ULTRAVIOLET SPECTROSCOPY OF TYPE IIB SUPERNOVAE: DIVERSITY AND THE IMPACT OF CIRCUMSTELLAR MATERIAL

We analyze four SNe IIb having high signal-to-noise ratio UV spectra obtained with HST within 30 days after the estimated explosion. This is an exhaustive list; to the best of our knowledge, no other early-time UV spectra of SNe IIb have been obtained with HST.

SN 2013df was discovered on 2013 June 7.78 (Ciabattari et al. 2013; UTC dates are used throughout this paper) in an outer spiral arm of NGC 4414 (distance d = 16.6 ± 0.4 Mpc; Freedman et al. 2001). A comparison of V-band LCs of SN 1993J, SN 2011dh, and SN 2013df led Van Dyk et al. (2014) to estimate an explosion date of 2013 June 4, which we adopt throughput this paper (see also Morales-Garoffolo et al. 2014). Van Dyk et al. (2014) identify the SN progenitor in pre-explosion HST images, arguing for an extended star with an effective radius of 545 ± 65 R_⊙ and an effective temperature of 4250 ± 100 K. The progenitor is likely a YSG with an initial mass of 13–17 M_⊙.

SN 1993J was discovered by amateur astronomer F. Garcia on 1993 March 28.9, at d = 3.6 Mpc in the spiral galaxy M81 (Freedman et al. 1994). We adopt an explosion date of 1993 March 27.5 (Filippenko et al. 1993). The SN 1993J progenitor was identified in ground-based pre-explosion images by Aldering et al. (1994). The progenitor is a YSG with a radius of ∼600 R_⊙ and a mass of 12–22 M_⊙ (Aldering et al. 1994; Maund et al. 2004; Fox et al. 2014). Maund et al. (2004) use photometric and spectroscopic observations taken 10 yr after the explosion to give evidence for a massive star at the SN location, claimed to be the binary companion of the progenitor.

SN 2001ig was discovered on 2001 December 10.43 by Evans et al. (2001) in the spiral galaxy NGC 7424 at d = 11.5 Mpc (Soria et al. 2006). We adopt an explosion date of 2001 December 3 based on radio LC modeling (Ryder et al. 2004). Using the same modeling, (Ryder et al. 2004) argue that the most likely progenitor of SN 2001ig is a W–R star with a radius of ∼10¹¹ cm. SN 2001ig is the only SN cIIb candidate for which early-time UV spectra were obtained with HST.

SN 2011dh was discovered by amateur astronomer A. Riou on 2011 May 31.83 (Arcavi et al. 2011) in the Whirlpool galaxy (M51) at d = 8.05 ± 0.35 Mpc (Marion et al. 2014). SN 2011dh was not detected down to g = 21.44 mag in Palomar Transient Factory images taken on 31.27 May 2011, and we adopt an explosion date of 31.5 May 2011 (Arcavi et al. 2011; Marion et al. 2014). Bersten et al. (2012) and Horesh et al. (2013) argue for an extended progenitor with a radius of ∼200 R_⊙. This was verified by Van Dyk et al. (2013), who determined that the YSG at the SN location vanished in post-explosion HST images; see also Maund et al. (2011) and Ergon et al. (2014a, 2014b).

We compare our LCs with those of two additional objects, SN 2008ax (Pastorello et al. 2008; Crockett et al. 2008) and SN 2011hs (Bufano et al. 2014), albeit these events have no early-time UV spectroscopy.

SN 2008ax was discovered independently by Mostardi, Li, and Filippenko on 2008 March 3.95, and by Nakano and Itagaki on 2008 March 5.12 (Nakano 2008), in the barred spiral galaxy NGC 4490 at d = 11.1 Mpc. Extremely tight limits on the explosion time of this event are placed by Arbour (2008), who monitored the host galaxy only 6 hr before detection. We adopt an explosion date of 2008 March 3.7 (Pastorello et al. 2008, and references therein). SN 2008ax is a candidate Type cIIb event (Chevalier & Soderberg 2010).

