The Spectacular Ultraviolet Flash from the Peculiar Type Ia Supernova 2019yvq

ZTF (Bellm et al. 2019b; Graham et al. 2019; Dekany et al. 2020) simultaneously conducts multiple time-domain surveys using the ZTF camera on the the Palomar Oschin Schmidt 48 inch (P48) telescope. SN 2019yvq was first detected by ZTF on 2019 December 29.46, as part of the ZTF public survey (see Bellm et al. 2019a). The automated ZTF pipeline (Masci et al. 2019) detected SN 2019yvq using the image-differencing technique of Zackay et al. (2016). The candidate passed internal thresholds (e.g., Mahabal et al. 2019), leading to the production and dissemination of a real-time alert (Patterson et al. 2019) and the internal designation ZTF19adcecwu. The public alert included the position, $\alpha ={12}^{{\rm{h}}}27^{\prime} 21\buildrel{\prime\prime}\over{.} 836$, δ = +64°47'5987 (J2000), and brightness, r_ZTF = 17.14 ± 0.05 mag, which, together with the Itagaki (2019) discovery report suggested the SN was fading. There was an ∼8 day gap in ZTF observations prior to its initial detection of SN 2019yvq, meaning ZTF nondetections cannot directly constrain the time of explosion, t_exp. Continued monitoring with ZTF, and follow-up with other telescopes, confirmed a spectacular decline in the early emission from SN 2019yvq (Figure 1).

Zoom In Zoom Out Reset image size

The field of SN 2019yvq was additionally observed by ZTF with nearly a nightly cadence as part of the ZTF partnership Uniform Depth Survey (ZUDS; D. Goldstein et al. 2020, in preparation). Using images obtained as part of the ZUDS program, we perform forced point-spread function (PSF) photometry at the location of SN 2019yvq following the procedure described in Yao et al. (2019).²⁶ The evolution of SN 2019yvq in the g_ZTF, r_ZTF, and i_ZTF filters is shown in Figure 1, and the ZTF flux measurements are summarized in Table 1.

UV observations of SN 2019yvq were obtained with the UVOT (Roming et al. 2005) onboard the Neil Gehrels Swift Observatory (hereafter Swift; Gehrels et al. 2004) following a time-of-opportunity request.²⁷ Pre-SN UVOT reference images are limited to the uvw1, uvm2, and uvw2 filters. Therefore, accurate estimates of the SN flux in the Swift u, b, and v filters are not possible.

Table 1.  ZTF P48 Photometry of SN 2019yvq

MJD	Flux	${\sigma }_{\mathrm{flux}}$	Filter
	($\mu \mathrm{Jy}$)	($\mu \mathrm{Jy}$)
58,846.4699	504.81	7.28	rZTF
58,846.5385	374.33	4.99	iZTF
58,846.5583	595.33	5.56	gZTF
58,849.4489	487.54	7.75	rZTF
58,849.5078	379.06	5.54	gZTF

Note. Observed fluxes in the ZTF passbands, no correction for reddening has been applied. Due to poor observing conditions, SN 2019yvq is not detected in one g_ZTF and one i_ZTF image from 2020 March 9, and we therefore do not provide a flux measurement for those epochs.

Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.

Download table as:  DataTypeset image

We estimate the flux in the uvw1, uvm2, and uvw2 filters using a circular aperture with a $3^{\prime\prime}$ radius at the SN position, and subtract the flux measured using an identical procedure in the pre-SN images, as summarized in Table 2. For clarity, we only show the Swift uvw1 and uvm2 light curves in Figure 1.²⁸ Swift/UVOT observations show that the initial decline seen in the optical is even more dramatic in the UV.

Table 2.  UVOT Photometry of SN 2019yvq

MJD	Flux	${\sigma }_{\mathrm{flux}}$	Filter
	($\mu \mathrm{Jy}$)	($\mu \mathrm{Jy}$)
58,846.8969	457.90	30.80	uvw1
58,846.9017	314.30	22.96	uvw2
58,846.9066	392.00	24.90	uvm2
58,846.9607	390.40	27.64	uvw1
58,846.9655	307.40	22.56	uvw2

Note. Host-subtracted fluxes in the UVOT passbands, no correction for reddening has been applied. Epochs with a signal-to-noise ratio (S/N) < 3 are shown as upper limits in Figure 1.

Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.

Download table as:  DataTypeset image

While absolute flux measurements in the UVOT u, b, and v filters are not available, if we assume the underlying flux from the host is constant in time we can estimate the time of B-band maximum, ${T}_{B,\max }$, from the relative b-band light curve. Using a second-order polynomial, we model the b-band light curve near peak (including observations between MJD > 58,855 and MJD < 58,871). From this fit we measure ${T}_{B,\max }\,=58$,863.33 ± 0.21 MJD. The UVOT b filter is slightly different from the Johnson B filter, with the latter typically being used to estimate ${T}_{B,\max }$. Using nine SNe with ${T}_{B,\max }$ estimates from Johnson B-band observations (Krisciunas et al. 2017), we repeat the above procedure on Swift b-band observations (data from Brown et al. 2014). We find that most of these SNe have ${T}_{B,\max }$ measurements consistent to within the uncertainties. On average, ${T}_{B,\max }$ estimates from Swift b-band observations occur later than those in the Johnson B-band, with a median offset of ∼0.26 day.

In parallel with the Swift/UVOT observations, Swift observed SN 2019yvq with the XRT (Burrows et al. 2005) between 0.3 and 10 keV in the photon counting mode from 2019 December 29 through 2020 February 27. We analyzed the data with the online tools of the UK Swift team²⁹ , which uses the methods described in Evans et al. (2007, 2009) and the software package HEASOFT³⁰ version 6.26.1 (HEASARC 2014).

To build the light curve of SN 2019yvq and test whether transient X-ray emission is present at the SN position, we stack the data of each Swift observing segment. In the pre-SN observations from 2012, we detect X-ray emission at the position of SN 2019yvq. The average count rate in the 2012 observations is $0.0026\pm 0.0008\,\mathrm{ct}\,{{\rm{s}}}^{-1}$ (0.3–10 keV). The detected count rate during observations of SN 2019yvq is marginally higher than in 2012, however, spectra of the two epochs show no differences to within the uncertainties. Therefore, the same source from 2012 dominates the ongoing emission at the position of SN 2019yvq.

In the first epoch of XRT observations of SN 2019yvq, corresponding to the time we would expect the X-ray flux to be largest if the UV/optical flash is due to the collision of the ejecta with either circumstellar material or a nondegenerate companion, we marginally detect emission at the position of SN 2019yvq with a count rate of ${0.0031}_{-0.0013}^{+0.0017}$ ct s⁻¹. This flux is identical to that measured in 2012 to within the uncertainties. To estimate an upper limit on the SN flux, we take the difference between the 2019 and 2012 flux measurements and arrive at a 3σ upper limit on the SN count rate of $\lt 0.0057$ ct s⁻¹. The upper limits in future epochs of XRT observations are less constraining than this first epoch.

To convert the count rate to flux, we extracted a spectrum of the 2019–2020 data set. The spectrum is adequately described with an absorbed power law where the two absorption components represent absorption in the Milky Way and the host galaxy. The Galactic equivalent neutral-hydrogen column density was fixed to $2.03\times {10}^{20}\,{\mathrm{cm}}^{-2}$ (HI4PI Collaboration et al. 2016). The best fit suggests negligible host absorption, though we note that the data do not constrain this parameter, and a photon index³¹ of ${\rm{\Gamma }}={1.9}_{-0.5}^{+1.0}$ (all uncertainties at 90% confidence; ${\chi }^{2}=30.8$, with 32 degrees of freedom assuming Cash statistics). From this fit the unabsorbed count-rate-to-energy conversion factor is $5\times {10}^{-11}\,\mathrm{erg}\,{\mathrm{cm}}^{-2}\,{\mathrm{ct}}^{-1}$.

