Early Blue Excess from the Type Ia Supernova 2017cbv and Implications for Its Progenitor

Type Ia supernovae (SNe Ia) are the thermonuclear explosions of carbon–oxygen white dwarfs and are standardizable candles vital for cosmological distance measurements. Despite intense study, the progenitor scenarios and explosion mechanisms for these events are still not understood and may have multiple pathways (see Howell 2011; Maoz et al. 2014 for reviews). The two main progenitor pictures are the single-degenerate (SD) scenario, where the white dwarf accretes material from a nondegenerate secondary star (Whelan & Iben 1973), and the double-degenerate scenario (DD), where two white dwarfs are present in the pre-supernova system (Iben & Tutukov 1984; Webbink 1984). However, the details of both scenarios are still under investigation (e.g., Pakmor et al. 2012; Kushnir et al. 2013; Shen & Bildsten 2014; Levanon & Soker 2017).

The early light curves of SNe Ia are promising ways to constrain the progenitor systems and the physics of the explosion. For instance, the collision of the SN ejecta with a nondegenerate companion star may manifest as an early blue or ultraviolet (UV) bump in the light curve, depending on the viewing angle (Kasen 2010). Early temperature or luminosity measurements can directly constrain the radius of the progenitor (Piro et al. 2010; Rabinak et al. 2012); observed limits have confirmed that the exploding star must be a white dwarf (e.g., Nugent et al. 2011; Bloom et al. 2012; Zheng et al. 2013). Recently, Piro & Morozova (2016) explored how early SN Ia light curve behavior depends on the amount and extent of circumstellar material (CSM) and the distribution of ⁵⁶Ni in the ejecta, which is expected to vary considerably with the location(s) of ignition in the progenitor. Finally, different explosion mechanisms may also produce distinct early light curves (Noebauer et al. 2017).

Very early SN Ia light curve observations are becoming more common, and recent studies have shown that they display a range of early behaviors (Hayden et al. 2010; Bianco et al. 2011; Ganeshalingam et al. 2011; Brown et al. 2012b; Zheng et al. 2013, 2014, 2017; Cao et al. 2015; Firth et al. 2015; Goobar et al. 2015; Im et al. 2015; Marion et al. 2016; Shappee et al. 2016a). iPTF14atg (Cao et al. 2015) and SN 2012cg (Marion et al. 2016) both showed early, UV/blue excesses in their light curves. iPTF14atg was an SN 2002es-like event, which are subluminous, do not follow the Phillips (1993) relation, have low velocities, and show Ti ii absorption (Ganeshalingam et al. 2012). Cao et al. (2016) found that the UV excess in iPTF14atg was consistent with the SN ejecta interacting with a companion star, although Kromer et al. (2016) find the SD scenario incompatible with the observed spectral evolution. In SN 2012cg, a normal SN Ia, the early blue bump was again interpreted as a signature of ejecta–companion interaction, consistent with a 6 M_☉ main-sequence companion 29 R_☉ from the white dwarf, using the Kasen (2010) formulation.¹² However, Shappee et al. (2016b) find that other probes of the progenitor system do not support the SD interpretation for SN 2012cg.

Here, we present the early light curve and spectra of SN 2017cbv, which show a clear blue excess during the first several days of observations. This excess may be a more subtle version of that seen in SN 2012cg and iPTF14atg, but seen more clearly here thanks to denser sampling.

SN 2017cbv (a.k.a. DLT17u) was discovered on MJD 57822.14 (2017 March 10 UT) at a magnitude of $R\approx 16$ by the Distance Less Than 40 Mpc survey (DLT40; L. Tartaglia et al. 2017, in preparation), a one-day cadence SN search using a PROMPT 0.4 m telescope (Reichart et al. 2005), and was confirmed by a second DLT40 image during the same night (Valenti et al. 2017). Our last nondetection was on MJD 57791. The SN is located on the outskirts of the nearby spiral galaxy NGC 5643. Within hours of discovery (MJD 57822.7), the transient was classified as a very young SN Ia with the robotic FLOYDS spectrograph mounted on the Las Cumbres Observatory (LCO; Brown et al. 2013) 2 m telescope in Siding Spring, Australia (Hosseinzadeh et al. 2017).

UBVgri follow-up observations were obtained with Sinistro cameras on LCO's network of 1 m telescopes. Using lcogtsnpipe (Valenti et al. 2016), a PyRAF-based photometric reduction pipeline, we measured aperture photometry of the SN. Because the SN is bright and far from its host galaxy, image subtraction and PSF fitting are not required. Local sequence stars were calibrated to the L101 standard field observed on the same night at the same observatory site using UBV Vega magnitudes from Stetson (2000) and gri AB magnitudes from the SDSS Collaboration (2016).

