The Exotic Type Ic Broad-lined Supernova SN 2018gep: Blurring the Line between Supernovae and Fast Optical Transients

SN 2018gep/ZTF18abukavn (Figure 1, top) was first discovered on UT 03:55:17 2018 September 9 ¹⁹ (JD = 2458370.6634) by Ho et al. (2018) as part of the partnership ZTF survey (Bellm et al. 2019) at (R.A., decl.) = (16:43:48.22, +41:02:43.37). Initially the spectrum of SN 2018gep was characterized by a blue, mostly featureless continuum that evolved (see Section 4 and Ho et al. 2019a) until approximately 10 days later on 2018 September 19. Burke et al. (2018), as part of the Global Supernova Project (GSP), obtained an optical spectrum (see Section 2.3) and classified the object as a broad-line Type Ic supernova (Ic-bl) with an ejecta velocity of ∼24,000 km s⁻¹ and a redshift of 0.032 which is consistent with the probable host galaxy identified by Ho et al. (2018), SDSS J164348.22+410243.3 with a z = 0.033 with an SN–host separation of ∼15.

Zoom In Zoom Out Reset image size

The ZTF public survey observed SN 2018gep between 2018 September 8 and 28 in the r-ZTF and g-ZTF filters. ZTF data were obtained from public alerts made available by the Las Cumbres Observatory MARS broker, which provides access to the publicly available background subtracted ZTF data products.

The GSP obtained additional Las Cumbres Observatory (LCO) BVgri-band follow-up data with the Sinistro and Spectral cameras on 1 m and 2 m telescopes, respectively. Using lcogtsnpipe (Valenti et al. 2016), a PyRAF-based photometric reduction pipeline, point-spread function (PSF) fitting was performed. Reference images were obtained with the Sinistro and Spectral Imager after the SN faded and image subtraction was performed using PyZOGY (Guevel & Hosseinzadeh 2017), an implementation in Python of the subtraction algorithm described in Zackay et al. (2016). BV-band data were calibrated to Vega magnitudes using the AAVSO Photometric All-Sky Survey (Henden et al. 2009), while gri-band data were calibrated to AB magnitudes using the Sloan Digital Sky Survey (SDSS, Aguado et al. 2019). Science observations were taken between 2018 September 22 and 2018 October 30 with template photometry taken between 2019 January 22 and 26.

Additional photometric observations were collected with the 0.6/0.9 m Schmidt telescope at Piszkesteto Mountain Station of Konkoly Observatory, Hungary, using the 4k × 4k FLI CCD equipped with Johnson–Cousins–Bessel BVRI filters. After the usual bias, dark, and flatfield corrections, PSF photometry was performed on the SN and a set of nearby stars used as tertiary standards. Photometric calibration was done using PS1 photometry on the local tertiary standards, after transforming the cataloged g_P , r_P , i_P magnitudes to BVRI ones via the calibration by Tonry et al. (2012). Finally, the flux contribution from the host galaxy was taken into account by computing aperture photometry on the host as it appeared on the PS1 frames and subtracting its fluxes from those obtained from PSF photometry on the Konkoly frames.

Observations with the Neil Gehrels Swift Observatory (Swift hereafter; Gehrels et al. 2004) Ultra-Violet/Optical Telescope (Roming et al. 2005) began on 14:02:56 2018 September 9 (∼0.5 day after discovery) using three optical (u, b, v) and three UV filters (uvw2, uvm2, uvw1: λ_c = 1928, 2246, 2600 Å, respectively; Poole et al. 2008) after being triggered by Schulze et al. (2018) and Ho et al. (2019a). Regular observations continued through 2018 October 3 with a final observation obtained on 2018 October 29. Data were reduced using the process described in Pritchard et al. (2014) with the final observation used for galaxy template subtraction. While there may be some small contamination from the SN at this time, any UV emission is far below the Swift sensitivity at this time frame and the optical observations from Swift are consistent with the other sources presented here (LCO, Konkoly). Data from these sources are presented in Figure 2 and made available in Table 1.

Zoom In Zoom Out Reset image size

We obtained optical spectroscopy of the SN as well as its host galaxy and list the journal of our spectroscopic observations in Table 2.

