Peculiar Spectral Evolution of the Type I Supernova 2019eix: A Possible Double Detonation from a Helium Shell on a Sub-Chandrasekhar-mass White Dwarf

SN 2019eix was discovered by ATLAS on 2019 May 1:59 UT, using the instrument ACAM1 at Haleakalā with a discovery magnitude of 17.0 in the cyan filter (Tonry et al. 2019). The last non-detection of the same object was on 2019 April 21:63 UT with a magnitude of 18.39 (orange filter) from the same telescope. A classification report was provided about a week after its discovery (2019 May 7), with a classification spectrum taken on 2019 May 5 with the 3 m Lick Shane reflector telescope suggesting that SN 2019eix was a Type Ic SN (Dimitriadis 2019). Due to the lack of early photometry, the explosion epoch 2458600 JD was adopted throughout the paper assuming a rise time of 16 days to its maximum light in the r band. A typical Ic rise time falls between 13 and 16 days (Taddia et al. 2017) and SN 2019eix appears to be on the wide side for SNe Ic as shown in Figure 6; thus, the choice for 16 days (and the fact that SN 2019eix has similar light curves to SNe Ic). The classification spectrum was downloaded from the Transient Name Server (TNS) ⁹ and was included in our analysis. The SN is located at R.A. 18^h42^m42^s89 and decl. $+40^\circ 22\buildrel{\,\prime}\over{.} 07\buildrel{\prime\prime}\over{.} 8$, and is situated in the host-galaxy NGC 6695 with a redshift of z = 0.018303 (Corrales 2015). In Figure 1, we show an image of the field of view, location, and host galaxy of SN 2019eix in the gri bands taken with the 1 m telescope of Las Cumbres Observatory (LCO; Brown et al. 2013) on 2019 May 4. The distance of the galaxy NGC 6695 is ≈83.80 Mpc according to the NASA Extragalactic Database (NED) ¹⁰ where the Tully–Fisher method was used to calculate the luminosity distance (Theureau et al. 2007). The Milky Way extinction value E(B − V) = 0.0604 was adopted from the Schlafly & Finkbeiner (2011) calibration of the Schlegel et al. (1998) dust maps.

Zoom In Zoom Out Reset image size

After the discovery, the Global Supernova Project obtained photometry in the BVgri bands using the 1 m telescopes of LCO. We reduced the data using lcogtsnpipe (Valenti et al. 2016). Point spread function fitting photometry was performed and the zero-points were calculated from the Landolt standard catalog (Landolt 1992) for the BV filters. Additionally, the gri filter zero-points were calculated by incorporating the Sloan magnitudes of field stars (Albareti et al. 2017). Besides LCO data, the public Zwicky Transient Facility (ZTF) ¹¹ data including their upper limits in the g and r bands, were used for comparison and are plotted in Figure 2. Note that all of the photometric data are available in Table 8.

Zoom In Zoom Out Reset image size

Additional photometric information is presented in Table 1, which shows the time of maximum, the peak absolute magnitude, and the decline rates for SN 2019eix. We measured the peak magnitude by fitting a polynomial through the data points and used their maximum value. For the B filter where rising data is not available, we used the peak magnitude from the B-band data as an estimate of maximum light. A comparison of Δm₁₅ with typical SNe against SN 2019eix is shown in Table 2, where we can see that SN 2019eix's decline rates fall within the average values for SNe Ic but are somewhat faster in the r and i bands (but not unprecedented; see Taddia et al. 2017).

Table 1. Time of Maximum Light, Absolute Magnitude, and Decline Rate (Δm₁₅, the Drop in Magnitudes between Peak and 15 Days After Peak) for SN 2019eix in Each Filter of Its Light Curve

Filter	${t}_{\max }(\mathrm{JD})$	${M}_{\max }$	Δm15
B	2458609.8	−17.2	1.2
g	2458607.8	−18.12	1.0
V	2458612.9	−18.35	0.84
r	2458615.5	−18.63	0.95
i	2458616.35	−18.33	0.72

Download table as:  ASCIITypeset image

As part of the Global Supernova Project follow-up, LCO obtained a spectroscopic series using the FLOYDS spectrograph mounted on the 2 m Faulkes Telescope North in Haleakalā Hawai'i. The spectra were reduced as described in Valenti et al. (2014). We also use the classification spectrum posted in TNS (described above); the KAST 3 m Lick Shane Reflector took the spectrograph on 2019 May 5:0 UT. All spectra were corrected for Milky Way reddening (E(B − V) = 0.0604) and host corrected by measuring the equivalent width of the Na-ID doublet as described in Section 3.1 and shown in Figure 3 and its dates of relative phases are available in Table 9.

Zoom In Zoom Out Reset image size

We investigated two techniques for determining host-galaxy reddening, which produced different results as shown in Table 3. In the first method, the equivalent width of the sodium line Na ID was measured to be EW_NaID = 0.139 [Å] from IRAF's ¹² task splot, using the equation ${A}_{{\rm{V}}}^{\mathrm{host}}$ = 0.78(± 0.15) × EW_NaID [Å] from Stritzinger et al. (2018). This gives ${A}_{{\rm{V}}}^{\mathrm{host}}=0.04\pm 0.15$ and an excess value of E(B − V) of 0.013 ± 0.002 assuming a reddening law of R_v = 3.1.

