iPTF 16hgs: A Double-peaked Ca-rich Gap Transient in a Metal-poor, Star-forming Dwarf Galaxy

iPTF 16hgs (=SN 2016hgs) was discovered by the intermediate Palomar Transient Factory (iPTF) (Law et al. 2009; Rau et al. 2009; Cao et al. 2016; Masci et al. 2017) and was detected first in r-band photometry taken with the CFH12K 96-Megapixel camera (Rahmer et al. 2008; Law et al. 2010) mounted on the 48 inch Samuel Oschin Telescope at Palomar Observatory (P48), on 2016 October 20.32 (MJD 57681.32)¹⁵ at J2000 coordinates α = 00^h50^m5139. δ = +27°22'480. The source was discovered at an apparent magnitude of r ≈ 18.9 mag, while it was not detected on 2016 October 8.01 (MJD 57669.01; 12.31 days before discovery) up to a limiting magnitude of r ≥ 20.8.

The transient was found in the outskirts of a nearly edge-on spiral host galaxy with a photo-z of 0.017, and at a projected offset of ≈17'' (≈1.9 R_eff) from the nucleus (Figure 1). We obtained a spectrum of the source on 2016 October 22 with the Discovery Channel Telescope (DCT), to find a Type Ib-like SN spectrum, similar to the pre-peak photospheric spectrum of the Ca-rich transient PTF 10iuv for the assumed photo-z of the host galaxy. A subsequent spectrum of the apparent host galaxy confirmed a redshift of z = 0.017, corresponding to a luminosity distance of D_L = 73.8 Mpc. Subsequent follow-up photometry and spectroscopy of the transient revealed that the source exhibited a faint peak absolute magnitude (M_r ≈ −15.5) and early transition (at ≈+30 days) to a nebular phase dominated by [Ca ii] emission, thus leading to its classification as a Ca-rich gap transient.

Zoom In Zoom Out Reset image size

We obtained r- and g-band photometry of iPTF 16hgs with the P48 CFH12K camera, along with gri-band photometry with the automated 60 inch telescope at Palomar (P60; Cenko et al. 2006). PTF image reduction is presented in Laher et al. (2014), and its photometric calibration and magnitude system are described in Ofek et al. (2012). P48 images were reduced with the Palomar Transient Factory Image Differencing and Extraction (PTFIDE) pipeline (Masci et al. 2017), which performs host-subtracted point spread function (PSF) photometry, while the P60 images were reduced using the pipeline described in Fremling et al. (2016). We correct all our photometry for galactic extinction for $E(B-V)=0.056$ from the maps of Schlafly & Finkbeiner (2011). We do not correct for any additional host extinction because we do not detect any Na D absorption at the host redshift in our spectra (Section 2.3).

We show the multicolor light curves for iPTF 16hgs in Figure 2, while the data are presented in Table 1. As shown, the source exhibited a double-peaked light curve in all photometric bands where we had early time coverage. For all subsequent discussions, we refer to the second peak of the light curve as the main peak and specify all observation phases with respect to this peak.

Zoom In Zoom Out Reset image size

We obtained spectroscopic follow-up of the transient, from ≈−8 days to ≈+59 days after r-band peak, using the DeVeny spectrograph on the Discovery Channel Telescope (Bida et al. 2014), the Double Beam Spectrograph (DBSP) on the 200 inch Hale telescope (Oke & Gunn 1982), and the Low Resolution Imaging Spectrograph (LRIS) on the Keck I telescope (Oke et al. 1995). Our last spectrum of the source obtained at +59 days after r-band peak was in the form of a spectroscopic mask observation intended to characterize the host environment of the transient. We present our sequence of spectra in Figure 3 (the spectroscopy epochs are indicated as dashed lines in Figure 2), while the spectroscopic observations are summarized in Table 2. We discuss the spectroscopic evolution of the source in Section 3.2. We also obtained a spectrum of the nucleus of the host galaxy of iPTF 16hgs with LRIS, which was found to exhibit prominent emission lines of Hα, Hβ, [N ii], [S ii], [O ii], and [O iii], as shown in Figure 4.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

All spectra and photometry will be made available by the WISeREP repository (Yaron & Gal-Yam 2012).¹⁶

We obtained X-ray follow-up of the transient with the Swift X-ray telescope (XRT) (Burrows et al. 2005). The source was observed on 2017 February 17 (MJD 57801; Phase ∼+107 days) for a total exposure time of 5 ks, and the data was processed with the HEAsoft package.¹⁷ No source was detected at the location of the transient to a 3σ upper limit of 2.2 ×  10⁻³ counts s⁻¹, corresponding to an unabsorbed 0.3–10 keV flux upper limit of 7.5 × 10⁻¹⁴ erg cm⁻² s⁻¹ for a photon index of Γ = 2.¹⁸ This constrains the unabsorbed X-ray luminosity from the source to <4.9 × 10⁴⁰ erg s⁻¹.

The Swift Ultraviolet/Optical telescope (UVOT) (Roming et al. 2005) also simultaneously observed the field in the UVW2 band. No source was detected at the transient location up to a 5σ limiting AB magnitude of 23.05.

2.5.1. Host Environment Spectroscopy

2.5.2. Late-time Imaging and Spectroscopy

2.5.3. Host IFU Observations

Owing to its apparent close location to its host galaxy and potential proximity to relatively dense ISM, we initiated deep radio follow-up of the transient, in order to constrain the presence of a radio counterpart, as potentially expected in some proposed models for Ca-rich gap transients (e.g., tidal detonations of white dwarfs that produce a relativistic jet, or a core-collapse explosion of a massive star).

2.6.1. AMI Observations

2.6.2. JVLA Observations

We obtained radio observations of the transient with the Very Large Array (VLA) under the Director's Discretionary Time (DDT) program (17A–427; PI: De). The VLA observed the source on 2017 June 24 (in C configuration; ∼250 days after r-band peak) at X band (centered at 10 GHz) for a total on source time of ≈1.8 hr, with the Wideband Interferometric Digital Architecture (WIDAR) correlator configured in continuum mode with 4 GHz bandwidth. The data were flagged and calibrated with the VLA calibration pipeline, while deconvolution and imaging was performed with standard routines in CASA. Source 3C48 was used as the flux and bandpass calibrator while source J0042+2320 was used as the phase calibrator. The final processed image has a noise rms of ≈2.6 μJy/beam.

Although we find a faint radio source very close to the transient location, its position is offset by 5'' from the source and coincident with a background galaxy in the late-time LRIS image—and hence not associated with the transient. We also obtained a spectrum of the background galaxy in our spectroscopic mask observation (Section 2.5.1). We found it to be consistent with an active galactic nucleus (AGN) at a redshift of 0.362, and clearly unrelated to the transient. No other source is detected at the transient location up to a 3σ limiting flux density of ≈7.8 μJy (Figure 5). At the redshift of the host galaxy, the flux upper limit constrains the 10 GHz radio luminosity to ≲5.1 × 10²⁵ erg s⁻¹ Hz⁻¹.

Zoom In Zoom Out Reset image size

2.6.3. uGMRT Observations

We first analyze the properties of the light curve of iPTF 16hgs, which we show to be unique in comparison to other known Ca-rich gap transients. The multicolor light curves of iPTF 16hgs are shown in Figure 2. The light curve of iPTF 16hgs shows clear evidence for two distinct components—an early declining phase (which was caught at discovery), followed by rebrightening to a second (main) peak that was followed up to late times. The early declining phase was detected in both the r and g bands, and was characterized by significantly bluer colors than the rest of the light curve, as evident from the rapid early decline in the g band.

Figure 6 compares the multicolor light curves of iPTF 16hgs to that of other known Ca-rich gap transients from PTF (Kasliwal et al. 2012; Lunnan et al. 2017). As shown, the second light curve peak of iPTF 16hgs is well matched to that of the light curves of the other Ca-rich transients, although the rapid decline from the first peak distinguishes it from the entire sample. At the same time, the i-band light curve of iPTF 16hgs is fainter than all the other events starting from ≈10 days after r-band peak. Because some of the previously known transients have very well-sampled early light curves, we can rule out the possibility that a first peak at similar timescales was missed in the cases of PTF 10iuv, PTF 11kmb, and PTF 12bho. However, a similar feature cannot be ruled out in the cases of PTF 09dav and PTF 11bij, which had sparse photometric coverage before peak light.

Zoom In Zoom Out Reset image size

Figure 6 also shows the g − r and r − i color evolution of iPTF 16hgs compared to the Ca-rich transient PTF 10iuv (which had good multicolor photometric coverage). As shown, iPTF 16hgs exhibited very blue colors at early times after discovery (with g − r ≈ −0.2), but subsequently exhibited rapid reddening over the next ≈20 days, evolving to g − r ≈ 1.5 at ≈10 days after the r-band peak. The r − i color also exhibits reddening with time, although the evolution is much slower. For comparison, the color curves of PTF 10iuv also exhibited similar but less rapid reddening with time, although iPTF 16hgs remained redder at similar light curve phases.