Table 2. Spectroscopic Observations of SN 2018gep and Its Host Galaxy

UT Date	JD	${t}_{V,\max }$ a	Tel. + Instr.	Wave. Range	P.A.	Airmass	Slit	Exp.
		(days)		(Å)	(°)		('')	(s)
2018-09-11.31	2458372.81	−3.7	OGG 2m+FLOYDS	3700–10000	95.3	1.74	2.0	1800
2018-09-19.24	2458380.74	+4.3	OGG 2m+FLOYDS	3700–10000	112.6	1.30	2.0	1800
2018-11-07.45 b	2458429.95	+54.5	CAHA+PMAS c	3750–7500	N/A	2.2	N/A	3 × 1200
2019-02-05.64 b	2458520.14	+144.6	Keck+LRIS	3200–9200	250	1.33	1.0	900

Notes.

^a Days with respect to V-band maximum. Note that SN 2018gep's rise time in the V-band is t_rise,V = 5.6 ± 0.5 days (see Section 3), such that our first two spectra were taken 1.9 and 9.9 days after the presumed date of explosion, respectively. ^b No SN light, only host galaxy. ^c IFU observations, thus long-slit information such as slit size and P.A. is not applicable here.

Download table as:  ASCIITypeset image

Additional observations of the location of SN 2018gep and its host galaxy were obtained by two different telescopes at 1.5–2 months after the explosion. One was obtained via Director's Discretionary Time (PI: Bensch) using the Potsdam MultiAperture Spectrophotometer (PMAS; Roth et al. 2005), which is an IFU instrument, mounted on the 3.5 m telescope at the Centro Astronómico Hispano en Andalucía (CAHA). The other was with the Low-Resolution Imaging Spectrometer (LRIS, Oke et al. 1995; McCarthy et al. 1998; Rockosi et al. 2010) at the 10 m W. M. Keck Observatory on Maunakea, Hawaii, as part of the LCO–GSP follow-up program (PI: Valenti), using a long-slit aperture.

The IFU observations using the PMAS instrument in PPak mode (Verheijen et al. 2004; Kelz et al. 2006) were carried out on 2018 November 7. We used the V500 grating with G_rot = 143.5, which covers a wavelength range between ∼3750 and 7500 Å at a resolution of 6.5 Å FWHM, corresponding to ∼350 km s⁻¹. The PPak IFU consists of 331 science fibers with diameters of 27. The science fibers are placed in a hexagonal parcel resulting in a filling factor of 65%, and cover a field-of-view (FOV) of 72'' × 64''. For sky subtraction 36 sky fibers are placed around the science fibers. An additional 15 fibers illuminated by internal lamps are used to calibrate the instrument. Three science exposures of 1200 s each were obtained at a signal-to-noise ratio (S/N) of ∼10 per angstrom for the spectral continuum. We used a dithering pattern consisting of three pointings to cover the entire FOV including the spaces in-between the fibers. In Figure 3 (bottom) we show the FOV of the PPak IFU and the region around the host which is plotted in the subsequent figures that display host-galaxy properties.

Zoom In Zoom Out Reset image size

To reduce the PMAS–PPak data we used a python-based pipeline that executes the following steps: identification of the position of the spectra on the detector along the dispersion axis; extraction of each individual spectrum; distortion correction of the extracted spectra; wavelength calibration; fiber-to-fiber transmission correction; flux-calibration; sky-subtraction; cube reconstruction; finally, differential atmospheric correction (for more details, see García-Benito et al. 2010, 2015; Husemann et al. 2013). These IFU data and their analysis are discussed as part of our host-galaxy study in Section 5.

The late-time long-slit Keck spectrum (see Table 2) was reduced in the standard way using the LPIPE pipeline (Perley 2019)—no SN emission was detected at the location of SN 2018gep in either the 1D or the 2D spectra. The spectrum is included as part of our host-galaxy study in Section 5.

