SN 2017dio: A Type-Ic Supernova Exploding in a Hydrogen-rich Circumstellar Medium^∗

Optical photometry was obtained in ugriz filters using ALFOSC⁴⁰ at the Nordic Optical Telescope (NOT), IO:O at the Liverpool Telescope (LT), and Spectral at the 2 m telescopes of Las Cumbres Observatory. JHK photometry was obtained using NOT/NOTCam at one epoch, 2017 July 27.

Standard reduction techniques were applied using Iraf,⁴¹ and aperture photometry was carried out for the SN and surrounding stars. Photometric zero points were derived by comparing the instrumental magnitudes of the stars in the field to photometry from Sloan Digital Sky Survey, Data Release 14.⁴² Additionally, pre-discovery V-band data points were obtained from the CRTS⁴³ archive and these were scaled to have the data point at 2017 May 01 match the g-band data point at the same epoch.

Optical slit-spectroscopy was obtained using ALFOSC at the NOT, EFOSC2⁴⁴ at the ESO New Technology Telescope (NTT, through the ePESSTO program), and SPRAT⁴⁵ at the LT. ALFOSC spectra cover 3300–9500 Å, with a resolution of 16 Å. With EFOSC2, grism #13, the coverage was 3500–9300 Å, at a 21 Å resolution. Additionally, in two epochs the EFOSC2 spectra were taken using grism #11 (16 Å resolution). SPRAT covers 4000–8000 Å, with a 18 Å resolution.

After the standard reduction procedures, the spectra were extracted, then wavelength and flux calibrated. Iraf, Foscgui,⁴⁶ and PESSTO pipeline⁴⁷ tools were used in the procedures.

At four days post-discovery, SN 2017dio was classified as an SN Ic. The subsequent spectra at +5 and +6 days (+0 days being the discovery date) are very similar in appearance to the classification spectrum (Figure 1). The spectra are relatively smooth, with broad features at wavelength ranges ∼4000–5500 and ∼8000–9000 Å. There are emission lines of H and He i at λλ5876, 7065. At +18 days, the continuum becomes almost featureless and after this point the spectrum starts to show evolving features. Both H and He emission lines are always present, and are resolved by our instruments, with Lorentzian full-width at half-maximum (FWHM) corresponding to an expansion velocity of ∼500 km s⁻¹ (corrected for instrumental broadening). These lines are therefore not originating from an underlying H ii region and are more likely attributed to the SN CSM.

Zoom In Zoom Out Reset image size

The early light curve (LC) of SN 2017dio was found to be rising (Figure 2). While the presence of narrow H emission features would put SN 2017dio into the SN IIn subclass, the underlying continuum does not show the typical blue featureless early spectrum (see, e.g., Kiewe et al. 2012; Taddia et al. 2013). At later phases, it became strikingly similar to some SNe IIn (Figure 1). The original SN Ic classification was obtained by omitting these emission lines and only inspecting the underlying spectrum. In the case of SNe IIn, the narrow emission lines are considered to be a sign of CSM, whereas the underlying SN in SNe IIn is hidden. SN 2017dio becomes quite clearly an SN IIn after +18 days, but at early phases shows broad absorption/emission features typical of an SE SN. As such, it provides a rare opportunity to investigate the underlying SN type in SNe IIn.

Zoom In Zoom Out Reset image size

A number of objects in the literature have been considered to be either SN Ia or Ic, such as SNe 2002ic (Hamuy et al. 2003; Benetti et al. 2006) and 2012ca (Fox et al. 2015; Inserra et al. 2016), due to the spectral similarities of SNe Ia and Ic particularly in the early phase. Nevertheless, the majority of these objects are believed to be SNe Ia embedded in a dense CSM (Silverman et al. 2013; Leloudas et al. 2015). PTF11kx (Dilday et al. 2012) is considered to be a clear example of a bona fide SN Ia with CSM interaction, as the SN was relatively interaction-free in the early phase. In Figure 1, it can be seen that there are similarities between SN 2017dio at the early phase with both SNe Ia and Ic; however, no exact match was found and its spectral appearance cannot be reproduced by any of those interacting SNe Ia.

