 2020), confirming our spectrophotometry with a measured offset of ${\rm{\Delta }}m=-0.005\pm 0.007$ mag.

In addition to the optical GMOS observations, we observed SN 2019yvq with NIRES on the Keck II telescope, which covers 0.95–2.45 μm with a resolution of $\lambda /{\rm{\Delta }}\lambda \approx 2700$. The NIRES observations are normalized by the flat field, and A–B observation pairs are used to remove the sky background. SN 2019yvq is too faint to trace across all the echelle orders so the standard star observation is used as a trace template. Arc-lamp exposures are used to derive the initial wavelength solution, which is then improved with sky emission lines. The standard star observation is used to correct the spectrum for instrumental response and broadband atmospheric absorption. However, NIR photometry of SN 2019yvq could not be obtained before it set for the season, so the NIRES spectrum is presented on a relative flux scale.

Miller et al. (2020) adopted a distance of 42.5 ± 2.1 Mpc to NGC 4441 derived from the 2M++ peculiar velocity model (Carrick et al. 2015), which is inconsistent with the surface brightness fluctuation (SBF) distance of ≈19 Mpc from Tonry et al. (2001). NGC 4441 is likely the merger of a spiral and an elliptical galaxy (Manthey et al. 2008) but SBF distance measurements are best for single-age stellar populations and can be skewed by spatial variations caused by dust or recent star formation (Bothun 1998), which may explain the difference between the kinematic and SBF distances. The SBF distance also implies a high peculiar velocity of ∼1300 km s⁻¹ for NGC 4441 so we attempt to verify the validity of the kinematic distance by comparing the 2M++ results to the peculiar velocity field derived by Graziani et al. (2019). Their approach involves taking measured velocities and distances from Cosmicflows-3 (Tully et al. 2016) and applying a hierarchical Bayesian model to derive the peculiar velocity field out to $z\sim 0.05$ ⁹ (Kourkchi et al. 2020). For NGC 4441, the model provides an expected distance of D = 43.5 Mpc, consistent within ∼2% of the 2M++ result of Carrick et al. (2015).

In addition to constraints from peculiar velocity reconstructions, we also check if NGC 4441 belongs to a nearby galaxy group. Based on the original SBF distance from Tonry et al. (2001), Kourkchi & Tully (2017) label NGC 4441 as a field galaxy with no other group members. Nearby in position and velocity, only one galaxy (NGC 4545) has a redshift-independent distance available in Cosmicflows-3, with a measured Tully–Fisher (Tully & Fisher 1977) distance of $D=34.5\pm 5.2$ Mpc and a preliminary Cosmicflows-4 measurement (Tully et al., in preparation) of $D=36.6\pm 7.3$ Mpc. New galaxy distances from Cosmicflows-4 also show several additional galaxies in the vicinity of NGC 4441 with Tully–Fisher distances ranging between 35 and 42 Mpc. The most likely scenario is that NGC 4441 is a member of a group with the nearby NGC 4521 as the brightest member. This reinforces the likelihood that the original SBF distance to NGC 4441 by Tonry et al. (2001) is erroneous. Thus, we adopt the same distance to NGC 4441 as Miller et al. (2020) for consistency, namely $D=42.5\pm 2.1$ Mpc and $\mu =33.14\pm 0.11$ mag.

Figure 1 shows the early g-band light curve including our new photometry, the ZTF g-band observations from Miller et al. (2020), and the discovery magnitude from Itagaki (2019) compared to the light-curve models from Miller et al. (2020) and Siebert et al. (2020). The photometry from Itagaki (2019) was obtained with a Clear filter, which we convert to an approximate g-band magnitude for easier comparison. This conversion assumes the Clear filter can be approximated as a combination of g and r filters, that the color evolution is negligible between the discovery observation and the first ZTF observations (i.e., a constant color of $g-r\approx -0.2\,$ mag), and a conversion between AB and Vega systems of ∼0.2 mag similar to comparable optical filters. As the early spectra rapidly evolve from a strong blue continuum (Miller et al. 2020), this likely slightly underestimates the true g-band flux. The pseudo-g magnitude is included only for instructive purposes, and we caution that the associated uncertainties are likely of order 20%.