From the count-rate conversion factor, we estimate an upper limit on the X-ray flux of $2.9\times {10}^{-13}$ erg cm⁻² s⁻¹ at the first epoch of Swift observations. At the distance of SN 2019yvq (see Section 3), this corresponds to an X-ray luminosity of ${L}_{X}\lt 6.2\times {10}^{40}$ erg s⁻¹. This luminosity is significantly lower than the ∼5 $\times \,{10}^{44}$ erg s⁻¹ estimate from Kasen (2010) for the interaction between the SN ejecta and a nondegenerate companion. However, this discrepancy is not surprising as the X-ray emission is only expected to last for minutes to hours, and the Swift observations occurred at least 1.1 days after explosion (based on the initial detection from Itagaki 2019).

Spectroscopic observations of SN 2019yvq were initiated because the transient passed the threshold criteria for both the ZTF Bright Transient Survey (Fremling et al. 2019) and the ZTF Census of the Local Universe experiment (De et al. 2020). Our first spectrum, obtained ∼1.8 days after the initial ZTF detection with the SPectrograph for the Rapid Acquisition of Transients (SPRAT; Piascik et al. 2014) on the 2 m Liverpool Telescope (LT; Steele et al. 2004), had a blue and nearly featureless continuum. Further spectroscopy was obtained with a variety of telescopes, including: the Spectral Energy Density machine (SEDM; Blagorodnova et al. 2018; Rigault et al. 2019) on the Palomar 60 inch telescope (P60), Binospec (Fabricant et al. 2019) on the 6.5 m MMT telescope, the Low-Resolution Imaging Spectrometer (LRIS; Oke et al. 1995) on the 10 m Keck I telescope, the Andalucia Faint Object Spectrograph and Camera (ALFOSC)³² on the 2.5 m Nordic Optical Telescope (NOT), and the Double Spectrograph (DBSP; Oke & Gunn 1982) on the Palomar 200 in Hale Telescope. The optical spectral evolution of SN 2019yvq is illustrated in Figure 2, with an accompanying observing log listed in Table 3.

Zoom In Zoom Out Reset image size

Table 3.  Spectroscopic Observations of SN 2019yvq

tobs	Phase	Telescope/	R	Range	Air
(MJD)	(days)	Instrument	$({\rm{\Delta }}\lambda /\lambda )$	(Å)	Mass
58,848.27	−14.9	LT/SPRAT	350	4020–7990	1.24
58,850.28	−12.9	LT/SPRAT	350	4020–7990	1.24
58,851.21	−12.0	LT/SPRAT	350	4020–7990	1.29
58,852.07	−11.2	LT/SPRAT	350	4020–7990	1.88
58,853.07	−10.2	LT/SPRAT	350	4020–7990	1.86
58,854.22	−9.0	LT/SPRAT	350	4020–7990	1.27
58,860.13	−3.2	LT/SPRAT	350	4020–7990	1.46
58,860.34	−3.0	P60/SEDM	100	3850–9150	1.64
58,863.38	+0.0	P60/SEDM	100	3850–9150	1.40
58,866.50	+3.1	MMT/Binospec	4000	4645–6155	1.19
58,872.61	+9.2	Keck I/LRIS	1100	3200–10250	1.41
58,873.30	+9.9	P60/SEDM	100	3850–9150	1.64
58,875.54	+12.1	P60/SEDM	100	3850–9150	1.19
58,878.09	+14.6	NOT/ALFOSC	360	3760–9620	1.41
58,880.39	+16.9	P60/SEDM	100	3850–9150	1.26
58,887.10	+23.5	LT/SPRAT	350	4020–7990	1.32
58,888.07	+24.5	LT/SPRAT	350	4020–7990	1.39
58,888.97	+25.4	LT/SPRAT	350	4020–7990	1.87
58,890.01	+26.4	LT/SPRAT	350	4020–7990	1.60
58,891.06	+27.5	LT/SPRAT	350	4020–7990	1.40
58,892.25	+28.6	P60/SEDM	100	3850–9150	1.66
58,900.22	+36.5	P60/SEDM	100	3850–9150	1.69
58,906.45	+42.7	P200/DBSP	700	3410–9995	1.19
58,908.32	+44.6	P60/SEDM	100	3850–9150	1.25
58,930.47	+66.5	Keck I/LRIS	1100	3200–10250	1.42

Note. Phase is measured relative to ${T}_{B,\max }$ in the SN rest frame. The resolution R is reported for the central region of the spectrum.

Download table as:  ASCIITypeset image

With the exception of SEDM, all observations were obtained with the slit positioned along the parallactic angle, and the spectra were reduced using standard procedures in IDL/Python/Matlab. SEDM is a low-resolution ($R\approx 100$) integral field unit (Blagorodnova et al. 2018), and the observations are reduced using the custom pysedm software package (Rigault et al. 2019). While SEDM was designed specifically for SN classification (e.g., Fremling et al. 2019), the quality for SN 2019yvq is high enough to provide detailed absorption line measurements (see Section 5.2).

We estimate the time of first light, t_fl, for SN 2019yvq following the procedure described in Miller et al. (2020). Briefly, Miller et al. (2020) model the early emission from an SN Ia as a power law in time, $f\propto {(t-{t}_{\mathrm{fl}})}^{\alpha }$, where f is the flux, t is time, and α is the power-law index. t_fl is assumed to be the same everywhere in the optical, allowing us to simultaneously fit observations in each of the ZTF filters.

An important caveat for SN 2019yvq is that the observed early decline in the light curve clearly does not follow the simple power-law model, and thus these observations must be masked when performing the fit. We conservatively exclude observations from the first two nights of ZTF detection from the fit (this choice is conservative as it is unclear when the mechanism that powers the initial bump in SN 2019yvq no longer significantly contributes to the flux in the g_ZTF and r_ZTF filters). From the fit we find t_fl$=-{17.5}_{-1.3}^{+1.0}$ days relative to ${T}_{B,\max }$.³⁵ We know that the time of explosion must be $\lt -17.4$ days based on the discovery detection in Itagaki (2019), and, by definition ${t}_{\mathrm{fl}}\geqslant {t}_{\exp }$, meaning a portion of the posterior distribution for our model cannot be correct. We find ${\alpha }_{g}={2.15}_{-0.36}^{+0.49}$ and ${\alpha }_{r}={1.91}_{-0.31}^{+0.42}$, which are typical of the normal SNe Ia studied in Miller et al. (2020). If we only exclude the first observation from the model fit we find significantly different results with a rise time that increases by ∼5 days and power-law indices that increase by $\gtrsim 1$. We note that such a long rise is unlikely, however, as our spectroscopic models (see Section 5.1) estimate the time of explosion, t_exp, to be ∼17.9 days prior to ${T}_{B,\max }$, fully consistent with our estimate of t_fl.

To estimate the luminosity and temperature of the initial UV/optical flash from SN 2019yvq, we model the broadband SED as a blackbody. The assumed distance and reddening toward SN 2019yvq are taken from Section 3. The ZTF optical and Swift UV observations were not simultaneous, so we interpolate the optical light curves to estimate the flux during the same epochs as Swift observations. While SNe Ia do not emit as pure blackbodies, our initial spectrum of SN 2019yvq shows a blue and nearly featureless continuum largely consistent with blackbody emission. The blackbody assumption is therefore reasonable for the early flash from SN 2019yvq, which is distinctly different from normal SNe.

