SN 2017ein and the Possible First Identification of a Type Ic Supernova Progenitor

SN 2017ein was constantly monitored photometrically in the bands BVR_CI_C with both the 0.76 m Katzman Automatic Imaging Telescope (KAIT; Filippenko et al. 2001) and the 1 m Nickel telescope at Lick Observatory, as well as with unfiltered KAIT images, beginning within 3 days of discovery and interrupted only by nights with poor observing conditions. We followed the SN until it was no longer accessible from either KAIT or the Nickel telescope. We show a KAIT image of the field in Figure 1, with the SN indicated. From our KAIT data we measured an absolute position for the SN of 11^h52^m5326 $+44^\circ 07^{\prime} 26\buildrel{\prime\prime}\over{.} 2$ (J2000). All images were reduced using a custom pipeline (Ganeshalingam et al. 2010). Point-spread function (PSF) photometry was then obtained using DAOPHOT (Stetson 1987) from the IDL Astronomy User's Library.¹³ We also tried to perform the photometry after using the image template subtraction method, in order to estimate the host contribution to the light. We find that the differences between using subtraction and not using subtraction are very small ($\lt 0.1$ mag) for all epochs, implying that the host contribution is minimal. We thus adopted the results without template subtraction.

Zoom In Zoom Out Reset image size

Six stars in the KAIT and Nickel fields were chosen as local calibrators; they are indicated in Figure 1. This field was observed by Pan-STARRS and included in the release of the Mean Object Catalog.¹⁴ We obtained from that release the mean PSF photometry for the six stars in the PS1 bands and used the transformations to BVR_CI_C provided by Tonry et al. (2012). We list the resulting magnitudes for the stars in Table 1 and use these for calibration of the SN photometry.

Apparent magnitudes of the SN and the six calibrators were all measured in the KAIT4/Nickel2 natural system. The final results were transformed to the standard system, using the local calibrators and color terms for KAIT4 (Ganeshalingam et al. 2010, their Table 4) and updated Nickel color terms (Shivvers et al. 2017). We present the early-time KAIT and Nickel photometry of SN 2017ein in Table 2.

Over a 2-month period beginning on 2017 June 2, eight optical spectra of SN 2017ein were obtained with the Kast Spectrograph mounted on the 3 m Shane telescope (Miller & Stone 1993) at Lick Observatory. They were taken at or near the parallactic angle (Filippenko 1982) to minimize slit losses caused by atmospheric dispersion. Data were reduced following standard techniques for CCD processing and spectrum extraction (Silverman et al. 2012) utilizing IRAF¹⁵ routines and custom Python and IDL codes.¹⁶ Low-order polynomial fits to comparison-lamp spectra were used to calibrate the wavelength scale, and small adjustments derived from night-sky lines in the target frames were applied. Observations of appropriate spectrophotometric standard stars were used to flux-calibrate the spectra.

We obtained 3 × 1200 s exposures with the Blue Channel spectrograph on the MMT on 2017 June 24 (JD 2,457,928.69). The data were taken with the 1200 line mm⁻¹ grating with a central wavelength of 6360 Å and a 10 slit width. The seeing was 12. Standard reductions were carried out using IRAF, and wavelength solutions were determined using internal He–Ne–Ar lamps. Flux calibration was achieved using spectrophotometric standards at a similar airmass taken throughout the night. A log of all spectroscopic observations is provided in Table 3.

We note that we cross-checked the calibrations of our photometry and our spectra at four nearly contemporaneous epochs (MJD 57,906.76, 57,925.77, 57,931.76, and 57,935.74) by comparing observed colors with colors synthetically generated from the spectra using pysynphot,¹⁷ and we found that the two data sets differed by at most ∼0.1 mag, within the uncertainties of the photometry; thus, we are confident that our calibrations are sufficient.

The SN 2017ein site is in publicly available archival images obtained with the HST Wide Field Planetary Camera 2 (WFPC2) on 2006 October 20 (by program GO-10877, PI: A. Filippenko, originally to observe SN 2005ay) in bands F555W (total exposure time 460 s) and F814W (700 s). The image mosaics were obtained from the Hubble Legacy Archive, whereas the individual WFPC2 frames were obtained from the HST Archive at MAST. The SN site is just off the edge of the field of view (NIC3) of archival Near-Infrared Camera and Multi-Object Spectrograph (NICMOS) images obtained on 2003 April 3 in bands F160W, F187N, and F190N, and thus we do not consider these.