We fit a third-order polynomial to the main peak of the r-band light curve (which is sampled best), to find a peak magnitude of M_r = −15.65 and a peak time of MJD 57691.59. All phases mentioned in this paper are with respect to this epoch. We note that the peak absolute magnitude is similar to that of several previously confirmed Ca-rich gap transients. Determining the explosion time for this transient is non-trivial due to the presence of the early declining emission, thus precluding a conventional t² fit to the early light curve. Additionally, the first rise is not sampled, due to a gap in coverage of ≈12 days between the last non-detection and the first detection. Nonetheless, we try to estimate the rise time by fitting a parabolic function to the main peak of the light curve in flux space, and find a best-fit rest frame rise time of ≈9.9 days. Based on the last non-detection, we can put an upper limit of ≈12.3 days on the rise time of the first peak.

Photospheric phase spectra taken near the second peak show typical features of Type Ib SNe, most notably prominent lines of He i, O i, Mg ii, and Ca ii. We compare the photospheric phase spectra of iPTF 16hgs to example spectra of other Ca-rich transients in Figure 7. The photospheric spectrum of iPTF 16hgs exhibits a number of good similarities to that of PTF 11kmb and PTF 10iuv, most notably in the strong He features. He features in photospheric spectra are indeed common in many Ca-rich transients (PTF 11kmb, PTF 10iuv, SN 2005E, and SN 2007ke), although it is not used to exclusively define this class. Indeed, as noted by Lunnan et al. (2017), the photospheric phase diversity may point to different progenitor channels of this observationally defined class. From the peak photospheric spectrum (taken at 0.72 days after r-band peak in the source rest frame), we measure a photospheric velocity of ≈10,000 km s⁻¹ for this source.

Zoom In Zoom Out Reset image size

We identify prominent features in the −4.28 day spectrum of iPTF 16hgs with SYNOW (Fisher 2000) in Figure 8 (note that SYNOW may not be self-consistent and is only suggestive for line identification). We constrain the fit by matching the visible features along with the relative strengths of the features in order to constrain the composition of the ejecta as well as the photospheric velocities of the lines . The most prominent features include He i, Mg ii, Si ii, O i, and Ca ii, along with weaker features of Fe ii and Al ii. The SYNOW model uses a continuum temperature of 8000 K and velocities in the range of 8000–12,000 km s⁻¹, with He i found to be at the highest velocity of ≈12,000 km s⁻¹. The velocities of other prominent ions include O i at 10,000 km s⁻¹, Mg ii at 8000 km s⁻¹, Si ii at 9000 km s⁻¹, Ca ii at 8000 km s⁻¹, Al ii at 10,000 km s⁻¹, and Fe ii at 8000 km s⁻¹. Overall, the SYNOW fit fairly reproduces all the absorption features in the spectrum.

Zoom In Zoom Out Reset image size

Figure 7 also shows a comparison of a nebular spectrum of iPTF 16hgs to that of other Ca-rich gap transients. These spectra exhibit weak continua superimposed with strong forbidden and permitted lines of Ca and O. In particular, iPTF 16hgs exhibits the characteristic nebular features of this class, i.e., strong [Ca ii] λλ7291, 7324 emission combined with relatively weak [O i] λλ6300, 6363 emission (with a [Ca ii]/[O i] ratio of ≈7). For comparison, we also show the photospheric and nebular phase spectra of SN 2009jf (Valenti et al. 2011) which exhibits significantly stronger [O i] emission than [Ca ii] in the nebular phase. The [Ca ii]/[O i] ratio has been used as a defining feature of the class of "Ca-rich transients," e.g., Milisavljevic et al. (2017) suggest that a [Ca ii]/[O i] ratio of >2 separates the class of Ca-rich transients from other Type Ib/c SNe. We analyze this issue further in Section 7.

In Figure 9, we show a comparison of the velocity profiles of the nebular [Ca ii] and [O i] emission lines. Interestingly, this last nebular spectrum taken at ≈+60 days shows clear evidence of host galaxy emission features of Hα and [O ii] λλ3727, 3729 (on the blue side). These were also detected in a late-time Keck LRIS spectrum taken after the transient faded away. Hence, as opposed to other Ca-rich gap transients that are found in the far outskirts of their host galaxies (Kasliwal et al. 2012; Lunnan et al. 2017), iPTF 16hgs shows evidence of being located inside its host galaxy. However, we cannot rule out a scenario where the transient was located significantly behind the host galaxy, such that the galaxy emission features arise from the foreground. We note that the narrow Hα feature is found superimposed on a broader underlying component (Figure 9), that could potentially be associated with hydrogen in the ejecta. A similar H feature was also observed in the nebular spectrum of PTF 09dav (Kasliwal et al. 2012), and potentially in PTF 11kmb as well (Milisavljevic et al. 2017). However, Milisavljevic et al. (2017) suggest that the feature could also be associated with Ca I] λ6572.

Zoom In Zoom Out Reset image size

Taken together, we have thus far demonstrated that iPTF 16hgs exhibited (1) a low peak luminosity of M_r ≈ −15.6, (2) rapid photometric evolution similar to other known Ca-rich gap transients, (3) normal photospheric velocities (∼10,000 km s⁻¹), (4) early transition to a nebular phase (at ≈30 days after r-band peak), and (5) a nebular phase spectrum dominated by [Ca ii] emission. Thus, despite its unique light curve, iPTF 16hgs falls squarely in the class of Ca-rich gap transients.

We begin our modeling by constructing a bolometric light curve for the transient. For epochs where we have contemporaneous photometry in the gri bands, we compute a pseudo-bolometric luminosity by performing a trapezoidal integration of the multicolor fluxes up to the edges of the g and i band. In order to account for additional flux below 4000 Å and above 8000 Å, we use the peak photospheric spectrum (at +0.72 days) to determine the fraction of flux missed between 3000 Å and 10000 Å. We find that the gri trapezoidal integration misses 23% of the total flux, and hence we scale all the photospheric phase fluxes the peak by a factor of 1.3. We also add a 5% uncertainty to the computed luminosities to account for potential uncertainties in the fraction of flux missed.

Although the bolometric luminosity for the first peak is important to understand its origin, we do not have photometry in the i band for the early peak. Hence, we use the only spectrum taken within the early decline (at −8.06 days) to find the fraction of flux missed between 3500 Å and 8000 Å in a gr trapezoidal flux estimate (≈33%), and scale all the gr trapezoidal fluxes within the first peak. Finally, we have one epoch of contemporaneous ri detection at ≈+35 days. For this epoch, we use the +28 day spectrum to similarly estimate the fraction of flux missed between 3000 Å and 10000 Å (≈50%), and scale the ri trapezoidal luminosity to account for it. Note that the pseudo-bolometric luminosity estimates above are strict lower limits on the total bolometric luminosity from the source.

The bolometric light curve thus obtained is shown in Figure 10, clearly exhibiting a rapidly declining and relatively luminous early peak, followed by a rise to a second (main) peak and subsequent decline. The main peak reaches a peak bolometric luminosity of ≈3 × 10⁴¹ erg s⁻¹, while the peak luminosity for the first component was at least ≈4 × 10⁴¹ erg s⁻¹. The last photometric data point around 30 days after r-band peak indicates that the source faded to ≲4 × 10⁴⁰ erg s⁻¹ by this epoch.

Zoom In Zoom Out Reset image size

We first try to understand the power source for the main peak of the light curve. Because the main peak is very similar to that observed in other Ca-rich gap transients, we consider a radioactively powered light curve for the main peak. We fit a simple Arnett model to the bolometric light curve, assuming that the decay of ⁵⁶Ni powers the main peak of the light curve. Note that the Arnett model has several simplifications, including homologous expansion, spherical symmetry, constant opacity, and centrally located ⁵⁶Ni in the ejecta. We do not include data points within the first peak (at <−5 days from r-band maximum), nor the last data point, because the Arnett model is only valid in the optically thick photospheric phase.

We use the analytic relations presented in Lyman et al. (2016a) and Valenti et al. (2014) for this fitting, for which the only free parameters are the diffusion time through the ejecta τ_M and the Ni mass M_Ni. Keeping the explosion time as an additional free parameter, we get the best-fit Arnett model as shown by the red dashed line in Figure 10. As shown, the model reasonably reproduces the bolometric evolution of the main peak, although there are clear discrepancies near peak light. This is not unexpected, as the Arnett model is very simplified; for instance, it ignores potential effects of ⁵⁶Ni mixing in the outer layers of the ejecta, which can significantly affect the rise of the light curve. The best-fit model indicates a ⁵⁶Ni Mass of 8 × 10⁻³ M_⊙ and diffusion time of ≈5.9 days. The best-fit explosion time is 9.96 days before r-band peak, which is very close to our initial estimate based on the t² law fitting to the r-band light curve.

We note that the last luminosity estimate from ≈45 days after the explosion is clearly much fainter than the predicted Arnett model luminosity. This is expected, as γ-ray trapping is likely to be inefficient at these late phases, given the low ejecta mass, and hence the Arnett model is not applicable at late times (Valenti et al. 2014). Using an optical opacity of κ = 0.07 cm² g⁻¹ (Cano 2013; Taddia et al. 2018) and ejecta velocity of 10,000 km s⁻¹ for the second peak, we derive an ejecta mass of 0.38 M_⊙ and explosion kinetic energy of ≈2.3 × 10⁵⁰ erg. Note that these estimates do not include the (yet) unknown power source of the first peak, which is clearly not consistent with the Arnett model presented here. Given that such an early emission component has never previously been observed in Ca-rich gap transients, we discuss several potential power sources for the first peak in the following sections.

