ZTF18aalrxas: A Type IIb Supernova from a Very Extended Low-mass Progenitor

The light curves (LCs) and spectra of stripped-envelope (SE) supernovae (SNe) can show a wide range of different behaviors. In particular, SNe IIb show intermittent signatures of hydrogen in their photospheric spectra (e.g., Filippenko 1997; Gal-Yam 2017). Furthermore, given early enough observations, the presence of hydrogen tends to be accompanied by an initial cooling phase in the optical LCs before the main radioactively powered peak (e.g., SN 1993J; Filippenko et al. 1993; Nomoto et al. 1993; Schmidt et al. 1993, SN 2011dh; Arcavi et al. 2011; Ergon et al. 2014, 2015 and SN 2013df; Van Dyk et al. 2014; Ben-Ami et al. 2015).

In some SNe IIb, the cooling signature dominates the early optical LCs during the first weeks following the explosion (SNe 1993J and 2013df). However, in SN 2011dh, there were strong hydrogen features present in early spectra, but the cooling phase lasted less than 5 days (Arcavi et al. 2011). A similar evolution was also seen in, e.g., SN 2008ax (Crockett et al. 2008; Pastorello et al. 2008) and PTF12os (Fremling et al. 2016). The early cooling emission is a result of the SN shock breaking out of the stellar envelope, where the strength and duration of the optical emission is driven by the radius and mass of the envelope material (e.g., Rabinak & Waxman 2011; Nakar & Piro 2014; Piro 2015). Thus, an important set of progenitor parameters can be directly probed by studying the early optical emission.

The varying strength of the cooling emission seen in SNe IIb can be explained if their progenitors have been stripped of their hydrogen envelopes to different degrees. This stripping could either be due to strong stellar winds from very massive stars, with zero-age main-sequence (ZAMS) masses, M_ZAMS > 20 M_☉ (e.g., Groh et al. 2013), or due to binary interactions (e.g., Yoon et al. 2010, 2017; Claeys et al. 2011), where the stars do not need to be as massive (typically M_ZAMS < 17 M_☉).

A few SNe IIb have been found to have broad and slowly evolving LCs, which are indicative of large ejecta masses that could be consistent with massive stars (see e.g., SN 2003bg; Hamuy et al. 2009; Rubin & Gal-Yam 2016; and Figure 6 in Fremling et al. 2018). However, for the majority of SNe IIb discovered thus far, sample studies tend to favor progenitor mass ranges that are more in line with the expectations for binary models (e.g., Drout et al. 2011; Cano 2013; Taddia et al. 2015, 2018; Lyman et al. 2016; Prentice et al. 2016). Furthermore, for SN 1993J a binary companion has likely been directly observed in post-explosion imaging (Maund & Smartt 2009 and Fox et al. 2014), providing the most direct evidence for the binary scenario.

Regardless of the origin of SNe IIb, it is interesting to investigate the main drivers for the differences in the strengths of the cooling signatures and derived progenitor radii (and extended envelope masses) at the time of explosion. Could it be the case that there is a relation between the ZAMS mass of the progenitor and the mass and extent of the envelopes at the time of explosion? In the case of SN 1993J, nebular models indicate a larger ZAMS mass by at least a few solar masses compared to, e.g., SN 2011dh and SN 2008ax (Jerkstrand et al. 2015); also, SN 1993J has one of the strongest cooling signatures observed in any SN IIb.²⁰

In this Letter we present a counter-example to this ZAMS mass-driven picture, and instead argue that the structure of the extended envelope is determined by the binary configuration and subsequent evolution. During science validation the Zwicky Transient Facility (ZTF; Bellm et al. 2019; Graham et al. 2019) discovered an SN IIb, ZTF18aalrxas, exceptionally close in time to the explosion. We observed a very strong cooling signature in the optical, almost identical to what was seen in SN 1993J. The cooling phase lasted approximately one week, and was followed by the emergence of a SN IIb LC that qualitatively behaves just like the low-mass SN 2011dh. Late-time spectra also indicate a similarly low ZAMS mass as seen in SN 2011dh (M_ZAMS = 12–13 M_☉), which puts ZTF18aalrxas among the least massive SNe IIb found to date, while still having a very strong cooling signature.

