iPTF 16asu: A Luminous, Rapidly Evolving, and High-velocity Supernova

iPTF 16asu was discovered by the intermediate Palomar Transient Factory (Law et al. 2009; Cao et al. 2016; Masci et al. 2017) and was first detected in data taken with the 48 Samuel Oschin Telescope at the Palomar Observatory (P48) on 2016 May 11.26 UT (UT dates are used throughout this paper) at coordinates R.A. = 12^h59^m0928, decl. = +13°48'092 (J2000.0) and at a magnitude of g = 20.54 mag. We obtained a spectrum with the Double Beam Spectrograph (DBSP; Oke & Gunn 1982) on the 200" Hale Telescope at the Palomar Observatory (P200) on 2016 May 14.3 that shows a blue continuum and narrow Hα and [O iii] lines from the host galaxy, setting the redshift to z = 0.187. A later spectrum taken on 2016 June 04 by the Deep Imaging Multi-object Spectrograph (DEIMOS; Faber et al. 2003) on the 10 m Keck II Telescope on 2016 Jun 04 shows SN features consistent with an SN Ic-BL. The spectroscopic evolution is discussed in Section 4.

iPTF 16asu was detected in a nightly cadence g band experiment with iPTF, and we therefore have P48 data covering the time up to explosion as well as the early rise. Subsequent photometry was obtained with the automated 60 telescope at Palomar (P60; Cenko et al. 2006) in the gri bands. Host-subtracted point-spread function photometry was obtained using the Palomar Transient Factory Image Differencing and Extraction pipeline (Masci et al. 2017) on the P48 images and the FPipe SEDM presented in Fremling et al. (2016) on the P60 images using Sloan Digital Sky Survey (SDSS; SDSS Collaboration et al. 2016) images as templates and also calibrating to SDSS. Our last photometric observation came from the 3.58 m Telescopio Nazionale Galileo (TNG) and was processed through the FPipe. The photometry was corrected for Galactic extinction following Schlafly & Finkbeiner (2011), with $E(B-V)=0.029$ mag, and all magnitudes in this paper are reported in the AB system. Table 1 lists all photometric data, shown in Figure 2.

Zoom In Zoom Out Reset image size

We obtained a sequence of eight low-resolution spectra for iPTF 16asu using the DBSP on P200, the Andalucia Faint Object Spectrograph and Camera (ALFOSC) on the 2.56 m Nordic Optical Telescope (NOT), the Device Optimized for the Low Resolution (DOLORES) on TNG, the Low-resolution Imaging Spectrometer (LRIS; Oke et al. 1995) on Keck I, and the DEIMOS on the 10 m Keck II Telescope. The times of the spectra are marked as dashed lines in Figure 2, and details of the spectroscopic observations are given in Table 2. Spectra were reduced using standard procedures using IRAF¹⁸ and IDL, including wavelength calibration using arc lamps and flux calibration using standard stars. The spectroscopic sequences for iPTF 16asu is shown in Figure 3, and the spectroscopic properties are analyzed and discussed in Section 4. All spectra will be made available in the Weizmann Interactive Supernova Data Repository (Yaron & Gal-Yam 2012).

Zoom In Zoom Out Reset image size

We observed the field of iPTF 16asu with the Karl G. Jansky Very Large Array (VLA) on two epochs (Program VLA/16B-043; PI: A. Corsi). The first observation was carried out starting on 2016 June 13, 01:18:22 UT (MJD 57552), with the VLA in its B configuration. The second observation was carried out with the VLA in its A configuration, starting on 2017 January 10, 09:43:06 UT (MJD 57763). Both these observations were carried out in the C band (nominal central frequency of ≈5 GHz), using the 8 bit configuration and 2 GHz nominal bandwidth. On both epochs we used 3C286 as a bandpass and flux density calibrator and J1300 + 1417 as a phase calibrator. The total observing time was about 1 hr (including calibration and overhead) per epoch.

VLA data were calibrated using the automated VLA calibration pipeline in CASA (McMullin et al. 2007). After visual inspection, additional flags were applied when needed. Images of the fields were produced using the CLEAN task (Högbom 1974).

We searched for a radio counterpart to iPTF 16asu within a 2'' radius circle centered on the iPTF position of iPTF 16asu. No radio source was detected within this region down to a 3σ limit of $\approx 17\,\mu \mathrm{Jy}$ at 6.2 GHz for both epochs.

At the time of the first spectrum, iPTF 16asu resembled a very young SLSN, with its already high luminosity and blue spectrum indicating a high temperature. We therefore triggered our Swift program for SLSNe (GI-1215281, PI: R. Lunnan), and three epochs of Swift UVOT (Roming et al. 2005) and XRT (Burrows et al. 2005) data were obtained at phases corresponding to 7.4, 13.4, and 19.2 days after explosion (see Section 3.1 for the calculation of the explosion date).

We reduced the Swift data using the HEASoft package provided by NASA.¹⁹ UVOT photometry was performed using the task UVOTsource with an aperture of 5''. iPTF 16asu was detected in all filters except the V band in the first observation and undetected in all UVOT filters in the subsequent two epochs, due to the rapid fading of the SN. All UVOT photometry is listed in Table 1.