Because the second peak of the light curve peak of iPTF 16hgs can be explained well by radioactive decay, we first consider a radioactively powered scenario for the first peak. In this case, the shape of the early emission with respect to the main peak puts strong constraints on the radial distribution of the radioactive material. Dessart et al. (2012) and Piro & Morozova (2016) show that a radial monotonically decreasing distribution of ⁵⁶Ni in the ejecta lead to smoothly rising light curves for radioactively powered Type Ib/c SNe. As the early decline in iPTF 16hgs is distinctly separated from the main peak, it is likely that the relevant radioactive isotope was strongly mixed into the surface of the progenitor, and separated from the radioactive material powering the second peak.

We can obtain approximate estimates for the amount of radioactive material and the ejecta mass above it by analyzing the early bolometric light curve. Taking the peak luminosity of the first peak to be >4 × 10⁴¹ erg s⁻¹, we estimate that ≳7 × 10⁻³ M_⊙ of ⁵⁶Ni in the outer layers would be required if the early component was powered by ⁵⁶Ni decay. However, we do not have strong constraints on the rise time of the first peak because the transient was discovered on the early declining phase. Given that the Arnett model fitting of the main peak suggests that the explosion occurred at ≈10 days before the r-band peak (which is almost at the epoch when the transient was first discovered), we consider the case where the observed width of the first component of ≈2 days corresponds to the diffusion time through the ejecta above this radioactive layer. In such a case, the mass above this layer would be ≳0.05 M_⊙, taking an opacity of κ = 0.07 cm² g⁻¹ and ejecta velocity of 12,000 km s⁻¹ (as measured from the first spectrum of the source).

We note that the derived ⁵⁶Ni estimates are similar to those of some other Type Ib/c SNe with early excess blue emission (Bersten et al. 2013; Drout et al. 2016), and where a radioactivity powered first peak was also suggested. Alternatively, it is possible that the first peak is powered by the decay of other radioactive species. For instance, modeling of He shell detonations in the context of faint and fast-evolving SNe by Waldman et al. (2011) and Sim et al. (2012) show that radioactive isotopes like ⁴⁸Cr, ⁵²Fe, and ⁴⁴Ti can be important in powering the light curves of these events, particularly at early times for short-lived isotopes such as ⁵²Fe.

A possible explanation for the early peak could be due to interaction of the SN ejecta with a non-degenerate companion (Kasen 2010). Such interaction signatures depend sensitively on the binary separation of the companion as well as the viewing angle of the observer, with the most prominent signatures arising when the source is viewed along the direction of the companion. These signatures have been previously suggested in Type Ia SNe (e.g., Cao et al. 2015; Hosseinzadeh et al. 2017), allowing one to estimate the separation of the companion and its radius (assuming Roche lobe overflow). For viewing angles oriented close to the direction of the companion, the analytic equations in Kasen (2010) yield good estimates of the expected emission.

We attempted to fit the early g- and r-band light curves with the Kasen (2010) models but could not get a reasonable match to the data. This is particularly because the colors of iPTF 16hgs on the declining phase are markedly different than that predicted in the companion interaction models. For instance, we observe colors of g − r ≈ 0 about one day after discovery when the bolometric luminosity is ≳2.5 × 10⁴¹ erg. Taking equation (22) in Kasen (2010) for t_day = 1, v₉ = 1.2 and M_c = 0.3, the bolometric luminosity requires the separation a to be ≳2.5 × 10¹¹ cm. At the same time, the g − r color suggests that the color temperature of the emission is ∼5500 K when assuming a blackbody spectrum, which gives a ∼ 2.4 × 10¹⁰ cm using equation Kasen's Equation (25). Note that this result is insensitive to the exact epoch of this observation because the model luminosity and temperature scale similarly (∝ t^−1/2) with time after explosion.

Hence, we find that a companion interaction scenario is unable to account for the early declining emission within the framework of the analytic equations presented in Kasen (2010). However, we note that viewing angle dependencies may affect our conclusions. For example, Figure 2 in Kasen (2010) shows that both the peak luminosity and morphology of the companion interaction light curve may be significantly affected along lines of sight away from the companion. In fact, the early drop of a factor of ≈1.7 in luminosity over the first two days in iPTF 16hgs is indeed reminiscent of the bolometric evolution predicted along directions ∼90° to the companion, as shown in Kasen (2010). As the simulations with varying viewing angles (in Kasen 2010) were specifically for Type Ia SN ejecta, future modeling will be required to understand whether these signatures would be similar in ejecta with compositions different compositions, as in iPTF 16hgs.

We now consider whether interaction with dense external circumstellar material (CSM) can explain the early blue emission in iPTF 16hgs. Given the projected location of iPTF 16hgs inside its host galaxy (and hence its likely proximity to dense CSM), such a scenario can potentially explain the uniqueness of the light curve of iPTF 16hgs with respect to the other members of this class. We can obtain rough estimates of the characteristics of this CSM by using the early bolometric light curve and the methods presented in Smith (2017). The interaction luminosity can be estimated using

$$\begin{eqnarray}&&L=\displaystyle \frac{1}{2}4\pi {R}^{2}\rho {V}^{3},\end{eqnarray} \tag{ 1 }$$

which depends on the velocity of the ejecta material V and the density of the medium ρ.

Assuming a constant density CSM, and using L ≳ 3 ×10⁴¹ erg s⁻¹, V ∼ 12,000 km s⁻¹ at t ∼ 1 day after the explosion, we find ρ ∼ 3 × 10⁻¹⁵ g cm⁻³. This corresponds to a particle density of 5 × 10⁸ cm⁻³ if the CSM was dominated by He at a distance of ∼10¹⁴ cm from the progenitor. If we instead assume a constant mass-loss wind like CSM density profile, using the same values as above, we estimate a $\dot{M}\sim 7\times {10}^{-5}\tfrac{{v}_{\mathrm{CSM}}}{100\,\mathrm{km}\,{{\rm{s}}}^{-1}}$ M_⊙ yr⁻¹. However, we note that there is no evidence for spectral signatures of circumstellar interaction (as seen in Type IIn and Type Ibn SNe) in iPTF 16hgs, which would argue against a CSM interaction scenario. Nevertheless, such signatures may be hidden in the case of asymmetric CSM configurations (e.g., in the form of a disk) where the interaction region is hidden by the expanding ejecta (Smith 2017).

The double-peaked light curves of some stripped-envelope core-collapse SNe have been explained using shock cooling emission of an extended envelope around the progenitor. Such extended envelopes have been shown to be particularly relevant for sources that exhibit an early peak in the redder r and i bands, because "normal" progenitors (i.e., progenitors without an extended envelope) cannot reproduce such light curves (Nakar & Piro 2014; Piro 2015; Sapir & Waxman 2017). Because iPTF 16hgs was found in a star-forming host galaxy (indicating a core-collapse supernova origin is a possibility, given the presence of a young stellar population), we examine if the early declining emission (which was detected in the g and r bands) can be explained by shock cooling emission of an extended progenitor.

We use the extended envelope models of Piro (2015) to fit the first peak of iPTF 16hgs. Adopting the ejecta mass and explosion energy as derived in the Arnett model, and an optical opacity of κ = 0.2 cm² g⁻¹, the only other free parameters in the model are the mass in the extended envelope M_e and the radius of the envelope R_e. Keeping the explosion time t₀ as an additional free parameter, we obtain the best-fit model as shown in Figure 11. As shown, an extended envelope model with M_e = 0.08 M_⊙ and R_e = 13.0 R_⊙ is able to reproduce the early peak for an explosion time t₀ = −12.7 days before the main peak of the r-band light curve. Although the model used here is very simplified, in that it ignores the crucial density structure of the envelope, the numbers derived are expected to be correct to an order of magnitude (as shown in more realistic simulations including density profiles; see Piro et al. 2017).

Zoom In Zoom Out Reset image size

We use the synchrotron self-absorption model of Chevalier (1998) to generate analytic radio light curves for a range of circumstellar environments. We follow the prescription given in Chomiuk et al. (2016), who present analytic equations for the expected radio light curves of Type Ia SNe based on Chevalier (1998) (but are also applicable to other hydrogen-poor SNe). We generate these light curves for both a constant wind mass-loss environment (where ρ = K r⁻², with $K=\tfrac{\dot{M}}{4\pi {v}_{w}}$) and a constant density environment (ρ = constant), using an outer ejecta density profile of ρ ∝ r⁻¹⁰, as appropriate for compact progenitor stars (Matzner & McKee 1999). By comparing the predicted radio light curves to those of our upper limits, we constrain the wind mass-loss parameter K and the external circumstellar density n₀ by obtaining the limiting cases for a 3σ detection, as shown in Figure 12.

Zoom In Zoom Out Reset image size

In the case of a wind mass-loss environment, the strongest constraints arise from the VLA (10 GHz) and uGMRT (1.2 GHz) observations, which we use to constrain the mass-loss rate to $\dot{M}\lesssim 2\times {10}^{-6}\tfrac{{v}_{w}}{100\,\mathrm{km}\,{{\rm{s}}}^{-1}}$ M_⊙ yr⁻¹ for _B = 0.1. Adopting _B = 0.01, the VLA and uGMRT observations constrain the mass-loss rate to $\dot{M}\lesssim 8\times {10}^{-6}\tfrac{{v}_{w}}{100\,\mathrm{km}\,{{\rm{s}}}^{-1}}$ M_⊙ yr⁻¹. For the case of a constant density environment, the strongest constraints arise from the uGMRT observations, which limit the circumstellar density to n₀ ≲ 125 cm⁻³ for _B = 0.1, and to n₀ ≲ 800 cm⁻³ for _B = 0.01.