Our analysis of ZTF18aalrxas spans from ≈0.7 days past explosion until ≈180 days past explosion. In Section 2 we describe our follow-up observations and give details on the data reduction involved. Section 3 presents the multi-band LCs and describes the construction of the bolometric LC of ZTF18aalrxas. A qualitative comparison to other SNe IIb shows that the bolometric properties of the SN are not consistent with the progenitor being very massive (i.e., M_ZAMS > 20 M_☉). In Section 4 we report our follow-up spectroscopy, confirm the classification of ZTF18aalrxas as an SN IIb, and provide velocity measurements of the SN ejecta. In this section we also use late-time spectroscopy (≈180 days past explosion) to constrain the amount of oxygen and hydrogen in the ejecta. Section 5 presents a Monte-Carlo (MC) based model for the LCs and spectra of ZTF18aalrxas, which we use to constrain the synthesized nickel mass and the total ejecta mass of the explosion, along with the helium core mass of the progenitor. Semi-analytical models are also used in this section to constrain the mass and radius of the extended envelope. Section 6 presents our conclusions, and contains a discussion where we put our observations of ZTF18aalrxas in context with other SNe IIb.

2.1. Detection and Classification

2.2. Optical Photometry

Following the discovery of ZTF18aalrxas, we obtained follow-up photometry during the photospheric phase in g and r with the ZTF Camera on the P48, in gri with the SEDM on the P60, in ugri through the Las Cumbres Observatory Global Telescope Network (Brown et al. 2013), in gr with the Kitt Peak Electron Multiplying CCD on the Kitt Peak 84 inch telescope (Coughlin et al. 2019), and in gri with IO:O on the Liverpool Telescope. The Nordic Optical Telescope (NOT) at La Palma was used to obtain late-time photometry in gri around 150 days past the discovery.

LCs from the P48 come from the ZTF production pipeline (Masci et al. 2019), with limiting magnitudes re-computed using forced point-spread function (PSF) fitting at the expected position of the SN (Y. Yao et al. 2019, in preparation). Stacked P48 LCs were also prepared using SkyPortal (Van der Walt et al. 2019), an online interface to the forced PSF-fit photometry. The stacked LCs were produced by taking the inverse-variance weighted average of the single-epoch PSF photometry in non-overlapping -day windows. LCs from the rest of our optical imaging data have been produced with the image-subtraction pipeline described in Fremling et al. (2016), with template images from the Sloan Digital Sky Survey (SDSS; Ahn et al. 2014). This pipeline produces PSF magnitudes, calibrated against SDSS stars in the field. We have also compared our results to PSF magnitudes calibrated against Pan-STARRS1 (Chambers et al. 2016), finding no significant offsets.

In our analysis we have corrected all photometry for galactic extinction, using the Milky Way (MW) color excess E(B − V)_MW = 0.0192 mag toward the position of ZTF18aalrxas (Schlafly & Finkbeiner 2011). All reddening corrections are applied using the Cardelli et al. (1989) extinction law with R_V = 3.1. No further host galaxy extinction has been applied, as there is no sign of any Na ID absorption in any of our spectra. The multi-color LCs of ZTF18aalrxas are shown in Figure 1.

Zoom In Zoom Out Reset image size

2.3. Swift-Ultraviolet Optical Telescope Photometry

2.4. Optical Spectroscopy

The P48 g- and r-band pre-explosion limits along with the UV and optical LCs of ZTF18aalrxas are displayed in Figure 1. Qualitatively, the behavior of ZTF18aalrxas is well matched to both SN 1993J and SN 2013df during both the first and second LC peaks. The second peak is also well matched to SN 2011dh, even though SN 2011dh lacked a strong initial cooling phase. For the explosion date of ZTF18aalrxas we have adopted t_exp = 2458227.1 ± 0.1 (JD), from our best-fitting model for the early cooling emission²⁵ (Section 5.1). The decline rates of the LCs after 50 days past explosion also appear very similar for these objects (dashed lines in Figure 1). We measure the r-band decline rate of ZTF18aalrxas to $1.83\tfrac{\mathrm{mag}}{100\,\mathrm{days}}$ by fitting a first-order polynomial to the data starting from 50 days past explosion.