The Swift satellite began observing SN 2017cbv on MJD 57822.52. Ultra-Violet Optical Telescope (UVOT; Roming et al. 2005) photometry is given in the UVOT Vega photometry system using the pipeline for the Swift Optical Ultraviolet Supernova Archive (SOUSA; Brown et al. 2014) and the zeropoints of Breeveld et al. (2010). Smaller hardware windows were used in the optical near maximum brightness in order to reduce coincidence loss and measure brighter magnitudes without saturation (Poole et al. 2008; Brown et al. 2012a). A few U- and B-band observations on the rising branch were saturated and are excluded.

Finally, we obtained absolute magnitudes by applying the distance modulus, μ = 31.14 ± 0.40 mag (16.9 ± 3.1 Mpc), of Tully (1988) and the Milky Way extinction corrections, $E(B-V)=0.15$ mag, of Schlafly & Finkbeiner (2011). We assume no additional host galaxy extinction, given the SN's position in the outskirts of NGC 5643 and a lack of narrow Na i D absorption in high-resolution spectra (D. J. Sand et al. 2017, in preparation).

Our photometry is shown in Figure 1. By fitting quadratic polynomials to the observed light curve, one around peak and one around +15 days, we find that SN 2017cbv reached a peak magnitude of B = 11.72 mag (${M}_{B}=-20.04$ mag) on MJD 57841.07, with ${\rm{\Delta }}{m}_{15}(B)=1.06$ mag. The decline rate of SN 2017cbv is near the average for normal SNe Ia (see, e.g., Figure 14 of Parrent et al. 2014). The peak absolute magnitude appears to be on the bright end of SNe Ia (Parrent et al. 2014), but this is uncertain due to the poorly constrained distance to NGC 5643. All figures use MJD 57821 as the nominal explosion date (see Section 4).

Zoom In Zoom Out Reset image size

The very early light curve of SN 2017cbv shows a prominent blue bump in the U, B, and g bands during the first five days of observation, also visible in its U − B and B − V colors (Figure 2). This indicates a high-temperature component of the early light curve in addition to the normal SN Ia behavior.

Zoom In Zoom Out Reset image size

There are several possibilities for the origin of the early light curve bump (see Section 5 for a discussion), but here we fit the analytic models of Kasen (2010) for a nondegenerate binary companion shocking the SN ejecta, as might be expected from the SD scenario. While in reality such a collision would be highly asymmetric, we use Kasen's analytic expressions for the luminosity within an optimal viewing angle. By further assuming that the shock consists of a spherical blackbody, Kasen arrives at the following equations for the photospheric radius and effective temperature:¹³

$$\begin{eqnarray}&&{R}_{\mathrm{phot}}=(2700\,{R}_{\odot }){x}^{1/9}{\kappa }^{1/9}{t}^{7/9},\end{eqnarray} \tag{ 1 }$$

$$\begin{eqnarray}&&{T}_{\mathrm{eff}}=({\rm{25,000}}\,{\rm{K}}){a}^{1/4}{x}^{1/144}{\kappa }^{-35/144}{t}^{-37/72},\end{eqnarray} \tag{ 2 }$$

where a is the binary separation in units of 10¹¹ m (144 R_☉), κ is the opacity in units of the electron scattering opacity (we fix $\kappa =1$ in our fits), and t is the time since explosion in days. In addition, we define

$$\begin{eqnarray}&&x\equiv \displaystyle \frac{M}{{M}_{\mathrm{Ch}}}{\left(\displaystyle \frac{v}{\mathrm{10,000}\mathrm{km}{{\rm{s}}}^{-1}}\right)}^{7},\end{eqnarray} \tag{ 3 }$$

where M is the ejected mass, ${M}_{\mathrm{Ch}}=1.4\,{M}_{\odot }$ is the Chandrasekhar mass, and v is the transition velocity between power laws in the density profile.

Our light curve model is the sum of the SiFTO template (Conley et al. 2008) to account for the normal SN Ia emission, and Kasen's shock model to account for the blue excess. We scale each band of the SiFTO template independently. In the U, B, V, and g bands, we fix the scaling factor to match the observed peak. In the r and i bands, we leave the scaling factor as a free parameter in order to account for any contribution from the shock in those bands around peak (predicted for some combinations of parameters). We also allow the time of B-band maximum light and the stretch (Perlmutter et al. 1997) of the SiFTO template to vary.