The second method dealt with comparing the color curves with a template of a typical Type Ic SN (Taddia et al. 2017). We considered using this method due to the similarities with Type Ic. Host-galaxy reddening values are estimated through the comparison of observed optical and near-infrared colors to intrinsic color-curve templates constructed from subsamples of minimally reddened CSP-I stripped-envelope SNe (Stritzinger et al. 2018). The templates can be found on the Carnegie Supernova Project (CSP) webpage. ¹³ Figure 4 shows the color curve B – V of SN 2019eix and the SN Type Ic template against time. The template is not an appropriate template for the data as it does not match the shape of the curve due to its peculiarity, it is easier shown when the template is raised by 4 standard deviations to match the SN 2019eix color data. In Figure 5, it can be seen that the color evolution of SN 2019eix is inconsistent with SN Ic, meanwhile, it exhibits a notable agreement with SN 2016hnk (a DD candidate).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The color excess from the weighted difference between the two is E(B − V) = 0.96. This value is inconsistent with the one measured from the equivalent width of the Na-ID absorption feature E(B − V) = 0.013. In addition, studies have shown that the V – R for an SNe Ic is about (V − R)_V10 = 0.26 ± 0.06 mag where V₁₀ corresponds to 10 days after V-band maximum (Drout et al. 2011) in contrast SN 2019eix has a value of (V − R)_V10 = 0.3612. This points to the possibility that SN 2019eix could be intrinsically redder than a normal SN Ic. This SN is inconsistent with a reddened normal Type Ic, so we conclude that the red color is primarily from the intrinsic red color of this peculiar event. We therefore adopted the extinction value from the sodium line to correct for reddening.

Figure 6 shows the light curves of SNe Ib, Ic, and Ia (with subtypes such as 91bg and DD candidates) along with SN 2019eix in the B, r, and i bands. All SNe are Milky Way and host extinction corrected. We observe that the peak absolute magnitude and shape are most similar to Type Ic's, including SN 1999ex (Stritzinger et al. 2002) and various others in gray. SN 2016dsg (a thermonuclear He-shell detonation candidate, Dong et al. 2022) and SN 1991bg (a subluminous SN Ia, Leibundgut et al. 1993) are the closest in peak magnitude, particularly in the i band (although fainter), with faster evolving light curves than SN 2019eix. For instance, SN 2016dsg has an i-band decline rate of around 0.077(0.003) mag day⁻¹ (or Δm₁₅ = 1.15), as opposed to Δm₁₅ = 0.72 for SN 2019eix. Generally, from Figure 6 it can be seen that SNe Ib/c appear to be a better match to SN 2019eix than the thermonuclear SNe in terms of their light curves.

Zoom In Zoom Out Reset image size

The striking features of the spectral sequence of SN 2019eix are concentrated in the suppression of the blue color λ < 5000 Å, the absence of an O i feature (This is peculiar for SNe Ic which have been shown to generally have a strong O i 7774 feature; however, an exception has been observed; Fremling et al. 2018; Williamson et al. 2022), and the strong evolution of Ca ii feature. In Figure 7, we plot SN 2019eix along with various types of SNe that match spectroscopically at different epochs. We see that the early spectrum at −5 days before maximum matches SNe Ic (e.g., SN 1994I and SN 2007gr; Gal-Yam 2017) with the exception of the lack of an O i absorption feature. We also compared SN 2019eix with SN 2019ewu, a peculiar Type Ic SN that, like SN 2019eix, lacks an O i absorption feature. However, SN 2019ewu displays distinct characteristics at this epoch, including an extreme blue continuum and high absorption velocities, which are not typically observed in SNe Ic (Williamson et al. 2022). These peculiarities are outlined and discussed in greater detail in Section 3.4. We refrain from comparing SN 2019ewu at later phases to SN 2019eix due to the significant observational differences already mentioned and the reappearance of the O i feature in the SN 2019ewu spectra.

Zoom In Zoom Out Reset image size

At −5 days we see strong emission features in the blue between 4000 and 5000 Å and a less pronounced Si ii feature, much less pronounced than the 91bg-like 2005bl. Moreover, as demonstrated in Figure 7, the characteristics observed between 5500 and 6500 Å do not align with those of SNe Ic. Furthermore, as indicated in Figure 12, this particular feature is primarily attributed to Si ii. Generally, before maximum SN 2019eix looks like SN 2005bl, but the similarities are not as apparent as they are to SNe Ic.

At maximum light, the 3500–5000 Å region becomes suppressed making it peculiar compared to the average Ic. SN 2019eix appears to transition in appearance to SN 2018byg a DD candidate and to the fast-moving Type Ic PTF12gzk (although the photospheric velocity is much slower than PTF12gzk, having photospheric velocities of v_ph ≈ 10,500 and ≈15,300 km s⁻¹ at peak, respectively; Ben-Ami et al. 2012). Around 13 days after maximum, the suppression in the blue along with the absence of the O i feature persists. We observe PTF12gzk to evolve like SN 2019eix up until this epoch in terms of its general shape (initially resembling a Type Ic and evolving with suppressed features in the blue). Nevertheless, we also noticed key differences, including an O i feature, a smoother profile, and a large blueshift most apparent in the Ca ii and Si ii absorption features in PTF12gzk. Similarly, the DD candidates SN 2016hnk and SN 2005bl (a 91bg-like) also appear to have a strong resemblance (however, SN 2005bl does not have as strong of a suppression in the blue), with the exception of the O i feature.