One way to reassess the classification is by removing the spectral component attributed to the CSM interaction, and then cross-matching the spectrum with a template of SNe with various types. For this purpose, the +18 day spectrum was used as a proxy for the CSM interaction spectrum, and subtracted from the +6 day spectrum to obtain a "pure" SN component. Before subtraction, the spectra were normalized by the average flux across the wavelength range. The first spectrum in Figure 3 (left) shows the subtraction result, compared to several other SNe. The SNID (Blondin & Tonry 2007) and GELATO (Harutyunyan et al. 2008) tools were used to compare this spectrum with those of known SNe. Both tools yield an SN Ic classification around the maximum light, and a better match is obtained with broad-lined SN Ic (hereafter SN IcBL) spectra than with normal SN Ic. However, a precise classification is difficult at these epochs (Prentice & Mazzali 2017). Other SN types (Ia, Ib, II) do not provide a better match compared to SN Ic.

Zoom In Zoom Out Reset image size

To check the validity of this method, a reverse procedure was applied: the +18 day spectrum representing CSM interaction was coadded to templates of known SN spectra. Leloudas et al. (2015) simulated various pure SN spectra coadded to a CSM interaction component to investigate the spectral identification of SNe IIn, and showed that the flux ratio of the underlying SN to the continuum f_V is the most important parameter determining if an underlying, intrinsic SN spectrum can be classified correctly. Following Leloudas et al. (2015), we experimented with varying f_V to match the appearance of the coadded spectra with the +6 day spectrum. We found that the +6 day SN 2017dio spectrum is best matched by type-Ic SNe coadded to the CSM interaction spectrum with ${f}_{V}\gtrsim 0.3$ (Figure 3, right). This is consistent with the finding of Leloudas et al. (2015) that a critical ${f}_{V}\gtrsim 0.3$ is needed in order to correctly classify an underlying pure SN Ic spectrum contaminated by CSM interaction. Below f_V ≈ 0.3, these events would be indistinguishable from typical SN IIn without a hint of the nature of the underlying SN.

These analyses suggest that SN 2017dio underwent a transition from type Ic to IIn. The SN Ic spectrum was prominent in the early phases, while at later times the CSM interaction component dominates. A signature of the increasing strength of CSM interaction is the broadening of the emission lines. Figure 4 shows the evolution of the strongest H and He lines, showing their widths increasing with time. The Lorentzian FWHM of the lines was typically around 500 km s⁻¹ at the beginning and increased to ∼2000 km s⁻¹ in the later spectra. The line profile was initially symmetric with wings indicative of electron scattering (e.g., Dessart et al. 2016), then evolved into an asymmetric profile with a broad blue component. This evolution is similar to that seen in SN 2010jl, where such an asymmetric profile may be caused by dust (Gall et al. 2014), or alternatively radiative acceleration or radiative transfer effects in the cool dense shell (Fransson et al. 2014). Nevertheless, in the early phase of SN 2017dio, such a profile is not observed. The observed optical color evolution showing SN 2017dio becoming bluer with time and the Hα/Hβ line ratio staying relatively constant in these epochs do not support dust formation (Figures 2 and 5). Furthermore, the observed near-infrared colors are consistent with those expected for an SN IIn without any significant excess emission from dust (Taddia et al. 2013). Therefore, we do not expect dust to affect our observations.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

As the spectral evolution points to SN 2017dio being an SN Ic interacting with CSM, the LC should also match this interpretation. The underlying SN Ic luminosity must be fainter compared to the observed LC that includes the CSM interaction contribution. In Figure 2, we plot template SN Ic LCs from Taddia et al. (2015) to match our photometry. The epochs of the template LCs are the same, and they are only shifted vertically. It is apparent that the template LC peak corresponds to the epoch of the first spectra. This is consistent with the spectral match to SNe Ic around the peak. In a few days, the observed LC continued rising and there was a flux offset between the observed and template LCs. Since then, the subsequent photometric evolution can be explained by the increasing contribution from the interaction. The Hα line luminosity is generally increasing, suggesting an increasing strength of the interaction (Figure 5). Interaction appears to have started about the time of the SN Ic LC peak, yielding extra flux in the observed LC at subsequent phases, and eventually broadening the emission lines.