A pseudo-bolometric LC for ZTF18aalrxas was constructed by fitting a blackbody (BB) model to the spectral energy distribution (SED) derived from the LCs as a function of time, and integrating the flux of the fitted BB (starting from 2600 Å). We have weighted our BB fits by the photometric errors. This de-emphasizes the data collected by Swift, which has very large uncertainties. The photometry has been interpolated to the dates of the r-band data points, in order to allow the construction of the bolometric LC. Errors have been calculated using a MC method where the BB fits are re-computed for the range of LCs that are allowed by shifting the LC points within the photometric uncertainties (we assume a normal distribution when resampling).

The BB parameters and the pseudo-bolometric LC are shown in Figures 2 and 3, respectively. In Figure 3 we also display the pseudo-bolometric LC of SN 2011dh (data from Ergon et al. 2014) both without scaling and with scaling to match the peak luminosity (nickel mass) of ZTF18aalrxas, as well as scaled pseudo-bolometric LCs of SN 2013df (data from Szalai et al. 2016) and SN 1993J (data from Richmond et al. 1994). The bolometric LCs of SN 2011dh and SN 2013df have been constructed in the same way as the bolometric LC of ZTF18aalrxas. When determining the BB parameters and bolometric LC of SN 2013df we have applied a color-based extinction correction with E(B − V) = 0.175 mag, based on the observed g − r color 10 days past the r-band peak, as described in Taddia et al. (2015). Because the explosion time for SN 2013df is uncertain, we have matched the phase of maximum light of the main peak in the bolometric LC to that of ZTF18aalrxas.²⁶ For SN 1993J there are no gri data. Instead we have determined the BB parameters and constructed the pseudo-bolometric LC via BB fitting of BVRI LCs. As SN 2013df has both gri and BVRI coverage, we have checked that the bolometric LC does not significantly change depending on the filter set in our method. We have adopted the lower value for the extinction (E(B − V) = 0.08 mag) used by Richmond et al. (1994) when performing BB fits to the spectral energy distribution (SED) of SN 1993J.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The peak luminosity of ZTF18aalrxas is higher by about a factor of 2.2 compared to SN 2011dh, indicating a ⁵⁶Ni mass²⁷ that should be higher by a similar amount, which turns out to be ≈0.17 M_☉. This is very close to the median nickel mass (${M}_{{}^{56}\mathrm{Ni}}=0.15\pm 0.07$ M_☉) of the SN IIb sample studied by Taddia et al. (2018). The initial cooling phase of ZTF18aalrxas remains very prominent in the bolometric LC, and there is a striking similarity to the cooling phase seen in SN 1993J. The luminosity at the initial peak is also similar to that observed in SN 2013df, but the duration of the cooling phase is shorter in ZTF18aalrxas. After about one week past explosion, the main ⁵⁶Ni-powered LC peak starts to dominate in ZTF18aalrxas, while the cooling dominates for almost one more week in SN 2013df. This indicates a higher mass in the envelope of the progenitor to SN 2013df compared to the progenitor of ZTF18aalrxas (see Section 5.1).

The width of the main bolometric LC peak appears to be very similar for ZTF18aalrxas, SN 1993J, SN 2013df, and SN 2011dh. A bolometric LC well matched to SN 2011dh strongly disfavors a progenitor scenario for ZTF18aalrxas, where the mass loss was dominated by the stellar wind from a very massive star, because the LCs of SN 2011dh are consistent with stars with M_ZAMS ≤ 13 M_☉. Under the assumption that a compact neutron-star remnant is formed, a massive progenitor (M_ZAMS ≳ 20 M_☉) should have a large ejecta mass and slower evolution of the bolometric LC.

From the BB fits to the SED of ZTF18aalrxas we have derived the evolution of R_BB, which can be interpreted as a rough approximation of the photospheric radius (bottom panel of Figure 2). R_BB evolves in a very similar way when compared to SN 2013df, and especially when compared to SN 1993J (shown as comparisons). Our earliest measurement is at 2.8 days past explosion, and indicates R_BB in the range 0.8 ±0.5 × 10⁴ R_☉, which later peaks around (2.5–3) × 10⁴ R_☉ at ≈40 days past explosion. T_BB (top panel of Figure 2) starts out at around 15 × 10³ K in ZTF18aalrxas at 2.8 days past explosion, but drops quickly to around 7.5 × 10³ K over the following few days. Again this is remarkably similar to the evolution of SN 1993J, including the rapid decline seen initially. In SN 2013df T_BB stayed around 10⁴ K for the first 10 days past explosion. Compared to the larger sample of SE SNe presented by Taddia et al. (2018), both T_BB and R_BB of ZTF18aalrxas appear to be very consistent with the averages for SE SNe.