In total, we have eight parameters:

• 1.  
  The explosion time;
• 2.  
  The binary separation, a;
• 3.  
  $x\propto {{Mv}}^{7}$ (Equation (3));
• 4.  
  A factor on the r SiFTO template;
• 5.  
  A factor on the i SiFTO template;
• 6.  
  The time of peak;
• 7.  
  The stretch; and
• 8.  
  A factor on the shock component in U (see below).

We fit this combined model to our UBVgri light curve from LCO using a Markov Chain Monte Carlo routine based on the emcee package (Foreman-Mackey et al. 2013). We cannot include the Swift data in the fit due to a lack of early UV SN Ia templates. In addition to the photometric uncertainty, we add a 2% systematic (0.02 mag) uncertainty in quadrature as an estimate for our calibration uncertainties. For each of our observations, we find the expected R_phot and T_eff at that time, calculate the corresponding average L_ν in each filter, and compare that to our measured L_ν.

One caveat to our approach is that the SiFTO template may not describe the early light curve behavior of normal SNe Ia correctly in all filters. Given the small number of events observed 10–20 days before peak, reliable templates do not exist at these phases. However, we are encouraged by the fact that our SiFTO-based results qualitatively agree with other template fitters that we experimented with—SALT2 (Guy et al. 2007), MLCS2k2 (Jha et al. 2007), and the observed light curve of SN 2011fe (Zhang et al. 2016)—even at these early times.

We find that the SiFTO+Kasen model provides an excellent fit to our ground-based data (reduced ${\chi }^{2}=8.6$ after including our 2% calibration uncertainty), correctly predicting the blue bump at early times (Figure 3, top). Although at first glance our light curves do not look peculiar in the redder bands, emission from the shock model several weeks after explosion contributes to the observed luminosity around peak. Specifically, our best-fit model indicates that 5% and 15% of the r- and i-band peak luminosities, respectively, come from the shock component.

Zoom In Zoom Out Reset image size

The best-fit binary separation is 56 R_☉, implying a stellar radius of ∼20 R_☉ (assuming Roche lobe overflow; Eggleton 1983). 56 R_☉ is among the largest binary separations for SD SN Ia progenitors from binary population synthesis calculations (Liu et al. 2015). However, this value is quite sensitive to the early color evolution of our SN Ia template. As these templates may not be valid 15–20 days before peak, this result should be treated with caution. Furthermore, our simplified spherical model ignores the degeneracy between binary separation (a) and viewing angle; a bright (large a) off-angle shock looks similar to a faint (small a) shock along the line of sight.

Given the strong dependence of x on the transition velocity (∝v⁷), which is not observable, we cannot robustly estimate the ejecta mass. However, taking our best-fit value of $x=3.84\,\pm 0.19$ and assuming a Chandrasekhar mass of ejecta, we find a reasonable transition velocity of $v\approx 12000$ km s⁻¹ (subject to uncertainty in the distance modulus). The best-fit explosion time is MJD 57821.9, about 7 hr before discovery, and the best-fit time of B-band peak for the SiFTO component is MJD 57840.2. The best-fit stretch from the SiFTO template is 1.04.

Despite the success of the binary companion shock model in the optical, we required a scaling factor of 0.61 on the U-band shock component in order to fit the data. The UVW1, UVM2, and UVW2 emission are even further overpredicted by the shock model (Figure 3, bottom). We discuss the potential causes of this discrepancy in Section 5.

We obtained several additional optical and near-infrared (NIR) spectra of SN 2017cbv with FLOYDS and the SpeX NIR spectrograph (Rayner et al. 2003) at the NASA Infrared Telescope Facility, a selection of which are presented in Figure 4. As the event is ongoing at the time of publication, the full data set and analysis will be presented elsewhere (D. J. Sand et al. 2017, in preparation). In summary, the spectrum of SN 2017cbv greatly resembles the spectrum of SN 2013dy (Zheng et al. 2013) during the two weeks before maximum light, with Si ii and Ca ii absorption features weaker than the prototypical Type Ia SN 2011fe but stronger than the somewhat overluminous SN 1999aa (Garavini et al. 2004). Despite the spectroscopic similarity to SN 2013dy, the light curve of SN 2017cbv declines slightly faster: ${\rm{\Delta }}{m}_{15}(B)\,=1.06$ mag for SN 2017cbv versus 0.92 mag for SN 2013dy (Pan et al. 2015).