The epochs 22 and 32 days after maximum of SN 2019eix seem to resemble 91bg-like SNe and DD Ia's as shown in Figure 7, but with even stronger and more blended features. Additionally, the Ca ii feature becomes so broad and large that it even appears to be saturated at 33 days after maximum. We also notice a weak O i feature previously not observed in the earlier spectra.

The spectral evolution presented in Figure 7 does not fully match a typical Type Ic. This is illustrated further when plotting SN 2019eix spectra with the mean spectra of SN Ic illustrated in Figure 8. The templates were used from the NYU SN group GitHub webpage ¹⁴ (Modjaz et al. 2016). The gray shaded area represents 1 standard deviation from the mean at various epochs. These standard deviation regions quantify the spectral diversity within the SN Ic class and facilitate the identification of anomalous features in SN 2019eix. From Figure 8, the early spectrum is near the mean Type Ic but not at later epochs. SN 2019eix reaches the upper limits of the standard deviation from the mean template and in some cases falls outside the 1 standard deviation regime. This along with the lack of an O i 7774 feature shows that SN 2019eix is not a typical SN Ic spectroscopically.

Zoom In Zoom Out Reset image size

To estimate the pseudo-bolometric luminosity we used the technique detailed in Howell et al. (2009). We warped spectral templates for Type Ic SN found from Peter Nugent's webpage ¹⁵ so that they matched our photometry. We integrated the warped template to get the pseudo-bolometric flux at a given epoch where we had data. A bolometric factor was applied in the end to account for the missing flux for those filters with no data and a bolometric luminosity was computed. We also made an effort to determine the bolometric luminosity using a subluminous Type Ia template. Nonetheless, the resulting luminosity showed only a slight alteration, increasing from 3.5 × 10⁴² to 3.6 × 10⁴². As a result, we decided that the estimate obtained using the Type Ic template was more appropriate and opted to use it in our analysis.

The bolometric luminosity of SN 2019eix is plotted along other SNe Types as shown in Figure 9. The luminosity for Type Ib/c SNe was acquired from the Carnegie SN project's website. ¹⁶ For the subluminous Type Ia comparison, we used the DD candidates including SN 2016hnk from Jacobson-Galan et al. (2020) and OGLE13-079 from Inserra et al. (2015); although whether SN 2016hnk is a DD or a 91bg-like SN it is still being debated (Galbany et al. 2019). For the other subluminous Type Ia case, we used SN 2005bl, a 91bg-like SN from Taubenberger et al. (2008). From Figure 9, we can see that the maximum bolometric luminosity for SN 2019eix is about 3.5 × 10⁴² erg s⁻¹ cm⁻² Å⁻¹. Comparing SN 2019eix with the various types of SNe in this plot, we can see that it is a better match to Type Ic. However, SN 2019eix is more luminous than most Type Ic, while much brighter than a subluminous SN Ia.

Zoom In Zoom Out Reset image size

An essential property of the physics of the ejecta used to measure important physical parameters, such as the ejecta and ⁵⁶Ni mass, is the photospheric velocity. One approach to estimate the photospheric velocity is by measuring the Fe ii 5018 line in the spectra. Because the Fe ii in our spectra was blended and difficult to identify as shown in Figure 10, we used the Si ii 6355 line instead. Figure 11 shows a comparison of the velocity with other types of SNe, including Ic, Ic-BL, and subluminous SNe Ia. SN 2019eix seems to have velocities slightly above a typical SN Ic and be slower evolving. The data from Figure 11 was taken from Ben-Ami et al. (2012), Taddia et al. (2016), and Gutiérrez et al. (2021) for SN Ic and SN Ic-BL. For the subluminous SN Ia, we used SN 2016hnk (calcium-rich transient from a proposed He detonation; Jacobson-Galan et al. 2020), SN 2018byg (DD He-shell Type Ia candidate; Jacobson-Galan et al. 2020), and SN 2005bl a 91bg-like SN (Taubenberger et al. 2008). It is important to notice that different absorption features can give different velocities, and in some cases, the identification of the Si ii 6355 line can be problematic (Parrent et al. 2016b). Nonetheless, from Figure 11 we can see that the Si ii 6355 velocities seem to decline at similar rates to the subluminous SNe. We also notice that Type Ic SNe decline faster and seem to have lower velocities than SN 2019eix.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Using the pseudo-bolometric luminosity calculated from Section 3.3, we fit Arnett's model presented by Arnett (1982) and Cano (2013); see Cano's Equation (1). We attempt to estimate physical parameters such as the mass of the ejecta and the mass of ⁵⁶Ni synthesized in the explosion:

$\begin{eqnarray}&&L(t)={M}_{\mathrm{Ni}}{{\rm{e}}}^{-{x}^{2}}\times \left(\left({\epsilon }_{\mathrm{Ni}}-{\epsilon }_{\mathrm{Co}}\right){\int }_{0}^{x}A(z){dz}+{\epsilon }_{\mathrm{Co}}{\int }_{0}^{x}B(z){dz}\right),\end{eqnarray} \tag{ 1 }$

where

$\begin{eqnarray}&&A(z)=2z{\exp }^{-2{zy}+{z}^{2}},B(z)=2{{ze}}^{-2{zy}+2{zs}+{z}^{2}},\end{eqnarray} \tag{ 2 }$