The XRT data were reduced with the Ximage software from the HEASoft package. No X-ray source was detected at the position of iPTF 16asu in either epoch. The $3\sigma$ upper limits correspond to $5.6\times {10}^{-3}$ counts s⁻¹, $2.9\times {10}^{-3}$ counts s⁻¹, and $3.9\times {10}^{-3}$ counts s⁻¹, respectively. Using WebPIMMS²⁰ and assuming a Galactic nH of $2.2\times {10}^{20}\,{\mathrm{cm}}^{-2}$, we find that $1\times {10}^{-3}$ counts s⁻¹ corresponds to $3.76\times {10}^{-14}\,\mathrm{erg}\,{\mathrm{cm}}^{-2}\,{{\rm{s}}}^{-1}$ (unabsorbed; 0.3–10 keV), assuming a power-law model with an index of 2. At a redshift of z = 0.1874, our X-ray count limits translates to flux limits of $2.5\times {10}^{43}\,\mathrm{erg}\,{{\rm{s}}}^{-1}$, $1.1\times {10}^{43}\,\mathrm{erg}\,{{\rm{s}}}^{-1}$, and $1.5\times {10}^{43}\,\mathrm{erg}\,{{\rm{s}}}^{-1}$, respectively.

We searched the Gamma-Ray Coordinates Network archives for any announced GRBs consistent with the location and best-fit explosion time of iPTF 16asu (Section 3.1). No announced GRB was consistent with the location and time of iPTF 16asu or when extending the search to include bursts detected between the last iPTF nondetection and the first detection of iPTF 16asu. However, our analysis of the Konus-Wind data (KW; Aptekar et al. 1995) reveals that a weak burst was detected by KW in the waiting mode (with a time resolution of 2.944 s) on 2016 May 10.41, which is consistent with our best-fit explosion time of 2016 May 10.53 ± 0.17 days (see Section 3.1). The burst was observed by the KW S2 detector pointing the northern ecliptic hemisphere (nothing was seen in the opposite S1 detector), which is also consistent with the position of iPTF 16asu, but the burst source position cannot be constrained more precisely from the KW data.

The burst emission was significant in the two softest KW energy bands: G1 (20–80 keV, $13\sigma$) and G2 (80–300 keV, $8\sigma$). The burst light curve shows a single emission episode with a duration of 126 s (T₅₀ = 56 ± 11 s and T₉₀ = 100 ± 11 s, both measured in the 20–300 keV energy band). Fitting the KW tree-channel time-integrated spectrum (measured from T0 to T0+126.592 s) by a simple power law yields the photon index of ${2.35}_{-0.14}^{+0.18}$, ${\chi }^{2}/\mathrm{dof}=2.7/1$. From this fit, the burst had an energy fluence of ${8.25}_{-0.86}^{+1.60}\times {10}^{-6}\,\mathrm{erg}\,{\mathrm{cm}}^{-2}$ and a 2.944 s peak energy flux, measured from T0 + 73.6 s, of ${2.41}_{-0.94}^{+1.02}\times {10}^{-7}\,\mathrm{erg}\,{\mathrm{cm}}^{-2}{{\rm{s}}}^{-1}$ (both in the 20–1200 keV energy range). At the distance of iPTF 16asu, this fluence would correspond to an equivalent isotropic energy ${E}_{\mathrm{iso}}$ of $8.2\times {10}^{50}\,\mathrm{erg}$. The fit with a power law with an exponential cutoff model yields only an upper limit on spectrum peak energy: ${E}_{{\rm{p}}}\lt 67\,\mathrm{keV}$.

During the KW burst, Swift was in the South Atlantic Anomaly and the position of iPTF 16asu was Earth-occulted. However, the position of iPTF 16asu was not occulted for Fermi (and six GBM detectors had incident angles of less than 60°). We analyzed the Fermi-GBM continuous data and found no emission in the 30–300 keV band coincident with the KW burst. Given that the background of Fermi-GBM is considerably lower than KW, this implies that the KW burst came from a source Earth-occulted to Fermi and therefore is not related to iPTF 16asu.

We also searched for a possible GRB in the INTEGRAL-SPI-ACS (SPI-ACS; von Kienlin et al. 2003) data covering the 75–8000 keV range and found no candidate event down to the 3 sigma level. Since KW and SPI-ACS were observing the whole sky during the interval of interest, upper limits on gamma-ray flux can be obtained. For the whole interval (excluding the KW burst), assuming a typical long GRB spectrum (the band function with $\alpha =-1$, $\beta =-2.5$, and ${E}_{{\rm{p}}}=300\,\mathrm{keV}$), the corresponding KW and SPI-ACS limiting peak fluxes estimates are $\sim (1\mbox{--}4)\times {10}^{-7}\,\mathrm{erg}\,{\mathrm{cm}}^{-2}\,{{\rm{s}}}^{-1}$, both in the 10 keV–10 MeV band at 3–10 s time scales.

We conclude therefore that there is no statistically significant evidence for an SN-associated GRB down to the threshold of ${10}^{-7}\,\mathrm{erg}\,{\mathrm{cm}}^{-2}\,{{\rm{s}}}^{-1}$. The associated isotropic peak luminosity limit is ${L}_{\mathrm{iso}}\lesssim {10}^{49}\mathrm{erg}\,{{\rm{s}}}^{-1}$ and the total energy is ${E}_{\mathrm{iso}}\lesssim {10}^{50}\,\mathrm{erg}$ (both calculated in the 10 keV–10 MeV energy range). Hence from these limits an accompanying low-luminosity GRB, like GRB 980425 (${L}_{\mathrm{iso}}\sim 5\times {10}^{46}\,\mathrm{erg}\,{{\rm{s}}}^{-1}$, ${E}_{\mathrm{iso}}\sim {10}^{48}\,\mathrm{erg};$ Galama et al. 1998b), cannot be excluded. We return to discuss possible GRB models for iPTF 16asu in Section 6.3.