Sell et al. (2015) proposed that Ca-rich gap transients could arise from tidal disruptions of low-mass He WDs by intermediate-mass black holes, based on the work of Rosswog et al. (2009) (see also Rosswog et al. 2008; MacLeod et al. 2014). In this scenario, when a low-mass WD comes within the tidal radius of a massive compact object (with a mass of ≲10⁵ M_⊙), the WD is tidally crushed, leading to a runaway thermonuclear detonation powering an optical transient. The accretion of the WD onto the compact object would then also power a super-Eddington X-ray flare, potentially leading to the launch of a relativistic jet (Sell et al. 2015; MacLeod et al. 2016). One direct prediction of such a model is that these transients should then also be associated with prominent X-ray and radio emission for suitably oriented observing angles.

MacLeod et al. (2016) presented simulations of disruptions of 0.6 M_⊙ WDs by an intermediate-mass black hole, including predictions for expected light curves and spectra of the thermonuclear transient. They show that the disruption of the WD, along with the explosive detonation, leads to less than half of the WD mass being accreted on to the BH. They also predict several characteristics of the radio emission that would be expected if these events produced relativistic jets that eventually interact with the surrounding interstellar medium (ISM). Thus, we use our deep radio limits on iPTF 16hgs to constrain the phase space of jet energy and ISM density for different viewing angles of the observer.

The super-Eddington flare of accretion in such a disruption event can lead to the launching of a relativistic jet that carries away some fraction of the rest mass energy of the accreted WD (MacLeod et al. 2016). We thus consider a range of jet energies from 10⁴⁷ erg to 2 × 10⁵¹ erg. If the disruption event in iPTF 16hgs involved a 0.4 M_⊙ He WD (as expected from the He-rich spectra), the range of jet energies corresponds to a fraction of ∼10⁻⁶–10⁻² of the rest mass energy of the accreted half of the WD. We then use the BOXFIT code (van Eerten et al. 2010) to generate simulated multi-frequency radio light curves for different surrounding ISM densities (in the range between 10⁻⁶–1 cm⁻³). As suggested in MacLeod et al. (2016), we also assume that the jet has an initial starting Lorentz factor of Γ ∼ 10 and an opening angle of 0.2 rad (i.e., a jet beaming factor of 0.02).

We show the contour plots (in the phase space of jet energy and circumstellar density) of the expected radio fluxes at the epochs of the VLA and GMRT observations for different viewing angles, along with our limits on the radio emission of this source in Figure 13. In each of the panels, the phase space ruled out by our observations are indicated by the hatched region. There are several interesting factors to note from the allowed phase space. First, for a jet nearly along the line of sight (≈10°), the GMRT upper limits completely rule out jet energies higher than about 10⁴⁹ erg for ISM densities as low as 10⁻⁶ cm⁻³. Note that the critical density of the universe is ∼10⁻⁶ cm⁻³ while the electron density inside the host galaxy should be at least an order of magnitude larger, ruling out the on-axis case completely. For a viewing angle of 45°, the GMRT and VLA upper limits together rule out ISM densities ∼10⁻⁴ cm⁻³ if the jet energy is larger than 10⁴⁹ erg. For lower jet energies, our radio limits do not constrain the ISM environment because the limits lie above the optically thick locus of the light curves. Finally, for a 90° observing angle, the radio limits are the least constraining, although they do rule out ISM densities ∼10⁻² cm⁻³ if the jet energy was larger than about 5 × 10⁴⁸ erg. However, they are not constraining if the jet energy was lower.

Zoom In Zoom Out Reset image size

We first estimate the gas-phase metallicity of the host galaxy using the emission lines fluxes in the spectrum of its nucleus and the pyMCZ code (Bianco et al. 2016). The measured emission line fluxes are presented in Table 4. The code calculates the host oxygen metallicity, and is based on the original code of Kewley & Dopita (2002) with updates from Kewley & Ellison (2008). Typical metallicity estimates derived using this method indicate 12 + log(O/H) metallicity of ${8.26}_{-0.03}^{+0.03}$ on the O3N2 scale of Pettini & Pagel (2004) and ${8.21}_{-0.02}^{+0.02}$ on the O3N2 scale of Marino et al. (2013). In general, we note that all the derived oxygen metallicity indicators suggest a significantly subsolar metallicity (where 12 + log(O/H)_⊙ ≈ 8.69; Asplund et al. 2009) of ≈0.4 Z_⊙ (Z ≈ 0.008) for the spectrum of the nucleus. The low metallicity estimate places the host galaxy in the lowest 10% of the distribution of host galaxy metallicities of Type Ib/c SNe, while it is on the lowest 30% of the range of the host galaxies of Type Ic-BL SNe (Sanders et al. 2012).

Table 3.  Redshifts of Galaxies Near iPTF 16hgs as Measured from Our Spectroscopic Mask Observation

Object name	α (J2000)	δ (J2000)	Redshift	Offset ('')
Obj1	00h50m5116	27°22'4519	0.362	4.12
Obj2	00h50m5329	27°22'3653	0.017	27.70
Obj3	00h50m5326	27°23'166	0.195	28.36
Obj4	00h50m4590	27°22'4964	0.160	73.17
Obj5	00h50m5256	27°24'370	0.290	77.30
Obj6	00h50m5520	27°23'5547	0.210	84.43
Obj7	00h50m5552	27°24'634	0.107	95.70
Obj8	00h50m5200	27°24'2622	0.107	98.57
Obj9	00h50m4533	27°21'4531	0.175	102.26
Obj10	00h50m3434	27°21'4560	0.170	235.51
Obj11	00h50m3334	27°21'3134	0.171	252.40
Obj12	00h50m2908	27°21'3340	0.170	306.36

Download table as:  ASCIITypeset image

Next, we use the integrated fluxes of the host galaxy to estimate the global properties of the stellar population in the host galaxy of iPTF 16hgs. The photometric fluxes in the SDSS ugriz, 2MASS JHK, and GALEX FUV/NUV bands were fit using the FAST code (Kriek et al. 2009). The fitting was performed assuming a Maraston (2005) stellar population, an exponentially declining star formation history, a Salpeter IMF, and a Milky Way-like extinction law. We also constrain the models to be at a subsolar metallicity (as indicated by the spectra of the host galaxy) of Z = 0.01 (0.5 Z_⊙), which is the model grid closest to the inferred metallicity.

Using this model, we obtain a best-fit stellar mass of ${6.45}_{-0.29}^{+0.31}\,\times {10}^{8}$ M_⊙ and mean stellar population age of ${1.99}_{-0.17}^{+0.05}\times {10}^{8}$ years. The integrated star formation rate is poorly constrained from the photometry only, and hence we estimate it from the Hα maps in our IFU observations (see Section 6.2). Integrating the Hα flux over the entire map where Hα emission is detected, we find a total Hα flux of ≈10⁻¹³ erg cm⁻² s⁻¹. Converting this to an equivalent star formation rate using the redshift of the host galaxy (total Hα luminosity of ≈6.5 × 10⁴⁰ erg s⁻¹) and the relations in Kennicutt (1998), we get an integrated star formation rate of 0.5 M_⊙ yr⁻¹, placing this galaxy in the lower half of the distribution of star formation rates found in the hosts of stripped-envelope SNe (Galbany et al. 2014).

We also use our Keck-LRIS spectrum of the nucleus of the host galaxy to estimate the stellar age and stellar phase metallicity of the older stellar populations in the galaxy. Similar to the analysis presented in Galbany et al. (2016), we fit the stellar continuum and absorption features in the host nucleus spectrum using the STARLIGHT code (Cid Fernandes et al. 2005). Using a Cardelli et al. (1989) dust extinction law and Bruzual & Charlot (2003) stellar population models at a range of metallicities (from Z = 0.001 to Z = 0.05), we find the best-fit spectrum as shown in Figure 15. The insets in Figure 15 also show the best-fit stellar population mixture computed by STARLIGHT. As shown, the stellar continuum is fit well by a mixture of both old (age ≳1 Gyr) and young (age ≲0.5 Gyr) stellar populations, where >65% of the stellar population in the STARLIGHT fit is in the former category.

Zoom In Zoom Out Reset image size

In terms of metallicity, more than 70% of the stellar population in the best STARLIGHT fit is at a subsolar metallicity (Z < 0.02), with ≈50% of the population at Z < 0.001. Note that the stellar metallicity reflects the galaxy metallicity when the stars were formed (which is likely to be lower than the current metallicity), while the estimates based on emission lines reflect the current gas-phase metallicity in the galaxy. Weighting over the entire population mixture produced by the model, we find a mean stellar age of 7.2 × 10⁸ years and a mean stellar metallicity of Z = 0.012. Thus, the mean stellar age estimated from the nuclear spectrum is older than that inferred for the whole galaxy from the broadband photometry, while the mean stellar metallicity is similar to the gas-phase metallicity estimated from the emission line fluxes in the nuclear spectrum. Taken together, this leads us to conclude that the host galaxy of iPTF 16hgs is a metal-poor, star-forming dwarf galaxy with a mixture of both young and old stellar populations.