Following interpolation to an epoch 1.24 days after t_fl (MJD = 58,846.93), and including the uvw2 filter, we estimate a blackbody luminosity $L=({1.7}_{-0.1}^{+0.2})\times {10}^{42}$ erg s⁻¹ and temperature ${T}_{\mathrm{eff}}={14.8}_{-1.2}^{+0.9}$ kK. This estimate represents a lower limit to the peak luminosity of the initial flash, as the UV flux was already decreasing at this time (Swift obtained two sets of UV observations separated by ∼90 minutes during the first epoch of observations, and the uvm2 and uvw1 flux is clearly decreasing during this time; see Figure 1).

At an epoch 3.15 days after t_fl, we estimate the luminosity and temperature to have fallen to $L=({7.0}_{-0.6}^{+0.9})\times {10}^{41}$ erg s⁻¹ and ${T}_{\mathrm{eff}}={8.7}_{-0.4}^{+0.5}$ kK, respectively. For this epoch we have excluded the uvw2 flux from the blackbody model due to the significantly lower temperature, and known red leak for that filter (see Section 2.2). This measurement of T_eff is consistent with our model spectrum from 2.6 days after t_fl (see Section 5.1). At a similar epoch, ∼4 days after explosion, Cao et al. (2015) estimated a UV flash luminosity of ∼3 $\times \,{10}^{41}$erg s⁻¹ in iPTF 14atg, a factor of ∼2 less than for SN 2019yvq.

Finally, if we conservatively assume that the early flash peaked 1 day after t_fl (i.e., at the epoch of the first Swift observation), and abruptly ended 3 days after t_fl (i.e., at the epoch of the second Swift observation), then the initial flash emitted a total integrated energy of ∼4$\times {10}^{42}$ erg. These assumed times are highly uncertain, however, it is likely that the SN exploded before t_fl (see, e.g., Sections 5.1 and 6), and the UV flux continues to decline >3 days after t_fl (Figure 1) suggesting the flash lasted longer than 3 days.

While the early emission from SN 2019yvq may be approximated as a blackbody, SNe Ia do not emit as blackbodies around maximum light. To estimate the bolometric luminosity of SN 2019yvq, we model changes in the observed flux in the uvm2, uvw1, g_ZTF, r_ZTF, and i_ZTF filters as a Gaussian process (Rasmussen & Williams 2006) using the Gaussian_process library in scikit-learn (Pedregosa et al. 2011). This allows us to interpolate flux measurements in each of these filters to a grid of times between 1 and 70 days after t_fl, while also estimating an uncertainty on the interpolation. From there, we can estimate the bolometric luminosity, L_bol, via trapezoidal integration of the SED.

There is emission blueward of the uvm2-band and redward of the i_ZTF-band that is not constrained by our observations, and for fast-declining SNe the near-infrared (NIR) provides a significant contribution to L_bol (e.g., Taubenberger et al. 2008). To estimate the NIR flux, we extrapolate redward from the i_ZTF-band to the K_s-band (${\lambda }_{\mathrm{eff}}=2.159\,\mu$m) by assuming the K_s-band flux is equal to the ratio of the i_ZTF-band to the K_s-band flux for a 8500 K blackbody. This choice of temperature is reasonable based on our TARDIS spectral models (see Section 5.1 and Table 4). While the true temperature is not constant, we find that varying the temperature between 6000 and 12,500 K changes the peak L_bol by ≲3%, which is significantly less than the total systematic uncertainty. The assumed emission redward of the i_ZTF-band results in a NIR contribution of ∼20% to L_bol near maximum light, and ∼35% at ${T}_{B,\max }$ $\approx +30$ days. This is similar to SN 2004eo, an SN with an intermediate decline rate like SN 2019yvq (see Section 4.4), and other fast-declining SNe Ia (Taubenberger et al. 2008).

Table 4.  TARDIS Input Parameters

Date	MJD	Phase	$t-{t}_{\exp }$	L	vboundarya	Tboundaryb
(UT)		(days)	(days)	($\mathrm{log}{L}_{\odot }$)	(km s−1)	(K)
2019 Dec 31.277	58,848.277	−14.9	3.0	8.55	25,000	8173
2020 Jan 03.217	58,851.217	−12.0	6.0	8.60	16,500	7925
2020 Jan 15.392	58,863.392	+0.0	18.0	9.29	10,500	9696

Notes. Phase is measured in rest-frame days relative to ${T}_{B,\max }$. The time of explosion, t_exp, is assumed to be 0.4 day before t_fl for the TARDIS model.

^aEjecta velocity at the inner boundary of the photosphere. ^bTemperature at the inner boundary of the photosphere. T_boundary is not explicitly an input parameter for TARDIS, it is derived from the luminosity, time since explosion, inner boundary velocity, and then iteratively updated throughout the simulation.

Download table as:  ASCIITypeset image

Similar to our procedure in the NIR, we estimate the flux in the far-UV by extrapolating between the uvm2-band and 1000 Å assuming a 12,500 K blackbody. While such a high temperature is only appropriate for the early UV flash from SN 2019yvq (see Section 4.2), the far-UV contribution to L_bol following this assumption is negligible (≲1%) around maximum light and later epochs.

In the near-UV, probed by the UVOT uvm2 and uvw1 filters, the SN is only marginally detected at epochs $\gt 26.5$ days after t_fl (see Figure 1). Given the low S/N in the Swift observations at these epochs, for $t\gt {t}_{\mathrm{fl}}+26.5$ days we interpolate the uvm2 and uvw1 flux by assuming their ratio relative to the g_ZTF flux, which is measured at a very high S/N, is fixed and set by their relative ratios at $t={t}_{\mathrm{fl}}+26.5$ days. Fixing the UV flux in this manner does not change our estimate of the ⁵⁶Ni mass, and does not have a significant effect on our estimate of the total ejecta mass.

The bolometric luminosity of SN 2019yvq is shown as a function of time in Figure 4. Statistical uncertainties in L_bol are estimated via bootstrap resampling of the interpolated flux at each epoch, and are typically on the order of a few percent. We estimate a systematic uncertainty of ∼10% based on the total procedure (including interpolation, extrapolation, and integration).

Zoom In Zoom Out Reset image size

As shown in Figure 4 the method compares favorably with a blackbody model (at early epochs) and spectroscopic modeling (during the SN rise). The maximum-light luminosity estimate from the TARDIS spectral model likely overestimates the flux in the NIR (see the third panel in Figure 7), due to the model assumption that there is a single, sharp photosphere that does not vary with wavelength. This explains the discrepancy between SED integration and the TARDIS model at that epoch.

From the SED integration we find that the bolometric luminosity of SN 2019yvq peaked 18.1 days after t_fl (∼0.6 day after ${T}_{B,\max }$) at ${L}_{\mathrm{bol},\max }=6.4\pm 0.1(\mathrm{statistical})\,\pm 0.6(\mathrm{systematic})\times {10}^{42}\,\mathrm{erg}\,{{\rm{s}}}^{-1}$. This peak luminosity is ∼70% larger than the peak luminosity of iPTF 14atg ($3.8\,\times {10}^{42}\,\mathrm{erg}\,{{\rm{s}}}^{-1};$ Kromer et al. 2016).