We also observed the transient itself with HST on 2017 June 12 using the Wide Field Camera 3 (WFC3) UVIS channel in subarray mode, as part of our Target of Opportunity (ToO) program (GO-14645, PI: S. Van Dyk), in F438W, with 10 s individual frame times and a total exposure time of 270 s. The F438W band was chosen in this case to attempt to avoid saturation by the SN, at which we were successful. All of these data had been initially processed via the default pipelines at STScI and were obtained from the HST Archive. The individual flc frames, which had been corrected by the pipeline for charge transfer efficiency losses, were individually combined into an image mosaic using AstroDrizzle within PyRAF.

We show the combined KAIT and Nickel light curves in Figure 2. Here we have also included the photometry at R from Im et al. (2017). We compared the light curves with those of a number of SNe Ic and found that the best match was with the light curves from Hunter et al. (2009) for the normal SN Ic 2007gr in NGC 1058. The light curves for SN 2017ein and SN 2007gr show quite similar behavior. We also compared the SN 2017ein light curves with those of SN 2004aw (Taubenberger et al. 2006), which may agree at early times but diverge post-peak, with SN 2004aw being generally more luminous at all bands. (We have made the comparison with SN 2004aw here, since, as we show in Section 3.5, SN 2017ein and SN 2004aw could be similar bolometrically.) One can see that we missed observations of maximum light at V [V(max)] for SN 2017ein, owing to poor observing conditions. Based on the comparison with SN 2007gr, we estimate, however, that V(max) occurred on JD ≈ 2,457,913.1 (i.e., June 8.6).

Zoom In Zoom Out Reset image size

That the SN was not detected to R > 18.4 mag in images obtained by Im et al. (2017) on May 24.63 (JD 2,457,898.13), but then was seen at R = 17.7 mag by these investigators on May 25.77 (JD 2,457,899.27) and later discovered by Arbour on May 25.99, would imply that the explosion date was likely sometime between JD 2,457,898 and 2,457,899. We analyzed the R-band light curve, including the Im et al. (2017) points, by fitting a simple analytic model for H-free SNe (Vacca & Leibundgut 1997; see Figure 2), and we found that the explosion could have been as early as JD 2,457,898.1, or May 24.6, which we adopt. The analytic model further implies that R-band maximum (again, missed by our observations) occurred on approximately JD 2,457,916, or ∼3 days after V-band maximum.

In Figure 3 we show the early-time B − V, V − R, and V − I color curves. We have initially corrected the observed color curves for Galactic foreground reddening (Schlafly & Finkbeiner 2011, via the NASA/IPAC Extragalactic Database [NED]). Again, comparing with SN 2007gr (Hunter et al. 2009), we can see that the color evolution of SN 2017ein follows much the same behavior, to within the uncertainties. We also found that V − R of SN 2017ein evolved at early times in a fashion similar to that of SN 2004fe (not shown; Drout et al. 2011).

Zoom In Zoom Out Reset image size

We show the early-time spectra of SN 2017ein in Figure 4. The first spectrum was obtained on June 2, about 1 week after discovery (and explosion). From the light-curve comparison with SN 2007gr, we estimate the ages of the spectra relative to V(max). In the figure we also show a comparison with spectra of SN 2007gr from Valenti et al. (2008) and Chen et al. (2014) and of SN 2004aw from Taubenberger et al. (2006), at approximately the same ages. The spectra of SN 2004aw and the Valenti et al. spectra of SN 2007gr were obtained from WISeREP¹⁸ (Yaron & Gal-Yam 2012), and the Chen et al. SN 2007gr spectra were obtained from the UC Berkeley Supernova Database¹⁹ (SNDB; Silverman et al. 2012). One can see that the spectra of SN 2017ein resemble quite closely those of SN 2007gr, as was also true for the light curves. SN 2017ein also somewhat resembles SN 2004aw, spectroscopically, although the latter may have had somewhat broader lines. Many of the spectral features, which we have indicated in the figure, are the same between SN 2017ein and SN 2007gr, and the line widths are comparatively narrow. SN 2017ein, like SN 2007gr, is clearly not a broad-lined SN Ic.

Zoom In Zoom Out Reset image size

In the pre-maximum spectrum, the features at 6350 Å and at ∼7000 Å, attributed to C ii λλ6580 and 7235 in SN 2007gr (Valenti et al. 2008), are evident in SN 2017ein, and the Ca ii λλ8542, 8662, 8498 near-infrared (NIR) feature and, interestingly, the ∼10,400 Å feature, attributed primarily to C i, both make an early appearance. In the later spectra, Ca ii, Ti ii, Mg ii λ4481, the well-resolved Fe ii λλ4924, 5018, 5169, Sc ii, Na i, O i λ7774, and C i are also seen, as in SN 2007gr. For both SN 2017ein and SN 2007gr at later times, weak interstellar Hα emission appears.