Zoom In Zoom Out Reset image size

The time of discovery from Itagaki (2019) coupled with the ASAS-SN nondetection ≈1.3 days prior to discovery constrain the likely first-light time t_fl to MJD $58\,844.4-58\,845.7$ (18.8 to 17.4 days before ${T}_{{\rm{B}},\max }$). This is mostly consistent with the photometric estimate from Miller et al. (2020) and confirms their decision to exclude the two early ZTF observations when fitting t_fl , although their t_fl is likely too late (i.e., too close to the Itagaki 2019 discovery) to be physically reasonable. The last ASAS-SN nondetection requires a minimum rise of ≳8.0 μJy hr⁻¹ whereas the photometric t_fl from Miller et al. (2020) implies a rise of ∼400 μJy hr⁻¹. Even if the Itagaki (2019) measurement is erroneous by over a magnitude, using g = 18 mag still implies a rise of ∼150 μJy hr⁻¹ for the photometric t_fl from Miller et al. (2020). Thus, the spectroscopic rise time of ∼18 days from Miller et al. (2020) is likely closer to the true rise time and is consistent with our new nondetections and early photometry.

We compare SN 2019yvq to the photometric properties of other SNe Ia near maximum light in Figure 2. SN 2019yvq is not well fit by standard SN Ia light-curve models (Miller et al. 2020) so we fit polynomials to measure the light-curve parameters and employ bootstrap resampling to estimate the associated uncertainties. Fitting quadratic polynomials to the ZTF g- and i-band light curves near peak, we find an offset of ${t}_{\max }^{i-g}=2.26\pm 0.08\,\mathrm{days}$ and ${g}_{\max }=14.82\pm 0.01$ mag for SN 2019yvq.

Zoom In Zoom Out Reset image size

Another metric for SNe Ia light curves is the stretch s_XY (Burns et al. 2014) measured from two filters X and Y, which utilize the color evolution instead of the decline rate to standardize light curves (e.g., s_BV and s_gr ; Ashall et al. 2020). We use spline fits to resample the ZTF r-band light curve (Miller et al. 2020) and uncertainties to associated with the epochs of the g-band light curve. Then, we fit a quadratic polynomial to the g − r light curve and derive ${s}_{{gr}}=0.87\pm 0.01$. These light-curve parameters place SN 2019yvq in an unoccupied region of parameter space in the left panel of Figure 2, reinforcing both the uniqueness and rarity of SN 2019yvq–like events.

The right panel of Figure 2 compares the g-band absolute magnitude and stretch of SN 2019yvq to SNe Ia from the Carnegie Supernova Project I (CSP-I; Krisciunas et al. 2017) and the ZTF 2018 sample (Yao et al. 2019). s_gr is measured directly from the g- and r-band light curves for the CSP SNe Ia. The g- and r-band light curves for the ZTF sample (Yao et al. 2019) often do not extend to $\gtrsim 30$ days after maximum so we fit these SNe Ia with templates in SNooPy (Burns et al. 2011), which computes a template-derived s_BV . We then convert s_BV to s_gr using the relation from Ashall et al. (2020). Due to having observations in only two filters, the ZTF objects have a significant uncertainty due to the limited constraints on the host-galaxy reddening. CSP and ZTF observe in slightly different g-band filters with an offset of ${g}_{\mathrm{CSP}}-{g}_{\mathrm{ZTF}}\approx -0.03$ mag for normal SNe Ia at peak light up to $z\lt 0.1$, and we correct for this offset in our comparison. SN 2019yvq is fainter than all SNe Ia within ${s}_{{gr}}\pm 0.1$ and fainter than all normal SNe Ia in the ZTF 2018 sample (see Figure 6 from Miller et al. 2020).