$\begin{eqnarray}&&{\tau }_{m}\approx {\left(\displaystyle \frac{\kappa }{\beta c}\right)}^{1/2}{\left(\displaystyle \frac{{M}_{\mathrm{ej}}}{{v}_{\mathrm{ph}}}\right)}^{1/2},\end{eqnarray} \tag{ 3 }$

and x ≡ t/τ_m , y ≡ τ_m /(2τ_Ni) and s ≡(τ_m (τ_Co - τ_Ni)/(2τ_Co τ_Ni)).

The assumptions used here are the same as the stripped-envelope case from Cano (2013). The constant opacity κ = 0.07 cm² g⁻¹ and the constant of integration β was set to 13.8. The energy released in one second per gram of ⁵⁶Ni and ⁵⁶Co are _Ni = 3.90 × 10¹⁰ erg s⁻¹ g⁻¹ and _Co = 6.78 × 10⁹ erg s⁻¹ g⁻¹ (Cappellaro et al. 1997), respectively, with decay times of τ_Ni = 8.77 days and τ_Co = 111.3 days (Taubenger et al. 2016, and references therein).

The fit is done using data prior to 30 days after explosion (the bolometric luminosity used is shown in Figure 9) using v_ph ≈ 10,500 km s⁻¹ measured from the Si ii feature and an assumed time of a maximum of 16 days. From the relationship between v_ph and the ejecta energy, the kinetic energy was calculated using ${E}_{K}/{M}_{\mathrm{ej}}=\tfrac{3}{10}{v}_{\mathrm{ph}}{({t}_{\max })}^{2}$ (Wheeler et al. 2014). Table 4 shows the resulting parameters of SN 2019eix from Arnett's fit, where the estimated ejecta mass is ≈2.5 M_⊙ and the nickel mass ⁵⁶Ni is ≈0.17 M_⊙. The rest of the values are taken from Table 8 (Taddia et al. 2017) for comparison, and we notice that they agree with a Type Ic SNe. In contrast, some of the Type Ic SNe with similar spectra had values that were more diverse, showing the extreme range of ejecta masses a Type Ic/Ic-BL can have. SN 1994I Ic had ejecta mass of M_ej = 0.5 ±0.2M_⊙ (Nicholl et al. 2015), PTF10vgv Ic-Pec had M_ej = 1.5 M_⊙, and, lastly, PTF12gzk Ic-pec had M_ej = 7.5 M_⊙ (Ben-Ami et al. 2012). SN 1994I had fast rise times and therefore less massive ejecta. On the other hand, PTF12gzk had long rise times, larger ejecta masses, and high ejecta velocities of about 30,000 km s⁻¹ days after explosion (Ben-Ami et al. 2012).

Table 4. Summary of Our Findings and Similarities between SN 2019eix in the R Band ¹

SN	${M}_{\max }^{R}$	Δm15(R)	Mej(M⊙)	MNi(M⊙)	Spiral Gal
19eix	−18.32	0.95	2.5(1.0)	0.09(0.05)	✓
Ic	−18.3(0.6)	0.73(0.27)	2.1(1.0)	0.13(0.04)	✓
Ic-BL	−19.0(1.1)	0.6(0.14)	4.1(0.9)	0.96(1.09)	×
91bg-like	−16.7 to −17.7(B)	1.8–2.1 (B)	0.5	0.05–0.1	×
Ultra-stripped (2005ek)	−17.26(0.15)	2.88(0.05)	0.3	0.03	×
Calcium-rich transient	−15.5 to −16.5	?	0.4–0.7	≈0.01	×
Ia-DD (2018byg)	−18.2	1.23	< 1.4	0.11	×

Note. The inputs with an "(B)" indicates the value in the B band as the R band is not provided and the values in parentheses are the errors.

¹ We converted r to R using the Vega–AB magnitude conversion (Blanton & Roweis 2007) and the different classes. From the table, we can see that Type Ic shares more similarities than the other classes.

Download table as:  ASCIITypeset image

From the measured parameters in Table 4 we observe that these masses (⁵⁶Ni and ejecta) and energies of SN 2019eix are more consistent with those of a typical Type Ic, since any thermonuclear event would require a lower mass than the 1.4 M_⊙ limit. Another method we used to estimate the mass of ⁵⁶Ni, is from the tail of the light curve. We used Equations (1) and (2) from Terreran et al. (2016) and fit the equations to our late photometry (past 50 days after maximum) and estimated a ⁵⁶Ni mass of ≈0.09 M_⊙. This mass is lower than the one approximated from the peak. For our purposes, we will use this value and add a 0.08 M_⊙ error estimate to include the measured value from the peak for consistency. Additionally, to test for inconsistencies we also applied the constant opacity κ = 0.1 cm² g⁻¹ commonly used for SNe Ia (Lyutykh et al. 2021). The resulting mass of ejecta slightly decreased from 2.5 to 2.1 M_⊙, bringing it closer to the Type Ia scenario, but it still remains significantly higher than the 1.4 M_⊙ threshold.