Due the small offset of iPTF 16hgs from its host galaxy, we obtained IFU observations of the host galaxy via the PCWI in order to study the spatially resolved ISM of the host galaxy, and in particular, the local ISM environment of iPTF 16hgs. For each pixel in the reduced and stacked spectral cube from the PCWI, we modeled the continuum emission using a low-degree polynomial, and subtracted it out to measure the emission line fluxes of the most prominent lines from the host galaxy. The lines that were within the spectral cube include Hα, [N ii] λ6584, [S ii] λ6716, and λ6731. For the strongest Hα emission line, we also measure the velocity by fitting a Gaussian profile to the continuum-subtracted Hα feature, its equivalent width, and the full width at half maximum (FWHM). The resulting maps are shown in Figure 16. For comparison, we also show a continuum image of the host galaxy, as obtained from the late-time LRIS observation in Figure 16.

Zoom In Zoom Out Reset image size

A few features are readily apparent from the maps in Figure 16. iPTF 16hgs occurred in a star-forming spiral galaxy, as indicated by the morphology of the galaxy in the continuum image and the prominent Hα emission extending through out the image. The multiple blobs of Hα emission are likely to be individual H ii regions in the disk of the galaxy. The Hα velocity map clearly shows evidence of ordered rotation of the spiral arms of the galaxy (reaching a velocity of ≈100 km s⁻¹ near the edges), such that the location of iPTF 16hgs is on the receding arm of the galaxy that is viewed close to the plane of the disk. In particular, we note that object Obj2 in the spectroscopic mask (which was classified as a separate galaxy in SDSS) is consistent with being an individual H ii region in the host galaxy, given that its velocity lies exactly on the rotation curve of the host.

The Hα equivalent width (EW) map also shows regions of very high EW (≳100 Å), suggestive of very young stellar populations in the host galaxy. The metal emission lines of S and N are significantly weaker than the bright Hα emission (as would be expected from a metal-poor galaxy), and are thus detected only near the nucleus and along the plane of the galaxy. In particular, we note that the metal emission line maps show evidence of bright emission regions co-located with the bright blobs of Hα emission near the nucleus of the host galaxy.

6.2.1. Star Formation Density

We use the Hα flux map from our observations to measure the equivalent star formation surface density by using the relations in Kennicutt (1998). Because the galaxy is viewed nearly edge-on, we caution that the star formation density will be subject to a projection effect. Nevertheless, we show the spatially resolved star formation density in Figure 17. As shown, the host galaxy of iPTF 16hgs exhibits multiple prominent blobs of star formation, reaching surface densities ≳0.1 M_⊙ yr⁻¹ kpc⁻². Interestingly, there is evidence for outlying H ii regions in the host galaxy, as evidenced by the bright Hα emission blob located to the southwest of the nucleus, which is very faint in the continuum image (and hence has a very high EW). Given that this blob lies on the rotation curve of the galaxy, it likely represents a very young star-forming H ii region as opposed to a companion dwarf galaxy. Additionally, there is clear evidence of a large star-forming region close to the location of iPTF 16hgs, at a projected offset of ≈3'' (≈1 kpc at the redshift of the galaxy; denoted by the black square) from the location of the transient.

Zoom In Zoom Out Reset image size

6.2.2. Metallicity Map

6.2.3. Stellar Population Age

In this section, we use both the IFU maps of the host galaxy as well as the deeper late-time LRIS spectrum at the location of the transient to measure the properties of the stellar population at the site of the explosion. First, the projected offset of the explosion site from the nucleus of the host galaxy (1.9 R_eff ≈ 6 kpc) is typical of the host nucleus offsets of core-collapse SNe found by PTF (Kasliwal et al. 2012; see also Galbany et al. 2014), while it is on the lower end of the distribution (cumulative fraction ≲50%) found in Type Ia SNe from PTF (Lunnan et al. 2017; L. Hangard et al. 2018, in preparation). The host-normalized offset is also on the lower end (cumulative fraction ≲50%) of the distribution of both Type Ia SNe (Lunnan et al. 2017) and short GRBs (Fong & Berger 2013).

Using the late-time LRIS spectrum, which exhibits a number of galaxy emission lines, we measure a Hα flux of ≈2.5 × 10⁻¹⁷ erg cm⁻² s⁻¹. This is consistent with the value estimated from the location of the transient in the IFU observations (Figure 16). At the redshift of the host galaxy, this corresponds to a Hα luminosity of ≈1.6 × 10³⁷ erg s⁻¹. Translated to an equivalent star formation rate using the relations in Kennicutt (1998), we infer a star formation rate of ≈1.3 × 10⁻⁴ M_⊙ yr⁻¹ within the 1'' slit used for the observations. The measured Hα luminosity is similar to the majority of H ii regions associated with Type Ib/c SNe (Crowther 2013; Kuncarayakti et al. 2018).

As noted earlier, we find that the Hα map of the host galaxy indicate the presence of a large H ii region with a young stellar population at a projected offset of ≈1 kpc from the location of the transient (see square symbols in Figure 16). The integrated Hα luminosity over this H ii region is ≈1.1 × 10³⁹ erg s⁻¹, corresponding to a star formation rate of ≈0.01 M_⊙ yr⁻¹. When compared to the typical Hα luminosities of H ii regions hosting core-collapse SNe, this association lies on the brightest end of the observed distribution of all types of core-collapse SNe (Kuncarayakti et al. 2018). Hence, if iPTF 16hgs was due to the core-collapse explosion of a massive star, it appears likely that the progenitor could have originated in this H ii region. Given the offset of ≈1 kpc from the explosion site, the required systemic velocity of the progenitor would be ∼50 km s⁻¹, assuming a progenitor lifetime of ∼20 Myr. Additionally, the observed offset would also be consistent (albeit at the higher end) with the distribution of offsets of stripped-envelope SNe from their likely parent H ii regions (Galbany et al. 2014).

The overall light curve of iPTF 16hgs (i.e., its main peak and subsequent decline) is quite consistent with the other members of this class. In particular, our modeling suggests that the main peak can be modeled well by a ⁵⁶Ni powered light curve with ≈0.4 M_⊙ of ejecta and ≈8 × 10⁻³ M_⊙ of radioactive ⁵⁶Ni. This is similar to the explosion parameters estimated for other members of this class (Kasliwal et al. 2012; Valenti et al. 2014; Lunnan et al. 2017; Perets et al. 2010). Despite its overall similarity to the class of Ca-rich gap transients, the double-peaked light curve of iPTF 16hgs is unique among the members of this class. Hence, we now discuss the progenitor channels relevant for the various power sources that could power the first peak of the light curve.

7.1.1. A Thermonuclear Detonation?

We considered a radioactivity-powered scenario for the early peak in iPTF 16hgs, and found that ≈0.01 M_⊙ of ⁵⁶Ni in the outer 0.05 M_⊙ of the ejecta can explain the early bump in the light curve. Interestingly, such configurations have been suggested for some physical scenarios relevant for the potential progenitors of Ca-rich gap transients. For example, double-detonation models for Type Ia SNe invoke explosive ignition of a He layer on the surface of a carbon-oxygen WD that leads to the formation of iron group radioactive isotopes near its surface and a subsequent explosive detonation of the entire star (Fink et al. 2010; Kromer et al. 2010, 2016). The presence of such radioactive material close to the surface has been shown to produce bluer colors at early times (Shen et al. 2010; Dessart et al. 2012; Piro & Morozova 2016), consistent with the observed early blue colors of iPTF 16hgs.

For the specific case of Ca-rich gap transients, a widely discussed progenitor channel involves the detonation of a He shell on the surface of a WD (Perets et al. 2010; Waldman et al. 2011; Dessart & Hillier 2015). Such a configuration could arise from a close binary system with a CO WD that accretes He-rich matter from a He WD or a He-rich nondegenerate companion (Bildsten et al. 2007; Shen et al. 2010; Waldman et al. 2011; Dessart & Hillier 2015). Numerical simulations for the expected optical signatures of these events were performed by Shen et al. (2010) and Sim et al. (2012), and we show a comparison of these models to iPTF 16hgs in Figure 18. Interestingly, as shown in Figure 18, Shen et al. (2010) did find double-peaked light curves in their simulations of He shell detonations for some combinations of core and shell masses. In particular, their models suggest that the first peak arises out of radial stratification of short-lived radioactive isotopes (⁵²Fe, ⁴⁸Cr, and ⁴⁴Ti) in the outer ejecta, whereas the second peak is powered by ⁵⁶Ni decay deeper in the ejecta.