SN 2019yvq also exhibits unique properties after maximum light, in particular the lack of a prominent NIR secondary maximum (Miller et al. 2020). Figure 3 shows the postmaximum TESS and ZTF i-band (Miller et al. 2020) light curves. The TESS light curve shows a slight inflection at a time that roughly coincides with the NIR secondary maximum of normal SNe Ia (e.g., Kasen 2006; Ashall et al. 2020). Only 91bg-like, 02cx-like and "super-Ch"-mass (03fg-like) SNe Ia lack prominent secondary maxima (e.g., González-Gaitán et al. 2014; Ashall et al. 2020) but SN 2019yvq is spectroscopically inconsistent with these subtypes of SNe Ia (Miller et al. 2020). The timing and presence of the NIR secondary maximum are thought to be caused by the recombination of Fe-group elements in the ejecta (Höflich et al. 2002; Kasen 2006; Jack et al. 2015). This can be seen in brighter SNe Ia, which have stronger and later NIR secondary maxima, whereas dimmer SNe Ia have no NIR secondary maximum (e.g., Taubenberger 2017). Similar SNe Ia having intermediate luminosities but a weak/absent NIR secondary maximum include SN 2006bt (Foley et al. 2010), SN 2002es (Ganeshalingam et al. 2012), and SN 2006ot (Krisciunas et al. 2017). However, these SNe Ia have low photospheric velocities, which are at odds with the high photospheric velocities observed in SN 2019yvq (Miller et al. 2020).

Zoom In Zoom Out Reset image size

We do not detect any H or He emission lines in our optical and NIR spectra, including Hα, Paα, Paβ, He i $\lambda \lambda 5876,6678$, and He i $1.083,2.059\,\mu {\rm{m}}$. Several previous studies place limits on nondetections on H/He in nebular spectra (e.g., Sand et al. 2019; Tucker et al. 2020), relying on radiative-transfer models to predict the observed emission (e.g., Mattila et al. 2005; Botyánszki et al. 2018). However, Dessart et al. (2020) model the effect of changing optical depth on the visibility of the H and He emission and find a nonnegligible time-dependent impact from line blanketing, especially on the higher-order Balmer lines. Thus, we place upper limits on the equivalent width W following the procedure of Leonard (2007; also see Leonard & Filippenko 2001). For both epochs of optical spectra, the equivalent width limit on Hα emission of ${W}_{{\rm{H}}\alpha }(3\sigma )\lt 0.6\,{\rm{\mathring{\rm A} }}$ excludes any reasonable nondegenerate companion (Dessart et al. 2020). This precludes the need for scaling the spectra to later epochs and assuming a nonvariable spectral shape to match the nebular spectra models of Botyánszki et al. (2018) as done by Siebert et al. (2020).

We also do not detect permitted or forbidden O i in the NIR (Figure 6) or optical spectra (Figure 7). Oxygen emission is expected for violent mergers of two C/O WDs as the secondary (lower-mass) WD is only partially burnt (Pakmor et al. 2012). SN 2010lp (Taubenberger et al. 2013) exhibited strong [O i] emission in its nebular spectra and a violent merger is the preferred explanation for this feature (Kromer et al. 2013; Taubenberger et al. 2013). We note that O emission is only expected for merging C/O WDs and may not be present in the merger of a C/O WD and a He WD, which we discuss further in Section 5.3.

Zoom In Zoom Out Reset image size

The sharp spectral feature at ∼7300 Å is almost certainly [Ca ii] emission (Siebert et al. 2020) although [Co ii], [Fe ii], and [Ni ii] contribute to its wings. We lack the spectral resolution for a full decomposition of this region as done by Siebert et al. (2020) and instead focus on the temporal evolution of the [Ca ii] emission between our two spectra because this has not been examined previously. We assume the [Ca ii] emission and its wings can be approximated by Gaussian profiles even though they are blends of multiple lines (e.g., Mazzali et al. 2015). We require the velocity shifts and widths to be roughly consistent between the two epochs, allowing for a slight redshifting and widening of the profiles in the later spectrum to account for the decreasing opacity of the ejecta (e.g., Black et al. 2016).

The results are shown in Figure 8. The absolute [Ca ii] flux decreases between the two epochs but increases relative to the nearby [Ni ii] and [Fe ii] emission lines. Interestingly, the [Ca ii] feature evolves differently than the surrounding [Fe ii] + [Ni ii] features and the strong emission line at ∼8700 Å, but similarly to the [Fe iii] emission line at ∼4600 Å with a near-constant [Ca ii]/[Fe iii] flux ratio of ∼0.4.