We measured (see Table 4) the photospheric expansion velocity of SN 2017ein from the Fe ii λλ4924, 5018, 5169 and O i λ7774 absorption lines, by fitting a Gaussian to each of the lines with the routine splot in PyRAF. The measurements from the Fe ii lines were averaged. We show these velocities for both sets of lines in Figure 5, with a comparison to SN 2007gr (Chen et al. 2014), SN 2004aw (Taubenberger et al. 2006), and several SNe Ic in the sample from Liu et al. (2016). The expansion velocities of SN 2017ein are not initially (soon after explosion) as high as those of SN 2007gr and SN 2004aw, and overall they appear to be intermediate between these two other SNe.

Zoom In Zoom Out Reset image size

Unlike SN 2007gr, though, as the Ca ii NIR triplet in SN 2017ein appeared to narrow in width with advancing age, a notch of additional absorption appeared in the red wing of the triplet feature. It is unclear to what to attribute this feature; it could be the Ca ii λ8664 line becoming more apparent as the triplet feature narrowed, or it could be the C i λ8727 line, which is normally blended with the Ca ii NIR triplet, as the other C i absorption features also become stronger.

Also notably exceptional for SN 2017ein is the presence of significant Na i D absorption in the June 2 spectrum (day −6.3), implying more interstellar extinction to SN 2017ein than to SN 2007gr. This is particularly seen in Figure 6, based on the MMT spectrum, in which the well-resolved Na i doublet is strong (at 5904.9 and 5910.6 Å, respectively), attributable internally to the host galaxy. The doublet is far weaker in the Galactic foreground, consistent with the assumed low reddening from this contribution. We have also indicated in Figure 6 the presumed location from the host galaxy of the diffuse interstellar band (DIB) feature at 5780 Å, the strength of which can be used to infer visual extinction A_V (Phillips et al. 2013) and which can change over time in a subset of SNe Ib/c (Milisavljevic et al. 2014).

Zoom In Zoom Out Reset image size

The extinction to SN 2017ein, as it turns out, is not well determined from our data. We first attempted to estimate the visual extinction from the SN spectra. Stritzinger et al. (2018) presented for SESNe a correlation between the equivalent width (EW) of the Na i D feature and host galaxy A_V. From the blended Na i D feature in the June 2 (day −6.3) spectrum, arising entirely from the host galaxy, we measured EW = 1.61 ± 0.06 Å. From the MMT spectrum, we measured from the resolved Na i D1 and D2 lines an EW of 0.67 ± 0.04 Å and 0.72 ± 0.01 Å, respectively. The total EW of the feature from this spectrum is then 1.39 ± 0.05 Å. Referring to Stritzinger et al. (2018, their Figure 17), we then can see that A_V is somewhere in the range of 0 to ∼2 mag. Their best fit to the relation, A_V = 0.78(±0.15) × EW(Na i D), results in host A_V = 1.26 ± 0.29 and A_V = 1.08 ± 0.25 mag from the blended feature in the Lick spectrum and the sum of the resolved lines in the MMT spectrum, respectively.

Unfortunately, the DIB λ5780 feature is not particularly distinct in the MMT spectrum. Following Phillips et al. (2013), we fit a Gaussian of 2.1 Å FWHM intensity to the spectrum and placed a 3σ upper limit of 203 mÅ on its EW, which, from their Equation (6) (which has a 50% systematic uncertainty), corresponds to A_V ≲ 1.06 mag.

We can also estimate the amount of extinction to SN 2017ein through a color-curve comparison. Drout et al. (2011), from their systematic study of SN Ib/c light curves, found that the intrinsic color (V − R)₀ = 0.26 ± 0.06 mag on V(10 day), 10 days past V(max) (see also Bianco et al. 2014; Dessart et al. 2016). For SN 2017ein, we estimate that V(10 day) was about JD 2,457,923.1, and on that day, V − R = 0.49 ± 0.27 mag. The Galactic foreground contribution to E(V − R) is quite small, 0.012 mag (Schlafly & Finkbeiner 2011). This would indicate that the host contribution to E(V − R) is 0.22 ± 0.28 mag for SN 2017ein, where the uncertainty includes both the measurement uncertainty in the SN 2017ein V − R color from KAIT and the uncertainty in the Drout et al. (2011) intrinsic color. Assuming a Fitzpatrick (1999) reddening law and R_V = 3.1, this corresponds to A_V = 1.07 ± 1.35 mag.