The combined UV–optical light curves for SN 2018gep are shown in Figure 2. From the Swift v-band data we calculate that the epoch of maximum light in the v-band is t_peak,V = 58375.7 ± 0.8 MJD using the Monte Carlo method outlined in Bianco et al. (2014). Using the same method we calculate the observed peak magnitude in the r-band, m_r,peak = 16.29 ± 0.03 mag. The relatively deep upper limits from the ZTF survey provide for a strong constraint on the rise time—using the last ZTF upper limit (Ho et al. 2019a) as an upper limit on the explosion date we calculate a rise time of t_rise,v = 5.6 ± 0.5 days and t_rise,r = 5.1 ± 0.5 days. From the observed redshift of z = 0.031875 ± 0.000075 (see Section 5) we calculate the absolute magnitude for SN 2018gep at peak to be M_V = −19.53 ± 0.23 mag and M_r = −19.54 ± 0.24 mag using the astropy (Astropy Collaboration et al. 2013; Price-Whelan et al. 2018) cosmology package and a flat ΛCDM model with H₀ = 74.22 km s⁻¹ Mpc⁻¹ (Riess et al. 2019) and Ω_m = 0.286, and corrected for Milky Way line-of-sight extinction using the values from Green et al. (2019). This cosmology model is used throughout the rest of this work for consistency.

Assuming an SN is powered by the typical ⁵⁶Ni-decay model, for a particular absolute magnitude and SN rise time, we may calculate an ejecta mass and nickel fraction as outlined in Arcavi et al. (2016) (Equations (1) and (2) and following from Arnett 1982; Stritzinger & Leibundgut 2005; Wheeler et al. 2015). It is important to note that this approach makes a number of simplifying assumptions including: spherical symmetry, a constant opacity, a central nickel concentration, that the photospheric velocity is characteristic of the ejecta velocity, and optical to bolometric corrections for both the rise-time and peak absolute magnitude. This relation is therefore more indicative than strict, and in Figure 4 we sketch out lines for a series of ejecta masses and two additional lines corresponding to objects in which the ejecta mass must be entirely composed of nickel to power their light curve. If an object lies above these lines, an additional source of energy injection or a different source of power is required. Since this limit also depends on ejecta velocity, we draw two lines: the lower line corresponding to a typical SN with ∼10,000 km s⁻¹ expansion and the top line corresponding to an SN with ∼24,000 km s⁻¹ expansion velocity as measured from the post-maximum spectrum of SN 2018gep. In this parameter space, SN2018gep is like the other luminous fast-rising transients shown, namely right on the border of what can be easily described with simple nickel-powered relations, and it is consistent with being an outlier from the other SNe Ic-bl which are comfortably below this relation. This implies that SN 2018gep most likely had to have an additional powering source besides the decay of ⁵⁶Ni (see also Ho et al. 2019a for a detailed model involving CSM interaction and pre-explosion mass loss).

Zoom In Zoom Out Reset image size

In Figure 5 we compare the light curve of SN 2018gep to a sample of SNe Ic-bl from Taddia et al. (2019) in the optical and a few select SESNe with Swift UV observations. The optical evolution of SN 2018gep is broadly similar to the most rapidly evolving SNe Ic-bl. The most similar objects are iPTF16asu (an outlier as noted in Taddia et al. 2019), PTF10vgv (Corsi et al. 2012), and SN 2006aj. While the optical evolution of SN 2018gep and iPTF16asu is similar to that of the SN Ic-bl population as a whole, the early color evolution is not, particularly in the UV and bluer filters. As we show in the middle and bottom panels of Figure 5, there is significantly more blue emission from these SNe at early times than from the rest of the Ic-bl sample, by more than a magnitude in the optical and almost 4 mag in the UV. By ∼10 days after r-band maximum the color curve of SN 2018gep becomes similar to that of the sample as a whole, albeit remaining somewhat on the blue side. Only SN 2018gep and iPTF16asu show this significant early blue excess. PTF10vgv and SN 2006aj have similar colors to the SNe Ic-bl sample as a whole. The only other similarly blue emission is that from SN 2006aj at early times, but it has a significantly faster evolution. The mechanism that drives this is still a topic of some debate (see Irwin & Chevalier 2016 for a discussion), and the SN component of the GRB/SN SN2006aj quickly returns to the "typical" behavior of the SNe Ic-bl sample as a whole—including during phases where SN 2018gep is still UV bright.