Zoom In Zoom Out Reset image size

The strong emission feature at ∼8700 Å is uncommon in nebular spectra of normal SNe Ia but is usually seen in near-maximum spectra and attributed to the absorption and emission of the permitted Ca ii triplet (e.g., Branch et al. 2006, 2008). However, this feature usually disappears ≲100 days after maximum light (e.g., $\lesssim 75$ days for SN 2011fe, Pereira et al. 2013; ≲90 days for SN 1991bg, Turatto et al. 1996; $\lesssim 80$ days for SN 1991T, Silverman et al. 2012; also see Branch et al. 2008) yet the feature is essentially unchanged between near-maximum spectra and our two spectral epochs at +128 and +150 days (Figure 9). Siebert et al. (2020) attribute this feature to Ca ii and the smooth evolution of this feature from maximum light through our observation reinforces this conclusion.

Zoom In Zoom Out Reset image size

The [Ca ii] λ7300 and the Ca ii NIR emission profiles are juxtaposed in Figure 10. The irregularities in the Ca ii NIR emission profile at the velocity shifts for [Ca ii] λ7300 and [Fe ii] derived by Siebert et al. (2019) suggest that the Ca-emitting material producing the [Ca ii] λ7300 feature does contribute to the Ca ii NIR feature but cannot explain the entire emission profile. There is a clear absorption trough centered at ∼8300 Å (Figures 4 and 9), so this profile is likely a complex blend of absorption and emission. Additionally, there are other lines present at this location typical of nebular SNe Ia such as [Fe ii] $\lambda \lambda 8617,8892$. However, the Fe transitions in the $4500\text{--}5000\mathring{{\rm{A}}}$ are roughly consistent with normal SNe Ia (Siebert et al. 2020) so this explanation for the entire $8700\mathring{{\rm{A}}}$ feature is unlikely. Thus, it is likely that this emission is dominated by the Ca ii NIR triplet but that it does not originate in the same region of the ejecta as the [Ca ii] λ7300 emission.

Zoom In Zoom Out Reset image size

The absorption component of the Ca ii NIR triplet is a common feature of SNe Ia near maximum light, which transitions into a complex blend of absorption and emission features in the weeks after maximum light (e.g., Maguire et al. 2014; Silverman et al. 2015; Siebert et al. 2019). The absorption component commonly has different polarization angles and strengths from the surrounding continua (e.g., Wang et al. 2003; Leonard et al. 2005; Wang et al. 2006), suggesting an asymmetric distribution of absorbing material outside the photosphere (e.g., Cikota et al. 2019). Additionally, there is a subset of SNe Ia with high-velocity absorption features (e.g., Mazzali et al. 2005) where the high-velocity component of the Ca ii NIR triplet is kinematically distinct from the photospheric component and thus likely originates in a different portion of the ejecta (e.g., Wang et al. 2003; Leonard et al. 2005; Wang et al. 2009). Considering the high Si ii velocities of SN 2019yvq (Miller et al. 2020), it is possible that this persistent Ca ii emission is related to high-velocity SNe Ia and would naturally explain why the kinematic parameters of the [Ca ii] λ7300 and Ca ii NIR triplet are dissimilar. However, other SNe Ia with high-velocity absorption features lack a prominent Ca ii NIR emission component >100 days after maximum (e.g., SN 2002dj, Pignata et al. 2008; SN 2009ig, Foley et al. 2012). Thus, it is unclear if these similarities are coincidental or if the differences are due to intrinsic variations of a common source such as line-of-sight variations, different CSM configurations, or some other aspect of the progenitor system and/or explosion mechanism.

The nebular-phase NIRES spectrum is shown in Figure 5 and possible emission lines are marked in Figure 6. Consistent with our optical spectra, we do not detect the expected emission lines from any potential stripped companion material such as Paα or He i $1.083\mu {\rm{m}}$. Stripped material models predict the NIR H and He emission lines to be the strongest features in the spectrum (Maeda et al. 2014; Botyánszki et al. 2018; Dessart et al. 2020) and our nondetections severely restrict any potential stripped mass or nearby CSM.