Our spectral sequence (Figure 4) starts out at 7 days past explosion with spectra dominated by the Balmer series. Comparison spectra of SN 2013df, which are very similar to spectra of SN 1993J (see e.g., Morales-Garoffolo et al. 2014; Szalai et al. 2016) are also shown in Figure 4. Our earliest spectrum from SEDM is of low S/N, but does show a broad feature where Hα is expected, which is also consistent with the NOT spectra taken at 12 and 16 days past explosion, where several of the Balmer lines are present. A spectrum of SN 2013df taken at 12 days past explosion is qualitatively very similar, but slightly bluer, which is consistent with the higher T_BB that we derived in Section 3.

Zoom In Zoom Out Reset image size

He i signatures appear rather late in ZTF18aalrxas; we observe the simultaneous appearance of He i λλ5016, 5876, 6678, 7065 in absorption at around 7500 km s⁻¹ in our spectra past 30 days, with the signatures being quite clear at 56 days past explosion. This evolution is similar to that of SN 2013df and SN 1993J, which also lacked clear He i for the first month. In particular, our LRIS spectrum of ZTF18aalrxas taken at 56 days past explosion is qualitatively very similar to a spectrum taken at a similar epoch of SN 2013df. This clearly solidifies the SN IIb classification of ZTF18aalrxas.

Our spectrum taken at 85 days past explosion shows the emergence of [O i] λλ5577, 6300, 6364 emission lines along with the O i λ7774 triplet, hinting that the ejecta are becoming optically thin. The latest spectrum, taken at 177 days past explosion, is clearly in the nebular phase, with prominent [O i] λλ6300, 6364 seen in emission. There is also broad flat-topped Hα emission emerging, similar to what was seen in nebular spectra of SN 2013df and SN 1993J. This nebular spectrum is further analyzed and modeled in Section 4.1 in order to constrain the O and H envelope masses.

Fe ii λ5169 is a decent tracer for the photospheric velocity (Dessart & Hillier 2005). Absorption from this line is present in our spectra taken later than 26 days past explosion. Velocities measured for Fe ii λ5169, Hα, Hβ, and He i λλ5876, 7065 are shown in Figure 5. These measurements were performed by smoothing the spectra and locating the relevant absorption minima. Uncertainties were estimated through a MC simulation where many simulated spectra were created, using smoothed spectra and randomly generated noise reflecting the observed S/N. The Fe ii λ5169 velocity of ZTF18aalrxas is in good agreement with the velocity derived from the evolution of the BB radius (Section 3), which gives v_ph ≈ 8000 km s⁻¹ around 26 days past explosion. This velocity is also similar to what can be seen in SN 1993J, SN 2013df, and SN 2011dh at a similar epoch.

Zoom In Zoom Out Reset image size

4.1. Nebular Spectroscopy and Oxygen Mass Constraints

More massive stars produce larger amounts of metals, and the oxygen nucleosynthesis in particular is a strong and monotonic function of M_ZAMS (Woosley & Heger 2007). This has been exploited by Jerkstrand et al. (2015), who used a grid of simulated spectra to show a strong dependence between the ratio of the [O i] λλ6300, 6364 line luminosity compared to the total ⁵⁶Co decay power, for SE SNe in the nebular phase.

We have used the model grid of Jerkstrand et al. (2015) to match our nebular spectrum of ZTF18aalrxas taken at 177 days past explosion, with a focus on the [O i] λλ6300, 6364 emission (Figure 6). The same model that was used by Jerkstrand et al. (2015) to model the nebular emission of SN 2011dh (model 12C in Jerkstrand et al. 2015) is also quite well matched to ZTF18aalrxas. The peak luminosity of the [O i] λλ6300, 6364 line is slightly higher in our observation, but not as high as what would be expected for a M_ZAMS = 13 M_☉ model. Thus, we conclude that the O mass should be in the range 0.3–0.5 M_☉ (O mass values from Jerkstrand et al. 2015), which for these models indicate M_ZAMS = 12–13 M_☉ for the progenitor of ZTF18aalrxas.