Zoom In Zoom Out Reset image size

In Sections 3.1 and 3.2 we show that while SN 2018gep shares similarities with other Ic-bl SNe, it is a notable outlier in terms of color, absolute magnitude, and rise time. In Section 3.1 we show that other objects that may behave similarly to SN 2018gep are the recently discovered class of Fast Evolving Transients first noted by Drout et al. (2014) and later in Arcavi et al. (2016), Rest et al. (2018), and Pursiainen et al. (2018). With many of these objects having poorly constrained rise times due to their rapid evolution, we focus on a comparison with the PS1 sample from Drout et al. (2014) which has a significant number of objects with a detected rise as well as multi-color observations. These are, however, found at a significantly large range of redshifts (z = 0.07–0.65); to compare we match the observed SN 2018gep band with the closest rest-frame band of a PS1 object, as shown in Figure 6. This is a rather coarse measurement, as the relative filter bandpass is different and a more detailed analysis would perform k-corrections to address this. However, we choose to avoid k-corrections as they are spectral energy distribution (SED) dependent and we have limited information about the SEDs of all PS1 objects, while we know that they undergo significant color evolution.

Zoom In Zoom Out Reset image size

As seen in Figure 6, both SN 2018gep and iPTF16asu have similar relative light curve shapes as the PS1 Fast Evolving Transient population as a whole. Furthermore, the observed g–r colors are similar to the sample as reported in Drout et al. (2014). There is some suggestion that there may be some longer-lived emission in some PS1 fast transients (as seen in the late-time rest-frame i-band comparison and u-band comparison), although these late-time deviations each come from a single PS1 object and it is not clear how homogeneous of a sample these objects is.

While the high luminosity of SN 2018gep (M_V = −19.53 ± 0.23 mag and M_r = −19.54 ± 0.24 mag) may lie close to that of superluminous SNe (SLSNe), which appear to range from −22 ≲ M_g ≲ −20 mag (De Cia et al. 2018; Lunnan et al. 2018; Kann et al. 2019), its light curve is distinctly different: The "fast" SLSNe-Ic have typical rise times of ∼28 days (Inserra 2019), namely ∼4–5 times larger than for SN 2018gep. Furthermore, while they may rise as quickly as in ∼15 days (Inserra 2019), SLSNe (both fast and slow) tend to decline slower by a factor of 2 compared to their rise time (Nicholl et al. 2015; De Cia et al. 2018), which is a trait not seen in SN 2018gep specifically, nor FBOTS in general (Drout et al. 2014).

This study represents the first IFU host-galaxy study of a Fast Evolving Transient. The PPak IFU spaxels in our final cube have an angular size of 1'' × 1''; however, the seeing during observations was only 18, hence the nominal spatial resolution is lower. For our spatially resolved analysis of the host galaxy we use custom-written IDL codes to extract emission-line maps and properties from the data cubes.

5.1.1. Emission-line Analysis

In order to obtain emission-line fluxes in each spaxel we sum the fluxes in the spectral direction around the redshifted position of each emission line and subtract the galaxy continuum. 2D maps of the main emission lines are shown in the Appendix, Figure 13. To study the properties of the region around the SN at different spatial resolutions, we extract 1D spectra from 1, 5, and 7 spaxels centered on the SN position using QFitsView. ²¹ The spectra are shown in Figure 8

Zoom In Zoom Out Reset image size

Using the integrated spectrum of the host galaxy we determine a precise redshift from the strong emission lines of Hβ λ4861 Å, [O iii] λ4959 Å, [O iii] λ5007 Å, [O i] λ6300 Å, Hα λ6563 Å. The mean value obtained from all emission lines yields z = 0.031875 ± 0.000075.

To obtain the interstellar extinction in the host galaxy, we use the Balmer decrement of Hα/Hβ according to Domínguez et al. (2013), adopting the Calzetti et al. (2000) attenuation curve with R_V = 4.05 which assumes a starburst attenuation law. We assume the standard recombination model for star-forming galaxies and Case B for H I recombination lines (Osterbrock & Ferland 2006). The intrinsic Balmer decrement at an electron temperature T = 10⁴ K and density n_e = 10² cm⁻³, is expected to be j_Hα /j_Hβ = 2.86. The values obtained for the reddening in the galaxy and the regions around the SN are listed in Table 4. The distribution of the extinction across the galaxy is shown in Figure 9. Curiously, the spectra of the SN line-of-sight region indicate some extinction while the extinction based on the integrated spectrum of the host is consistent with zero. As we increase the aperture of extraction from the SN position (see Table 4) we see the calculated extinction drop until for the total host-galaxy-integrated IFU spectrum the overall extinction is low and consistent with E(B − V) = 0 mag. This is also consistent with the zero to low value obtained from the Keck LRIS host-galaxy spectrum. One explanation for this apparent discrepancy may be that the overall galaxy emission has little extinction and, while extinction is present throughout the galaxy it is not homogeneously distributed. Therefore the integrated spectrum is dominated by regions with little extinction but more emission, explaining the overall low E(B − V). This is reflected in the IFU map in Figure 9 which shows that the distribution of the extinction is not uniform. We observe that the high extinction in the SN line-of-sight region could either imply a considerable amount of dust at the SN site or it is dust behind the SN. The latter is favored by the fact that the SN is observed to be UV bright, ∼4 mag bluer in the UV than other SNe Ic-bl at early times. If we de-reddened the SN data using extinction values based on the 1'' × 1'' section around the SN in the IFU map (E(B − V) = 0.5–0.6 mag, A_V = 1.8 mag) the intrinsic peak luminosity in the UV would be unreasonably large. We therefore conclude that if there is dust, it is located behind the SN, hence not showing up in the NaD absorption in the sightline but can affect the Balmer decrement measured from the projection of the excited gas at and around the SN site.