The only major detected feature occurs at ∼1.03 μm, which we attribute to a blend of [Co iii] and [S ii]. Siebert et al. (2020) derived velocity shifts for the ⁵⁶Ni decay products on the order of ∼−4000 km s⁻¹, so this feature is likely dominated by [S ii] because a [Co ii] would need to be redshifted. We lack the signal-to-noise to decompose the individual contributions. This feature is stronger than any other feature in the NIR spectrum by a factor of ≈3.

The 1.15–1.4 μm region exhibits little flux with the exception of a feature at ≈1.27 μm. This coincides with the location of Paα, but it is unlikely to stem from stripped companion material for several reasons. The host spectrum from SDSS (Figure 4, York et al. 2000; Abolfathi et al. 2018) has strong Hα emission indicating star formation, which is generally accompanied by Paschen emission in NIR spectra (e.g., Hill et al. 1996). Additionally, the emission profile is narrower than expected for stripped companion material (e.g., Boehner et al. 2017; Botyánszki et al. 2018) but has a velocity width consistent with the optical host-galaxy emission lines (e.g., Cid Fernandes et al. 2005). The broader component is likely a blend of [Co iii] and [Fe ii], as our epoch (+173 days after maximum light) is when the NIR spectrum is transitioning from Co dominated to Fe dominated (e.g., Sand et al. 2016; Diamond et al. 2018).

The $1.5\mbox{--}1.8\,\,\mu {\rm{m}}$ region has nonzero flux, which can be attributed to a complex blend of [Fe ii], [Fe iii], [Co iii], and [Ni iii] (Diamond et al. 2015; Maguire et al. 2016; Diamond et al. 2018). Due to the low signal-to-noise ratio of this region, we cannot model the various contributing emission components.

Double-detonation explosions occur when a surface shell of He ignites, driving a shock wave into the WD interior and igniting the C/O core (Livne 1990; Livne & Glasner 1991; Livne & Arnett 1995). The He shell can be acquired through accretion from a nearby He star (i.e., the SD scenario; Bildsten et al. 2007) or accretion of a tidally disrupted lower-mass WD (i.e., the DD scenario; Fink et al. 2007). The SD channel is unlikely due to our nondetection of H or He from the stripped material (Section 4.1) but the DD channel is still a possibility, especially because it can produce an early excess flux when the surface He ignites and creates ⁵⁶Ni in the outer ejecta (Livne & Arnett 1995). Miller et al. (2020) fit the early light curve with double-detonation models from Polin et al. (2019), resulting in a best-fit ${M}_{\mathrm{He}}=0.04\,{M}_{\odot }$ and ${M}_{\mathrm{tot}}=0.96\,{M}_{\odot }$ with the caveat that the associated model spectra predict lower photospheric velocities and more line blanketing than are observed in the early spectra of SN 2019yvq.

Siebert et al. (2020) derive a higher total mass of ${M}_{\mathrm{tot}}=1.15\,{M}_{\odot }$ with a shell mass of ${M}_{\mathrm{He}}=0.05\,{M}_{\odot }$ by comparing the observed nebular spectrum to the nebular spectra models of Polin et al. (2021). However, this result is driven by the [Ca ii]/[Fe iii] ratio (see Figures 4 and 7 from Polin et al. 2021), and this model both severely overpredicts the peak luminosity and fails to reproduce the UV/optical "bump" in the early light curve (i.e., Figure 1). Siebert et al. (2020) attribute the discrepancy between the photometric and spectroscopic modeling results to viewing-angle effects, as double-detonation explosions are expected to be asymmetric and have observed properties that depend on viewing angle (e.g., Fink et al. 2010; Kromer et al. 2010; Sim et al. 2012; Gronow et al. 2020).