We compared the SN 2017ein B − V curve, corrected for the low Galactic foreground reddening (0.019 mag), with that of SN 2007gr, for which the Galactic reddening is E(B − V) = 0.055 mag (Schlafly & Finkbeiner 2011; Chen et al. 2014) and the host reddening is assumed to be quite low (Hunter et al. 2009; Drout et al. 2011; Chen et al. 2014); see Figure 3. Additionally, Stritzinger et al. (2018), based on their sample from the Carnegie SN Project (CSP), provided color templates, after correction for Galactic foreground reddening, for SESNe, specifically for SNe Ic, from 0 to +20 days relative to V(max). The relevant template to apply here is for B − V, which we show compared to the color curve for SN 2017ein in Figure 3. (Two other templates from Stritzinger et al. 2018, V − r and V − i, might have been applicable here; however, it is not readily evident how to transform accurately for SNe Ic between the Sloan Digital Sky Survey r and i bands used by the Carnegie Supernova Project (CSP) and the Johnson–Cousins R_C and I_C bands that we used for this study.) Both the template and the SN 2007gr B − V color curve imply that E(B − V) = 0.34 ± 0.07 mag for SN 2017ein. From a minimum χ² fitting of the SN 2017ein color curves to the corresponding SN 2007gr ones, we estimate that E(V − R) = 0.24 ± 0.06 and E(V − I) = 0.57 ± 0.06 mag for SN 2017ein (the Galactic foreground contribution to E(V − I) is 0.026 mag). The value of E(V − R) for the entire color curve is consistent with that inferred from the Drout et al. (2011) fiducial value at V(10 day).

We performed an analysis of the three color excesses in a manner similar to that of Stritzinger et al. (2018), assuming a Fitzpatrick (1999) reddening law. We fit the values of host A_V and R_V that were best constrained by all three excesses. We found a rather large range in both parameters, A_V = 1.3–1.9 mag for R_V = 3.3–6.8. Immediately, these imply that the visual extinction to SN 2017ein is appreciable, consistent with the large range in A_V inferred from the Na i D feature strength, and that the dust is dissimilar from the diffuse Galactic component (for which R_V = 3.1). Stritzinger et al. (2018) have indeed argued that R_V = 4.3 is typical for SNe Ic, consistent with their general location in dusty environments presumably composed of larger dust grains (we note, however, that this inference is based on only one of the events in their sample, which had more extreme reddening). Values of R_V larger than 3.1 are typical for massive star-forming regions; for example, Smith (2002) found that the local R_V = 4.8 for clouds in the Carina Nebula. This effect likely arises from the strong ultraviolet radiation from young O-type stars that destroys some of the smallest grains, leading to a flatter reddening law.

The high values of A_V and R_V are most strongly driven in our analysis by the value of E(V − I). If we relax our analysis and consider only E(B − V) and E(V − R), and if we further assume that R_V = 4.3 (see above), then A_V would be in the range of 1.3–1.7 mag, consistent with the range in A_V when including all three color excesses. If, however, we assume that R_V = 3.1, similar to Galactic diffuse interstellar dust, then A_V would be distinctly lower, 1.0–1.2 mag. For the sake of discussion below, specifically when we analyze the nature of the SN progenitor in Section 4.3, we will consider all three ranges in extinction and their implications.

We summarize all of the host galaxy extinction estimates for SN 2017ein in Table 5. If a value for R_V is either estimated or assumed, it is indicated in the table as well.

Another quantity for SN 2017ein that is not well known is its distance, D. Several estimates exist for the distance to the host galaxy, NGC 3938. Poznanski et al. (2009), based on assuming that SNe II-P are standardizable candles, estimated that the distance modulus to SN 2005ay is μ = 31.27 ± 0.13 mag (D = 17.9 Mpc). Rodríguez et al. (2014), invoking a similar, color-based standardization of the absolute brightness of SNe II-P, found that μ = 31.75 ± 0.24 mag (22.4 Mpc) for their method in the V band and 31.70 ± 0.23 mag (21.9 Mpc) in I. Several early Tully–Fisher estimates (Bottinelli et al. 1984, 1986) resulted in far shorter distances, with μ ≈ 28.7 ± 0.7 mag (D ≈ 5.7 Mpc), although Tully (1988) lists μ = 31.15 ± 0.40 mag (17.0 Mpc).²⁰

We therefore consider hereafter a range in possible distance moduli to SN 2017ein of 31.15–31.75 mag (D = 17.0–22.4 Mpc). Given this distance range and the inferred SN extinction, above, for SN 2017ein, M_V(max) ≈ −16.9 to −17.5 mag, which is consistent with −17.2 mag for SN 2007gr (Hunter et al. 2009), but less luminous than −18.0 mag for SN 2004aw (Taubenberger et al. 2006).