Zoom In Zoom Out Reset image size

Table 4. Emission-line Fluxes Corrected for Galactic and Host-galaxy Extinction, and Calibrated with SDSS Photometry

Emission Line	λ (Å)	Host Galaxy	λ (Å)	SN Region	λ (Å)	SN Region	λ (Å)	SN Region
				1'' × 1''		3'' × 3''		3'' × 3''
						(five spaxels)		(nine spaxels)
[O ii] λ3727 Å	3727.929	57.705 ± 1.249	3725.427	7.208 ± 2.241	3725.927	23.161 ± 4.500	3726.198	19.215 ± 3.995
Hβ λ4861 Å	4861.477	36.445 ± 0.547	4861.556	3.649 ± 0.640	4861.404	14.952 ± 1.952	4861.405	16.402 ± 2.172
[O iii] λ4959 Å	4959.094	54.767 ± 0.612	4959.239	6.105 ± 1.030	4959.112	25.487 ± 3.214	4959.088	26.438 ± 3.385
[O iii] λ5007 Å	5007.047	155.614 ± 0.894	5007.136	16.767 ± 2.773	5007.057	73.183 ± 9.054	5007.027	76.130 ± 9.565
[O i] λ6300 Å	6297.685	8.126 ± 0.704	⋯	⋯	⋯	⋯	⋯	⋯
Hα λ6563 Å	6562.696	84.348 ± 1.171	6562.640	10.326 ± 1.274	6562.660	42.315 ± 3.954	6562.650	46.437 ± 4.413
[N ii] λ6584 Å	6584.315	3.892 ± 0.584	6584.007	0.556 ± 0.096	6583.500	1.760 ± 0.309	6583.506	2.070 ± 0.445
[S ii] λ6717 Å	6716.741	7.075 ± 0.552	6714.937	1.790 ± 0.252	6714.704	4.303 ± 0.554	6714.525	5.982 ± 0.766
[S ii] λ6731 Å	6732.335	5.710 ± 0.573	6732.570	1.107 ± 0.161	6732.847	4.210 ± 0.505	6732.899	5.581 ± 0.676
E(B − V) [mag]:		0.000		0.493 ± 0.040		0.403 ± 0.030		0.246 ± 0.030
SFR [M⊙ yr−1]:		0.139		0.017		0.070		0.076

Note. All fluxes are in 10⁻¹⁶ erg s⁻¹ cm⁻².

Download table as:  ASCIITypeset image

The emission-line fluxes of the spectra were measured using SPLOT in IRAF. Statistical errors were calculated following Pérez-Montero & Díaz (2003). We found an offset between the SDSS photometry and the magnitude derived from the integrated spectra of m – m₀ = 0.26 mag and therefore calibrated the emission-line fluxes using the SDSS g' and r' filters. Fluxes were corrected for Galactic extinction (A_v = 0.0286 mag), and extinction in the host galaxy as estimated in each corresponding spectrum. We list the final extinction-corrected, SDSS-calibrated emission-line values as extracted for different parts of the galaxy in Table 4 and the Appendix.