The light curves of iPTF 16asu are shown in Figure 2. The rise and peak are only sampled in the g band, so we fit a second-order polynomial to the g band light curve near peak brightness to determine a best-fit explosion date, time of peak, and peak luminosity. The fit is shown in Figure 4, and the explosion and best-fit peak dates are MJD 57518.53 ± 0.17 and MJD 57523.25 ± 0.14, respectively. Corresponding calendar dates are 2016 May 10.53 and 2016 May 15.25. Thus, the rise time (time of peak—time of explosion) is 3.97 ± 0.19 days in the rest frame. The last optical upper limit prior to the first detection was MJD 57513.31, setting an upper limit to the rise time of 9.94 days.

Zoom In Zoom Out Reset image size

Using our series of spectra, we calculate K-corrections from the observed filters at z = 0.187 to the rest-frame filters, listed in Table 3. At this redshift the observed gri corresponds most closely to the rest-frame BVr, and the wavelength coverage of our spectra allows us to also calculate K-corrections to the u, g, and i filters. Applying this, we find that the time of peak corresponds to a peak absolute magnitude of ${M}_{{\rm{B}}}=-20.4$ mag (AB).

Table 3.  K-corrections Derived from Spectra

Phase	${K}_{\mathrm{rr}}$	${K}_{\mathrm{gr}}$	${K}_{\mathrm{gg}}$	${K}_{\mathrm{ug}}$	${K}_{\mathrm{ii}}$	${K}_{\mathrm{ri}}$	${K}_{\mathrm{Bg}}$	${K}_{\mathrm{Vr}}$
(rest-frame days)	(mag)	(mag)	(mag)	(mag)	(mag)	(mag)	(mag)	(mag)
−0.73	−0.368	−0.041	−0.270	−0.194	⋯	−0.173	−0.192	−0.269
+0.75	−0.357	−0.062	−0.260	−0.215	⋯	−0.164	−0.177	−0.258
+8.25	−0.136	−0.286	0.056	−0.372	−0.285	−0.206	−0.078	−0.175
+10.27	−0.107	−0.280	0.062	−0.438	−0.325	−0.204	−0.079	−0.158
+17.03	−0.109	⋯	⋯	⋯	⋯	−0.199	⋯	−0.121
+19.53	−0.043	−0.467	0.410	−0.688	−0.255	−0.195	0.102	−0.112
+22.11	−0.050	−0.487	0.422	−0.768	−0.214	−0.187	0.127	−0.110

Notes.

^aK-correction is defined as in Hogg et al. (2002), so that for filters Q and R, ${m}_{{\rm{R}}}={M}_{{\rm{Q}}}+{DM}+{K}_{\mathrm{QR}}$. ^bPhase is in rest-frame days relative to the bolometric maximum light (MJD 57523.25).

Download table as:  ASCIITypeset image

iPTF 16asu inhabits an unusual location in rise time versus luminosity parameter space (see Figure 1). In this section, we compare its light curve in more detail to objects in the literature that have been noted for their fast timescales and/or high luminosities. Unfortunately, K-corrections are not available for the majority of our comparison objects due to a lack of spectroscopic coverage. For the purposes of comparison, we corrected iPTF 16asu and all comparison objects for redshift using the following equations:

$$\begin{eqnarray}&&\lambda =\displaystyle \frac{{\lambda }_{\mathrm{obs}}}{(1+z)},\end{eqnarray} \tag{ 1 }$$

$$\begin{eqnarray}&&M={m}_{\mathrm{obs}}-5\,{\mathrm{log}}_{10}\left(\displaystyle \frac{{D}_{L}}{10\,\mathrm{pc}}\right)+2.5\,{\mathrm{log}}_{10}(1+z).\end{eqnarray} \tag{ 2 }$$

We then chose filters with rest wavelengths as closely corresponding to those of iPTF16asu as possible in order to facilitate comparison. Comparing the approximation used in Equation (2) to the actual K-corrections calculated for iPTF 16asu (Table 3), we expect this approximation to introduce errors on the order of 0.1 mag. Figure 5 shows comparisons to the g band (left) and r band (right) light curves.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

First we compare against the light curves of SNe noted for both their high luminosities and their rapid timescales. These include SN 2011kl (Greiner et al. 2015), an SN associated with the ultralong gamma-ray burst GRB 111209A (plotted in black), and PTF 10iam (blue), SNLS04D4ec (blue), SNLS05D2bk (green), and SNLS06D1hc (cyan) from Arcavi et al. (2016). In the g band as seen in Figure 5 (left), iPTF 16asu reaches a higher peak luminosity than these transients by over half a magnitude. Measuring from the rest-frame phase at ${M}_{\mathrm{peak}-1\mathrm{mag}}$ to ${M}_{\mathrm{peak}+1\mathrm{mag}}$, iPTF 16asu's timescale is about two times shorter with ${\tau }_{\mathrm{peak}-1\mathrm{mag}}=10$ days. iPTF 16asu displays both a steeper rise and steeper decay than the Arcavi et al. (2016) objects and SN 2011kl in the g band. However, iPTF 16asu resembles these objects more closely in the r band, shown in Figure 5 (right). The peak r band magnitude of iPTF 16asu is approximately the same as PTF 10iam, SNLS05D2bk, and SNLS06D1hc and the slope of decay runs nearly parallel to that of SNLS06D1hc. Although we have no data on the rise in the r band, iPTF 16asu has a peak magnitude similar to that of to the Arcavi et al. (2016) objects and decays on the same timescale as SNLS06D1hc.