We constructed an early-time bolometric light curve of SN 2017ein from our observed BVR_CI_C light curves. Given the relative sparseness in the photometric coverage and the large ranges in both distance and reddening, high precision was not a particular concern. For that reason, rather than performing our own detailed blackbody fitting of each set of photometric points, we utilized the relations for SESNe from Lyman et al. (2014) for estimating bolometric corrections to the absolute B and V light curves based on the reddening-corrected colors, in order to generate the SN 2017ein bolometric light curve. We show this curve in Figure 7. Although there are uncertainties in both the photometry and the bolometric corrections, these are dwarfed by the overall uncertainties in both the distance and the extinction to the SN.

Zoom In Zoom Out Reset image size

For comparison we also show in Figure 7 the bolometric light curves of SN 2007gr (Hunter et al. 2009; Chen et al. 2014) and SN 2004aw (Taubenberger et al. 2006). The bolometric luminosity of SN 2017ein is consistent at peak, to within the large uncertainties, with that of SN 2007gr; however, overall, the former is generally more luminous than the latter. SN 2004aw, which was more luminous than SN 2007gr, is in agreement with SN 2017ein, to within the uncertainties.

We have modeled the bolometric luminosity via a semianalytical light curve, following the method of Arnett (1982; see also Valenti et al. 2008; Cano 2013; Taddia et al. 2015; Lyman et al. 2016; Prentice et al. 2016; Arnett et al. 2017). The best-fit model to the bolometric luminosity implies that the maximum bolometric luminosity, L_bol(max), is (1.7–3.4) × 10⁴² erg s⁻¹, and the maximum occurred around day 14 since explosion. We show the mean of the best-fit model in Figure 7.

The Arnett (1982) model (see, e.g., Equations (A1) and (A2) of Valenti et al. 2008; Equations (1) and (2) of Cano 2013) fits two parameters, the nickel mass M(⁵⁶Ni) and the diffusion timescale τ_m, where ${\tau }_{{\rm{m}}}^{2}=(C{\kappa }_{\mathrm{opt}}{M}_{\mathrm{ej}})/(\beta {{cv}}_{\mathrm{sc}})$. In this expression, M_ej is the SN ejecta mass, c is the speed of light, β is a constant of integration of ∼13.8 (Arnett 1982), v_sc is the scale velocity, κ_opt is the optical opacity (often assumed to be 0.07 cm² g⁻¹), and C is a constant of proportionality. Note that the M_ej estimate is just for the mass involved in the diffusion of SN light; if nonionized ejecta mass also exists, the value of M_ej could be larger (Wheeler et al. 2015). The quantity v_sc is generally assumed to be the photospheric velocity (v_ph) at maximum light. Note that, for example, Lyman et al. (2016) assumed that C = 2, whereas Chatzopoulos et al. (2012) assumed C = 10/3 (since, generally, the mean photospheric velocity is adopted, rather than the peak velocity, at optical depth 1/3). The kinetic energy of the ejecta is ${E}_{K}=(3/10){M}_{\mathrm{ej}}{v}_{\mathrm{sc}}^{2}$. From our fitting, the value of τ_m = 11.6 days. Based on our measurement of velocities from the Fe ii lines (assumed to be nearly photospheric) in the observed spectra, we have set v_ph = 9700 km s⁻¹.

The results of the modeling are that M(⁵⁶Ni) is in the approximate range of 0.09–0.18 M_⊙. If κ_opt = 0.07 cm² g⁻¹, then for SN 2017ein M_ej ≈ 1.45 M_⊙ and, thus, E_K ≈ 8.1 × 10⁵⁰ erg. For comparison with SN 2007gr, M(⁵⁶Ni) ≈ 0.06 M_⊙ (Chen et al. 2014) to ∼0.08 M_⊙ (Valenti et al. 2008; Hunter et al. 2009), and Hunter et al. (2009) estimated that M_ej ≈ 2.0–3.5 M_⊙ and E_K ≈ (1–4) × 10⁵¹ erg. From nebular spectra of SN 2007gr, Mazzali et al. (2010) estimated a lower M_ej ≈ 1 M_⊙. In contrast, Taubenberger et al. (2006) found for SN 2004aw significantly higher values of M(⁵⁶Ni) = 0.25–0.35 M_⊙, M_ej = 3.5–8.0 M_⊙, and E_K = (3.5–9.0) ×10⁵¹ erg (see also Mazzali et al. 2017). We note that Taddia et al. (2015), from their sample of SNe Ic, estimated mean values of M(⁵⁶Ni) = 0.33 M_⊙, E_K = 1.75 × 10⁵¹ erg, and large M_ej = 5.75 M_⊙. We summarize this comparison in Table 6.