Zoom In Zoom Out Reset image size

4.2. Hydrogen Mass Constraints

5.1. Modeling the Cooling Emission

Figure 7 shows the gri-band LCs of ZTF18aalrxas up until ≈8 days past explosion. We detected ZTF18aalrxas while the g-band luminosity was still rising toward the initial peak in the LC. This early emission can be modeled as the result of the SN shock breaking out from a progenitor surrounded by an extended envelope (e.g., Rabinak & Waxman 2011; Nakar & Piro 2014; Piro 2015). In these models, the SN shock heats the extended material to very high temperatures (T > 10⁵ K), after which the material rapidly starts to cool. This means that what looks like an initial rise to a peak in the optical is actually a temperature effect. The SED of the SN gradually moves into the optical range as the ejecta cool, resulting in rising LCs, but the total bolometric luminosity is actually declining. Regardless, having data points on the optical rise to the first peak is a highly significant observation, as it allows a very accurate explosion time estimate, which is rare for double-peaked SNe IIb. For SN 2016gkg, which had a comparable cooling phase duration to SN 2011dh, there is an extremely early discovery (0.1–0.2 days past the explosion; Bersten et al. 2018). However, in both SN 1993J and SN 2013df the first points on the LCs are on the decline from the initial LC peaks.

Zoom In Zoom Out Reset image size

In order to estimate the properties of the extended material that gives rise to the initial LC peak of ZTF18aalrxas, we have used the Markov Chain MC (MCMC) method in the Python emcee package (Foreman-Mackey et al. 2013), and the model by Nakar & Piro (2014). This is a one-zone model for the extended envelope, with the mass in the envelope M_env concentrated at a radius R_env. For the opacity of the envelope material we assume κ = 0.3 cm² g⁻¹. The core mass, M_c, the time of explosion t_exp, and the velocity of the extended material, v_e, also enter as parameters. For M_c we use the He core mass from our model of the main bolometric LC peak (Section 5.2). For the explosion energy we use E = 10⁵¹ erg, which gives v_e that is reasonable³⁰ compared to the velocities derived from the evolution of R_BB and the Hα velocities in our earliest spectra (Figure 5). The result of the MCMC simulation is ${M}_{\mathrm{env}}={4.3}_{-0.13}^{+0.14}\times {10}^{-2}$ M_☉, ${R}_{\mathrm{env}}={0.73}_{-0.02}^{+0.03}\times {10}^{14}$ cm $[{1050}_{-30}^{+40}$ R_☉], and ${t}_{\exp }={2458227.1}_{-0.1}^{+0.1}$ (JD). Uncertainties represent 68% confidence intervals. The range of models allowed are visualized in Figure 7. These models indicate that the contribution from the cooling envelope to the LCs drops rapidly after 6 days past explosion.

We have also used the model by Nagy & Vinkó (2016) to model the early bolometric LC (see Figure 3). This model is fit by eye, and we are able to achieve an excellent match to the early declining LC for, M_env = 5.4 × 10⁻² M_☉, R_env =0.55 × 10¹⁴ cm [≈790 R_☉], with an initial kinetic energy E_k = 0.1 × 10⁵¹ erg, and initial thermal energy E_th = 0.05 ×10⁵¹ erg, and κ = 0.3 cm² g⁻¹. While these values give the nicest looking fit for the bolometric LC, models that are closer to the result from the MCMC fit based on the Nakar & Piro (2014) model are still generally within the error bars on our early bolometric data points. The expansion velocity of the outer ejecta in the best-fitting Nagy & Vinkó (2016) model is ≈20,000 km s⁻¹, which is consistent with the early BB evolution and the Hα velocity in our earliest spectrum (Figure 5). Thus, we conclude that both of these models are generally in agreement. The early emission of ZTF18aalrxas can be very well reproduced as the cooling emission from a low mass and very extended envelope.