From Arnett's rule (Arnett 1982), which states that the peak luminosity is equal to the instantaneous rate of radioactive energy released by ⁵⁶Ni, we estimate the total mass of ⁵⁶Ni, ${M}_{{56}_{\mathrm{Ni}}}$, synthesized in the explosion. Using Equation (19) from Nadyozhin (1994, see also Stritzinger et al. 2006; Howell et al. 2009; Scalzo et al. 2014a), we find ${M}_{{56}_{\mathrm{Ni}}}=0.31\pm 0.05\,{M}_{\odot }$, where the uncertainty is dominated by the (assumed) systematic uncertainty on L_bol. We note that if our previous assumption about the NIR contribution to L_bol at maximum light is revised downward from ∼20% to ∼5%, as is typical for normal SNe Ia (e.g., Suntzeff 1996; Contardo et al. 2000), the total ⁵⁶Ni mass still agrees with the above estimate to within the uncertainties. This yield is low for a normal SN Ia as typical explosions yield ∼0.4–0.8 ${M}_{\odot }$ of ⁵⁶Ni (e.g., Scalzo et al. 2014b).

Following Jeffery (1999), we can estimate the total mass ejected by SN 2019yvq, M_ej, by calculating the transparency timescale, t₀, from the decline of the bolometric light curve (see also Stritzinger et al. 2006; Scalzo et al. 2014a; Dhawan et al. 2018). Briefly, t₀ is a parameter that governs the time-varying γ-ray optical depth of an SN, and it is related to M_ej as follows (Jeffery 1999; Dhawan et al. 2018):

$$\begin{eqnarray}&&{M}_{\mathrm{ej}}=1.38\left(\displaystyle \frac{1/3}{q}\right){\left(\displaystyle \frac{{v}_{e}}{3000\mathrm{km}{{\rm{s}}}^{-1}}\right)}^{2}{\left(\displaystyle \frac{{t}_{0}}{36.80\mathrm{days}}\right)}^{2}{M}_{\odot },\end{eqnarray} \tag{ 1 }$$

where v_e is the e-folding velocity of an exponential density profile, and q is a form factor that describes the distribution of ⁵⁶Ni in the ejecta (q = 1/3 corresponds to an evenly distributed ⁵⁶Ni profile). In Equation (1), the γ-ray opacity has been assumed to be 0.025 cm² g⁻¹.

For SN 2019yvq we estimate ${t}_{0}=42.0\pm 1.0$ days (the uncertainty is dominated by the uncertainty on the rise time, for which we adopt 1 day). Assuming ${v}_{e}=3000\pm 180$ km s⁻¹ and that q = 0.45 ± 0.08, we find ${M}_{\mathrm{ej}}=1.33\pm 0.27\,{M}_{\odot }$. Unlike our estimate of ${M}_{{56}_{\mathrm{Ni}}}$, the adopted NIR correction does affect the measurement of t₀. In addition to the total luminosity, Figure 4 shows the SED-integrated luminosity assuming no NIR flux (i.e., flux = 0 for all λ > 1 μm). From this light curve we estimate t₀ = 38.1 ± 1.0 days, corresponding to ${M}_{\mathrm{ej}}\,=1.10\pm 0.22\,{M}_{\odot }$. Given the overall uncertainty in the NIR correction, our observations broadly bracket the total mass of ejecta to be somewhere between ∼0.9 and 1.5 ${M}_{\odot }$.

While the rise time and power-law indices of SN 2019yvq are similar to other normal SNe Ia (see Section 4.1), the full photometric evolution does not resemble a normal SN Ia. The photometric evolution of SN 2019yvq is highlighted in Figure 5, where SN 2019yvq is compared to 121 normal SNe Ia from Yao et al. (2019).³⁶ SN 2019yvq is somewhat underluminous (${M}_{g,\max }\approx -18.5$ mag), declines rapidly $[{\rm{\Delta }}{m}_{15}(g)={1.30}_{-0.02}^{+0.01}$ mag, uncertainties represent the 68% credible region], and does not exhibit a "shoulder" in the r_ZTF or a strong secondary maximum in the i_ZTF light curves post maximum. The slightly underluminous peak and moderately fast decline of SN 2019yvq are very similar to SN 1986G-like SNe Ia, which represent a transitional group between normal SNe Ia and the underluminous SN 1991bg-like class (e.g., Taubenberger 2017 and references therein). While the photometric evolution of SN 2019yvq is similar to 86G-like SNe, we show that SN 2019yvq is spectroscopically distinct from these transitional SNe (Section 5.3).

Zoom In Zoom Out Reset image size

We also find that standard SN Ia fitting techniques do not provide good matches to the evolution of SN 2019yvq. For example, a SNooPY (Burns et al. 2011) fit to the optical light curve requires significant host-galaxy extinction $[E{(B-V)}_{\mathrm{host}}\approx 0.4$ mag, see Section 3] to match the observed red colors, while predicting a secondary maximum in the i_ZTF-band and a fast evolution after peak that is not seen in SN 2019yvq. A SALT2 (Guy et al. 2007) fit leads to similar inconsistencies to those in SNooPy. These inconsistencies support our conclusion above that the photometric evolution of SN 2019yvq does not match normal SNe Ia.

SN 2019yvq is further distinguished from normal SNe Ia by its unusual color evolution (Figure 6). The lower panel of Figure 6 shows the g_ZTF−r_ZTF evolution of 62 spectroscopically normal SNe Ia with ZTF observations within 5 days of t_fl (see Bulla et al. 2020), with the color evolution of SN 2019yvq over-plotted. The initially blue colors in SN 2019yvq rapidly evolve to the red over the first few days of observation before gradually becoming bluer in the time leading up to ${T}_{B,\max }$ (this behavior is deemed an early "red bump" in Bulla et al. 2020). Similar red bumps are only seen in 6 of the 62 normal SNe Ia (∼10%) in the ZTF sample (Bulla et al. 2020), and they are typically less pronounced than what is observed in SN 2019yvq.

Zoom In Zoom Out Reset image size

Furthermore, while normal SNe Ia exhibit a large scatter in g_ZTF−r_ZTF shortly after t_fl they evolve to form a tight locus between ∼10 and 30 days after t_fl. SN 2019yvq is redder at peak than each of the normal SNe Ia in the Bulla et al. (2020) sample. Post maximum, only one normal SN Ia, ZTF 18abeegsl (SN 2018eag), exhibits a similarly rapid decline in g_ZTF–r_ZTF color. The g_ZTF−r_ZTF color evolution of SN 2019yvq is again intermediate between normal SNe Ia and underluminous 91bg-like SNe. Figure 6 shows that normal SNe Ia are reddest at ∼+30 days, while 91bg-like SNe are reddest between ∼+10 and 15 days (Burns et al. 2014). SN 2019yvq reaches a g_ZTF−r_ZTF maximum at an intermediate time of ∼+20 days.

The offset in the g_ZTF−r_ZTF color evolution of SN 2019yvq relative to normal SNe Ia would be reduced if the reddening toward SN 2019yvq has been significantly underestimated. A color excess of $E(B-V)\approx 0.25$ mag, rather than the 0.05 mag adopted in Section 3, would roughly align the g_ZTF–r_ZTF color of SN 2019yvq with the tight locus between ∼5 and 20 days after t_fl seen in Figure 6. Such a correction would also bring the peak optical brightness of SN 2019yvq in line with normal SNe Ia [for $E(B-V)\approx 0.25$ mag, ${M}_{g}\,\approx -19.25$ mag and ${M}_{r}\approx -19.1$ mag for SN 2019yvq].