Table 6.  Comparison of Light-curve-derived Parameters for SN 2017ein

SN	M(56Ni)	Mej	EK	Source
	(M⊙)	(M⊙)	(1051 erg)
2017ein	0.10–0.17	0.96–1.76	0.54–0.99	This worka
⋯	0.05–0.2	${1.2}_{-0.2}^{+0.3}$	${0.7}_{-0.1}^{+0.2}$	This workb
2007gr	0.061 ± 0.014	⋯	⋯	Chen et al. (2014)a
⋯	0.076 ± 0.020	2.0–3.5	1–4	Valenti et al. (2008), Hunter et al. (2009)a
⋯	⋯	∼1.0	⋯	Mazzali et al. (2010)c
⋯	${0.07}_{-0.01}^{+0.01}$	${1.2}_{-0.4}^{+0.6}$	${0.8}_{-0.3}^{+0.3}$	Drout et al. (2011)b
2004aw	0.25–0.35	3.5–8.0	3.5–9.0	Taubenberger et al. (2006),a Mazzali et al. (2017)c
⋯	${0.27}_{-0.05}^{+0.05}$	${4.5}_{-1.3}^{+2.1}$	${2.8}_{-0.8}^{+1.3}$	Drout et al. (2011)b
SN Ic sample	0.33 ± 0.07	5.75 ± 2.09	1.75 ± 0.24	Taddia et al. (2015)a
⋯	0.24 ± 0.15	${1.7}_{-0.9}^{+1.4}$	${1.0}_{-0.5}^{+0.9}$	Drout et al. (2011)b

Notes.

^aFrom the quasi-bolometric light curve. ^bFrom multiband light curves. ^cFrom analysis of nebular spectra.

Download table as:  ASCIITypeset image

We can also analyze these parameters, M(⁵⁶Ni), M_ej, and E_K, from the observed light curve, following Drout et al. (2011). From the simple analytic model we constructed for the R-band light curve (see Section 3.1), we found that M_R(max) ≈ −17.1 to −18.6 mag for SN 2017ein, given the distance and extinction range. From this model, the quantity Δm₁₅, the decrease in brightness between maximum and 15 days post-maximum, is 0.87 mag. This Δm₁₅ value is consistent with that of SN 2007gr, to within the uncertainties (Drout et al. 2011). From M_R(max) and Δm₁₅, we found that M(⁵⁶Ni) ≈ 0.05–0.2 M_⊙, which agrees well with that found from the bolometric curve. From the simple model, the characteristic width of the light curve is ${\tau }_{c}=10\,\pm 1$ days (again, consistent with SN 2007gr; Drout et al. 2011). For ${v}_{\mathrm{ph}}=9700$ km s⁻¹, we can then infer that M_ej ≈ 1.2 M_⊙ and E_K ≈ 7 × 10⁵⁰ erg for SN 2017ein. Again, these are completely consistent both with what we found (above) for the bolometric light-curve analysis and with SN 2007gr. We note that Drout et al. (2011) found much higher values for these various parameters for SN 2004aw: τ_c ≈ 19, M(⁵⁶Ni) ≈ 0.27 M_⊙, M_ej ≈ 4.5 M_⊙, and E_K ≈ 2.8 × 10⁵¹ erg.

To possibly identify a progenitor candidate in the 2006 HST images, we astrometrically registered the 2006 WFPC2 F555W image mosaic to the 2017 WFC3 image mosaic. Using 27 fiducial stars, we were able to register the images to 0.26 WFPC2/WF pixel (1σ rms; 26 mas). As a result, SN 2017ein would be at (798.55, 1766.58) in the WFPC2 image mosaic. An object can be seen at (798.33, 1766.54). This is a difference of 0.22 pixels, which is within the rms uncertainty in the astrometry. We therefore consider this object to be a candidate for the progenitor of SN 2017ein. Note that we first preliminarily identified this object in Van Dyk et al. (2017). We show the WFC3 and WFPC2 images to the same scale and orientation in Figure 8.