It is important to note a major limitation in applying semi-analytic modeling techniques (Arnett's model) to stripped-envelope light curves is the assumption of a constant opacity, κ, ambiguity regarding the velocity, and lack of early photometry. It was shown by Dessart et al. (2016) how different values of κ can lead to different results for the progenitor parameters, and that ultimately the assumption of constant opacity is quite poor for stripped-envelope SNe. Parrent et al. (2016b) showed that the Si ii velocity was problematic, while the Fe ii 5018 can be hard to identify as shown in Figure 10. Therefore, there is no true velocity to use as both methods come with uncertainties and the estimates of the ejecta mass highly depend on this velocity. Earlier photometry would have allowed for a more accurate rise time and as a result more accurate ejecta mass estimates from the light curve, but lacked these early observations. Additionally, we assumed no ⁵⁶Ni mixing, if SN 2019eix is a DD a significant amount of ⁵⁶Ni would be synthesized on the surface of the He shell. This is equivalent to some degree of ⁵⁶Ni mixing that could possibly affect our estimates of the ejecta and ⁵⁶Ni masses. Overall, the ⁵⁶Ni mass error is dominated by the error on the SN distance, explosion epoch estimate, the fit uncertainty, and the possible ⁵⁶Ni mixing synthesized in the He shell (if SN 2019eix is a DD).

Due to the complex nature of the spectra of SN 2019eix, we experimented with whether it was possible to reproduce some of the unusual features using an SN Ic as a base model (as originally classified). By employing TARDIS, a Monte Carlo radiative-transfer code that tracks the number of photons that propagate through the SN ejecta, we simulated the spectra of SN 2019eix. The version of TARDIS utilized in this study is TARDIS 2022.11.21 (Kerzendorf & Sim 2014), which is based on previous works (Abbott & Lucy 1985; Lucy & Abbott 1993; Mazzali & Lucy 1993; Lucy 1999, 2002, 2003). The user must input the density structure, velocity, and abundance of elements. TARDIS then solves the ionization and excitation states of the plasma assuming a homologous expansion. The evolution in the early photospheric phase is such that the photosphere can be approximated to be optically thick and thus the blackbody approximation is appropriate. Additionally, TARDIS approximates the photosphere by assuming its position in velocity space. For the simulations presented in this paper, we use the Kurucz atomic data set version 1.0 (i.e., line list) for calculating the bound-bound transitions. We also used the initial parameters listed in Table 1 from Williamson et al. (2020) and their input files for SN 1994I, which can be found on the GitHub webpage. ¹⁷

Given that SN 2019eix is inconsistent with the spectral evolution of typical Type Ic, we modeled SN 2019eix spectra using an SN 1994I (SN Ic) spectral model at day 16 as our base model. We adjusted the abundance (Hachinger et al. 2012), density, velocity, and temperature of radiation according to different epochs and set the damping constant to 0.5. We changed these parameters in order to see whether the spectral features of SN 2019eix could be reproduced, as discussed below. The epochs modeled include 4 days before maximum, 1 day after maximum, and 13 days after maximum light as shown in Figure 12. All of our models were well converged in the temperature of the radiation space.

Zoom In Zoom Out Reset image size

In the Appendix, we plot the changes made to the temperature and density profiles, and in Figure 16, we show the changes made to the abundances per epoch. Note in Figure 12 that the models fail to reproduce the O i region as they are unable to attenuate it. Attempts were made to deplete the O i feature throughout all epochs, by either raising the temperature (in an attempt to ionize the O i) or lowering the abundance in the outer layers. We found that raising the temperature to ionize the O i was unable to fully eliminate the absorption feature in the model. Additionally, the temperature increase caused the spectra to gain bluer features that did not match our observations. Lowering the abundance on the other hand (various combinations were attempted including depleting the outer shells, reducing O i in all shells, and linearly decreasing the O i in all shells) did not seem to affect the spectra as shown in Figure 12.

The phase 4 days before maximum light was assumed to be 13 days after explosion in our T ARDIS model, with an accompanying photospheric velocity of 12,200 km s⁻¹ (chosen based on the measured Si ii velocity and the velocity values that best-matched TARDIS models to observations). We used the reference model with a slight adjustment in the density and temperature, which were all increased. Additionally, Ti was increased in the earlier shells and Si was decreased. Although not perfect, this combination yielded the best results that matched our observations. One issue with it was the overprediction of the Si ii feature. Attempts were made to enhance the feature, including decreasing its abundance, but failed to reproduce it.

At around maximum light (+1 day), we used a time of explosion from SN 2019eix of 19 days and a photospheric velocity of 11,700 km s⁻¹. We lowered the density profile from our reference model as the ejecta expands and kept the same temperature as the base model. A slight increase in Ti and Fe gave the best match to our data. One approach we used to determine which elements to alter in our models was comparing the absorption and emission features from Figure 12 with the data. From this figure, we were able to identify which elements could be responsible for creating particular features. Therefore, altering the abundance of these elements could help achieve the desired results and uncover the tomography of the SN at specific epochs. When contrasted with the results from Hachinger et al. (2012) at 22 days after the explosion, it becomes apparent that SN 2019eix has considerably higher concentrations of Ti and V in its inner layers. On the other hand, the spectral model for 1994I (Hachinger et al. 2012) at 22 days after the explosion shows more significant amounts of Ni and Fe in the inner layers, indicating a substantially faster evolution rate than that of SN 2019eix.