Zoom In Zoom Out Reset image size

We measure a Si ii λ6355 velocity of 22,800 km s⁻¹ in the initial spectrum of SN 2017cbv, 19 days before maximum light. The absorption feature just redward of Si ii could either be a lower-velocity component (9400 km s⁻¹) of the same line or C ii λ6580 at 19,800 km s⁻¹. We prefer the latter interpretation, as (1) it correctly predicts the tentative position of C ii λ7234 in the same spectrum and (2) 9400 km s⁻¹ would be uniquely low among very early SN Ia spectra (Silverman et al. 2012). The detection of unburned carbon can directly discriminate between the proposed SN Ia explosion mechanisms but is rarely seen in optical spectra even at these early times (Parrent et al. 2011; Silverman & Filippenko 2012), except for in super-Chandrasekhar SNe Ia (Howell et al. 2006). In SN 2017cbv, this feature disappears by day −13, reinforcing the need for early spectroscopy to fully account for unburned carbon.

The Si ii absorption feature at −13 days clearly shows signs of two velocity components at 16,500 and 10,500 km s⁻¹. It is likely that the earlier spectra of SN 2017cbv are dominated by the high-velocity component of Si ii, but we cannot fully trace the transition to low-velocity Si ii due to our lack of FLOYDS spectra between −13 and −2 days. We note that −13 days was also the approximate epoch at which SN 2012fr, another SN Ia with a prominent high-velocity Si ii component, began showing low-velocity Si ii (Childress et al. 2013). A similar multi-component velocity structure is evident for the Ca ii H&K feature in our pre-maximum spectra.

We fit the early velocity evolution of Ca ii H&K and the Si ii high-velocity component (the only one we can clearly identify at early times) to a ${t}^{-0.22}$ power law, as suggested by Piro & Nakar (2013) for finding the explosion time for SNe Ia. To do this, we mimicked the methodology of Piro & Nakar (2014) and allowed the power-law dependence to vary between ${t}^{-0.20}$ and ${t}^{-0.24}$ to estimate our uncertainties. This fit implies an explosion on MJD 57821.0 ± 0.3, 1.1 days prior to discovery and 0.9 days before the implied explosion time from our binary shock + standard SN Ia model presented in Section 3.

If SN 2017cbv had an SD progenitor, we might expect to see hydrogen in its late-time spectra (Mattila et al. 2005; Leonard 2007; Maeda et al. 2014). However, we do not detect Paβ emission in an NIR spectrum taken 34 days after maximum light (Figure 4, bottom). Following the method of Sand et al. (2016), we calculate a rough limit of ≲0.1 M_☉ of hydrogen by comparing to a spectrum of ASASSN-14lp at a similar phase, although this limit depends on the viewing angle.

The companion-shocking models provide a good fit to our optical data, but not to our UV data. This discrepancy is not likely to be a reddening effect (unless the reddening varies very quickly with time) because the UV luminosities around peak are not unusual for SNe Ia. However, it could stem from several simplifying assumptions in our model:

• 1.  
  Blackbody: The analytic models assume a blackbody spectrum for the shock component. However, the observed spectral energy distribution (SED) during the bump deviates significantly from a blackbody spectrum in the UV (Figure 5). A Swift grism spectrum taken during the bump is similar to UV spectra of other SNe Ia, showing significant absorption relative to a blackbody continuum (D. J. Sand et al. 2017, in preparation). This UV suppression is likely due to line blanketing (e.g., from iron lines). Any alternative model, companion shocking or otherwise, will need to account for this deviation from a blackbody spectrum.
• 2.  
  Constant Opacity: We fixed the opacity to be that of electron scattering throughout the first 40 days of evolution, whereas in reality the opacity should change over time as the ejecta cool. Opacity and/or line blanketing that vary with time could potentially explain the discrepancy between the models and our UV data.
• 3.  
  Density Profile: The shock models we quote here assume a broken power-law density profile for the ejecta: ${\rho }_{\mathrm{inner}}\propto {r}^{-1}$ and ${\rho }_{\mathrm{outer}}\propto {r}^{-10}$. The earliest emission, which should peak in the UV bands, depends strongly on the density of the outermost ejecta layers. In particular, a steeper density profile could suppress the early luminosity.
• 4.  
  Spherical Symmetry: In order to make the problem analytically tractable, we have ignored the asymmetry that must be present in a binary system. The analytic predictions are roughly equivalent to numerical predictions for a favorable viewing angle (see Figure 2 of Kasen 2010). If in reality we are viewing the collision off-axis, we might invoke a larger binary separation, ejecta mass, or ejecta velocity to match our observations. However, 3D numerical modeling of the ejecta–companion interaction would be needed to disentangle the early SN color diversity from the angular dependence of the shock component's color.