Figure 3 shows that the duration of the first LC peak is markedly shorter in ZTF18aalrxas (≈6 days) compared to SN 2013df (≈12 days). A significant difference in cooling phase duration, in the context of these semi-analytical models, can most easily be explained by a difference in the envelope mass around the progenitor at the time of explosion. Using the Nagy & Vinkó (2016) model, we find that M_env = 11 ×10⁻² M_☉ and R_env ≈ 370 R_☉ is required to model the early (unscaled) bolometric emission of SN 2013df.³¹ We are using a higher value for E(B − V), but our values are still comparable to the values (M_env = 8.0 × 10⁻² M_☉, R_env ≈ 160 R_☉) derived using the same model for SN 2013df by Szalai et al. (2016). By similar reasoning it can be argued that the extended envelope mass and radius of SN 1993J should be very similar to that of ZTF18aalrxas; qualitatively the early LCs behave very similarly, with the turnover from cooling to ⁵⁶Ni power happening at a very similar epoch for both SNe. By fitting the unscaled bolometric LC of SN 1993J, we derive M_env ≈ 6 ×10⁻² M_☉, R_env ≈ 500 R_☉.

In conclusion, the hydrogen envelopes of ZTF18aalrxas and SN 1993J appear to be very similar, while the envelope of SN 2013df is slightly more massive. In Section 4.1 we arrived at the same qualitative conclusion, based on our nebular spectroscopy. However, the hydrogen masses were somewhat larger. This is likely a calibration issue of the semi-analytical LC models; we consider the nebular constraints as more indicative of the total H mass at the time of explosion, because these estimates are significantly less model dependent (e.g., they do not depend on assumptions about kinetic or explosion energies).

5.2. Modeling LCs and Spectra

We have performed a thorough analysis of the optical emission from ZTF18aalrxas, finding that the progenitor of this SN had, in the context of SE SNe (e.g., Taddia et al. 2018), a low mass (M_ZAMS = 12–13 M_☉) and was surrounded by a very extended envelope (R_env = 790–1050 R_☉), with a mass (M_env = 0.04–0.15 M_☉) at the time of explosion.

Our ZAMS mass results are based on the 12C model, originally presented by Jerkstrand et al. (2015), and here extended in time with a new MC NLTE code (JEKYLL; Ergon et al. 2018). The 12C JEKYLL model is able to reproduce the main bolometric LC of ZTF18aalrxas very well, while at the same time also reproducing the [O i] λλ6300, 6364 emission in our nebular spectrum of the SN. These two constraints are usually obtained from separate modeling codes: one that simulates the bolometric LC, and one that simulates the nebular phase spectra. In general, these simulations do not use the same progenitor model. Our new 12C JEKYLL model offers strong evidence for a very low ZAMS-mass progenitor for ZTF18aalrxas in the 12–13 M_☉ range, which is consistent with the ZAMS mass predicted for SE SNe by binary evolution modeling (e.g., Yoon et al. 2017), and not consistent with SE SNe produced through wind-driven mass loss from massive stars (e.g., Groh et al. 2013). Thus, it is very likely that the progenitor of ZTF18aalrxas was part of a binary system.

Furthermore, our nebular spectrum of ZTF18alrxas shows an excess around the Hα line, which can be explained through interaction between the SN ejecta and a dense CSM (Section 4.2). Incidentally, all of the well-studied SNe IIb in the literature that show very strong cooling signatures in their early LCs (due to their large radii at the time of explosion) also show signs of CSM interaction in their nebular spectra (SN 1993J, SN 2013df). While not as well studied, this is also likely the case for SN 2011fu (Kumar et al. 2013; Morales-Garoffolo et al. 2015, although the interpretation of this excess is different in Morales-Garoffolo et al. 2015). In contrast, no dense CSM (no significant excess around the Hα region in nebular spectra) is generally seen in SNe IIb that are more compact at the time of explosion and lack strong cooling signatures (e.g., SN 2011dh, SN 2008ax; see Jerkstrand et al. 2015, and PTF12os; see Fremling et al. 2016). Thus, as already suggested by Chevalier & Soderberg (2010), it appears that there are indeed two distinct classes of SNe IIb; one compact class, with less massive H envelopes without a dense CSM, and one extended class with more massive H envelopes and a dense CSM (see also Ben-Ami et al. 2015). Here we have added another member, ZTF18aalrxas, to the extended envelope SN IIb family.