Observing a thick He-shell double detonation directly along the pole as suggested by Siebert et al. (2020) can reconcile the lower peak luminosity with the high Si ii velocity, as the He burning creates high-velocity He ashes above the expanding SN Ia ejecta, which in turn suppresses the escape of photons from the inner ejecta (Kromer et al. 2010). However, this then produces the largest effects on the early photometric and spectroscopic evolution (e.g., Kromer et al. 2010; Gronow et al. 2020) as the lower peak luminosity is driven by large amounts of line blanketing from the He ashes (Kromer et al. 2010; Polin et al. 2019). Additionally, the extra absorption from the He ashes leads to other observable consequences such as extending the rise time to maximum and decreasing ${\rm{\Delta }}{m}_{15}$ (e.g., Figure 8 from Kromer et al. 2010) with the effects increasing for higher He-shell masses. Because SN 2019yvq has a rise time to maximum consistent with normal SNe Ia (see Figure 1), a rapid decline rate (Miller et al. 2020; Siebert et al. 2020), and a lack of strong line blanketing in the early optical spectra (Miller et al. 2020), a pole-on viewing angle seems unlikely to reconcile the discrepant double-detonation modeling results from Miller et al. (2020) and Siebert et al. (2020).

For completeness, we also briefly consider thin He-shell double-detonation models. Recent simulations have shown that very low amounts of surface He can induce a detonation (e.g., Shen & Moore 2014), although the minimum He-shell mass needed for core detonation is still unclear (Glasner et al. 2018). If very low mass He shells can induce a core detonation, the effects on observed near-peak properties may be minor (e.g., Townsley et al. 2019) although this conclusion has not been extensively tested in the nebular phase. The main issue with a thin He-shell double detonation producing SN 2019yvq is the lack of a prominent early flux excess in these models, and thin He-shell double detonations are generally considered a potential explosion mechanism for normal SNe Ia (e.g., Shen & Moore 2014; Townsley et al. 2019) instead of peculiar events like SN 2019yvq. Additionally, these models exhibit a strong correlation between peak magnitude (a proxy for the WD mass) and Si ii velocity at maximum light (Shen et al. 2018; Polin et al. 2019). SN 2019yvq does not conform to this relation, exhibiting very high Si ii velocities (Miller et al. 2020) but having a peak magnitude fainter than predicted (i.e., observed ${M}_{g}\approx -18.5$ mag versus the predicted ${M}_{B}\approx -19.5$ mag from Polin et al. 2019 for ${v}_{\mathrm{SiII}}\approx -{\rm{15,000}}$ km s⁻¹). Thus, both thin and thick He-shell double-detonation models are unable to qualitatively reproduce the observed characteristics of SN 2019yvq.

An interesting avenue of speculation for double-detonation models, especially the thick He-shell model preferred by Siebert et al. (2020), is the presence or absence of He i $1.083\,\,\mu {\rm{m}}$. Unless the He shell is entirely consumed during the initial surface ignition, residual He material will remain in the system and could produce observable signatures. For example, Boyle et al. (2017) predict He i λ1.083  μm absorption near maximum light whereas Dessart & Hillier (2015) predict strong optical and NIR He i emission at ≳50 days after maximum. These studies employ very different treatments of the He mass distribution so comparisons are limited, but this highlights the uncertainty of both (1) the amount of He material remaining after surface ignition and (2) the effect of any surviving He on the optical and NIR spectral evolution. Helium is a notoriously difficult element to model as it requires a full non-LTE treatment (e.g., Lucy 1991; Hachinger et al. 2012), so we consider this a promising avenue of further study. Any double-detonation model for SN 2019yvq not consuming all available surface He must account for its absence in near-maximum spectra (Miller et al. 2020), late-phase optical (this work; Siebert et al. 2020), and nebular NIR (this work) spectra.

Finally, the presence of strong [Ca ii] is not a direct consequence of a double detonation as suggested by Miller et al. (2020; and subsequently by Siebert et al. 2020), but instead an indicator that the Ca-rich material is located in the same region of the ejecta as the ⁵⁶Ni decay products (e.g., Wilk et al. 2020). [Ca ii] is an extremely effective coolant due to its large oscillator strength, and therefore [Ca ii] emission can be generated by any explosion model producing Ca⁺ in the ejecta regardless of the explosion mechanism itself. Several models in the literature produce strong [Ca ii] emission in the nebular phase without requiring a double-detonation explosion (e.g., Mazzali & Hachinger 2012; Blondin et al. 2017; Botyánszki & Kasen 2017; Galbany et al. 2019; Wilk et al. 2020). These conditions can be created by a double-detonation explosion (e.g., Polin et al. 2021), as the He burning naturally intersperses ⁵⁶Ni with intermediate-mass elements such as Ca, but this effect is not exclusive to double-detonation explosions. Instead, the detection of Ca provides insight into the chemical distribution and bulk ionization state of the ejecta.