While the spectral appearance of SN 2019yvq is similar to some normal SNe Ia (see Section 5.3), the observed rapid decline in the g_ZTF filter provides strong evidence that SN 2019yvq is not a normal luminosity SN Ia. Phillips (1993) showed that in the optical SNe Ia follow a brightness–width relation, whereby brighter explosions decline less rapidly. Thus, with a typical peak in the optical, as would be implied with $E(B-V)\approx 0.25$ mag, the fast decline in SN 2019yvq $[{\rm{\Delta }}{m}_{15}(g)=1.3$ mag] would be largely unprecedented.³⁷ This conclusion is further corroborated by the rapid evolution of the g_ZTF−r_ZTF color to the red after ${T}_{B,\max }$ and the lack of a secondary maximum in the i_ZTF-band, each of which is typical of lower luminosity SNe Ia (see Taubenberger 2017, and references therein). We therefore conclude that the color excess toward SN 2019yvq is not underestimated, and that the SN is instead intrinsically red in the optical.

Even if one ignores the striking initial bump in the light curve of SN 2019yvq, we can still conclude that SN 2019yvq is not a normal SN Ia based on its other photometric properties (e.g., relatively faint peak optical brightness, moderately fast decline, lack of an NIR secondary maximum, and red appearance at peak).

To determine the structure of the ejecta and relative contributions of different ions at early and maximum-light phases, we modeled the spectra at −14.9, −12.0, and +0.0 days using the one-dimensional (1D) Monte Carlo radiative transfer code TARDIS (Kerzendorf & Sim 2014). We note that TARDIS assumes a single, sharp photosphere between the optically thick and thin regions. Therefore, if there is a contribution to the spectrum from an underlying quasi-blackbody (at early times this could be due to interaction, for example; see Section 6.1), this will impact the ability of TARDIS to fully reproduce the observations. Nevertheless, our models should provide a reasonable approximation of the plasma state within the ejecta. The parameters of our TARDIS models are given in Table 4.

The first spectrum of SN 2019yvq at −14.9 days (2.6 days after t_fl, 3.0 days after the TARDIS-inferred t_exp) shows shallow features consistent with IMEs moving at extremely high velocities (>20,000 km s⁻¹, Figure 2). The best-fitting TARDIS model is shown in Figure 7, along with the contribution of individual elements to the spectral features. For this model, we have assumed a uniform composition of O, Mg, Si, and S. Our model demonstrates that the shallow absorption features observed at this phase can be reproduced solely by IMEs (predominantly Si ii), and that the presence of IGE is not required to match the data. Our model also confirms the high velocities of the ejecta—we find the spectral features and temperature are best reproduced with a photospheric velocity of ∼25,000 km s⁻¹.

Zoom In Zoom Out Reset image size

Similarly, for the −12.0 day spectrum we find that a model that does not contain IGE above ∼16,500 km s⁻¹ reproduces the majority of the spectroscopic features. Again, our model contains a uniform composition of O, Mg, Si, and S, and is shown in Figure 7. At this phase the model suggests the photospheric temperature has not significantly changed, however, the features have become much more prominent. Compared to modeling of the spectroscopically similar SN 2002bo (see Section 5.3) at the same epoch (Stehle et al. 2005), we find SN 2019yvq has a lower photospheric temperature (∼8000 K, compared to ∼9500 K for SN 2002bo).

While the early spectra of SN 2019yvq are dominated by IMEs, there is no evidence in the observed spectra for C ii absorption. However, in our TARDIS models even if we increase the C abundance in the outer ejecta to large amounts (50%), the model spectra still lack any significant C ii features at the time of our observations. Our ability to detect C ii in the spectra of SN 2019yvq is likely affected by the extremely high ejecta velocities, which leads to blending with Si ii. Therefore, despite the lack of a C ii detection in the observed spectra, we are unable to place meaningful constraints on the C abundance in the very outermost ejecta.

Given that our +0 day maximum-light spectrum occurs 12 days after our previous model spectrum, we assume a composition for the inner ejecta (<16,500 km s⁻¹) similar to that found for SN 2002bo (Stehle et al. 2005). A more detailed ejecta structure could be achieved through modeling additional pre-maximum spectra, but is beyond the scope of the work presented here. As shown in Figure 7, our model is able to broadly reproduce many of the features observed. Notable exceptions include the features around ∼4200 and 4900 Å, which we attribute to Fe. Better spectroscopic agreement could potentially be achieved if SN 2019yvq had a lower abundance of IGE within the inner ejecta relative to SN 2002bo.

Overall, our TARDIS modeling demonstrates that SN 2019yvq is consistent with a low (or zero) fraction of IGE in the outer ejecta (i.e., there is little mixing in the SN ejecta). Additionally, the earliest phases show little change in temperature (see Table 4), as expected from the color evolution.

We have measured the velocity of the Si ii λ6355 absorption feature following the procedure in Maguire et al. (2014, see their Section 2.5). We have also estimated the pseudo-equivalent widths (pEWs) of the Si ii λλ5972, 6355 features, allowing us to measure their ratio, also known as ${ \mathcal R }$(Si ii); see Hachinger et al. (2008) for the updated definition relative to Nugent et al. (1995).

The velocity evolution of Si ii λ6355 is shown in Figure 8, compared to measurements for the Palomar Transient Factory (PTF) SN Ia sample from Maguire et al. (2014) and the median velocity evolution of SNe Ia belonging to the four different classes (Shallow Silicon, Core Normal, Broad Line, and Cool) identified in Branch et al. (2006);³⁸ hereafter, the Branch class. The Si ii λ6355 velocity in SN 2019yvq is ∼15,000 km s⁻¹ around ${T}_{B,\max }$.

Zoom In Zoom Out Reset image size

At ${T}_{B,\max }$, the pEW measurements for the Si ii λ6355 and λ5972 features are 183 ± 1 Å, and 13 ± 2 Å, respectively, unambiguously classifying SN 2019yvq as a Branch Broad Line SN Ia. SN 2019yvq stands out in Figure 8 with some of the highest Si ii velocities that have ever been observed. Within the PTF sample, only SN 2010jn (PTF 10ygu) exhibits a Si ii absorption velocity as high as SN 2019yvq at every phase in its evolution. Furthermore, we also find that the Ca ii infrared triplet velocities are high (we first detect this feature in the −3.0 day SEDM spectrum; see Table 3), with a photospheric component velocity of ∼13,200 km s⁻¹ and a clear high-velocity component at ∼23,500 km s⁻¹. Within the PTF sample only one other SN (PTF 09dnp) has a Ca ii high-velocity component with a similarly large velocity at the same phase.

As first noted by Nugent et al. (1995), and later confirmed by Hachinger et al. (2008), ${ \mathcal R }$(Si ii) is a luminosity indicator, with larger values of ${ \mathcal R }$(Si ii) corresponding to lower luminosities. This correlation is driven by the ionization balance of Si ii/Si iii, with cooler objects having stronger Si ii λ5972 features. In Figure 9 we show the evolution of ${ \mathcal R }$(Si ii) as a function of time for SN 2019yvq, compared to SN 2011fe, SN 2002bo, SN 2017erp, and five SNe with multiple measurements over a long baseline from the PTF SN Ia spectral sample. Figure 9 shows that most SNe Ia have a relatively flat evolution in ${ \mathcal R }$(Si ii) in the time leading up to ${T}_{B,\max }$ (see also Riess et al. 1998). SN 2019yvq and SN 2002bo, however, feature a very different evolution with initially large values of ${ \mathcal R }$(Si ii) that rapidly decrease to very low values between ∼10 and ∼5 days before ${T}_{B,\max }$.