At phase 12 days after maximum light a time of 29 days after explosion was assumed for the model and a photospheric velocity of 10,370 km s⁻¹. The density was significantly reduced along with the temperature and the Fe was increased by a factor of 5 from our reference model. When comparing this model to SN 1994I (Hachinger et al. 2012) at day 30 after the explosion we notice it to be significantly more evolved than ours. Their velocity at this point has dropped by a factor of 6 from day 16 and has significantly more iron group elements (IGE) and intermediate-mass elements (IME) (such as Fe, Ni and Si, S, respectively) in the inner shells than we observe for our SN 2019eix model. SN 1994I is a fast-evolving Ic, whereas SN 2019eix appears to not change much in terms of its abundance or photospheric velocity from before to after maximum light.

From our TARDIS models of SN 2019eix in comparison to that in Hachinger et al. (2012), we note that the velocity does not evolve as fast as SN 1994I (Hachinger et al. 2012) drops by a factor of 6 in 14 days, and in SN 2019eix models it drops by a factor of 1.2. The temperature has the opposite effect and appears to drop faster for SN 2019eix than that in Hachinger et al. (2012). Additionally, SN 2019eix also seems to be denser than that in Hachinger et al. (2012) throughout all the epochs. SN 2019eix seems to be a slow evolver in terms of velocity and reaction rates. A possible explanation for these peculiar features we observe in SN 2019eix could be a density and temperature effect, as those were the main parameters we varied the most and appeared to successfully reproduce the spectra as shown in Figure 12. ¹⁸

SN 2019eix shares numerous similarities with DD candidates from He-shell Type Ia as shown in Figures 6 and 7. Spectroscopically, SN 2019eix shows significant red colors as well as suppression in emission lines below 5000 Å; similar to SN 2016hnk (Jacobson-Galan et al. 2020), SN 2018byg (De et al. 2019), and SN 2016dsg starting from maximum following to 30 days after maximum light. The strong blanketing in the DD on the He-shell scenario can be attributed to the large amount of iron group elements created in the outer ejecta from the He burning. The substantial reddening can be explained by the ashes produced by the He-shell detonation causing redder colors throughout its evolution. Photometrically, SN 2019eix appears to be brighter and with a somewhat wider light curve than the other DD He-shell candidates as shown in Figure 6, with the exception of SN 2016hnk, which shows a slower evolution than SN 2019eix post maximum. However, in terms of the color evolution, SN 2016hnk seems to be the best match as shown in Figure 5 from the extensive comparison with multiple SNe of various types.

Due to the similarities of SN 2019eix with this class, we compare the spectra and the light curves to the DD models from Kromer et al. (2010), Sim et al. (2012), and Polin et al. (2019). Kromer et al. (2010) explored the observable properties of DD models. The simulations presented were carried out in 2D using radiative transport from ARTIS. To initialize models, they chose estimated values of temperature, the central density of the CO core, and the temperature and density at the base of the He layer. They ignited an initial He detonation in a single point at the base of the He shell and eventually created a shock wave that propagates into the core. These models have He-shell masses ranging from 0.0035–0.0126 M_⊙ previously considered by Fink et al. (2010).

Similar to Kromer et al. (2010), Sim et al. (2012) created their models using ARTIS and investigated small (0.45 M_⊙ WD + 0.21 M_⊙ He) and large (0.58 + 0.21 M_⊙ He) masses for both the single- and double-detonation scenarios. The two methods investigated were the He detonation wrapping around the CO WD, modeled as the convergence of shocks in the first method, and the He detonation igniting an inward-propagating shock at the CO core's edge, called the edge-lit core detonation in the second approach. Additionally, Polin et al. (2019) examined the explosions of WDs varying from 0.6–1.2 M_⊙ with He-shell masses of 0.01, 0.05, and 0.08 M_⊙. Polin et al. (2019) found that thicker shell models showed early time flux excess, redder colors, and higher line blanketing in the UV through the blue regime of the spectrum. Their models were created using the Eulerian hydrodynamics code Castro. After the SN ejecta reaches homologous expansion, a multidimensional time-dependent radiation transport code (SEDONA) is used to create the synthetic spectra and light curves.

In our study, the models we chose to compare to SN 2019eix are the DD models Polin0.9+0.08-D, Polin0.76+0.15-D, Polin0.76+0.15-0.2-D, Kromer0.81+0.126-D (model 1), and Sim.0.58+0.21-D (CSDD-S) as shown in Table 5. We chose these models as they appeared to be more consistent with our light curves and spectra of SN 2019eix. However, we also considered Polin0.8+0.08, Sim.0.45+0.21-D, and the only single detonation, Sim0.58+0.21-S. From Figure 13, we can see that photometrically, the more massive model with the thicker He shell such as Sim0.58+0.21-D (CSDD-S) overestimates the brightness in both bands. The other models Polin0.76+0.15-D, Polin0.76+0.15-0.2-D, and Kromer0.81+0.126-D (model 1) seem to be a better match to SN 2019eix, especially Polin0.76+0.15-D. Nonetheless, Polin0.76+0.15-D (unmixed) our best model underestimates the magnitude in the r band and declines much faster than SN 2019eix, but the i band is more consistent from maximum light up until 10 days after. Additionally, some of the challenges in modeling the light curves of SN 2019eix were attributed to the lack of early photometry. With earlier photometry, a more detailed comparison to DD He-shell models that predict an early bump in the light curve as illustrated in Figure 13.