Zoom In Zoom Out Reset image size

Thus, qualitatively, the double-peaked light curve of iPTF 16hgs does appear to be consistent with some predictions of this model, which would suggest that the first peak is likely powered by a different radioactive isotope than ⁵⁶Ni. However, the .Ia detonation models presented in Shen et al. (2010) involved detonations of typically low-mass shells (≲0.3 M_⊙) so that, overall, their light curves evolve faster (rising over ≈7–10 days) and are brighter (M_peak < −17) than the light curve of iPTF 16hgs. Even if the first peak is somewhat reproduced, the low shell masses lead to much faster-evolving second peaks than iPTF 16hgs. On the other hand, Sim et al. (2012) did not find such double-peaked light curves in their 2D models. Although the timescales of the overall light curve of iPTF 16hgs are similar to the Sim et al. (2012) models, they are also brighter than that of iPTF 16hgs. We thus find that, while some features of the light curve of iPTF 16hgs are reproduced in these models, they would require larger shell masses (than the Shen et al. 2010 models) and smaller amounts of synthesized radioactive isotopes to explain the lower luminosity.

There are also a number of spectroscopic differences between the predictions of the He shell detonation models (as in Shen et al. 2010 and Sim et al. 2012) and our observations. Spectroscopically, they find that the peak photospheric spectra are likely to be dominated by absorption lines of incomplete He-burning products, such as Ti ii and Ca ii (see also Holcomb et al. 2013), and most notably, lack lines of Si and Mg. This is different from our observations, where the spectra can be modeled well by prominent features of He i, Mg ii, Si ii, and Ca ii. While their spectra did not show He i lines, they did suggest that non-thermal excitation will likely lead to the production of He lines in these events (see also Waldman et al. 2011; Dessart & Hillier 2015). For comparison, the .Ia detonation candidate OGLE-2013-SN-079 (Inserra et al. 2015) did exhibit prominent Ti ii and Ca ii lines near peak light, unlike the Ca-rich transient SN 2005E, as noted in Inserra et al. (2015). While the highlighted differences may appear problematic to this interpretation, we caution that the .Ia detonation models shown for comparison involved simple simulations in 1D spherical symmetry and the nucleosynthetic outcome may differ in more realistic 3D simulations.

7.1.2. A Low-luminosity Core-collapse Explosion?

We showed that the early peak of iPTF 16hgs can be modeled well by shock cooling of an extended progenitor star at the time of explosion. If the first peak was powered by shock cooling emission, the extended progenitor at the time of explosion would strongly argue for a core-collapse origin of the explosion. In fact, a core-collapse origin is plausible even if the early peak was radioactively powered, e.g., Bersten et al. (2013) invoked a similar ⁵⁶Ni clump near the surface to explain the double-peaked light curve of SN 2008D, while Drout et al. (2016) also suggested outward ⁵⁶Ni mixing to explain the early blue bump in SN 2013ge.

In the case of a shock cooling first peak, the inferred parameters of the extended envelope (M_e ≈ 0.08 M_⊙ and R_e ≈ 13 R_⊙) provide important clues to the nature of the progenitor star. Because the Piro (2015) model used for this analysis is simplified and ignores the density structure of the envelope, these numbers are likely to be correct only to an order of magnitude (Piro et al. 2017). We note that such extended envelopes have been previously inferred in several other stripped-envelope SNe (e.g., Ben-Ami et al. 2015; Taddia et al. 2016; Arcavi et al. 2017), and are suggested to be associated with elevated mass loss prior to explosion, or formed due to binary interactions. In fact, studies of the pre-SN evolution of He stars suggest that they are capable of swelling significantly before core collapse (up to radii ∼10–100 R_⊙), consistent with such a scenario (Woosley et al. 1995; Yoon et al. 2010).

The main peak of the light curve suggests an ejecta mass of 0.4 M_⊙ and ⁵⁶Ni mass of ≈8 × 10⁻³ M_⊙, which is unusually low compared to the normal population of stripped-envelope core-collapse SNe (Drout et al. 2011; Lyman et al. 2016a; Taddia et al. 2018). In particular, the low inferred ejecta mass would require significantly more stripping than observed in the typical population of stripped-envelope SNe, due either to the presence of a compact companion or to stripping by a companion in a very close orbit. Hence, we compare iPTF 16hgs to models of ultra-stripped SNe arising from highly stripped massive star progenitors in close He star—neutron star (NS) binaries (Tauris et al. 2013, 2015).

Moriya et al. (2017) presented the expected light curves and spectra of ultra-stripped (Fe core-collapse) SNe in the context of systems that lead to double neutron star systems. However, their models did not explore SN explosions with ejecta masses as large as 0.4 M_⊙ (as in iPTF 16hgs), although such ejecta masses are allowed by binary population synthesis models (Tauris et al. 2015). Thus, if iPTF 16hgs originated in an ultra-stripped SN explosion from a He star—compact object binary, this would require either an initially more massive He star or a wider He star—NS binary separation than the systems simulated in Moriya et al. (2017), in order to explain the larger progenitor mass at the time of explosion. Nevertheless, we note that simulations of ultra-stripped explosions in Suwa et al. (2015) did explore systems that produced ≈0.4 M_⊙ of ejecta, and found synthesized ⁵⁶Ni masses of ≈8 × 10⁻³ M_⊙ in the explosion, very similar to our estimates for iPTF 16hgs.

While the majority of ultra-stripped SNe are expected to be of Type Ic, more massive progenitors (as would be the case for iPTF 16hgs) with larger He layers (M_He ≳ 0.06 M_⊙) may lead to He-rich Type Ib SNe (Hachinger et al. 2012; Moriya et al. 2017). Recently, Yoshida et al. (2017) also showed that the nucleosynthesis in ultra-stripped explosions may produce ejecta that are particularly rich in isotopes of Ca, suggesting that the ultra-stripped interpretation may explain the Ca-rich nebular spectra as well. It is also important to note that the larger ejecta mass and He-rich spectra of iPTF 16hgs would also be consistent with a core-collapse explosion in a close binary system of two non-degenerate massive stars (Yoon et al. 2010), where stripping by a close non-degenerate companion can lead to a similar highly stripped progenitor that retains a large amount of He in its outer layers.

Interestingly, the low peak luminosity of iPTF 16hgs and the associated low inferred ⁵⁶Ni mass, together with the peculiar signatures of nucleosynthesis (i.e., Ca-rich nebular spectra) is also reminiscent of models of electron capture SNe. Such SNe are initiated by the loss of pressure due to electron captures on to ²⁴Mg and ²⁰Ne in a degenerate O–Ne–Mg core of a massive star (Nomoto 1984). While only single stars in the mass range of 8–12 M_⊙ are expected to undergo such an outcome, the mass range may be significantly extended when considering binary interactions (Podsiadlowski et al. 2004). Because stars in this mass range do not produce sufficiently massive winds to remove their outer H layers, a stripped-envelope SN such as iPTF 16hgs would necessarily require a binary scenario to explain the observed SN. A similar scenario was also used to explain the Ca-rich SN 2005cz (Kawabata et al. 2010), although Perets et al. (2011) argue against a massive star scenario for this event on the basis of the old stellar population in its environment and stringent limits on nearby star formation.

Kitaura et al. (2006) presented simulations of such electron-capture SNe and found explosion energies and ⁵⁶Ni mass yields of ∼10⁵⁰ erg and ∼10⁻³ M_⊙. These are consistent with the properties of iPTF 16hgs within a factor of a few. We also compare iPTF 16hgs to the simulations of Moriya & Eldridge (2016) (hereafter M16), who investigated the expected signatures of stripped-envelope ECSNe with binary population synthesis models at solar and subsolar metallicity. Specifically, they performed population synthesis simulations of binary massive stars that led to ECSN progenitors either via the merger of two initially less-massive stars or due to close (Case B or Case C) stripping of an initially massive star by a nondegenerate companion. Note that Tauris et al. (2015) also found ECSN progenitors in their simulations of He star—NS binaries, although the large stripping by the NS in a close orbit led to explosions that had significantly lower ejecta masses (≲0.2 M_⊙) than that inferred for iPTF 16hgs or those presented in M16.

At a subsolar metallicity of Z = 0.004, the ejecta mass of iPTF 16hgs is typical of the ejecta masses expected in these explosions, lying in the lower half of the distribution presented in M16. However, the peak bolometric luminosity (≈3 ×10⁴¹ erg s⁻¹) is higher than that predicted by any of the models presented in M16, as the light curves presented in their work were much fainter (peak of ≈10⁴¹ erg s⁻¹). We show a comparison of the bolometric light curves of stripped-envelope ECSNe presented in M16 to iPTF 16hgs in Figure 19, for a model with 0.42 M_⊙ of ejecta (as derived from our Arnett fit) at a metallicity of Z = 0.004. As indicated earlier, the models presented in M16 included too little ⁵⁶Ni (2.5 × 10⁻³ M_⊙ for the shown light curves) to account for the (main) peak luminosity of iPTF 16hgs, so the model prediction (solid blue curve) is much fainter than the data. We also show a comparison model to account for a higher ⁵⁶Ni mass by scaling the luminosity of the original model by a factor of ≈3.5. After adding a shock cooling component to the light curve (for M_e = 0.05 M_⊙ and R_e = 10 R_⊙), to account for the early declining emission, we find that the total predicted luminosity of the model (solid red curve) is consistent with the observations.