In binary model systems for SE SNe, the mass-loss rate (and the potential for the presence of a dense CSM) is driven by the binary interactions (e.g., Smith 2014). Based on this, Maeda et al. (2015) proposed that the extended and compact classes of SNe IIb represent two different binary evolution paths. Extended SNe IIb (such as SN 1993J) would be exploding when strong binary interaction and subsequent mass loss is ongoing, while compact SNe IIb (such as SN 2011dh) should have a significant delay between the mass-loss episode where most of the H envelope is lost and the explosion.

More recently, Yoon et al. (2017) have constructed a large grid of binary models for SNe IIb, and the results from this work at least partly disagree with the picture suggested by Maeda et al. (2015). Compact SNe IIb are in Yoon et al. (2017) produced from systems in a tight orbit undergoing early Case B mass transfer, which means there is indeed a significant delay between the strong interaction and mass-loss phase and the explosion. However, Yoon et al. (2017) also showed that extended SNe IIb could be produced through wider orbits that result in late Case B systems, where there is also a significant delay until the explosion. Some of these models for extended SNe IIb also seem to produce strong enough winds to be roughly consistent with the presence of CSM interaction (the high mass-loss rate is a function of the radius and final mass at the time of explosion). However, we note that in the current grid, stars with ZAMS masses in the range suitable for ZTF18aalrxas (M_ZAMS = 12–13 M_☉) will not have strong enough mass loss at the time of explosion to explain the CSM signature; a model with M_ZAMS = 13 M_☉, at solar metallicity, and a final H mass of roughly 0.15 M_☉ (consistent with our nebular spectra), will result in a progenitor with R ≈ 640 R_⊙ and $\dot{M}\sim 4\times {10}^{-6}$M_☉ yr⁻¹. The radius of this model is indeed close to what we derive from the early cooling emission of ZTF18aalrxas, but the wind mass loss is around an order of magnitude too low -when compared to the mass-loss rates derived for SN 2013df and SN 1993J. Suitable mass-loss rates are only possible for progenitors with M_ZAMS > 16 M_☉ in the grid by Yoon et al. (2017), which is not consistent with our LCs and nebular spectrum of ZTF18aalrxas. It is possible that a large enough parameter space has not yet been explored; and ongoing mass transfer might really be needed to have high enough mass loss in the lower-mass progenitor systems. It is possible to have binary systems that produce SN IIb explosions after multiple episodes of mass transfer during both He and later burning stages (see e.g., Benvenuto et al. 2013, although this model produces a compact SN IIb).

We propose that nebular spectroscopy can be used to test this issue; if a Case B scenario like in Yoon et al. (2017) dominates the production of extended SNe IIb, then there should be a relation between the ZAMS mass, the CSM density, and the final envelope mass and radius; and a continuum in the strength of the CSM interaction should be seen. If an episode of strong binary interaction must be ongoing at the time of explosion, a high CSM density should be possible for both low- and high-progenitor ZAMS masses, and the CSM interaction should either be non-existent or strong; there should not be a continuum between compact and extended SNe IIb in terms of the CSM driven late-time Hα emission.

The ZAMS mass that we find for ZTF18aalrxas is similar to that found for SN 2011dh, which is on the lower end for what has been observed in SE SNe, based on their bolometric LCs (e.g., Taddia et al. 2018), and also their late-time oxygen emission (Jerkstrand et al. 2015). The duration of the first LC peak is somewhat shorter in ZTF18aalrxas (≈6 days) compared to SN 2013df (≈12 days), and we interpret this difference as a lower extended envelope mass for the progenitor of ZTF18aalrxas at the time of explosion compared to SN 2013df (Section 5.1). We also found that SN 1993J appears to have an extended envelope that is nearly identical in mass and extent to that of ZTF18aalrxas. Furthermore, a similar conclusion was also reached based on the late-time Hα emission of these SNe (Section 4.2). The progenitor of SN 1993J is likely the most massive among the progenitors to these extended SNe IIb (based on late-time O emission; Jerkstrand et al. 2015), but it did not have the strongest cooling signature. SN 2011fu is well matched to a 13 M_☉ ZAMS mass model (Morales-Garoffolo et al. 2015), and had a cooling signature even stronger than SN 2013df. Thus, while the sample is small, it appears that there is currently no relation between ZAMS mass and the strength or duration of the cooling emission in extended SNe IIb; the ZAMS masses instead appear to be similar in extended SNe IIb and compact SNe IIb. This could be argued to be in support of the need for an ongoing mass-loss episode at the time of explosion. However, the current sample of objects is far too small to make any real claims; a systematic study of the nebular spectra of a larger sample of SNe IIb is needed.