5.1.2. Derived Host Properties

We also obtained one long-slit spectrum of the host using LRIS/Keck. The LRIS spectrum is a light-weighted average of a 1'' × 4'' size region centered on the "nucleus" of the galaxy (i.e., the one with the strongest trace/continuum). Fluxes were measured using SPLOT in IRAF and errors calculated in the same way as for the integrated regions from the PMAS data. The fluxes are presented in Table 9 in the Appendix. We corrected all fluxes for Galactic extinction (A_V = 0.0286 mag). We determined the intrinsic extinction using the Balmer decrement as described above and found no extinction based on this spectrum. This result is consistent with the value of the extinction based on the IFU integrated galaxy spectrum, but is not consistent with the extinction deduced from IFU data at the SN position, which indicates a large Balmer decrement in that region. We only see high extinction at the SN region as we explain in Section 5.1.1. where we speculated that it may be due to dust that is accumulated in a small area behind the SN. Hence, without clear emission lines, the extracted LRIS spectrum with area of 1'' × 4'' centered on the galaxy "nucleus," may miss some light from the SN region.

In the Keck spectrum we detect the same lines as in the integrated IFU spectrum, and additionally we measure the [S iii] lines at λ9069 and 9532 Å. We then also derive metallicities using the pyMCZ code as described above, and present the results in the Appendix, Table 10. The results from the Keck spectrum are consistent with the metallicities found for the same calibrators in the integrated galaxy spectrum of the PMAS data.

The host galaxy is a blue dwarf galaxy, with an observed SDSS mag of g' = 18.87 mag, and with a diameter of ∼10''. We use the Le Phare code to perform SED fitting of the host galaxy of SN 2018gep using broadband data from the SDSS. The physical parameters were calculated using Bruzual & Charlot (2003) population synthesis models as galaxy templates. We use the photometry (corrected for the Galactic extinction of A_V = 0.0286 mag) presented in Table 6.

Our best fit has a reduced χ² of ∼1 (χ² = 4.86). In Figure 10 we show the SED fit of the host galaxy, and the physical parameters derived are listed in Table 7.

Zoom In Zoom Out Reset image size

Table 7. Physical Parameters of the Host of SN 2018gep Derived Using SED Fitting to Source Photometry

Parameter [Unit]	Value
Age [Gyr]	${0.32}_{-0.05}^{+0.01}$
M [107 M⊙]	${7.75}_{-1.22}^{+2.44}$
SFR [M⊙ yr−1]	${0.048}_{-0.010}^{+0.054}$
SSFR [Gyr−1]	${0.622}_{-0.043}^{+0.244}$
LNUV [107 L⊙]	5.357
LR [107 L⊙]	5.498
LK [107 L⊙]	1.088

Download table as:  ASCIITypeset image

Using this SED fitting method we infer the SFR to be SFR = ${0.048}_{-0.010}^{+0.054}$ M_⊙ yr⁻¹, while the values of the SFR based on the emission-line analysis ranges from 0.017 to 0.139 M_⊙ yr⁻¹, for the SN region (1'' × 1'' area) and the whole galaxy, respectively. The SED reveals the total mass to be equal to M = $({7.75}_{-1.22}^{+2.44})\times {10}^{7}$ M_⊙, and implies that it is a young galaxy with an age of ${0.32}_{-0.05}^{+0.01}$ Gyr. Our values are consistent within our 1σ errors with those of Ho et al. (2019a) (who do not report uncertainties), though lower than theirs (probably because we did not include NIR data, which they did).

Most star-forming galaxies follow the fundamental mass–metallicity relationship (e.g., Tremonti et al. 2004) in which higher-mass galaxies also have high metallicity. Thus comparing the host galaxy of SN 2018gep to those of other transients and to the general population of star-forming galaxies as traced by the SDSS (Kewley & Ellison 2008) may give us clues about the stellar population that preferentially produces those explosions.

In Figure 11 we compare the host mass and metallicity in the KD02 (Kewley & Dopita 2002) scale against the values for hosts of other SNe Ic-bl, GRB–SNe and Fast Evolving Transients. The hosts of SN 2018gep and iPTF16asu are low-mass, low-metallicity dwarf galaxies that lie beneath the observed SDSS population and its standard deviation (Kewley & Ellison 2008). The host galaxies of SN 2018gep and iPTF16asu have masses and metallicities that are broadly consistent with both the SN Ic-bl sample and the GRB–SN sample (the hosts of which are also comparable to each other; Modjaz et al. 2020). The host of iPTF16asu has both a mass and metallicity close to the average of these two samples while the host of SN 2018gep is on the very low-mass end while having a metallicity similar to the average. Comparing the hosts of SN 2018gep and iPTF16asu with those of the fast-transient hosts, we show that their host properties are on the extreme end of the observed distribution of fast-transient hosts. The host galaxies of SN 2018gep and iPTF16asu have metallicities comparable to that of the lowest measured host from the PS1 Fast Evolving Transient sample and with the SN 2018gep host galaxy having a mass similar to the least massive and most metal-poor hosts from the PS1 sample simultaneously. In general the population of host galaxies of Fast Evolving Transients contains objects with masses and metallicities higher than those of SNe Ic-bl or GRB–SNe.