Next, we consider ⁵⁶Ni mixing as a potential explanation for both the early flux excess and the presence of [Ca ii] in the nebular spectra. The source of the early flux excess is the same for both ⁵⁶Ni-mixing and double-detonation models: the presence of ⁵⁶Ni in the outer ejecta provides excess heating and thus excess flux (Piro 2012; Piro & Morozova 2016). However, the outer Ni in double-detonation models is produced by the He-burning process whereas ⁵⁶Ni mixing requires a transport mechanism to move the inner ⁵⁶Ni material to the outer ejecta. Previously invoked mechanisms include irregular/asymmetric deflagrations (Khokhlov et al. 1993; Hoeflich & Khokhlov 1996), GCDs (Plewa et al. 2004; Piro 2012), or the direct collision of two WDs (Rosswog et al. 2009; Raskin et al. 2009). Each theory has observational inconsistencies with normal SNe Ia, but these problems may not apply to peculiar SNe Ia such as SN 2019yvq.

It is plausible for SN 2019yvq that the outer ⁵⁶Ni clumps responsible for the early flux excess can also account for the detection of [Ca ii] in the nebular phase. Outward mixing of ⁵⁶Ni may boost the [Ca ii] emission due to the energy from local positron deposition in the Ca-rich region and ensuing radiative cooling. However, there are no published models that address this scenario and how ⁵⁶Ni mixing affects the nebular spectra. The presence of [Ca ii] emission in the nebular spectra models of Wilk et al. (2020) is highly dependent on the amount of clumping in the ejecta, providing a potential avenue for ⁵⁶Ni mixing to address both the early flux excess and the nebular [Ca ii] emission. Furthermore, the ⁵⁶Ni mixing models of Kasen (2006) predict no secondary I-band maximum for a fully mixed composition, a low ⁵⁶Ni yield, or a combination of these two. This could explain the extremely weak NIR secondary maximum in SN 2019yvq seen in Figure 3. Miller et al. (2020) estimate a low ⁵⁶Ni mass of $0.31\pm 0.05\,{M}_{\odot }$, which is lower than for normal SNe Ia ($0.4\mbox{--}0.8\,{M}_{\odot }$, Scalzo et al. 2014). Thus, ⁵⁶Ni mixing is a promising avenue for reproducing some of the major aspects of SN 2019yvq, and studies of normal SNe Ia have produced promising results for the existence of shallow ⁵⁶Ni (e.g., Piro & Nakar 2014).

However, there is a complex interplay between the necessary synthesized ⁵⁶Ni to power the optical light curve (e.g., Arnett 1982), the amount of mixing needed to reproduce the early flux excess (e.g., Piro & Morozova 2016; Magee & Maguire 2020), the amount of mixing needed to suppress the secondary NIR maximum (e.g., Kasen 2006), and limitations on outer ⁵⁶Ni from the lack of unusual spectroscopic features (i.e., line-blanketing effects) in the early spectroscopic evolution (Miller et al. 2020). These features are not self-consistently addressed by any model in the literature nor are there predictions for the effect on nebular spectra. New models are needed to determine if the right combination of ⁵⁶Ni mixing can produce all aspects of SN 2019yvq. Additionally, ⁵⁶Ni mixing is a result of the WD explosion, not a cause, and an explosion mechanism must still be invoked to actually destabilize the WD. All explosion models struggle to reconcile the low luminosity of SN 2019yvq with the high Si ii velocity, a hindrance for any model invoking ⁵⁶Ni to explain the peculiarities of SN 2019yvq.