Next we compared the light curve to PS1-10bjp, PS1-11qr, PS1-12bv, and PS1-12brf, a sample of rapidly evolving transients from the Pan-STARRS1 Medium Deep Survey (Drout et al. 2014). The objects shown are the four most luminous objects from the "gold" sample and are plotted in yellow in Figure 5. They have rise times and decay slopes similar to those of iPTF 16asu but are much fainter. In the g band, PS1-11qr and PS1-12bv are the brightest of the Pan-STARRS1 objects, reaching a peak magnitude of about −19.5 mag; thus iPTF 16asu is a magnitude brighter at peak. As seen in Figure 5, the shape of iPTF 16asu's light curve is quite similar to that of PS1-10bjp and PS1-11qr. Early in the decay of iPTF 16asu, the slope is nearly parallel to that of PS1-10bjp; however, at late times PS1-10bjp decays more sharply than iPTF 16asu. Comparing these objects to the r band data is less instructive because iPTF 16asu's rise was not captured in the r band and most of the Pan-STARRS1 objects do not have late-time data.

Finally we compared iPTF 16asu to the SLSNe PTF11rks (Inserra et al. 2013) and PS1-10bzj (Lunnan et al. 2013), which are both on the lower-luminosity end of SLSNe. In the g band, iPTF 16asu reaches about the same peak absolute magnitude as PTF11rks. In the r band iPTF 16asu's peak luminosity is about 0.5 mag dimmer than that of PTF11rks. However, the SLSNe have timescales several times longer than that of iPTF 16asu, as seen by the much broader peaks. Thus, while iPTF 16asu reaches luminosities similar to those of some SLSNe, it evolves on a very different timescale. iPTF 16asu stands out as a unique and surprising event, even amongst similar transients from the literature.

We fit a blackbody to all epochs where we have observations in at least three filters, using Scipy least-squares optimization routines (Jones et al. 2001) as well as to our two earliest spectra. Only the day with Swift/UVOT detections (+3 days past peak) has data in more than three filters. The fit to the Swift photometry is shown in Figure 6. From this fit we obtain T = 10,800 ± 250 K and R = (2.6 ± 0.2) × 10¹⁵ cm. This corresponds to a total blackbody luminosity of $(6.4\pm 1.6)\times {10}^{43}$ erg s⁻¹.

Zoom In Zoom Out Reset image size

Figure 7 shows the resulting derived temperatures and radii at all epochs. The overall trends show a cooling blackbody temperature and an increasing radius. In fitting a straight line to the blackbody radii, we get a best-fit slope of 34,500 ± 5400 km s⁻¹, indicating high average velocities.

Zoom In Zoom Out Reset image size

We construct a pseudobolometric light curve for iPTF 16asu by summing the observed flux on days where we have observations in at least three filters. We integrate over the observed spectral energy distribution using trapezoidal integration, interpolating to the edges of the observed bands. Since this only accounts for the observed flux, it constitutes a strict lower limit on the true bolometric luminosity.

Pre-peak photometry is only available in the g band so we approximate the rise of the pseudobolometric light curve by assuming a constant ratio of g band flux to total flux, i.e., a constant bolometric correction. This assumption is equivalent to assuming that the temperature on the rise is constant and equal to the temperature measured from the earliest multiband data. Similarly, for the late-time observations with data only in the r band we estimate the total flux by using the same bolometric correction as from the latest date with data in $\geqslant 3$ filters.

We caution that in constructing a pseudobolometric light curve from optical data only, we are implicitly assuming that the ratio of optical to bolometric flux is approximately constant over the time period of interest. In particular, at late times as the effective temperature falls we would expect the near-IR (NIR) fraction of the bolometric luminosity to increase, which is unconstrained from observations, so assuming a constant bolometric correction is likely an underestimate. Unfortunately, NIR data are also not available for other fast-evolving SNe so we cannot use them as a basis for comparison. However, given that iPTF 16asu resembles a normal SN Ic-BL–like SN 1998bw from ∼20 days past explosion onward (Sections 4 and 6.1), we expect its late-time evolution of the optical-to-bolometric flux ratio to resemble normal SNe Ic-BL. Comparing to the light curve samples analyzed in Lyman et al. (2014), we estimate that this adds an uncertainty on the level of 10%–20% at late times.

Figure 8 (left) shows the resulting pseudobolometric light curve. Using trapezoidal integration over time we calculate an estimated total radiated energy of $(4.0\pm 0.6)\times {10}^{49}$ erg and a peak luminosity of $(3.4\pm 0.3)\times {10}^{43}$ erg s⁻¹.

Zoom In Zoom Out Reset image size

The shape of the decline of the bolometric light curve sheds light on what physical processes may be powering this event. Figure 8 (right) shows the best-fit power law and exponential fits to the post-peak light curve. Clearly the decline of the light curve does not follow a power law; however, it fits an exponential well. The power law has a best-fit decay of $L\,\propto \,{t}^{-1.06\pm 0.14}$ and the exponential decays are on a timescale of $\tau =13.56\pm 0.56$ days. The power-law decay parameter is similar to those found for the objects in Arcavi et al. (2016). Also similar, two of the four Arcavi et al. (2016) objects are better fit by an exponential. The implications of these results are discussed in Section 6.

We obtained eight spectra of iPTF 16asu between 2016 May 14 and 2016 Jul 6. The spectra are shown in Figure 3. In this section, we look at the spectroscopic evolution in more detail and compare the spectroscopic properties of iPTF 16asu to similar objects from the literature.