Zoom In Zoom Out Reset image size

The host galaxies from Pursiainen et al. (2018) are not shown in Figure 11, as these galaxies had no reported metallicities. However, recent results from Wiseman et al. (2020) using the host galaxies from Pursiainen et al. (2018) have found that the host galaxy Dark Energy Survey (DES) sample of Rapidly Evolving Transients lie in a similar space as the SNe Ic-bl and GRB–SNe samples. The metallicity metrics used by Wiseman et al. (2020) are different from those used here (PP04-O3N2 versus KD02). Interestingly, their transient sample (from Pursiainen et al. 2018) does not require a strictly blue color, some of their objects are red, and for example could include objects such as PTF10vgv (see Figure 5), which lacks the strong blue colors but does evolve quite rapidly. The significant, systematic offset between the host galaxies of the PS1 sample and the DES sample likely implies either different intrinsic objects or a bias due to detection/selection method (Wiseman et al. 2020); and the host of SN 2018gep is not a clear match to either of these samples.

As we have discussed in Sections 3 and 4 SN 2018gep, while possessing the broad lines with high absorption velocities that are the defining characteristics of an SN Ic-bl, also appears to be an outlier in the general population of SNe Ic-bl as it exhibits an anomalous early blue rise and is on the luminous end of the SN Ic-bl absolute magnitude distribution. We conclude that not only is SN 2018gep different observationally from other observed SNe Ic-bl, but that it also requires a different (or at least additional) source of energy injection which is consistent with its location in Figure 4.

We compare the observed SN 2018gep light curve with simple semi-analytic model fits using the MOSFiT package (Guillochon et al. 2018) in Figure 12. For the MOSFiT Ic SN model, we see that this standard model (Ni-powered explosive SNe; Pankey 1962; Arnett 1982; Nadyozhin 1994) has a difficult time reproducing the rapid blue rise seen in the observed data. If we add an additional source of energy injection, here magnetar spin-down (Kasen & Bildsten 2010; Woosley 2010; Nicholl et al. 2017) or CSM interaction (Chatzopoulos et al. 2013), we see that the early fit improves significantly. This is overall consistent with our previous conclusion that SN 2018gep is both different from a typical SN Ic-bl and most likely has an additional, or different, source of energy injection. We caution that models like these are generally best used for population studies of objects that are well fit by standard models, and want to emphasize that the work here is most illustrative of how this is not a "typical" SNe Ic-bl and that the inclusion of additional physics is necessary.

Zoom In Zoom Out Reset image size

The best-fit model parameters for the three discussed models can be seen in Table 8. As these are simple semi-analytic models, the physical inference possible in such a unique case is somewhat limited. Overall, the standard Ic model requires a significant overabundance of Ni but most closely matches the ejecta velocity and explosion date inferred from the obtained data. The Ni+ energy injection models tend to have a more realistic Ni fraction while undershooting the ejecta velocity and being on the edge of allowed explosion dates.

Table 8. Best-fit Model Parameters in the MOSFIT Package for Powering the UV–Optical Light Curves of SN 2018gep (Figure 12)