The observed LC of SN 2017dio peaked at around early 2017 July (MJD  ∼ 57930), about 60 days after the discovery and during the interaction dominated phase. The peak absolute magnitude was ${M}_{g}={M}_{r}=-18.8$ mag. Leloudas et al. (2015) estimated the peak luminosity needed for CSM interaction to hide an SN Ic/IcBL. A SN IIn needs to reach at least approximately −19 mag to hide an SN Ic peaking at approximately −17.5 mag. In Figure 2, the underlying photospheric SN Ic LC might have reached M_g = −17.6 mag at peak, which is consistent with the above scenario. A second solution to the LC peak can be placed between the last non-detection and the first photometric point (Figure 2); nevertheless, this would yield a similar, perhaps even fainter, peak magnitude. The corresponding photospheric peak magnitude of M_r = −18.1 mag is well within the observed distributions of SNe Ic and IcBL (${M}_{R}=-18.5\pm 0.8$ mag and ${M}_{R}=-19.0\pm 1.1$ mag, respectively; Drout et al. 2011). SNe Ia with CSM interaction are found to peak brighter than ${M}_{V}\sim -19.5$ mag (Silverman et al. 2013; Leloudas et al. 2015), which is constrained by the underlying SN Ia LC peaking between −18.5 and −19.7 mag. These luminosity ranges cannot accommodate both the observed LC peak and the underlying SN Ic photospheric LC peak in SN 2017dio. In SNe Ia with CSM interaction, the underlying SN is always spectroscopically similar to the SN 1991T-like objects, which is a brighter subgroup of SNe Ia. They are spectroscopically distinguishable from SN Ic (e.g., Leloudas et al. 2015) and the spectra of SN 2017dio do not show similarities with these objects. The extinction toward SN 2017dio appears to be negligible (i.e., no sign of Na I absorption or high Balmer decrement, and normal colors), ruling out a higher intrinsic luminosity that would be required by the scenario for an SN Ia with CSM interaction.

The consistency between the spectral and photometric evolutions suggests that the interpretation of SN 2017dio being an SN Ic interacting with a CSM and transitioning into an SN IIn is robust. Alternative explanations such as SN Ia, II, or Ib with CSM interaction are not plausible.

The field of SN 2017dio is in the SDSS database. The host galaxy, SDSS J113627.76+181747.3, has u = 23.36, g =21.12, r = 20.67, i = 20.55, and z = 20.12 mag. Considering the distance modulus and foreground reddening, the absolute magnitude is M_g = −15.1 mag. The galaxy appears to be small without a discernible shape. Its diameter of ∼3'' corresponds to ∼2 kpc at the distance of ∼160 Mpc. This suggests that the host is a dwarf galaxy smaller and fainter than the Small Magellanic Cloud (${M}_{V}=-16.8$ mag, McConnachie 2012).

In the spectra of SN 2017dio, the interstellar emission lines from the galaxy are hidden by the SN. Therefore, no strong-line analyses such as estimates of star formation rate or metallicity can be done. Employing the galaxy luminosity–metallicity relation of Tremonti et al. (2004) yields an estimate of 12 + log(O/H) ≈ 8.0 dex for the galaxy. Comparing this luminosity to the sample of SN Ic and IcBL hosts (Modjaz et al. 2008), the host of SN 2017dio falls at the faint end of the distribution. Dwarf galaxies with the size of SN 2017dio host predominantly produce SNe IcBL rather than normal SNe Ic (Arcavi et al. 2010). This is also consistent with the possibility that SN 2017dio is an SN IcBL.

The spectra and LC show that the CSM interaction was relatively weak at the earliest epochs. Therefore, the bulk of the CSM is not located at the immediate vicinity of the progenitor star. The SN ejecta must have traveled outward in a rarefied environment before encountering the denser CSM parts. The Hα/Hβ line ratio was around 3 in the early phases and two times higher after two weeks, consistent with increasing CSM densities. After this peak, the Hα/Hβ ratio drops and rises again, indicating fluctuations in the CSM density (Figure 5). In a $\rho \sim {r}^{-2}$ spherically distributed CSM created by steady-state stellar winds (Chevalier & Fransson 1994), such behavior is not expected.

At the early epochs, the emission lines are narrow with symmetric wings. They are likely to arise from continuous ionization by the interaction of the shock with the CSM, and effects of electron scattering within the CSM. The later spectra show asymmetric lines with broad blue wings (Figure 4). The asymmetric line profile may be caused by the interaction region being occulted at the receding side of the optically thick ejecta, thus causing the red wing to be suppressed. At the earliest epochs, the CSM interaction was weak, and thus the SN Ic spectral characteristics were visible. Later, when the CSM interaction dominates, the SN Ic features are hidden by the CSM interaction component.