In conclusion, it appears clear that ZTF18aalrxas, SN 1993J, and SN 2013df are all very similar; they all show strong long-lasting cooling signatures in their early LCs, and over time they develop very similar CSM-powered flat-topped Hα profiles in their spectra—indicative of the presence of dense CSM. They likely originate from binary systems with very similar configurations, that result in strong mass loss during the final centuries before the explosions (either as a result of strong winds, or ongoing interaction). We note that as CSM can already be identified clearly at 177 days, it is also likely affecting earlier observations; the result of a dense CSM is generally shallower P-Cygni profiles (see e.g., Dessart et al. 2017), which is exactly what is seen in SN 1993J, SN 2013df, and ZTF18aalrxas, when compared to SN IIb models without any CSM (Figure 4). This also implies that the CSM could be affecting the bolometric luminosity of these SNe, which would lead to overestimated ⁵⁶Ni masses when they are modeled without taking this into account. This possibility deserves significant future consideration.

This work is based on observations obtained with the Samuel Oschin Telescope 48 inch and the 60 inch Telescope at the Palomar Observatory as part of the Zwicky Transient Facility project. ZTF is supported by the National Science Foundation under grant No. AST-1440341 and a collaboration including Caltech, IPAC, the Weizmann Institute for Science, the Oskar Klein Center at Stockholm University, the University of Maryland, the University of Washington, Deutsches Elektronen-Synchrotron and Humboldt University, Los Alamos National Laboratories, the TANGO Consortium of Taiwan, the University of Wisconsin at Milwaukee, and Lawrence Berkeley National Laboratories. Operations are conducted by COO, IPAC, and UW.

This work was supported by the GROWTH project funded by the National Science Foundation under PIRE grant No. 1545949. The Oskar Klein Centre is funded by the Swedish Research Council.

Part of this research was carried out at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration.

This work is also partially based on observations made with the Nordic Optical Telescope, operated by the Nordic Optical Telescope Scientific Association at the Observatorio del Roque de los Muchachos, La Palma, Spain, of the Instituto de Astrofisica de Canarias. Some of the data presented here were obtained with ALFOSC, which is provided by the Instituto de Astrofisica de Andalucia (IAA) under a joint agreement with the University of Copenhagen and NOTSA.

Some of the data presented herein were obtained at the W. M. Keck Observatory, which is operated as a scientific partnership among the California Institute of Technology, the University of California, and NASA; the observatory was made possible by the generous financial support of the W. M. Keck Foundation.

This Letter is partly based on observations made with the Italian Telescopio Nazionale Galileo (TNG) operated on the island of La Palma by the Fundación Galileo Galilei of the INAF (Istituto Nazionale di Astrofisica) at the Spanish Observatorio del Roque de los Muchachos of the Instituto de Astrofisica de Canarias. It is partially based on observations obtained with the Apache Point Observatory 3.5 m telescope, which is owned and operated by the Astrophysical Research Consortium.

The Liverpool Telescope is operated on the island of La Palma by Liverpool John Moores University in the Spanish Observatorio del Roque de los Muchachos of the Instituto de Astrofisica de Canarias with financial support from the UK Science and Technology Facilities Council.

The SED Machine is based upon work supported by the National Science Foundation under grant No. 1106171.

A.G.-Y. is supported by the EU/FP7 via ERC grant 307260, "The Quantum Universe" I-Core program by the Israeli Committee for planning and budgeting, by ISF, GIF, and Minerva grants, and by the Kimmel award.

This work is part of the research programme VENI, with project No. 016.192.277, which is (partly) financed by the Netherlands Organisation for Scientific Research (NWO).

Software: JEKYLL (Ergon et al. 2018), emcee package (Foreman-Mackey et al. 2013), SkyPortal (Van der Walt et al. 2019), HEAsoft (v6.25; HEASARC 2014), LPipe reduction pipeline (D. A. Perley et al. 2019, in preparation).