Zoom In Zoom Out Reset image size

If the companion-shocking scenario is correct, but the model is inaccurate in the UV, then the companion sizes found by Marion et al. (2016) and Cao et al. (2015) and the constraints of Brown et al. (2012b) would have to be reevaluated. However, if the UV overprediction is only due to line blanketing, which depends on temperature, the models may be more susceptible to failure in relatively low-temperature events. Alternatively, there could be another cause of early bumps in the UV or optical.

The observed bump could also be interpreted as a collision with CSM, rather than a collision with a companion star, as has recently been modeled by Piro & Morozova (2016). Kasen (2010) estimates that 0.01–0.1 M_☉ of CSM would be needed for a significant effect, and that the material would have to be located at distances comparable to the binary separation. Stellar winds would be unlikely to produce such a configuration, but a DD system could potentially eject enough mass to large enough distances during the pre-supernova accretion phase (Kasen 2010) or during the merger itself (Levanon & Soker 2017). A large mass of CSM would likely have decelerated the ejecta below the velocities we measure in Section 4, although an unshocked high-velocity component could still persist.

A third possibility is that the bump arises from a bubble of radioactive nickel that escaped most of the ejecta, allowing it to radiate light away faster than the typical diffusion timescale. Piro & Morozova (2016) explored various nickel distributions and their effect on early SN light curves. Shallow nickel distributions result in steeper, bluer early light curves. The early light curves in Piro & Morozova (2016) that included both CSM interaction and significant nickel mixing bear a qualitative resemblance to the light curve of SN 2017cbv. Likewise, Noebauer et al. (2017) find an early blue bump in their sub-Chandrasekhar double-detonation model, in which the initial helium detonation on the surface produces a small amount of radioactive material. We cannot rule out these possibilities, but more detailed modeling of this event in particular would be necessary to distinguish them from the companion-shocking case.

We have presented early photometry and spectroscopy of the Type Ia SN 2017cbv, which was discovered within ∼1 day of explosion. Its light curve shows a conspicuous blue excess during the first five days of observations. We find a good fit between our UBVgri data and models of binary companion shocking from Kasen (2010), but the fit overpredicts the observed UV luminosity at early times. This discrepancy might be due to several simplifying assumptions in the models. Alternatively, the excess emission could be due to interaction with CSM or the presence of radioactive nickel in the outer ejecta.

We observe no indication of ejecta interaction with hydrogen-rich material stripped from a companion star in the spectra of SN 2017cbv. However, more deep optical and near-infrared spectra out to the nebular phase are needed to confirm this finding. Intriguingly, we do detect unburned carbon in the earliest spectra at a level rarely seen in normal SNe Ia. A connection between an early light curve bump and the presence of unburned carbon could provide an important clue about SN Ia progenitors, but the scarcity of events with either of these observations prevents us from drawing any conclusions now.

Our analysis demonstrates the importance of (1) discovering and announcing SNe as early as possible and (2) obtaining extremely well-sampled follow-up light curves and spectral series. The DLT40 survey and LCO follow-up network are uniquely suited to find very young SNe and follow them with sub-day cadence for long periods. Such observations, which will become increasingly common in the coming years, greatly enhance our ability to confront theoretical models.

We thank Kyle Barbary for his help with sncosmo and Alex Conley and Santiago González Gaitán for providing the SiFTO templates. This work makes use of observations from the LCO network as part of the Supernova Key Project. A portion of the Swift observations were obtained by time allocated to the Danish astronomy community via support from the Instrument Center for Danish Astrophysics (PI: Stritzinger). G.H., D.A.H., and C.M. are supported by the National Science Foundation (NSF) under grant No. 1313484. D.J.S. and L.T. are supported by the NSF under grant No. 1517649. Support for S.V. was provided by NASA through a grant (program number HST-GO-14925.007-A) from the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Incorporated, under NASA contract NAS5-26555. Support for I.A. was provided by NASA through the Einstein Fellowship Program, grant PF6-170148. E.Y.H., S.D., and M.S. are supported by the NSF under grant No. AST-1613472 and by the Florida Space Grant Consortium. M.D.S. acknowledges support by a research grant (13261) from the Villum Fonden. D.J.S. is a visiting astronomer at the Infrared Telescope Facility, which is operated by the University of Hawai'i under contract NNH14CK55B with NASA. This research has made use of the NASA/IPAC Extragalactic Database, which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with NASA.