SN 2011hs was discovered on 2011 November 12.5 in the galaxy IC 5267 at d = 26.4 Mpc. We adopt an explosion date of 2011 November 6.5 based on a comparison of early-time photometric data with the SN 1993J LC (Bufano et al. 2014, and reference therein). Bufano et al. (2014) argue for an extended progenitor similar to the one observed for SN 1993J; see Section 3.1.

Table 1 lists all of these objects.

We observed SN 2013df with HST on four epochs: 2013 June 17.2, 22.1, 26.1, and 30.2 (respectively ∼13, 18, 22, and 26 days after the estimated explosion date; Program GO-13030, PI A. V. Filippenko). At each visit, three spectra of the SN were obtained: a UV spectrum using STIS/MAMA centered at 2376 Å, and two spectra using STIS/CCD centered at 4300 and 7751 Å. Data were reduced using the calstis software (Bostroem & Proffitt 2011).

A UV+optical spectrum of SN 1993J was obtained on 1993 April 15.4 using HST-FOS (∼19 days after the estimated explosion date; Program GO-4528, PI R. P. Kirshner; Jeffery et al. 1994). Data were reduced using the calfos software (Keyes et al. 1995).

UV+optical spectra of SN 2001ig were taken on two different epochs, 2001 December 14 and 22.6, using HST-STIS (∼11 and ∼19 days after the estimated explosion date; Program GO-9114, PI R. P. Kirshner; Marion et al. 2014). Data were reduced using the calstis software (Bostroem & Proffitt 2011).

A UV+optical spectrum of SN 2011dh was obtained with HST-STIS/CCD on 2011 June 24.1 (∼24 days after the estimated explosion date; Program GO-12540, PI R. P. Kirshner; Marion et al. 2014). Data were reduced using the calstis software (Bostroem & Proffitt 2011).

Table 2 summarizes all HST spectroscopic observations used in our analysis. Figure 1 shows the spectra described above, with insets highlighting the UV. Digital versions of these spectra are available from WISeREP (Yaron & Gal-Yam 2012).

Zoom In Zoom Out Reset image size

In order to set the context for the analysis of the UV emission, we first inspect the bolometric LCs of SNe IIb. We collected the LC of SN 1993J from Richmond et al. (1996), the LC of SN 2008ax from Pastorello et al. (2008), and the LC of SN 2011hs from Bufano et al. (2014).

In the case of SN 2011dh, the bolometric LC is taken from Marion et al. (2014). This LC starts four days after estimated explosion and does not include the initial adiabatic cooling phase. However, Arcavi et al. (2011) show that such a phase was detected using mostly unfiltered photometry from amateur astronomers (see Figure 2 inset). The duration of the shock-cooling phase for SN 2011dh is less than 3 days (Arcavi et al. 2011); see also Ergon et al. (2014a, 2014b).

Zoom In Zoom Out Reset image size

We construct a bolometric LC for SN 2013df from observations¹⁴ obtained with Swift-UVOT (UVW2, UVM2, and UVW1 bands), KAIT (B and V bands), and RATIR (r, i, z, J, and H bands). Our LC samples the period between 2013 June 13 and 2013 July 25. Magnitudes were corrected for host-galaxy and Milky Way extinction assuming A_V = 0.30 mag and using the Cardelli et al. (1989) reddening law (see Van Dyk et al. 2014 for a thorough discussion of extinction estimates for SN 2013df). Magnitudes were converted to quasi-monochromatic flux at the effective wavelength of each filter. A quasi-synthetic LC at each band was generated by fitting a spline to the measurements. In cases where the observations started on 2013 June 14 (KAIT B and V bands) and 2013 June 16 (RATIR zHK), an artificial data point was added on 2013 June 13 using backward extrapolation of the initial three measurements in each band. Varying this artificial point value by 20%, we find that this method introduces an error of <10% to the bolometric flux at this time. The monochromatic LCs were then summed using the trapezoidal integration method, with bins equal to the FWHM of each filter to derive a bolometric LC. In view of our well-sampled LC, and based on blackbody fits to the observed HST spectra, we estimate a coverage of ≥95% of the total bolometric flux, and no attempt was made to correct for emission below 2100 Å or above 17,800 Å. The bolometric LCs are plotted in Figure 2.