Zoom In Zoom Out Reset image size

As mentioned, SN 2019eix shows substantial line blanketing on the blue side of the spectra. This is reproduced by the majority of the models at maximum and past maximum light as shown in Figure 14. Reproducing the data from the models consistently at various epochs was difficult. This is evident in Polin.0.9+0.08, which indicates that a reasonable result at one epoch was observed and not the others. Nevertheless, in Sim0.58+0.21-D (CSDD-S) we observe consistent behavior throughout the epochs, although the light curve of this model severely overestimates the flux. However, it is a known issue that the light curves in Sim et al. (2012) are brighter due to the pure He-shell assumption, but, in actuality, the He layer can be polluted by the WD core material. By contrast, the Polin0.76+0.15-D (unmixed) light-curve model is the best match, but spectroscopically appears to be a mismatch before maximum light but overall a reasonable fit.

Zoom In Zoom Out Reset image size

Generally, it appears that the light-curve models do not match SN 2019eix consistently. However, some of this mismatch could be attributed to the degree of mixing in the outer layer (Kromer et al. 2010; Shen et al. 2010; Sim et al. 2012; Polin et al. 2019) in addition to the viewing angle (Kromer et al. 2010). The models used for comparison are all from local thermodynamic equilibrium (LTE), which takes into effect about 20 days before maximum light. Therefore, it is important to compare the models to the data at early times. In order to fully compare the light curve at all epochs, we must use non-LTE models.

In Figure 6, we showed that the light curves for SN 2019eix are the most similar to SNe Ic in terms of peak magnitude (Table 1), bolometric luminosity (illustrated in Figure 9), and decline rates (Table 2). However, from the spectral and color evolution standpoint we do not observe such consistency. Spectroscopically, before maximum SN 2019eix resembles SNe Ic with the exception of the absence of the O i feature. However, we observe significant discrepancy in the spectra at later times as shown in Figure 8 (albeit peculiar SNe Ic show some level of consistency as we will discuss below). Additionally, the color evolution of SN 2019eix appeared to be much redder than SNe Ic as depicted in Figure 5.

For peculiar SN Ic PTF12gzk, the color evolution at early times seems to agree with that of SN 2019eix (Figure 5), but a full comparison of colors was not conducted as we lacked photometry at later times. Additionally, the spectrum of PTF12gzk before and after maximum matches SN 2019eix as it starts to develop a blue suppression after maximum light, but does not seem to be depleted of the O i 7774 feature. PTF12gzk appears to also have more blended features than SN 2019eix suggesting higher velocities. Ben-Ami et al. (2012) reported PTF12gzk to exhibit large expansion velocities of ≈30,000 km s⁻¹ measured days after the explosion from the Si ii line. Additionally, the mass of ejecta was measured to be M_ej = 7.5 M_⊙, about a factor of 3 larger than our estimates for SN 2019eix. This suggests a higher initial progenitor mass of 25–30 M_⊙ (Ben-Ami et al. 2012) and a lower-mass progenitor for SN 2019eix.

We considered the ultra-stripped SNe (USSNe) scenario due to the lack of O i in SN 2019eix, as these SNe can be stripped of their O in addition to H and He layers from a compact binary (Dessart et al. 2020; Woosley et al. 2020). We dismissed this scenario, since neither the spectra nor the light curves are comparable (e.g., USSNe SN 2010X and SN 2005ek show no blue suppression and have O i features unlike SN 2019eix, as shown in Figure 15). The fast rise times of USSNe (typically between 5 and 10 days and magnitude of −16; Moriya et al. 2016) and fast decline rates (Δm₁₅ ≈ 3; Drout et al. 2013), suggest a smaller progenitor than SN 2019eix.

Zoom In Zoom Out Reset image size

From the observations, we note that peculiar Type Ic's (PTF12gzk) could roughly produce spectra that look similar to SN 2019eix. Due to the peculiar spectral evolution of SN 2019eix, we modeled the spectra using TARDIS. From our models, we showed that spectra similar to SN 2019eix could be reproduced by using the abundance models from SN 1994I type Ic (Iwamoto et al. 1994; Hachinger et al. 2012), but again the O i failed to be reproduced as discussed previously in Section 4.1. Overall, the light curves appear to be consistent with SN Ic; however, the spectral evolution appears to be atypical for this class, in particular, the strong reddening, lack of O i, and strong Ca ii at later times.

As shown in Figure 7, SN 2019eix starts to resemble subluminous SNe Ia in its later spectra, although it is notably different before maximum. Photometrically, 91bg-like SNe reach peak magnitudes between −16.7 and −17.7 in the B band similar to SN 2019eix. Overall, despite the similarities between SN 2019eix and 91bg-like SNe in terms of the spectra and the absolute magnitude, they also display key differences.