The first two spectra, taken within a day before and after peak, show a featureless blue continuum with no discernible broad features. The spectrum is well fit by a blackbody, as shown in Figure 9. Such spectra dominated by blue continua have also been observed at early phases in other SNe, typically while the luminosity is powered by the cooling of the stellar envelope following shock breakout (see, e.g., SN 1993J; Woosley et al. 1994; Matheson et al. 2000). Interestingly, the rapidly evolving SNe from Pan-STARRS1 (Drout et al. 2014) also showed featureless blue continua. Figure 10 shows a comparison of PS1-12bv at peak compared to iPTF 16asu at peak. Unfortunately, comparison at late times is not possible as there is no further follow-up spectroscopy on the Pan-STARRS1 events. Based on the limited spectroscopic data available we cannot rule out that they were caused by the same phenomenon as iPTF 16asu.

Zoom In Zoom Out Reset image size

The next two spectra, taken at phases 8 and 10 days past maximum, still show an underlying blue continuum but with broad features emerging. Such an evolution is reminiscent of GRB SNe. To illustrate this we show a comparison to SN 2006aj/GRB 060218 (Modjaz et al. 2006) in Figure 11. SN 2006aj is of particular interest here because it is one of the few GRB SNe that would not be ruled out by our radio and X-ray limits. We discuss GRB models for iPTF 16asu in detail in Section 6.3.

Zoom In Zoom Out Reset image size

The three spectra taken at phases +17, +19, and +22 days post maximum are dominated by distinct broad-line features, leading us to classify iPTF 16asu as an SN Ic-BL. Figure 12 shows a comparison of iPTF 16asu at +23 days after explosion (+19 days past peak) to SN 1998bw at +18 days after explosion, and features commonly identified in SNe Ic-BL are marked. Interestingly, the spectra of these events look very similar at roughly the same time after explosion, suggesting that iPTF 16asu may have a normal-timescale SN component hidden underneath the luminous and rapidly evolving peak.

Zoom In Zoom Out Reset image size

It is also worth noting that the spectroscopic evolution of iPTF 16asu is different from the few objects in Drout et al. (2014) and Arcavi et al. (2016) with spectra at later phases: PS1-12bb showed a featureless continuum at phase +33 days, while PTF 10iam showed broad Hα emission at phase +28 days. This spectroscopic diversity suggests that there are likely multiple physical mechanisms giving rise to light curves in this part of the transient phase space.

Our final spectrum, taken at phase +44 days past peak, is dominated by host galaxy light. We discuss the host galaxy properties in Section 5.

Measuring velocities from SNe Ic-BL spectra is challenging since the lines are often blended due to the high velocities. In addition, different lines can give different velocities because these elements are formed and found at different radii in the expanding ejected material. For iPTF 16asu we chose the strongest lines: the Si ii $\lambda 6355$ Å line and the Fe ii $\lambda 5169$ Å line.

In the case of the Si ii $\lambda 6355$ Å line we fit a parabola to find the minimum of the broad absorption feature. The corresponding wavelength is then used to determine velocities using the relativistic Doppler shift. The measured velocities are listed in Table 4.

Table 4.  Spectral Line Velocities of iPTF 16asu

Observation Date	Phasea	Fe ii Velocity	Fe ii Broadening	Si ii Velocity
	(rest-frame days)	(1000 km s−1)	(1000 km s−1)	(1000 km s−1)
2016 May 24.97	+8.25	${28.3}_{-1.3}^{+1.1}$	${5.5}_{-1.2}^{+1.0}$
2016 May 27.36	+10.27	${29.5}_{-1.4}^{+1.0}$	${5.9}_{-1.3}^{+1.0}$	23.3
2016 Jun 04.39	+17.03	${25.7}_{-0.3}^{+0.3}$	${5.1}_{-0.3}^{+0.3}$	19.8
2016 Jun 07.36	+19.53	${21.6}_{-0.4}^{+0.4}$	${4.4}_{-0.5}^{+0.4}$	19.2
2016 Jun 10.42	+22.11	${22.0}_{-1.3}^{+1.0}$	${4.3}_{-1.3}^{+1.3}$	16.8

Note.

^aPhase is in rest-frame days relative to the bolometric maximum light (MJD 57523.25).

Download table as:  ASCIITypeset image

In the case of the Fe ii $\lambda 5169$ Å line, similar to other SNe Ic-BL, this line is blended with the neighboring Fe ii λ4924 and Fe ii λ5018 lines. Thus we cannot simply fit the minimum of this feature to derive velocities. Instead we use the convolution method developed by Modjaz et al. (2016) and Liu et al. (2016) to extract velocities from the Fe ii $\lambda 5169$ Å line. The measured velocities are listed in Table 4. Figure 13 shows the Fe ii velocities from iPTF 16asu compared to the sample of SNe Ic and Ic-BL from Modjaz et al. (2016), with velocities derived using the same method (and code). The Fe ii velocities we measure for iPTF 16asu are high compared to the objects in this sample, closest to the Fe ii velocities of SNe Ic-BL associated with GRBs. We note that phase in this figure is measured with respect to maximum light—if iPTF 16asu has a normal SN component hidden underneath the blue luminous peak, then the SN maximum will be later and iPTF 16asu will move left in this plot, but the basic conclusion that the velocities are comparable to SNe Ic-BL with associated GRBs will be unchanged.

Zoom In Zoom Out Reset image size

Most SNe Ic/Ic-BL are powered by the release of energetic photons from the radioactive decay of ⁵⁶Ni into ⁵⁶Co and finally ⁵⁶Fe. Since the late-time spectra of iPTF 16asu look very similar to the spectra of other SNe Ic-BL (Section 4), we first consider whether the light curve of iPTF 16asu can be explained purely by the decay of ⁵⁶Ni.