Comparing the bolometric LCs, we find that, with the exception of SN 2011hs, a similar width around peak magnitude is observed; see Figure 2. As noticed by Pastorello et al. (2008) for SN 2008ax and SN 1993J, this suggests a similar value for the $M_{{\rm ej}}^{3}$/E_k ratio according to Arnett's law (Arnett 1982). The difference in luminosity is attributed to small changes in the amount of ⁵⁶Ni ejected in the explosion (e.g., Roming et al. 2009; Maurer et al. 2010; Perets et al. 2010; Ergon et al. 2014a).

The most significant difference between the observed LCs is the early contribution from the shock-heated envelope. Of the four events for which the shock-cooling phase duration can be estimated, SN 2011dh exhibits the fastest decay, followed by SN 2011hs, SN 1993J, and SN 2013df, in agreement with Morales-Garoffolo et al. (2014).

Zoom In Zoom Out Reset image size

Nakar & Piro (2014) argue that for progenitors with a loosely bound envelope and an envelope mass much smaller than the core mass, the cooling-phase duration is proportional to the amount of mass concentrated at the stellar radius. This suggests that the envelope of SN 2013df is more massive than that of, for example, SN 1993J. However, this does not agree with our findings from spectral modeling (see Figures 3 and 4), as well as with the conclusions of Morales-Garoffolo et al. (2014). We therefore suggest that perhaps an extended shock-cooling phase might also be caused by the presence of CSM at the vicinity of the progenitor; see Sections 3.2 and 4. This scenario is also supported by the analysis of the Hα emission in nebular spectra presented by Morales-Garoffolo et al. (2014). In the case of SN 2008ax, the absence of a shock-cooling phase, despite early photometric measurements taken <6 hr from the time of explosion, led Pastorello et al. (2008) and Chevalier & Soderberg (2010) to conclude that it had a compact progenitor (R ≈ 1 R_⊙). We note that this also likely means it did not have dense CSM, and may also argue against a nearby binary companion (Crockett et al. 2008).

Zoom In Zoom Out Reset image size

Two events that sample the lower and upper energy regimes in our sample are SN 2011hs and SN 2001ig (Ryder et al. 2004, 2006; Bufano et al. 2014). In the case of SN 2011hs, the estimated ⁵⁶Ni mass and the energy are 0.04 M_⊙ and 8 × 10⁵⁰ erg, respectively. An extended progenitor (R ≈ 550 R_⊙) is most likely based on the observed initial decline in the LC (Bufano et al. 2014). For SN 2001ig, no reliable photometric data were found. We therefore integrate over the HST spectra in the wavelength range 5000–5800 Å to create two data points in the V band, and compare this result to the V-band LC of SN 2008ax and SN 2013df (Figure 2, inset). SN 2001ig is the most luminous event in our sample in V. Silverman et al. (2009) suggest that 0.13 M_⊙ of ⁵⁶Ni was ejected by this event, slightly lower than predicted by the peak V versus Ni-mass correlation of Perets et al. (2010), if we assume the second data point represents the SN peak magnitude. The effect of the higher luminosity on the UV emission from this event is discussed in Section 3.2.