Parameter [Unit]	Ic	Ni+Mag	Ni+CSM
log Mej [M⊙]	$-{0.12}_{-0.04}^{+0.03}$	$-{0.58}_{-0.06}^{+0.08}$	$-{0.31}_{-0.30}^{+0.19}$
log fNi	$-{0.01}_{-0.01}^{+0.002}$	$-{0.30}_{-0.17}^{+0.17}$	$-{0.56}_{-0.25}^{+0.32}$
texp [days]	$-{5.42}_{-0.74}^{+0.64}$	$-{2.4}_{-0.75}^{+0.53}$	$-{1.9}_{-0.67}^{+0.50}$
log vej [km s−1]	${4.50}_{-0.06}^{+0.04}$	${4.45}_{-0.70}^{+0.4}$	${4.7}_{-0.27}^{+0.21}$
log κ [cm2 g−1]	$-{0.98}_{-0.02}^{+0.03}$	$-{0.15}_{-0.18}^{+0.23}$	$-{0.24}_{-0.33}^{+0.33}$
log nH,host	${17.66}_{-1.13}^{+0.93}$	${20.49}_{-1.92}^{+0.53}$	${17.94}_{-1.40}^{+1.52}$
log σ	$-{0.08}_{-0.02}^{+0.03}$	$-{0.26}_{-0.03}^{+0.03}$	$-{0.27}_{-0.02}^{+0.03}$
log ${T}_{\min }$ (K)	${3.63}_{-0.03}^{+0.03}$		${3.75}_{-0.02}^{+0.02}$
log B		${0.97}_{0.04}^{+0.02}$
MNS(M⊙)		${1.04}_{-0.03}^{+0.07}$
Pspin (ms)		${8.16}_{-3.02}^{+1.43}$
θPB (rad)		${1.34}_{-0.17}^{+0.15}$
log MCSM			$-{0.97}_{-0.02}^{+0.04}$
log ρ			$-{11.27}_{-0.06}^{+0.06}$

Note. Best-fit values and 2σ errors for model parameters. See Guillochon et al. (2018), Chatzopoulos et al. (2013), Nicholl et al. (2017), and Nadyozhin (1994) for parameter details. We caution that models like these are generally best used for population studies of objects that are well fit by standard models, and want to emphasize that the work here is most illustrative of how this is not a "typical" SNe Ic-bl and that the inclusion of additional physics is necessary.

Download table as:  ASCIITypeset image

Of the two models with some additional non-⁵⁶Ni energy injection, the magnetar model requires a large magnetic field, B ∼ 10¹⁴ G, which is comparable to that required for SLSNe by similar models (Nicholl et al. 2017).

Could SN 2018gep be a transitional object between "ordinary" SNe Ic-bl and SLSNe Ic, given its high luminosity and blue colors which are similar to those of SLSNe? There are certainly some similarities, but also some significant differences. SN 2018gep does show some similarities to those of SLSNe in the spectral features, with the detection of the "W" feature in the spectra at early times (Ho et al. 2019a). On the other hand, the spectra of SN 2018gep are much bluer than those of SLSNe Ic at both phases that we cover. Photometrically, the peak luminosity of SN 2018gep is comparable to that of some SLSNe and does lie between that of the SN Ic-bl sample and the SLSN sample. However, the light curve shape of SN 2018gep evolves much faster, with rise times significantly shorter than both SNe Ic-bl and SLSNe (i.e., SN 2018gep does not lie between the locus of the SN Ic-bl and SLSN Ic samples), and does not conform to proposed light curve scaling relations for even the fastest SLSNe Ic.

While there is significant flexibility in these models, the large required value of the magnetic field most likely disfavours this energy injection method without a compelling argument for a similar compact object arising from the stellar progenitor. This would make the ⁵⁶Ni+CSM interaction model the most favored of the three models, which is consistent with the results from Ho et al. (2019a). This picture is also qualitatively similar to the theoretical explosion models of Aguilera-Dena et al. (2018) which suggest a population SNe Ic originating from high-mass progenitors that may interact with a C/O-rich CSM, although these models would similarly predict an associated long-duration GRB which is disfavored in this case (Ho et al. 2019a).

The work done by Ho et al. (2019a) on SN 2018gep shows the detection of pre-explosion variability and inferred mass loss by the progenitor star and the subsequent interaction between the pre-explosion ejected mass and the SN shock. This is a well-substantiated and physically motivated model for SN 2018gep that is overall consistent with our more general and data-driven discovery that some additional source of energy injection needs to be present early on in the light curve.

The similarity between SN 2018gep and the PS1 and DES Fast Evolving Transients, while also noted by Ho et al. (2019a), is not studied in significant detail by them as we do here, including our light curve and environment studies and folding iPTF16asu into this as well. We find some similarity to both SLSNe (though SN 2018gep has an even bluer spectrum pre-max than SLSNe) and GRB–SNe (in the light curve and spectra) which is consistent with the Ho et al. (2019a) findings of potential SLSNe spectral features and the high velocities only seen otherwise in GRB–SNe.