Using the equations from Section 2 of Lyman et al. (2016), we compare our pseudobolometric light curve from Section 3.4 to the theoretical model for a ⁵⁶Ni decay–powered light curve in Arnett (1982). The model takes two input parameters: diffusion time and ⁵⁶Ni mass. The ⁵⁶Ni mass predominantly affects the luminosity of the light curve, as a larger ⁵⁶Ni mass would indicate a larger total energy input and the diffusion time controls the timescale over which the energy diffuses out (i.e. width of the peak). Figure 15 shows the bolometric light curve of iPTF 16asu plotted against an Arnett (1982) model with the parameters ${M}_{\mathrm{Ni}}=0.55$ M_⊙ and ${\tau }_{\mathrm{diff}}=1.5$ days, assuming an opacity of $\kappa =0.1$ cm² g⁻¹. As seen in Figure 15 , the ⁵⁶Ni decay model does not fit both the sharp peak and the steep decay well, though we caution that the lack of NIR data could mean that our late-time bolometric light curve is systematically underestimated (Section 3.4).

Zoom In Zoom Out Reset image size

An ejecta mass of ${M}_{\mathrm{ej}}=0.06$ M_⊙ was calculated using this diffusion time along with an estimate of the kinetic energy. Since our spectra near peak are featureless, and thus we cannot measure a velocity, we used the average velocity (35,000 km s⁻¹) derived from the evolution of the blackbody radii to calculate this kinetic energy. The ejecta mass is notably about 10 times smaller than the amount of ⁵⁶Ni required to power this light curve, which is unphysical: the ⁵⁶Ni mass cannot be larger than the total ejecta mass, since it is necessarily part of the ejecta. Thus we rule out spherically symmetric radioactive ⁵⁶Ni decay as the dominant energy source for iPTF 16asu.

The Arnett model considered above assumes spherical symmetry and a central energy source, i.e., that all the nickel is in the center. Therefore we cannot rule out the possibility of ⁵⁶Ni-powered models for iPTF 16asu in a highly mixed or strongly asymmetric scenario (e.g., a jet), though we note that assymmetry is not expected to have a large effect on the observed luminosity (Barnes et al. 2017). More sophisticated modeling is outside of the scope of this paper.

Although ⁵⁶Ni decay alone cannot explain the light curve of iPTF 16asu, it may still contribute. Figure 16 shows iPTF 16asu's light curve compared to other SNe Ic-BL from the literature. The light curve of SN 1998bw in the g and r bands is a good match to that of iPTF 16asu from ∼15 to 40 days, which interestingly also corresponds to the time when their spectra are very similar (Figure 12), suggesting that the light curve of iPTF 16asu could plausibly be dominated by a normal SN component at these times. Their late-time slopes deviate, mainly constrained by our last r band point at 60 days (Figure 16)—however, the decay rates of stripped-envelope SNe are heterogeneous and could be explained by differences in opacity and/or asymmetry, affecting the degree of gamma-ray trapping (Wheeler et al. 2015; Dessart et al. 2017). Fainter SNe Ic-BL such as SN 2006aj and SN 2002ap are below the light curve of iPTF 16asu at all times (Figure 16). Since the light curve shows only a single, smooth peak, any ⁵⁶Ni decay contribution to the total luminosity must be subdominant to whatever is powering the main peak.

Zoom In Zoom Out Reset image size

Finally we note that other radioactive species, such as ⁴⁸Cr and ⁵²Fe, have been proposed to power a class of fast-and-faint thermonuclear transients from He-shell detonations, so-called.Ia SNe (Bildsten et al. 2007; Shen et al. 2010). However, given that iPTF 16asu is 3–5 magnitudes brighter than these models predict, the spectrum at peak is blue and featureless without the expected strong Ti ii features, and the late-time spectrum is an excellent match to SNe Ic-BL, suggesting a core-collapse explosion, we do not consider these models relevant for iPTF 16asu.

During the core collapse of a massive star, a highly magnetized ($B\approx {10}^{14}\mbox{--}{10}^{15}$ G) rapidly spinning neutron star called a magnetar can be formed. As the newborn magnetar spins down, rotational energy is released and can significantly boost the luminosity of the SN if the spin-down time of the magnetar is comparable to the diffusion time through the ejecta (e.g., Kasen & Bildsten 2010; Woosley 2010; Metzger et al. 2015). Magnetar models have been suggested to explain highly luminous transients, including many SLSNe as well as SN 2011kl (Greiner et al. 2015; Bersten et al. 2016). iPTF 16asu has a luminosity similar to that of to SN 2011kl and some relatively low-luminosity SLSNe (Figures 1 and 5), so we examine whether a magnetar model can explain the peculiar light curve of iPTF 16asu.

As described in Kasen & Bildsten (2010), the hydrodynamic simulations for their magnetar model makes the simplifying assumption that all of the injected energy is thermalized spherically at the base of the ejecta (ignoring the possibility of anisotropic jet-like injection). They further assume homologous expansion, a shallow power-law structure for interior density, and the dominance of radiation pressure. An expanding bubble with a thin shell of swept-up ejecta and a low-density interior are formed due to central overpressure in the SN remnant but rarely affect the outer layers of the SN ejecta. At late times the energy injected by the magnetar continues to heat the ejecta, as in ⁵⁶Ni decay, but it is no longer dynamically important. This process significantly affects the SN light curve.