Zoom In Zoom Out Reset image size

At face value, the ${ \mathcal R }$(Si ii) evolution in Figure 9 suggests that the effective temperature of SN 2019yvq increases significantly as it rises to maximum light. Both the optical colors, which become bluer in the ∼14 days leading up to ${T}_{B,\max }$ (see Figure 6), and the TARDIS modeling (see Table 4) confirm an increase in temperature over the period in which ${ \mathcal R }$(Si ii) decreases. While the UV−optical colors become redder over the same time period, this is likely due to the increasing IGE fraction, and hence increased UV blanketing, as the photosphere recedes (see Section 5.1), and not a decrease in temperature.

This behavior is similar to, though less extreme than, SN 2002bo, which increases in temperature from ∼9500 K at −12.9 days to ∼14,000 K at maximum light (Stehle et al. 2005). The maximum-light temperature of SN 2002bo is similar to Branch Core Normal SNe, such as SN 2011fe, which typically have temperatures of ∼14,500–15,000 K at maximum light (Mazzali et al. 2014).

Benetti et al. (2004) interpreted these competing effects as the result of significant Si ii mixing in the ejecta of SN 2002bo. Mixing or synthesized Si in the outermost layers of the ejecta would (i) lead to larger Si ii velocities, (ii) produce Si ii line ratios that indicate cool temperatures (because there is less radioactive material to heat the ejecta), before eventually (iii) producing low values of ${ \mathcal R }$(Si ii) as the photosphere recedes through the ejecta to higher temperature regions. This picture is consistent with the Stehle et al. (2005) models of SN 2002bo. In those models, Si completely dominates the species at velocities above ∼23,000 km s⁻¹, while there is very little (∼1%) IGE above 1.35 ${M}_{\odot }$ in radial mass coordinates. A similar physical scenario likely explains the changes in Si ii absorption seen in SN 2019yvq. Although the temperature change in SN 2019yvq is less dramatic than in SN 2002bo, this may reflect slight differences in the ejecta composition as we find SN 2019yvq is consistent with no IGEs in the outer layers of the SN ejecta.

In Figure 10, we compare the spectral evolution of SN 2019yvq to two Branch Broad Line SNe Ia, SN 2002bo and SN 2010jn, and two Branch Cool SNe Ia, SN 1986G and SN 2004eo (Cristiani et al. 1992; Benetti et al. 2004; Pastorello et al. 2007; Silverman et al. 2011; Hachinger et al. 2013; Maguire et al. 2014) at four phases, pre-maximum, maximum, ∼1 week post maximum, and ∼6 weeks post maximum. The evolution of SN 2019yvq and SN 2002bo is remarkably similar at all phases. The only significant difference between the two is the absorption trough at ∼4800 Å in the pre-maximum and maximum-light spectra. This feature, which is typically attributed to a combination of Fe ii, Fe iii, and Si ii, is extremely narrow in SN 2019yvq. This is in agreement with the TARDIS modeling results where no Fe is required in the outer ejecta of SN 2019yvq to match the observed spectra at early times. SN 2010jn, which exhibits large Si ii velocities like SN 2019yvq, shows weaker IME absorption and stronger IGE absorption than SN 2019yvq. While the Branch Cool SNe 1986G and 2004eo show lower velocities than SN 2019yvq, there is strong agreement in the relative Si ii line strengths of SN 1986G and the earliest spectra of SN 2019yvq.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The maximum-light spectra shown in the second panel of Figure 10 reveal a higher temperature for SN 2019yvq, as the Si ii λ5972 absorption has nearly disappeared around ${T}_{B,\max }$ [see discussion of ${ \mathcal R }$(Si ii) in Section 5.2]. This increase in temperature is consistent with the change in optical colors (Figure 6) and TARDIS spectral modeling (Section 5.1). The appearance of SN 2019yvq, SN 2002bo, and SN 2010jn are all similar at this epoch, with the exception of the 4800 Å feature mentioned above. SN 2004eo has a similar appearance to SN 2019yvq, though it has lower velocities and cooler temperatures (as traced by Si ii λ5972).

The $+9.2$ day spectrum of SN 2019yvq, shown in the third panel of Figure 10, shows absorption due to IGE. Additional differences between SN 2019yvq and SN 2002bo can be seen at this phase. There is stronger absorption in SN 2019yvq blueward of Ca ii H and K, and the S ii "W" absorption feature is still present in SN 2019yvq and it cannot be identified in SN 2002bo or SN 2010jn. SN 2004eo maintains an appearance that is somewhat similar to SN 2019yvq, though as before, the temperature [as indicated by ${ \mathcal R }$ (Si ii)] and velocities are lower.

Spectra obtained ∼6 weeks after maximum light are shown in the fourth panel of Figure 10. By this time, as the SNe are transitioning into a nebular phase, the appearance of each spectrum is similar modulo some minor differences in the relative line strengths of different features.

For SD progenitors of SNe Ia, the SN ejecta will shock on the surface of the nondegenerate companion giving rise to a short-lived transient in the days after explosion. Kasen (2010) provided models for the appearance of this interaction, which is primarily dependent upon the binary separation of the system (assuming Roche lobe overflow for the nondegenerate companion). The observed emission following the ejecta-companion collision is dependent upon the orientation of the binary system relative to the line of sight (Kasen 2010).

An analytic formulation for the luminosity and effective temperature of the emission from the companion shock is given in Equations (22) and (25) in Kasen (2010). Brown et al. (2012) provide an analytic function to approximate the fractional decrease in the observed flux as a function of the orientation of the system. We combine the equations from Kasen (2010) and Brown et al. (2012) to model the early emission from SN 2019yvq as an ejecta-companion collision. We assume the interaction emits as a blackbody, and that the electron scattering opacity ${\kappa }_{e}=0.2$ cm² g⁻¹ (as in Kasen 2010). Assuming ${z}_{\mathrm{SN}}=0.0094$, $E{(B-V)}_{\mathrm{MW}}=0.018$ mag, and $E{(B-V)}_{\mathrm{host}}=0.032$ mag, we compare observed flux measurements with those predicted by the Kasen (2010) model at epochs with MJD < 58,849.2 (i.e., the first ∼2.5 days after discovery when emission from the companion interaction is significantly brighter than the luminosity due to radioactive decay).³⁹ The model parameters, summarized in Table 5, are constrained via a Gaussian likelihood and flat priors using an affine-invariant (Goodman & Weare 2010) Markov Chain Monte Carlo ensemble sampler (Foreman-Mackey et al. 2013).

Table 5.  Model Parameters Θ and Their Priors and Posteriors

Θ	Description	Prior	Posterior
a	Companion separation	${ \mathcal U }({10}^{10},{10}^{13})$	$(9.1\pm 0.7)\times {10}^{11}$ cm
Mej	Ejecta mass	${ \mathcal U }(0.6,1.5)$	$1.1\pm 0.3\,{M}_{\odot }$
vej	Ejecta velocity	${ \mathcal U }(5\times {10}^{8},3\times {10}^{9})$	$2.2{\pm }_{0.3}^{0.5}\times {10}^{9}$ cm s−1
θobs	Binary viewing angle	${ \mathcal U }(0,180)$	40 ± 28°
texp	Time of explosion	${ \mathcal U }({t}_{0}-5,{t}_{0})$ a	58,845.82 ± 0.04 (MJD)

Notes. Marginalized 1D posterior values represent the 68% credible region. M_ej is largely unconstrained by the observations. The posterior distribution on θ_obs is effectively flat between 0° and ∼70°, and ∼0 for all angles above ∼85°. There is a strong covariance between a and t_exp and also between v_ej and θ_obs.