Zoom In Zoom Out Reset image size

Taken at face value, we find that the explosion properties of iPTF 16hgs are consistent (within a factor of a few) with the expected properties of highly stripped-envelope ECSNe in binary systems. Although the association may be reasonable given the uncertainties in the explosion properties of ECSNe (Woosley & Heger 2015), we reiterate that current models of these explosions cannot explain the relatively high luminosity of iPTF 16hgs. Nevertheless, Kawabata et al. (2010) argued that such low-luminosity core-collapse explosions (as SN 2005cz) from lower-mass progenitors are likely associated with unique nucleosynthetic signatures—in particular, a higher abundance of Ca with respect to O, consistent with the Ca-rich nebular phase spectra of SN 2005cz. Hence, the high [Ca ii]/[O i] ratio in the nebular phase spectra of iPTF 16hgs would be consistent with a low-luminosity electron capture explosion of a stripped massive star (Woosley & Heger 2007; Nomoto et al. 2013; Wanajo et al. 2013; Sukhbold et al. 2016).

We have used the non-detection of radio emission in iPTF 16hgs to place stringent constraints on the environment of the progenitor, using models of a spherical SN shock. Additionally, we have also used our radio limits to constrain the presence of a relativistic jet, as expected in some progenitor models for Ca-rich gap transients.

7.2.1. A Low-density SN Environment

We discuss the CSM environment of the progenitor in the case where iPTF 16hgs was purely powered by a spherical SN explosion. In this context, we note that core-collapse SNe exhibit a wide range of radio emission properties that reflect their diverse circumstellar mass-loss environments (e.g., Soderberg et al. 2005; van der Horst et al. 2011; Weiler et al. 2011; Cao et al. 2013). However, no Type Ia SNe that likely share progenitor systems (i.e., WDs) similar to those of Ca-rich gap transients have been detected to date in the radio band to very stringent limits. For instance, Chomiuk et al. (2016) presented radio limits on a sample of Type Ia SNe and found that the radio non-detections constrained some of their environments to stringent limits of $\dot{M}\lesssim {10}^{-9}\tfrac{{v}_{\mathrm{CSM}}}{100\,\mathrm{km}\,{{\rm{s}}}^{-1}}$ M_⊙ yr⁻¹. Chomiuk et al. (2016) also presented radio limits on the Ca-rich gap transients SN 2005E (Perets et al. 2010) and PTF 10iuv (Kasliwal et al. 2012) (in addition to some other "Ca-rich" transients), and constrained their mass-loss environments to ≲4 × 10⁻⁷ M_⊙ yr⁻¹ and ≲10⁻⁵ M_⊙ yr⁻¹, respectively.

Taking standard values of the energy density fraction in the magnetic field of _B = 0.1, we find that our radio limits constrain the mass-loss environment of the progenitor to $\lesssim 2\times {10}^{-6}\tfrac{{v}_{\mathrm{CSM}}}{100\,\mathrm{km}\,{{\rm{s}}}^{-1}}$ M_⊙ yr⁻¹ for a wind-like CSM density profile, and to n_e ≲ 100 cm⁻³ for a constant density environment. While these limits are comparable to those obtained previously for Type Ia SNe and Ca-rich gap transients, they also imply a very low-density environment in the context of a core-collapse explosion. We show a comparison of the radio limits on iPTF 16hgs to the population of stripped and relativistic core-collapse SNe in Figure 20. As shown, these limits rule out radio emission similar to the majority of radio-detected Type Ib/c SNe (see also Figure 16 in Drout et al. 2016 and Figure 16 in Milisavljevic et al. 2017), which typically exhibit radio luminosities of ≳10²⁶ erg s⁻¹ Hz⁻¹. However, these limits are consistent with the low-density environments of stripped-envelope SNe, like the radio-faint SN 2007gr (Soderberg et al. 2010) and SN 2002ap (Berger et al. 2002), along with the Type Ib/c SN 2013ge (Drout et al. 2016), that remained undetected in the radio band. Additionally, these limits also rule out radio emission similar to that of late-time interacting events like SN 2007bg (Salas et al. 2013) and SN 2014C (Anderson et al. 2017).

Zoom In Zoom Out Reset image size

In the context of inferring the CSM environment of the progenitor, we also considered a scenario where the early peak in the optical light curve is powered by interaction of the ejecta with a dense CSM envelope around the progenitor. In order to power the early luminosity, we find that an ISM density ρ ∼ 3 × 10⁻¹⁵ g cm⁻³ (n_e ∼ 10⁹ cm⁻³) is required for a constant CSM density environment, while a mass-loss rate of $\dot{M}\sim 7\times {10}^{-5}\tfrac{{v}_{\mathrm{CSM}}}{100\,\mathrm{km}\,{{\rm{s}}}^{-1}}$ M_⊙ yr⁻¹ would be required for a constant mass-loss environment. We note that this is at odds with CSM environment constraints derived from our late-time radio observations, which constrain the mass-loss environment to $\dot{M}\lesssim {10}^{-5}$ M_⊙ yr⁻¹ and n_e ≲ 10³ cm⁻³ for a constant density environment, even for a _B = 0.01. Indeed, this would be consistent with the elevated pre-explosion mass-loss episodes inferred from very early observations of both H-rich and H-poor core-collapse SNe (Gal-Yam et al. 2014; Yaron et al. 2017).

7.2.2. Limits on a Tidal Disruption Scenario

Tidal detonations of low-mass He WDs have been previously proposed as a potential progenitor channel for Ca-rich gap transients (Sell et al. 2015). In this context, Foley (2015) argued that the host offsets of Ca-rich gap transients appear to show a correlation with their radial velocity offsets (as inferred from the late-time nebular spectra) from their host galaxy, suggesting a nuclear origin for the progenitors of Ca-rich gap transients. This would be consistent, for instance, if a WD binary system (where the companion is a NS/BH) were hardened due to interaction with a central supermassive BH in the host galaxy, eventually leading to the tidal disruption of the WD (Sell et al. 2015).

However, Milisavljevic et al. (2017) argue that the radial velocity argument (based on the blueshifts of the [Ca ii] emission lines in Foley 2015) may be flawed, because the [O i] velocities do not show these systematic offsets even if the [Ca ii] do exhibit them. In Figure 9, we show the profiles of the [Ca ii] λλ7291, 7324 lines and the the [O i] λλ6300, 6364 lines in the late-time spectra of iPTF 16hgs, as a function of velocity offset from 7306 Å and 6300 Å, respectively. It is interesting to note that the nebular emission features in iPTF 16hgs show no evidence of a systematic offset in either the [Ca ii] or [O i] lines.

We also compare the predictions of the models of MacLeod et al. (2016) to our data; they predicted the photometric and spectroscopic signatures of tidal disruptions of 0.6 M_⊙ WDs by an IMBH. In general, there are major differences between their predictions and our data, specifically that the timescales of the predicted light curves are longer while the predicted spectroscopic signatures are inconsistent. This is not surprising, given that the detonation of a 0.6 CO M_⊙ WD produces more massive ejecta (and hence a slower-evolving light curve) and different nucleosynthetic signatures (dominated by Si), compared to the He-rich spectra of iPTF 16hgs. Future modeling will be required to understand the properties of tidal disruptions of He-rich WDs and whether they can reproduce the observed signatures of iPTF 16hgs. Nevertheless, the first peak of the light curve remains to be explained in such a scenario, but it could potentially be associated with an optical flare arising out of the initial rapid fall-back accretion on to the compact object.

As observed in some tidal disruption events of nondegenerate stars (e.g., Gezari et al. 2012), the production of a relativistic jet in the disruption was suggested to be a potential signature of such a progenitor channel (Sell et al. 2015; MacLeod et al. 2016). Based on our multiwavelength data set on iPTF 16hgs, we place stringent constrains on the presence of a relativistic jet in Section 5.2. In particular, we rule out the presence of a relativistic jet for a near on-axis jet, unless the jet energy was very low (≲10⁴⁹ erg, equivalent to ≲10⁻⁵ of the rest-mass energy of a 0.4 M_⊙ WD).

Independently, we can try to place limits on the mass of the compact disrupting object by using our X-ray limits on iPTF 16hgs. The accretion flare coincident with the WD disruption is expected to be several orders of magnitude above the Eddington rate (Rosswog et al. 2008; MacLeod et al. 2014, 2016), and hence the X-ray emission along the axis of the jet is likely to be super-Eddington and tracing the ∝t^−5/3 accretion rate evolution of the BH (Phinney 1989). Along all other directions, the X-ray luminosity is likely to be at least the Eddington luminosity L_Edd, where

$$\begin{eqnarray}&&{L}_{\mathrm{Edd}}\approx {10}^{41}\displaystyle \frac{{M}_{c}}{{10}^{3}{M}_{\odot }}\,\mathrm{erg}\,{{\rm{s}}}^{-1}\end{eqnarray} \tag{ 2 }$$

and M_c is the mass of the compact object. Hence, the nondetection of X-ray emission in our Swift XRT observations constrains the mass of the compact object to be M_c ≲ 490 M_⊙. However, such an interpretation may be complicated by possible effects of reprocessing layers in the accreting system, which can reprocess the high-energy emission into optical/UV light (e.g., Gezari et al. 2012; Miller 2015; Hung et al. 2017).

Recent transient surveys have discovered a diverse population of events that can be classified as "Ca-rich" based on their nebular phase spectra. Apart from the members of the class of Ca-rich gap transients, notable examples include iPTF 15eqv (Milisavljevic et al. 2017), which exhibited a high [Ca ii]/[O i] ratio in its nebular phase spectra and hence was suggested to be "Ca-rich," while its high luminosity and slow decline exclude it from the class of Ca-rich gap transients. Filippenko et al. (2003) and Perets et al. (2010) presented a set of events that appeared to be Ca-rich based on their nebular spectra, although no photometric information was available for comparison to Ca-rich gap transients.