The shape of the light curve in magnetar models depends on three parameters: P, the initial spin period; B, the strength of the magnetic field; and ${\tau }_{\mathrm{diff}}$, the diffusion timescale that is proportional to ${M}_{\mathrm{ej}}^{1/2}$. Using the magnetar model fitting code from Kangas et al. (2017) we recover the parameters $B=(3.25\pm 0.44)\times {10}^{14}$ Gauss, $P=10.40\pm 0.62$ ms, and ${\tau }_{\mathrm{diff}}=1.59\pm 0.06$ days. Manually tweaking the parameters slightly to obtain a better visual fit, we show the resulting fit to the bolometric light curve in Figure 17 with the parameters $P=9.95$ ms, $B=3.15\times {10}^{14}$ G, and ${\tau }_{\mathrm{diff}}=1.8$ days, and assuming an opacity of $\kappa =0.1$ cm² g⁻¹. As done in the ⁵⁶Ni model, the diffusion time and average velocity from the blackbody fits are used to calculate an ejecta mass of ${M}_{\mathrm{ej}}=0.09$ M_⊙. The parameters allow for the energy and the timescale to essentially be tuned separately, making the magnetar model quite flexible and generating a tight fit to both the peak and the decay of the bolometric light curve.

Zoom In Zoom Out Reset image size

Although the magnetar model produces a light curve that fits iPTF 16asu, the derived ejecta mass of our best fit is very low. Arcavi et al. (2016) derived similarly small ejecta masses for their rapidly rising SNe events, which caused them to conclude that the magnetar model was unlikely, while Greiner et al. (2015) concluded that a magnetar was a likely explanation for SN 2011kl despite their low derived ejecta mass. For an SN Ic-BL caused by the core collapse of a massive star, a magnetar model with such a low ejecta mass would require an extreme stripping scenario to reduce the core mass. Furthermore, the Kasen & Bildsten (2010) magnetar model was tuned to an ejecta mass of ${M}_{\mathrm{ej}}=5$ M_⊙ and it is not clear that the assumptions of this model would remain valid in this low-mass regime.

Another way for a magnetar model to produce a fast timescale peak, similar to that of iPTF 16asu, is to use a small period and a high magnetic field, thereby decreasing the spin-down time. When constraining the ejecta mass to be ${M}_{\mathrm{ej}}=1.0$ M_⊙, we find the best-fit parameters $P=6.0$ ms, $B=4.4\times {10}^{15}$ Gauss, and ${\tau }_{\mathrm{diff}}=8.7$ days. This fit is shown as a red line in Figure 17 and can also reproduce the fast rise and luminous peak of iPTF 16asu. However, this model declines too quickly to explain the entire light curve, so one would need a two-component model (e.g., with the late-time powered by ⁵⁶Ni, as was considered by Bersten et al. 2016 for SN 2011kl). Thus despite the compelling light curve fit, we conclude that a magnetar model is unlikely to be the sole power source of iPTF 16asu but remains a candidate for powering the peak emission if the late time light curve is powered by ⁵⁶Ni.

Long GRBs are often associated with SNe Ic-BL, though not every SN Ic-BL has an accompanying GRB (see, e.g., Woosley & Bloom 2006 for a review of the GRB-SN connection). GRBs are extremely energetic, relativistic, and highly beamed explosions characterized by an initial flash of gamma-rays followed by an "afterglow" of radiation typically seen at wavelengths ranging from the X-ray to the radio. iPTF 16asu's spectra and velocities are similar to those of SNe Ic-BL associated with GRBs (Section 4, Figures 11 and 12), so we examine whether the excess blue emission at peak could be explained as a GRB afterglow.

Nondetections of iPTF 16asu in the X-ray and radio strongly constrain the allowable GRB parameter space. The upper limits from three epochs of Swift data are shown in the left panel of Figure 18. While data at earlier times would have been more constraining, the upper limits rule out the bulk of observed X-ray afterglows with ${E}_{\mathrm{iso}}\gt {10}^{52}$ erg; however, weak or off-axis GRBs are not excluded by the X-ray data alone. Similarly, the right panel shows the upper limits from our two epochs of VLA data. As evident from this figure, we can exclude a radio counterpart to iPTF 16asu as luminous as SN 1998bw or SN 2009bb, but we cannot exclude a lower-luminosity and/or faster-evolving radio counterpart such as SN 2006aj and SN 2010bh. If iPTF 16asu is associated with a GRB, then these limits suggest that it must be a faint (${E}_{\mathrm{iso}}\lt {10}^{50}$ erg) event. These constraints are consistent with the analysis from all-sky gamma-ray monitors (Section 2.6), as an on-axis burst at the distance of iPTF 16asu with (${E}_{\mathrm{iso}}\gt {10}^{50}$ erg) would have been seen by KW or SPI-ACS.

Zoom In Zoom Out Reset image size

The most unusual characteristic of iPTF 16asu is its abrupt four-day rise time in the optical. Such a rise time is extremely short in an SN context but would be unprecedentedly long in an SN-GRB context, even though optical afterglow light curves do sometimes show a rise (e.g., GRB 970508; Galama et al. 1998a). To explain the shape of the optical light curve as a GRB afterglow, we therefore consider off-axis GRB models.

From the New York University Afterglow Library data set of off-axis long GRBs at an observed wavelength of 3000 Å (10¹⁵ Hz), the models can reproduce a three- to six-day rise for an observer angle of between 23° and 17°, respectively (van Eerten et al. 2010). The data set assumes a jet energy of 2 × 10⁵¹ erg, a jet half opening angle of 115, and a homogeneous circumburst number density of 1 cm⁻³. The parameters for an observer angle of 17° produce a light curve with roughly the same peak magnitude as iPTF 16asu; however, in changing to an observed wavelength of 30 mm (10 GHz), these parameters produce a radio light curve that is orders of magnitude brighter than our radio limits. Similarly, considering the low-energy models from van Eerten & MacFadyen (2011), we find that parameters that satisfy the radio limits are inconsistently faint in the optical. The coarse grid of parameters used in van Eerten et al. (2010) does not allow us to make precise comparisons to their model but indicates that while a four-day optical rise could be constructed, our optical light curve and radio upper limits cannot be simultaneously satisfied by current models. A more thorough exploration of energy and density parameter space than is available in these model grids is necessary to determine whether GRB models can account for both the bright optical emission and the lack of X-ray and radio emission.