A third possibility for excess flux soon after explosion is the presence of a dense CSM surrounding the exploding WD. The near-peak spectra from Miller et al. (2020) and our nebular spectra exclude any H or He emission at high confidence, so any CSM must be H and He depleted. This can be created by a WD merger where the more massive WD tidally disrupts and accretes the lower-mass WD (Fink et al. 2007; Pakmor et al. 2010). This process is not completely efficient and some material will escape into the surrounding ISM instead of being consumed by the explosion (e.g., Shen et al. 2013). H- and He-rich surface layers comprise only a small fraction of the total C/O WD mass (${M}_{{\rm{H}}}/{M}_{\mathrm{WD}}\lesssim {10}^{-4}$ and ${M}_{\mathrm{He}}/{M}_{\mathrm{WD}}\sim {10}^{-2};$ e.g., Romero et al. 2012) so inefficient accretion/mass transfer can readily produce a H- and He-deficient CSM.

Miller et al. (2020) showed that the $0.9+0.76\,{M}_{\odot }$ merger model from Kromer et al. (2016) can qualitatively reproduce the photometric evolution of SN 2019yvq, although the early flux excess itself was not modeled due to the numerous possible CSM configurations. The ejecta–CSM interaction models from Piro & Morozova (2016) can produce an early flux excess with duration $\lt 4$ days, blue colors, and peak V-band luminosity of ${M}_{V}\sim -15$ mag for a CSM radius of $\sim {10}^{12}\,\mathrm{cm}$, similar to the early flux excess observed in SN 2019yvq. However, the primary issue with a violent merger of two C/O WDs producing SN 2019yvq is the lack of prominent O emission lines in our optical and NIR spectra. While ignition of the primary WD produces ⁵⁶Ni and iron-group elements, the secondary lower-mass WD is only partially burnt (Pakmor et al. 2011). For a C/O secondary WD, this results in unburnt oxygen near the center of the ejecta, which produces strong O emission lines in the nebular spectra as seen in SN 2010lp (Taubenberger et al. 2013; Kromer et al. 2013).

Interestingly, Kromer et al. (2013) and Taubenberger et al. (2013) state that [O i] is not expected for a merger between a C/O WD and a He WD, as there is no unburnt oxygen material to produce the nebular emission lines. There are no models directly assessing the result of a C/O WD merging with a He WD, but we consider two simplistic outcomes: the He WD is partially burnt or the He WD experiences no burning. If the He WD is partially burnt, by-products such as O, Ca, and Ti should be produced and reside near the center of the ejecta. This matches the detection of [Ca ii] λ7300, but the burning process must be highly efficient otherwise residual O or He remains in the ejecta and should produce nebular emission lines. If the He WD experiences little or zero burning, a large amount of unburnt He material remains near the center of the ejecta. This is reminiscent of material stripped from a He donor star and should produce strong NIR He emission lines (e.g., Botyánszki et al. 2018) which we exclude at high confidence.

Thus, a H- and He-deficient CSM produced in the merger of two WDs is an unlikely progenitor system to explain SN 2019yvq. A merger of two C/O WDs should produce O emission, which we do not detect, and a merger of a C/O WD with a He WD would likely produce strong emission lines of O and/or He. The exception is if the He WD experiences significant burning and converts all He to elements heavier than O, which would produce nebular [Ca ii] emission but no O or He emission lines. We consider this scenario unlikely but numerical simulations should confirms these qualitative considerations.

Finally, there is the direct collision of two WDs (e.g., Raskin et al. 2009; Rosswog et al. 2009). This scenario is thought to produce double-peaked emission lines of ⁵⁶Ni decay products (e.g., Dong et al. 2015) as the ensuing velocity distribution is inherently bimodal. However, double-peaked or asymmetric emission lines can also be produced by off-center explosions (e.g., Gall et al. 2018; Vallely et al. 2020). We do not find any evidence for double-peaked or flat-topped emission profiles in our optical spectra of SN 2019yvq, but our optical spectra are not late enough in the nebular phase for a definitive conclusion. Direct collision models in the literature are scarce, and important questions, such as the presence of unburnt material near the center of the ejecta, are currently unanswered. If unburnt material survives the collision and subsequent explosion, we would expect similar spectral signatures to those in the merger scenario (i.e., He or O emission lines). We encourage further modeling of this scenario to understand the chemical composition of the surviving material and its effect on the nebular spectra.