Thus, the geometry of the CSM embedding SN 2017dio progenitor does not seem to be consistent with that generated from spherically symmetric mass loss via winds. Instead, it could have a clumpy, toroidal, or even bipolar distribution. With the averaged expansion velocity of 10,000 km s⁻¹ for SNe Ib/c (e.g., Cano 2013), the SN ejecta took 80 days to reach the part of the CSM corresponding to the peak Hα luminosity. This translates into a distance of ∼500 au. Presumably, this part of the CSM was created through mass loss with a velocity of 500 km s⁻¹. The luminosity of the Hα line can be used to estimate the mass-loss rate of the progenitor star responsible for the CSM build-up in the vicinity, as Hα luminosity is proportional to the kinetic energy dissipated per unit time. The observed peak Hα luminosity of SN 2017dio is ∼1.5 × 10⁴¹ erg s⁻¹. The peak Hα luminosity can be used to estimate the progenitor mass-loss rate by employing the relation

$$\begin{eqnarray}&&\dot{M}=2\displaystyle \frac{{L}_{{\rm{H}}\alpha }}{{\epsilon }_{{\rm{H}}\alpha }}\displaystyle \frac{{v}_{\mathrm{wind}}}{{({v}_{\mathrm{shock}})}^{3}}.\end{eqnarray} \tag{ 1 }$$

Assuming a progenitor wind velocity of 500 km s⁻¹ from emission line FWHM at the early epochs, and an efficiency factor ${\epsilon }_{{\rm{H}}\alpha }$ of 0.01 (Chevalier & Fransson 1994), the pre-SN mass-loss rate of the progenitor star was estimated to be $\sim 0.02{({\epsilon }_{{\rm{H}}\alpha }/0.01)}^{-1}({v}_{\mathrm{wind}}/500$ km s⁻¹) $({v}_{\mathrm{shock}}/$10,000 km s⁻¹)⁻³ M_⊙ yr⁻¹. The shock velocity must be lower compared to the velocity of the part of the ejecta that is not interacting with the CSM, but this is hidden by the electron scattering wings, which extend up to ∼10,000 km s⁻¹ (Figure 4). The derived progenitor mass-loss rate is comparable to some SNe IIn (Kiewe et al. 2012; Taddia et al. 2013). It is a few times lower compared to SN 2010jl ($\dot{M}\sim 0.1$ M_⊙ yr⁻¹, and consistently, so are the peak bolometric⁴⁸ and Hα luminosities (∼10⁴³ erg s⁻¹ and ∼10⁴¹ erg s⁻¹, respectively, compared to ∼3  ×  10⁴³ erg s⁻¹ and ∼10⁴² erg s⁻¹; Fransson et al. 2014).

The progenitor must therefore have experienced major mass loss a few decades before the SN. Then, the rest of the envelope was removed through less vigorous mass loss in the final years, until being removed completely prior to the SN. This is reminiscent of the inferred progenitor behavior of the type-Ib/IIn SN 2014C (Milisavljevic et al. 2015). High pre-SN mass loss may be caused by eruptions (e.g., Smith & Arnett 2014) or stripping by a close binary companion (e.g., Yoon et al. 2017). With a wind-driven mass loss, it is inconceivable for the progenitor to still retain the H-rich material nearby while the He layer is depleted, as observed in SN 2017dio. In the binary progenitor scenario, the primary transfers material to the secondary through a Roche-lobe overflow and evolves into a C+O star. With an increased mass, the secondary's evolution is accelerated and it leaves the main sequence before the explosion of the primary. The secondary then experiences a luminous blue variable (LBV) phase, or becomes a giant and starts a reverse mass transfer to the primary. At this point there is a large amount of H-rich CSM around the primary and it explodes as an SN Ic. This path could have occurred within the standard binary evolution, and the progenitor system may be somewhat similar to some Wolf-Rayet+LBV binaries such as HD 5980 (Koenigsberger et al. 2014). The prototypical type Ibn SN 2006jc experienced a giant outburst two years prior to the SN (Pastorello et al. 2007). A possible lower-mass companion star detected in late-time observations together with the lack of further outbursts after the explosion suggest that the outburst originated from the SN progenitor itself (Maund et al. 2016). Given that SN 2017dio could be an SN IcBL, we note that binary evolution may also play an important role for SNe IcBL.