SN 2013df spectra longward of ∼4000 Å exhibit a continuum with strong absorption lines, as expected in the photospheric phase of CC SNe. The spectra are dominated¹⁵ by H Balmer lines (6563, 4861, 4340 Å), He i lines (5875, 4471, 3888 Å), Ca ii H&K (3968, 3934 Å), the Ca ii near-infrared triplet (8662, 8542, 8498 Å), the O i triplet (7774 Å), and many features from iron-group elements (e.g., Fe ii lines at 5316, 5276, 4583 Å and Co ii at 4660 Å). Below ∼4000 Å, all spectra show an excess of emission with respect to a single-temperature model. The SED at these wavelengths is smooth, with a strong, broad, asymmetric line at around 2800 Å. Using SYNOW (Branch et al. 2007; Parrent et al. 2010), we can model the observed line profile with Mg ii, Co ii, Fe ii, and Ti ii at a temperature of 10,000 K. However, this exercise introduces other lines into the modeled spectra below 2800 Å which we do not observe.

The SN 1993J spectrum taken on 1993 April 15 is dominated by H Balmer lines at 11,000 ± 1000 km s⁻¹, He i λλ4471, 4921 at 9000 ± 1000 km s⁻¹, and the Ca ii H&K lines at 9000 km s⁻¹. Iron-group element lines (e.g., Fe ii, Ti ii, and Cr ii) are also present all across the observed spectrum. SN 1993J exhibits a smooth continuum below ∼3000 Å similar to the one observed in SN 2013df, with some features of iron-group elements identified above the continuum; for further details, see Section 3.3 as well as Baron et al. (1993).

The SN 2011dh spectrum taken on 2011 June 24 is dominated by Ca ii H&K and the Ca ii near-IR triplet at 9000 ± 400 km s⁻¹, H Balmer lines at velocities of 10,000 ± 250 km s⁻¹, He i lines (3888, 4471, 4921, 5015, 5875, 6678, and 7065 Å) at velocities of 7600 ± 200 km s⁻¹, and the O i triplet (7774 Å) at a velocity of 6700 km s⁻¹. Contributions from Fe, Ti, and other iron-group elements are also evident all across the observed spectrum at velocities of 7500 ± 500 km s⁻¹. At UV wavelengths, the SED is smooth and featureless as in SN 2013df and SN 1993J, but the flux excess is significantly lower than in the latter two cases (Figure 1, bottom-left panel).

SN 2001ig spectra taken on 14 and 2001 December 22 (11 and 19 days past the estimated explosion date) are dominated by He i lines (3888, 4471, 4921, 5012, 5875 Å), Fe ii λ5169, and H Balmer lines. The contribution from Ca ii to the feature around 3900 Å can also be inferred from the analysis done on the optical and near-IR spectra (Silverman et al. 2009). At shorter wavelengths, absorption by iron-group elements, mainly Fe ii and Co ii lines, dominates the SED. All of the observed features strengthen in the later spectrum. Heavy-element lines point to a high expansion velocity of the ejecta (∼10,000 km s⁻¹), while the H and He lines are observed at velocities of ∼13,000 km s⁻¹.

In order to study the ejecta structure of SN 2013df (abundances, density profile), we calculate radiative-transfer models. In addition to modeling the four UV-optical spectra we have of SN 2013df, we also modeled the other SNe in our UV spectroscopic sample.

Modeling SN 2013df, we assume approximate spherical symmetry and follow the "abundance tomography" approach of Mazzali et al. (2014). We assume a density profile for the SN and then infer an abundance stratification from photospheric spectra in the phase of homologous, force-free expansion. Doing this, we exploit the fact that the photosphere recedes with time. Early-time spectra show the imprint of the outermost layers of the envelope, and later spectra are dominated by material further inward. Thus, starting from reasonable initial values, the abundances can be inferred step by step. The earliest spectrum is used to determine optimum-fit abundances in the outer layers, and later spectra to subsequently find the abundances in the layers below, while keeping the abundances in the outer layers fixed. This process requires iteration in order to be optimized. We have repeated the fitting procedure several times with different density profiles in order to ensure that the profile finally used reasonably describes the actual envelope.