^at₀ is the time of the first ZTF observation (MJD = 58,846.469942).

Download table as:  ASCIITypeset image

The results of this procedure are shown in Figure 11, where it is clear that the model presented in Kasen (2010) does an adequate job of explaining the early UV/optical emission from SN 2019yvq (constraints on the model parameters are reported in Table 5).

While the interaction models roughly approximate the SN emission in the ∼3 days after explosion, they significantly overestimate the flux immediately after the fitting window as shown in Figure 11. This problem is exacerbated by the fact that the models do not include emission associated with radioactive decay, meaning the true discrepancy between what is predicted and what is observed is even larger than what is shown in Figure 11. If we extend the fitting window to include the optical observations obtained ∼13.7 days before ${T}_{B,\max }$, the interaction models still overpredict the optical flux at this epoch. This overprediction of the optical flux poses a challenge for the companion-interaction scenario; an inability to simultaneously match both UV and optical observations has been noted for other SNe Ia with early bumps or linear rises (Hosseinzadeh et al. 2017; Miller et al. 2018).

Kasen (2010) notes several assumptions and approximations in the derivation of the equations used to estimate the emission from the companion shock. It is possible that the inclusion of more detailed physics, or additional complexity in the analytic formulation of the models,⁴⁰ could better reconcile companion-interaction models with SN 2019yvq. Such improvements are beyond the scope of this paper, leading us to explore other explanations for the early flash.

Following arguments from Kromer et al. (2016), the evolution of SN 2019yvq after the UV flash also poses a challenge to the companion-interaction scenario. In the SD scenario the WD explodes at, or very near, the Chandrasekhar mass. The leading mechanism for such an explosion is the delayed detonation scenario, in which the burning starts as a subsonic deflagration and later transitions to a supersonic detonation (Khokhlov 1991). This scenario was explored in detail via numerous 3D explosion models in Seitenzahl et al. (2013), with radiative transfer calculations of the resulting emission presented in Sim et al. (2013). While the faintest models presented in Sim et al. (2013) have a similar luminosity at peak as SN 2019yvq, they feature Si ii velocities that are significantly lower than SN 2019yvq. The Sim et al. models with high Si ii velocities are far too luminous to explain SN 2019yvq. In addition to the delayed detonation scenario, Chandrasekhar mass WDs can explode as pure deflagrations. While the ⁵⁶Ni yield and peak luminosity of pure deflagrations is more in line with SN 2019yvq than delayed detonation explosions, pure deflagrations result in low expansion velocities and relatively weak Si ii absorption (e.g., Fink et al. 2014) meaning they too provide a poor match to SN 2019yvq.

SN 2018oh was observed with an exquisite 30 minute cadence by the Kepler spacecraft and showed a clearly delineated linear rise in flux followed by a "standard" t² power law ∼4 days after t_fl. Models with extended clumps of ⁵⁶Ni just below the WD surface have been proposed as a possible explanation for the initial linear rise in SN 2018oh (Dimitriadis et al. 2019; Shappee et al. 2019). The models considered in Shappee et al. (2019) and Dimitriadis et al. (2019), which build on the work of Piro & Morozova (2016), only cover the first ∼10 days after explosion and assume relatively simple gray opacities. To further investigate this possibility, Magee & Maguire (2020) recently performed more detailed radiative transfer calculations for SNe Ia with extended clumps of ⁵⁶Ni. They then compared these models to SN 2018oh and SN 2017cbv, another event with a clearly resolved bump in the early light curve (Hosseinzadeh et al. 2017).

For SN 2019yvq we follow the procedure in Magee & Maguire (2020) to model the early flash and rise of the SN. As described in the beginning of Section 6, we generate a "baseline" model that replicates the rise of SN 2019yvq after the first two epochs of ZTF optical detections. Following the generation of this "baseline" model, we add clumps of ⁵⁶Ni to the outer layers of the SN ejecta, and perform full radiative transfer calculations using TURTLS (Magee et al. 2018).

We find that a model with a 0.02 ${M}_{\odot }$ clump of ⁵⁶Ni adequately matches the early optical evolution of SN 2019yvq in the g_ZTF- and r_ZTF-bands, as shown in Figure 12. The model flux in the i_ZTF-band is overestimated, however, meaning the model is redder than what is observed. In Figure 12, the Ni-clump models have been offset by −0.1 mag to better match the observations. This offset is necessary as the Ni clump provides some blanketing around maximum light, and the parameter space is too large to simultaneously optimize both the central Ni mass and the clump Ni mass.

While a clump of ⁵⁶Ni can produce an optical bump in the light curve, the same challenges identified in Magee & Maguire (2020) apply to SN 2019yvq. In particular, an extended clump of ⁵⁶Ni dramatically alters the appearance of the SN at maximum light. Figure 12 shows a comparison of the observed spectra with our calculated models. The Ni-clump models feature strong blanketing in the blue-optical region of the spectrum, which is simply not present in the observed spectra of SN 2019yvq. We therefore conclude that Ni clumps cannot explain the early flash seen in SN 2019yvq.

WDs that accrete a thin shell of He can explode via a "double detonation" whereby explosive burning in the He shell drives a shock into the C/O core of the WD. This shock can ignite explosive C burning and trigger a detonation that disrupts the entire star (e.g., Nomoto 1982a, 1982b; Woosley & Weaver 1994). Such explosions are even possible in C/O WDs that are well below the Chandrasekhar mass (see Fink et al. 2007, 2010 and references therein).

Recent models of double-detonation explosions presented in Polin et al. (2019a) show that such explosions can replicate several of the peculiar properties of SN 2019yvq, including the early UV/optical flash, a blue to red to blue color transition, the moderately faint optical peak, red colors at maximum, and a lack of IGE in the early spectra.

The appearance of double-detonation SNe is effectively determined by two properties: the mass of the C/O core and the mass of the He shell. The total mass of the system determines the central density of the WD and thus the amount of synthesized ⁵⁶Ni. The ⁵⁶Ni mass directly controls both the peak luminosity and the kinetic energy of the explosion. High mass WDs ($\gtrsim 1.1\,{M}_{\odot }$) create enough ⁵⁶Ni (${M}_{\mathrm{Ni}}\gtrsim 0.5\,{M}_{\odot }$) to produce large (≳14,000 km s⁻¹) photospheric velocities and reach normal brightness for an SN Ia, while low mass WDs ($\lesssim 0.9\,{M}_{\odot }$) exhibit lower photospheric velocities (≲10,000 km s⁻¹) and produce less ⁵⁶Ni, therefore peaking at fainter luminosities (Polin et al. 2019a). That we see both a high Si ii velocity and a low peak luminosity in SN 2019yvq presents a challenge for the Polin et al. (2019a) double-detonation models (see their Figure 11). Furthermore, thick He shells (${M}_{\mathrm{He}}\gtrsim 0.05\,{M}_{\odot }$) produce more pronounced UV/optical flashes shortly after explosion, particularly in conjunction with lower mass WDs, while thin He shells (${M}_{\mathrm{He}}\lesssim 0.02\,{M}_{\odot }$) produce a more extreme color inversion in the days after explosion.