Some of these differences include bluer colors as shown in Figure 5 (91bg-like SNe reach colors of ${(B-V)}_{\max }\approx 0.5\mbox{--}0.6;$ Taubenberger et al. 2008; Sullivan et al. 2011) being much bluer than SN 2019eix (${(B-V)}_{\max }\approx 1.8$); faster decline rates (Δm₁₅(B) ≈ 1.8–2.1) than SN 2019eix (Δm₁₅(B) ≈ 1.2); and spectroscopic disagreement in some absorption features (especially the Si ii, O i, and Ca ii features and generally the spectra is not as suppressed in the blue as SN 2019eix); thus, the inferred ⁵⁶Ni mass and ejecta mass for 91bg-like SNe were found to be between ∼0.05 and ∼0.10 M_⊙ (Mazzali et al. 1997; Sullivan et al. 2011) and ≈0.5 M_⊙ (Stritzinger et al. 2006), respectively, significantly lower than SN 2019eix. For these reasons, it is unlikely that SN 2019eix shares a similar progenitor as the 91bg-like objects. However, it is important to note that the M_ej mass estimate could be overestimated for SN 2019eix, as it depends on the photospheric velocity and width of the light curve.

Calcium-rich transients are peculiar as their explosion mechanisms are not well understood and a variety of scenarios can explain their observations. Due to the strength of the Ca ii absorption feature of SN 2019eix, we briefly compare these transients to SN 2019eix, but rule them out due to significant incompatibilities. Calcium-rich SNe are mostly characterized by peak magnitudes of −14 to −16.5, with fast-evolving light curves (much dimmer and faster evolving than those of SN 2019eix). From Figure 15, we can see that SN 2005E (a calcium-rich transient) is a mismatch to SN 2019eix in terms of its spectral evolution (e.g., it does not display the extreme suppression in the blue seen in SN 2019eix), but instead develops a [Ca ii] 7291, 7323 line 18 days after peak (not observed in SN 2019eix, but a common feature for these transients). Additionally, these transients also possess moderately red colors ${(B-V)}_{\max }\approx$ 0.6 mag (Taubenberger 2017), unlike the significant red colors observed in SN 2019eix. The majority of these objects display low ⁵⁶Ni and ejecta masses between 0.4 and 0.7 M_⊙ (Kasliwal et al. 2012). Due to the low luminosity, different color evolution, and incompatible spectra evolution; SN 2019eix is unlikely to share the same progenitor as these transients.

The observations of He-shell DD candidates (OGLE-2013-SN-079, SN 2016hnk, SN 2016dsg, and 2018byg) display significant spectroscopic resemblance to SN 2019eix. Before maximum light, we do not observe many similarities and thus are not plotted. However, during and after maximum light the spectra of SN 2016dsg and SN 2018byg are found to be extremely suppressed in the blue and have a minimum to no absorption O i feature, similar to SN 2019eix. During and after maximum light the spectra continue to have considerable line blanketing at λ < 5000 Å and show stronger Ca ii absorption as shown in Figure 7.

An explanation for the strong line blanketing in the blue for these DD candidates is that the detonation of a helium shell on the surface of a C/O WD pollutes the outer layers of the ejecta. These IME and IGE burning products cause this blue blanketing and significant reddening (Kromer et al. 2010; Shen et al. 2010; Sim et al. 2012; Polin et al. 2019). The reddening is further illustrated in Figure 5, where we see SN 2019eix sharing impressive similarities with SN 2016hnk as they both seem to be redder than the rest of the SN types. DD models also explain the strong Ca ii feature. This is a result of the He burning creating a large amount of IME in the outer layers (Fink et al. 2010; Kromer et al. 2010). SN 2019eix develops Ca ii into a deep and high-velocity feature after maximum light, as shown in Figure 3. Therefore, a large amount of IME created in the DD models can explain such an anomaly observed in SN 2019eix.

We compare existing DD models to SN 2019eix in Figures 13 and 14. The models are roughly able to reproduce the light curve and spectral features. The best DD model of SN 2019eix is Polin.0.76+0.15 (unmixed), as it is able to be a better match to the light curves (although it still declinHes faster in the r band) and it is able to reproduce the most prominent features during and after maximum light (although it fails to match the features before maximum light). Notice in Figure 14 how the DD spectral models reproduce the lack of an O i feature, a pronounced feature in SN 2019eix. The physics of the absence of the O i feature in the case where the He shell is thick is justified by the ash of the He-burning products created in the outermost layers of the ejecta. It is expected to cover the unburned oxygen from the core underneath the He products limiting them to a restricted velocity range (Kromer et al. 2010; Sim et al. 2012; Polin et al. 2019). This could explain why the O i feature for SN 2019eix is not observed until roughly 30 days after maximum.

Figure 6 shows that DD candidates show some level of agreement with SN 2019eix, particularly in the r band. Existing candidates seem to be dimmer and faster declining than SN 2019eix (Figure 9), although not many of these objects have been found in the literature. Therefore, SN 2019eix remains a great candidate for the DD He-shell model (with a possible WD mass close to 0.76 M_⊙ and a thick He shell close to 0.15 M_⊙), as it is able to physically explain the abnormalities from the spectra and the color evolution unlike other classes. Therefore, for those reasons, the DD scenario is the favored scenario for SN 2019eix.