The spectral synthesis is performed with our spherically symmetric Monte Carlo radiative-transfer code (Mazzali et al. 1993; Lucy 1999; Mazzali 2000a), including abundance stratification (Stehle et al. 2005) and a module that calculates the ionization of H and He in full non-local thermodynamic equilibrium (NLTE; Hachinger et al. 2012). The code calculates the radiation field within a spherically symmetric SN expanding "atmosphere," producing synthetic spectra. As input parameters, it reads the abundance and density stratification, as well as the luminosity of the SN, and the radius/velocity of the "photosphere" (the lower boundary of the atmosphere) at each epoch t. The code simulates the propagation of energy packets, starting from the photosphere where a blackbody spectrum is assumed to be emitted. The packets interact with atomic lines and free electrons. The state of the plasma, which is needed to calculate these interactions, is computed in NLTE for H and He. Our NLTE module (Hachinger et al. 2012) takes into account heating by Compton electrons produced by γ-rays from ⁵⁶Ni and ⁵⁶Co decay, using the methods of Lucy (1991). The required heating rates are computed using a modified version of the LC code developed by P. A. Mazzali and used by Cappellaro et al. (1997). In order to arrive at a consistent solution for both the radiation field and plasma, the radiation Monte Carlo routine and the plasma state routine of the code are iteratively called in turn. After convergence, an output spectrum is calculated from the formal integral of the transfer equation.

The density model we use for SN 2013df is derived from model 13C of Woosley et al. (1994), who set up a model sequence to explain the prototypical SN 1993J. Model 13C has been successfully used by Tsvetkov et al. (2009) to model SN 2008ax. It turns out that the high-velocity tail of 13C produces somewhat too strong H lines. Therefore, we tried different ad-hoc modifications of the model (mostly downscaling the density above a certain threshold velocity/radius). A model that describes SN 2013df well was obtained by scaling the density above 11,000 km s⁻¹ to one half of its original value; see Figure 4. The original 13 C model (Woosley et al. 1994) assumes a progenitor mass of 3.7 M_⊙ (1.5 M_⊙ of which end up in the remaining neutron star), and our model is only slightly less massive (by 0.04 M_⊙). The explosion energy in 13 C was 1.2 × 10⁵¹ erg, which is reduced to 1.1 × 10⁵¹ erg in our model. Our estimate of the progenitor mass at the time of explosion is higher than the one derived using the relation between the He-core mass and main sequence mass (Sugimoto & Nomoto 1980) by Morales-Garoffolo et al. (2014). We argue that the overall characteristics of SN 2013df are more in agreement with SN 1993J than with SN 2011dh (e.g., SED and duration of shock-cooling phase), and therefore a progenitor mass similar to that of SN 1993J is more plausible. Moreover, our explosion-energy estimate is similar to that derived by Morales-Garoffolo et al. (2014), and assuming a lower progenitor mass at the time of explosion results in higher expansion velocities in our models than the expansion velocities observed.

The abundances we assume as a starting point are taken from adequate zones of the models for SN 2008ax presented by Hachinger et al. (2012). Only the outermost 0.1–0.2 M_⊙ of our envelope are H-enriched; the material below mostly consists of He (and C/O in the deeper layers). Taking into account the lower density profile we adopt (Figure 4), this suggests a lower H-envelope mass for SN 2013df in comparison to SN 1993J. Under these constraints, we optimized the abundances, inferring the best-fitting values. The synthetic spectra obtained using our radiative-transfer code are given in Figure 5. Our models suggest a blackbody temperature of 8200 K and an effective photosphere temperature of 5650 K in the spectrum taken on 2013 June 17, dropping to a blackbody temperature of 6800 K and an effective photosphere temperature of 5300 K in the spectrum taken on 2014 June 30, in agreement with Morales-Garoffolo et al. (2014). The photospheric velocity also drops from 9600 to 5900 km s⁻¹. Velocity-space stratification for the model describing the spectrum taken on 2013 June 30 is given in Figure 3. Only the outermost 0.1–0.2 M_⊙ of our envelope are H-enriched; the material below mostly consists of He and C/O in the deeper layers. Traces of iron-group elements are also present at high velocities (i.e., above 10,000 km s⁻¹).