Based on the general peculiarity of transients that exhibit strong [Ca ii] emission in their nebular phase spectra, we searched the sample of spectroscopically classified stripped-envelope SNe (Type IIb, Ib, or Ic) discovered by PTF and iPTF (based on the compilation by Fremling et al. 2018) to look for objects that exhibit a high [Ca ii]/[O i] ratio in their nebular phase spectra. We show our results in a plot of [Ca ii]/[O i] ratio of the transient as a function of the phase from light curve peak in Figure 21 (similar to Figure 11 in Milisavljevic et al. 2017 and Figure 13 in Valenti et al. 2014). A major fraction of the events in this sample are Type Ic SNe because these events were followed up to late times in the nebular phase. As shown, all the stripped-envelope SNe in this sample exhibit a low [Ca ii]/[O i] ratio at all phases, consistent with the suggestion that the Ca-rich classification is not sensitively dependent on the phase of the transient (Lunnan et al. 2017).

Zoom In Zoom Out Reset image size

The Ca-rich transients clearly stand out in this plot, occupying the phase space of high [Ca ii]/[O i] ratio at early phases. Denoted by a star, iPTF 16hgs is also consistent with the other members of this class. A notable exception is iPTF 15eqv (denoted in yellow), which, as mentioned earlier, exhibited a Ca-rich nebular phase spectrum but was not a Ca-rich gap transient photometrically. The lowest [Ca ii]/[O i] ratio in the sample of Ca-rich gap transients is that of SN 2012hn (Valenti et al. 2014), which exhibited a [Ca ii]/[O i] ≈2.3 at a phase of ≈150 days. In this context, Milisavljevic et al. (2017) suggested that a [Ca ii]/[O i] ratio of 2 separates the Ca-rich transients from the population of other stripped-envelope SNe.

The class of Ca-rich gap transients, as defined by Kasliwal et al. (2012), includes a unique subset of the broader class of Ca-rich transients that was empirically defined to include many faint and fast-evolving SNe that exhibited conspicuous [Ca ii] emission in their early nebular phase spectra. In particular, the definition did not constrain the properties of the photospheric phase spectra of these sources, which is likely to be an important indicator of the nature of the explosion. Hence, objects like PTF 09dav (Sullivan et al. 2011) and PTF 12bho (Lunnan et al. 2017) that exhibit significantly different photospheric phase spectra may arise from different progenitor channels than the majority of this class, which appear to exhibit He-rich Type Ib-like spectra near peak light, similar to the prototype event SN 2005E (Perets et al. 2010).

The discovery of iPTF 16hgs, with its distinct double-peaked light curve, adds further diversity to this class. If we were to invoke a progenitor channel for iPTF 16hgs similar to the old inferred progenitor systems of other Ca-rich gap transients, we would find that a thermonuclear detonation event from a WD binary system could still be a viable explanation that unifies this explosion and the rest of the class. In such a case, the light curve of iPTF 16hgs provides the first evidence of significant differences in the mixing of radioactive material in the detonation, providing strong constraints on the explosion mechanism. As the environment of the transient suggests its association with a young stellar population, this could perhaps also suggest that thermonuclear detonations occurring with shorter delay times lead to such highly mixed radioactive material in the explosion.

We also considered the scenario where the early emission in iPTF 16hgs could be powered by the interaction shock between the ejecta and a nondegenerate companion. An attractive aspect of this interpretation is that it would naturally explain the uniqueness of iPTF 16hgs in the class of Ca-rich gap transients, as this signature is expected to be prominent only if the observer is oriented along the direction of the companion. Based on this, Kasen (2010) estimates that only 10% of Type Ia SNe should exhibit these early signatures. If Ca-rich gap transients arise from accreting WDs with nondegenerate companions, we would also expect only 1 in 10 of these transients to exhibit early excess emission, consistent with the discovery of iPTF 16hgs. However, our analysis of the companion interaction model suggests that the color and luminosity evolution of iPTF 16hgs are inconsistent with the models presented in Kasen (2010). Future modeling will be required to understand whether the assumptions of the companion interaction models for the massive ejecta in Type Ia SNe hold in low ejecta mass (and hence low optical depth) events with different ejecta compositions, as in iPTF 16hgs.

Another widely discussed channel for the broader class of Ca-rich gap transients that may be relevant for iPTF 16hgs is the disruption of a low-mass WD by an NS or BH (Metzger 2012; Sell et al. 2015; MacLeod et al. 2016; Margalit & Metzger 2016). However, it has also been suggested that such a merger likely leads to an associated radio transient powered by the interaction of either a relativistic jet or a fast wind outflow with the surrounding CSM (Metzger 2012; MacLeod et al. 2016; Margalit & Metzger 2016). Based on our deep radio limits, we find that such a radio interaction signature can be hidden only if the explosion took place in a relatively clean environment (n_e ≲ 10⁻² cm⁻³) or if the jet energy was particularly low (E ≲ 10⁴⁹ erg), noting that the constraints are particularly stringent in the case of an on-axis event (n_e ≲ 10⁻⁶ cm⁻³). If iPTF 16hgs arises from a progenitor channel similar to those of other Ca-rich gap transients, then our observations provide strong evidence against the formation of relativistic jet-like outflows in these explosions.

Recent theoretical studies of the formation and evolution of systems leading to NS-WD mergers suggest that their rates are likely to be much lower than that inferred for Ca-rich gap transients (∼30–90% of the Type Ia SN rate) after accounting for the survey biases of PTF (Frohmaier et al. 2018; Toonen et al. 2018). The inferred rates would also be unusually high for scenarios involving tidal disruptions of low-mass He WDs by an NS or BH (Sell et al. 2015), in which case it would also be difficult to understand their preference for old stellar populations. Modeling of the physical properties of NS-WD mergers also suggest that they may not be able to explain the He-rich spectra of most Ca-rich gap transients and old environments (Zenati et al. 2018; see also Margalit & Metzger 2016).

As noted in Lunnan et al. (2017), the host environments of Ca-rich gap transients are striking in light of their preference for hosts with old stellar populations in group and cluster environments. In particular, these differ substantially from the observed host properties of Type Ia SNe (which occur in both early- and late-type galaxies; see discussion in Lunnan et al. 2017). In general, the older environments may be indicative of a delay-time distribution that extends out to much longer delay times than in the case of Type Ia SNe. While iPTF 16hgs was also found in a sparse galaxy group, its star-forming dwarf host galaxy makes it stand out in this sample. Interestingly, the only other Ca-rich gap transient hosted in a star-forming host galaxy was PTF 09dav, which was also a spectroscopically peculiar event compared to the other members of this class. Unlike the majority of Ca-rich gap transients, which do not show any evidence of underlying host systems down to stringent limits that rule out the presence of a dwarf galaxy or globular cluster (Lyman et al. 2014, 2016b; Lunnan et al. 2017), iPTF 16hgs clearly occurred inside its host galaxy, and is thus robustly associated with a host system.

Regardless, the low metallicity and young stellar population of the host galaxy still provide important clues to the nature of the progenitor star. In this context, we note that the remote locations of these transients in the outskirts of galaxies (where the metallicity is likely lower) have been used to argue for low-metallicity progenitors of these explosions (Yuan et al. 2013). If the low metallicity is indeed a driving factor leading to these peculiar explosions, it is not surprising that the majority of Ca-rich gap transients are found in the metal-poor halos of old galaxies. Although iPTF 16hgs was discovered at the smallest host offset of any known Ca-rich gap transient, it was found in a very low-metallicity galaxy, and is thus consistent with the preference of these events for metal-poor environments. In this context, it is also interesting to note that the host galaxy of SN 2005E (Perets et al. 2010), which was also discovered at a small projected offset, was a low-metallicity galaxy (12 + log(O/H) ≈8.4; Milisavljevic et al. 2017), further strengthening the association of these transients with metal-poor environments.

Nevertheless, given that there are examples of core-collapse SNe, such as iPTF 15eqv, that exhibit high [Ca ii]/[O i] ratio in their nebular phase spectra, and given that iPTF 16hgs occurred in a star-forming environment, it is indeed plausible that iPTF 16hgs is associated with a core-collapse explosion of a massive star. However, we reiterate that a core-collapse scenario can be ruled out for most members of this class based on their environments, and thus if a core-collapse scenario is true for iPTF 16hgs, it would require the existence of multiple progenitor channels for producing Ca-rich SNe, even though they appear quite homogeneous in terms of the SN properties. In this case, the low ejecta mass and low peak luminosity of the main peak make it striking in the context of core-collapse explosions, suggesting its association with a highly stripped progenitor similar to those discussed in the context of stripped-envelope electron-capture explosions (Kitaura et al. 2006; Yoon et al. 2010; Tauris et al. 2015; Moriya et al. 2017). Although our radio limits suggest an unusually low-density environment for the progenitor, the low final mass of the progenitor and the low metallicity of the progenitor environment may play a role in suppressing the formation of dense CSM environments by line-driven winds (Smartt 2009; Langer 2012; Crowther 2013).