Zoom In Zoom Out Reset image size

We have analyzed this object further by performing PSF-fitting photometry of the HST WFPC2 images using Dolphot (Dolphin 2000a, 2016). Prior to this step, we processed the individual WFPC2 frames with AstroDrizzle, to attempt to flag cosmic-ray hits. We set the Dolphot parameters FitSky = 3 and RAper = 8, using the TinyTim PSFs provided with Dolphot and setting the parameters InterpPSFlib and WFPC2useCTE to "true." The SN site is contained in the WFPC2 chip 2. We find 24.56 ± 0.11 Vega-mag and 24.58 ± 0.17 Vega-mag in F555W and F814W, respectively. Dolphot also outputs "object type" = "1" for the source, which indicates that the routine considers it to be a "good star." Other indicators also point to a star-like character: the sharpness and crowding parameters in each band are 0.002, −0.007 and 0.030, 0.076, respectively. The signal-to-noise ratio in each band for the star is appreciable, ∼10 and ∼6 in F555W and F814W, respectively. The output photometric quality flag from Dolphot is also "0" for both bands, meaning (from the Dolphot documentation) that the star was recovered very well in the image data. The observed color of the source is F555W−F814W = −0.02 ± 0.20 mag.

We can estimate the metallicity at the SN 2017ein site from the oxygen abundance gradient for NGC 3938 presented by Pilyugin et al. (2014). Oxygen abundance is often used as a proxy for metallicity. In that study the central oxygen abundance is $12+\mathrm{log}({\rm{O}}/{\rm{H}})=8.79\pm 0.05$ and the gradient is −0.0413 ± 0.0066 dex kpc⁻¹. We deprojected an image of the host galaxy from the Digitized Sky Survey, assuming the position angle (29°) and inclination (18°) from Jarrett et al. (2003), and calculate that the SN is 44'' from the host nucleus. Given the range in host galaxy distance (above), this corresponds to a ∼3.7–4.8 kpc nuclear offset, and with this range we estimate, from the published abundance gradient, that $12+\mathrm{log}({\rm{O}}/{\rm{H}})\approx 8.59$–8.64.

Nebular Hα and [N ii] λ6583 emission is also seen in the MMT spectrum from June 24. Although not an ideal strong-line indicator, the ratio of the line fluxes, [N ii] λ6583/Hα, known as the N₂ index (Pettini & Pagel 2004), provides an estimate of metallicity. Dereddening the spectrum assuming A_V = 1 mag (with R_V = 3.1), we obtain N₂ = −0.30. If the extinction in the SN environment were as high as A_V = 2 mag, the index would vary slightly, to −0.31. Following the calibration of this index by Marino et al. (2013), this would imply that 12 + log(O/H) ≈ 8.60.

If we adopt the oxygen abundance for the Sun as $12+\mathrm{log}({\rm{O}}/{\rm{H}})\approx 8.69$ (Asplund et al. 2009), we can infer that the metallicity at the SN site is somewhat subsolar (i.e., [Fe/H] ≈ −0.10).

Assuming that the extinction to the SN we have estimated applies to the progenitor candidate as well, and given the assumed range in distance, we find that the object is consistent with being quite luminous and blue. (Note that we are only considering here interstellar extinction, which provides a lower limit on the total extinction in the presence of any circumstellar extinction destroyed by the SN breakout; although not typically expected for hot blue stars, some binary WC stars are surrounded by dust formed by colliding winds; e.g., Williams 2008.) We show in Figure 9 three possible ranges of loci for the object in a color–magnitude diagram (CMD), based on our assumptions about the amount of reddening, as discussed in Section 3.3.

Zoom In Zoom Out Reset image size

We compared these loci in Figure 9 with single-star evolutionary tracks from the Modules for Experiments in Stellar Astrophysics (MESA) Isochrones and Stellar Tracks (MIST; Paxton et al. 2011, 2013, 2015; Choi et al. 2016) v1.1 with rotation at v_rot/v_critical = 0.4, solar-scaled abundances (assuming Z = 0.0142; Asplund et al. 2009), and subsolar metallicity [Fe/H] = −0.10. The tracks have been interpolated to the WFPC2 Vega-mag system by the MIST online interface.²¹ Among the tracks, we find that the candidate's locus, assuming the lowest reddening, is roughly consistent with, although somewhat blueward of, the terminus of a star with initial mass of 48–49 M_⊙. The net effect of rotation is to lead to a bigger He core mass, and thus more luminous progenitor, for the same initial mass. If the star was not a rapid rotator, then the implied initial mass (if single) would be even higher. We note that some evolutionary models for single stars with these initial masses at solar metallicity experience "failed" explosions, with the cores collapsing directly to black holes (e.g., Sukhbold et al. 2016).