A similar conclusion can be reached by comparing the observed spectral properties of iPTF 16asu to typical GRB afterglows, which are well described by synchrotron radiation resulting in both a light curve and a spectrum consisting of several power-law segments with associated indices (e.g., Sari et al. 1998). If the featureless blue spectra of iPTF 16asu are due to a GRB afterglow then we expect the spectrum to follow a power law (${F}_{\nu }\propto {\nu }^{-\beta }$), with typical values of the power-law index β around 0.5–0.6 (e.g., Kann et al. 2010). Fitting our first spectrum (at +3 days after explosion) with a power law, we find a best-fit index of $\beta =-0.5$, i.e., ${F}_{\nu }\propto {\nu }^{+0.5}$, which is inconsistent with a GRB-like spectrum. In contrast, the spectrum is well fit by a blackbody (Figure 6). Similarly, if we compare our earliest X-ray upper limit to the corresponding point on the r band light curve, then we derive a limit on the optical- to-X-ray spectral index ${\beta }_{\mathrm{OX}}\gt 1.24$ whereas typical GRB afterglows show ${\beta }_{\mathrm{OX}}\sim 0.5\mbox{--}1.0$ (Gehrels et al. 2008). We also note that the decline of the light curve is better fit by exponential decay than by a power law (Section 3.4, Figure 8 (right)).

While the properties of the luminous blue peak of iPTF 16asu do not seem to resemble a classical GRB afterglow (on- or off-axis), it is worth noting that low-luminosity GRBs like 060218 and 100316D have shown thermal emission in addition to the weak synchrotron component (e.g., Campana et al. 2006; Starling et al. 2011). Thus it is still possible that iPTF 16asu could be a related phenomenon but with a significantly brighter thermal component. The origin of the thermal emission in low-luminosity GRBs is debated, though one possibility is that it is associated with shock breakout. We consider next whether such a model can also explain iPTF 16asu.

The short timescales and blue colors of iPTF 16asu are reminiscent of shock cooling transients, where the early light curve of an SN is powered by the cooling of the envelope following the breakout of the SN shock, usually followed by a second peak from the SN itself (e.g., SN 1993J; Wheeler et al. 1993). Such a shock cooling phase should be present in all SNe (Nakar & Sari 2010), but both the duration and the luminosity will depend on the structure of the progenitor star. A peak in both the red and blue bands, as we see in iPTF 16asu, is generally associated with shock breakout from extended material (Nakar & Piro 2014). Shock cooling models have been considered for other rapidly evolving transients (e.g., Ofek et al. 2010; Drout et al. 2014) as well as low-luminosity GRBs (e.g., Nakar 2015), so we consider here whether iPTF 16asu could be explained by a shock cooling scenario.

Since the peak is seen in all bands, we consider the extended envelope model of Nakar & Piro (2014). Here the mass in the extended envelope scales as ${M}_{e}\propto {\kappa }^{-1}{{vt}}_{\mathrm{peak}}^{2}$ and the effective radius of the material scales as ${R}_{e}\propto \kappa {L}_{\mathrm{peak}}{v}^{-2}$. For the rise time and peak luminosity measured for iPTF 16asu, this suggests an envelope mass of around ∼0.5 M_⊙ and a lower limit on the effective radius of the material of $\sim 2\times {10}^{12}$ cm, still assuming $\kappa =0.1$ cm² g⁻¹.

Piro (2015) developed this extended envelope further, and Figure 19 shows a fit of the model with the parameters ${M}_{\mathrm{ej}}=0.45$ M_⊙, ${R}_{{\rm{e}}}=1.7\times {10}^{12}$ cm, and $E=3.8\times {10}^{51}$ erg. In general there is a degeneracy between the initial radius of the material and the energy deposited by the shock, but our high observed velocities suggest that we are in the regime of a smaller radius and higher energy (Piro 2015). Since the energy deposited into the extended material is just a fraction of the total SN energy, if this model is correct then it would imply a very high explosion energy, likely requiring a central engine.

Zoom In Zoom Out Reset image size

Unlike many shock-breakout SNe, iPTF 16asu exhibits only a single peak, so if the main light curve peak is powered by shock cooling then it must completely dominate the contribution from the underlying normal SN light curve. The model shown in Figure 19 approximates the shock cooling light curve with a Gaussian and is not expected to capture the decline of the light curve, which would depend on the density structure of the material. It does, however, demonstrate that an extended envelope model can produce a peak with a rise time and luminosity compatible with iPTF 16asu. As seen in Figure 16, an SN Ic-BL slightly less luminous than SN 1998bw could be hidden underneath a luminous shock-breakout peak forming one continuous peak by the merging of the second SN peak with the decay of the first peak.

In Nakar (2015), SN 2006aj/GRB 060218 was modeled by a shock breakout from energy deposited into an extended ($\gt 100\,{R}_{\odot }$) low-mass ($\sim 0.01$ M_⊙) envelope by a low-luminosity GRB. Thus iPTF 16asu could have a similar explosion mechanism but with a significantly higher-mass envelope, producing a longer-duration and luminous peak. The presence of circumstellar material would also be consistent with the constraints on high-energy emission; indeed, it has been suggested that low-energy soft GRBs like 060218 and 100316D have extended circumstellar material (Margutti et al. 2015).