We have attempted to model the evolution of SN 2019yvq as a double-detonation explosion, following the procedure in Polin et al. (2019a). We have specifically focused on matching the photometric evolution (as noted above no models create high-velocity ejecta and underluminous optical peaks), with particular attention to the colors during the early flash and at maximum light. We find that a model with ${M}_{{\rm{C}}/{\rm{O}}}=0.92\,{M}_{\odot }$ C/O core and a ${M}_{\mathrm{He}}=0.04\,{M}_{\odot }$ He shell best match SN 2019yvq, as shown in Figure 13.

Zoom In Zoom Out Reset image size

While this model adequately matches the evolution of SN 2019yvq in the r_ZTF filter, the predictions in the g_ZTF- and i_ZTF-bands do not match what is observed. We show for the first time that there is an expected UV flash associated with these double-detonation models, however, our best-fit model underestimates the flux that was observed in the UV.

Synthesized spectra from our double-detonation model exhibit features that are not seen in SN 2019yvq. The model spectra are dominated by Si ii absorption, and show high-velocity absorption due to O i and Ca ii, similar to SN 2019yvq. For our best-fit model, however, the Si ii velocities are too slow, the Si ii λ5972 absorption is too strong, and the S ii absorption is too weak. Nuclear burning in the He shell creates heavy elements in the outermost ejecta of double-detonation explosions, leading to deep Ti ii troughs and other blanketing in the blue-optical portion of the spectrum. Our model exhibits a strong Ti ii absorption trough blueward of ∼4400 Å (see the ${t}_{\exp }+9.25$ day spectrum in Figure 13). As was the case for models with extended clumps of ⁵⁶Ni, the lack of such absorption in SN 2019yvq poses a challenge for the double-detonation model.

With observations that probe a previously unexplored phase in the evolution of such explosions, SN 2019yvq provides an opportunity to determine where the double-detonation models must improve. It is possible that such improvements could lead to better agreement with SN 2019yvq. For instance, the nuclear reaction networks and 1D models in Polin et al. (2019a) always burn the He shells to nuclear statistical equilibrium. It is not unreasonable to think that 2D or 3D models, with a more sophisticated nuclear reaction network, would create more IMEs and less IGEs in the He shell, and that the ratio of the two created in the shell could be highly dependent upon the line of sight. For example, Townsley et al. (2019) modeled the explosion of a ${M}_{{\rm{C}}/{\rm{O}}}=1.0\,{M}_{\odot }$ C/O core with a ${M}_{\mathrm{He}}=0.02\,{M}_{\odot }$ He shell and found a higher ratio of IME to IGE than the analogous 1D model presented in Polin et al. (2019, though see also Gronow et al. 2020, which presents a 3D double-detonation explosion with nuclear yields that are only mildly different from 1D models). This could explain the lack of IGEs and strong Si ii absorption seen in the early spectra, while less IGEs in the outer layers would also reduce some of the line blanketing seen around maximum light. This would lead to less reprocessing of blue photons, perhaps creating better agreement between the models and photometry, particularly in the g_ZTF-band. The velocity discrepancy could also potentially be explained as a line-of-sight effect. If the ignition of the WD occurred off center, then the ejecta aligned with the site of the initial He ignition may receive a boost in velocity (e.g., Kromer et al. 2010). The discrepancies in the UV are less worrisome. While we show a qualitative UV flash, the magnitude of this flash will be highly sensitive to the precise temperature and composition in the very outermost ejecta, and thus any of the changes discussed above could easily boost the model flux in the UV.

Piro & Morozova (2016) show that circumstellar material in the vicinity of a WD at the time of explosion can give rise to an early flash or bump in the SN Ia light curve. Using a 1D toy model, with an assumed circumstellar density profile ∝r⁻³ and gray opacities, Piro & Morozova (2016) found that the peak of the early emission is roughly proportional to the extent of the circumstellar material, while the duration of the flash is proportional to the square root of the circumstellar mass. While the brightest model from Piro & Morozova (2016) has a flash brightness that peaks at ${M}_{V}\approx -15$ mag, circumstellar material that extends beyond ∼10¹² cm could give rise to a flash that peaks at ${M}_{g}\lesssim -16.4$ mag, as is observed in SN 2019yvq.

There are few proposed WD explosion models that produce dense circumstellar material in the vicinity of the WD at the time of explosion. A notable exception is the violent merger, so called because the thermonuclear explosion happens while the merger is still ongoing, of two C/O WDs (Pakmor et al. 2010, 2011, 2012). DD mergers should produce a wide variety of circumstellar configurations, depending on the initial parameters of the inspiralling binary, which would, in turn, produce different signals shortly after explosion (e.g., Raskin & Kasen 2013; Levanon & Soker 2019).⁴¹

Given the vast parameter space populated by different circumstellar configurations, we are going to proceed under the (potentially poor) assumption that such interaction could reproduce the UV/optical flash seen in SN 2019yvq. Following this assumption, a relevant question is—can violent mergers reproduce the properties of SN 2019yvq in the days before and weeks after ${T}_{B,\max }$?

In Kromer et al. (2016), the violent merger of two C/O WDs with masses of 0.9 and 0.76 ${M}_{\odot }$ produced a similar rise and maximum-light properties to iPTF 14atg, the other SN Ia with an observed early UV flash. A comparison of SN 2019yvq to the low-metallicity model from Kromer et al. (2016), which provides a good match to iPTF 14atg, is shown in Figure 14.⁴² We show that model here to illustrate the qualitative behavior of such a merger; it is not meant to provide an optimal match to SN 2019yvq. The Kromer et al. model was not designed to fit the early UV flash in iPTF 14atg.

Zoom In Zoom Out Reset image size

The photometric evolution of this violent merger model qualitatively matches SN 2019yvq: (i) a moderately faint peak in the optical ($-17.6\,\mathrm{mag}\gtrsim {M}_{g}\gtrsim -18.2$ mag, depending on the viewing angle), (ii) red g − r colors at peak, and (iii) a lack of a secondary maximum in the i-band. Furthermore, the spectra lack significant IGE absorption in the days after explosion (right panel of Figure 14), as is observed in SN 2019yvq. Interestingly, the violent merger model does show a decrease in the relative strength of the Si ii λ5972 absorption with time, similar to SN 2019yvq and unlike the other models considered here. A critical difference between SN 2019yvq and violent merger models, is that the merger models tend to produce relatively low expansion velocities (e.g., Pakmor et al. 2010; Kromer et al. 2013, 2016). Indeed, this is one of the stark differences between SN 2019yvq and iPTF 14atg, as iPTF 14atg had a Si ii λ6355 absorption velocity of ∼7500 km s⁻¹ at peak, or roughly half that observed in SN 2019yvq. It is also clear from Figure 14 that the violent merger model from Kromer et al. (2016) exhibits weaker IME absorption than what is seen in SN 2019yvq.

It is clear that additional modeling, likely of a different WD binary configuration, is needed to better match SN 2019yvq. For example, it is known that a higher mass primary WD can produce more ⁵⁶Ni, and hence a brighter optical peak (e.g., Pakmor et al. 2012), which would be more in line with SN 2019yvq. If, at the same time, the mass of the secondary were slightly decreased, then the kinetic energy of the ejecta would increase, perhaps bringing the model velocity of Si ii and other IMEs in line with SN 2019yvq. It would also be beneficial to track the unbound material following the DD merger, to see if the collision between this material and the SN ejecta can replicate the early UV/optical flash seen in SN 2019yvq. If this feature can readily be recreated, it is possible that a violent merger is responsible for SN 2019yvq.