Zoom In Zoom Out Reset image size

We find a good match between the observed and synthetic spectra at wavelengths above ∼5000 Å. At shorter wavelengths, our models simulate most of the observed lines (mainly Fe ii and Ca ii), but the observed spectrum seems to include an additional component that is not affected by line absorption and is responsible to most of the emitted flux. This trend is consistent in all four simulated spectra, and is stronger at shorter wavelengths. In an attempt to increase the UV flux, which in our models is absorbed by iron-group elements in the UV (Fe, Ti, Cr), we have lowered the metal content in the outer layers of the ejecta, but were unable to reproduce the observed iron-group element lines above ∼4000 Å. Increasing the metal content in the outer layers to allow higher probability for reverse fluorescence induces line structure in the UV that is not present in the observed spectra. We therefore conclude that the source of the UV emission is above the photosphere, and is most likely associated with CSM shocked by the SN blast wave. This argument is supported by the flat UV LC after the first peak, as well as by the nebular spectral analysis, both presented by Morales-Garoffolo et al. (2014).

The models for SNe 1993J, 2001ig, and 2011dh are set up as minimal modifications of our SN 2013df model. We have chosen this strategy, partly justified by the observed similarities in the bolometric LCs (Section 3.1), so as to bring out the principal physical changes from one to the other SN—even if more detailed differences in the density structure, etc, between the SNe may thus not be captured in our study.

With this approach, one spectrum of each comparison SN is sufficient in order to allow for a basic comparison. We started out rerunning the SN 2013df model with the modified values for epoch and basic SN parameters (such as distance and reddening), and a starting guess for photospheric velocity and luminosity. Usually, we arrived at a well-fitting model upon further optimizing the latter two parameters, and only slightly changing the abundance stratification.

For SN 1993J, a detailed analysis of the spectrum based on the Woosley et al. (1994) 13 C model we have used can be found in Woosley et al. (1994). The model suggests a blackbody temperature of 8200 K, a photosphere effective temperature of 5900 K, and a photosphere velocity of 8250 km s⁻¹. For SN 2011dh, our models suggest a blackbody temperature of 7470 K and an effective photospheric temperature of 5900 K, slightly lower than the results of Ergon et al. (2014a). The photospheric velocity we infer is 6750 km s⁻¹, in agreement with Ergon et al. (2014a).

For both SN 1993J and SN 2011dh, we find a similar trend to the one observed for SN 2013df, in which an additional component not affected by line absorption is responsible for most of the emission below ∼4000 Å. The relative amount of flux attributed to this component is weakest in the case of SN 2011dh, and similar in the case of SN 1993J, as seen in the insets of Figure 6.

Zoom In Zoom Out Reset image size

In the case of SN 2001ig, our models suggest a photosphere velocity of 8750 km s⁻¹, a photospheric temperature of 7000 K, and a blackbody temperature of 12,000 K, in agreement with Silverman et al. (2009). The UV flux ratio below 3000 Å is much lower. Iron-group element lines are clearly seen in this wavelength range, and we obtain a good match between our synthetic spectrum and the observed one at all wavelengths; see Figure 7. In addition, the line strength in the UV is directly proportional to the flux at longer wavelengths, as seen by comparing the spectra taken on 2001 December 14 (11 days after estimated explosion) and on 2001 December 22 (19 days after estimated explosion); see the bottom-right panel of Figure 1. We conclude that the SN 2001ig UV flux originates from reverse fluorescence of photons at longer wavelengths by excited iron-group elements in the expanding ejecta, and is similar to the observed UV spectrum of SNe Ia (e.g., Panagia 2007; Hachinger et al. 2012; Mazzali et al. 2014); see also Silverman et al. (2013) for a discussion of SNe Ia interacting with CSM. As mentioned in Section 3.1, this SN is intrinsically the brightest event in our sample, and it synthesized the most ⁵⁶Ni (Silverman et al. 2009).

Zoom In Zoom Out Reset image size