Additionally, we compared in Figure 9 with the endpoints of binary-star models from BPASS v2.1 (Eldridge et al. 2017) at two different metallicities, solar and somewhat subsolar (note that for BPASS, Z = 0.020 is considered solar metallicity). What is shown in the figure is the combined light of both the primary and the secondary in each of the binary models. (Here we have assumed that the F555W and F814W bandpasses used for BPASS are generic enough to apply for WFPC2.) We have considered only systems for which the primary's surface H mass fraction is equal to 0, the surface He mass fraction is ≤0.27, and the ratio of the mass of remaining He to the total ejecta mass is ≤0.1 at the model endpoint. These criteria should be sufficient to approximate an He-stripped SN Ic progenitor (J. J. Eldridge 2018, private communication). The systems allowed by the color and luminosity range for the progenitor, again assuming the lowest reddening, are to the lower right corner of this color–luminosity range. At Z = 0.020 the lone system has a 60 M_⊙ primary, a mass ratio of the two components q = 0.9 (i.e., the system consists of two stars of nearly the same mass), and an initial orbital period log P(day) = 1.0. The primary of this system ejects M_ej ≈ 10 M_⊙ and leaves behind a remnant with ∼6 M_⊙. At Z = 0.014 there is one system with these same parameters, as well as systems with a 80 M_⊙ primary, q = 0.7, and a range of log P = 0.8–2.0. These primaries eject M_ej ≈ 6 M_⊙ and have remnants with ∼7–15 M_⊙. At Z = 0.010 the systems have primaries that, similarly, range from 60 to 80 M_⊙, with q = 0.7–0.8 and log P = 1–2.8. Ejecta masses are in the range of 5–8 M_⊙, and remnant masses are ∼8–12 M_⊙. The remnants for all of these systems are most likely to be black holes.

It may seem counterintuitive that the model binary systems should possess a primary that is more massive than the single-star models, with both binary- and single-star models having essentially the same observed luminosity in F814W at their endpoints. However, it should be remembered that what we are showing in Figure 9 is a CMD and not a Hertzsprung–Russell diagram (HRD). For all of the binary models considered here, the mass transfer is generally nonconservative: the more massive primaries in the models all lose substantial amounts (≳75%) of their initial mass, while the secondaries increase little in mass as the systems evolve. On an HRD the primaries end up being less luminous bolometrically, although far hotter (T_eff > 10⁵ K) than they were on the zero-age main sequence; the secondaries generally increase monotonically in luminosity, while changing little in effective temperature, ending up more luminous than the primaries. In fact, the model primaries in the binary systems end up less luminous than the endpoints of the single-star models. In the CMD shown in the figure, the secondaries all dominate the observed light of the systems at their endpoints; as an additional effect, the hot, stripped primaries contribute comparatively far less light at F814W.

It appears that the two highest-reddening scenarios, A_V = 1.3–1.9 mag for R_V = 3.3–6.8 and A_V = 1.3–1.7 mag at R_V = 4.3, imply that the object would be too blue and possibly too luminous for single stars or binary-star systems. The lowest-reddening scenario also may be just on the verge of being too blue and luminous.

In addition, we cannot rule out that the progenitor candidate is actually a compact cluster. At the distance of NGC 3938, such a cluster could be unresolved. To that end, we considered stellar clusters in a galaxy similar to NGC 3938, those in NGC 628 (M74; also a nearly face-on Sc galaxy, but closer to us at 10.19 Mpc; Jang & Lee 2014), which Adamo et al. (2017) studied (see also Grasha et al. 2015). We considered the "default" or "reference" deterministic catalog, for which the cluster photometry is an averaged aperture correction method, and for which Milky Way extinction and Padova stellar evolutionary tracks for age dating are adopted.²² We selected from the catalog only those clusters that are symmetric and compact, and we include these for comparison in Figure 9. (Here we neglected differences in the F555W and F814W bandpasses between WFC3/UVIS and WFPC2.) We have dereddened the clusters, using the "best E(B − V)" in the catalog and assuming R_V = 3.1. Note that ∼5 of these clusters are also roughly consistent with the possible loci for the progenitor. These particular clusters are all quite blue and young, with estimated ages ≲4 Myr and masses ≲2000 M_⊙. Based on the MIST tracks, the turnoff masses for such young clusters are ≳65 M_⊙.

We can also potentially constrain the initial mass of the SN 2017ein progenitor by estimating the age of the stellar population, or populations, in its immediate environment. Here we consider a radius from the SN position for the environment of ∼100 pc, a typical size for an OB association, within which stars should generally be coeval with the progenitor. We had already extracted the photometry from the archival HST WFPC2 images for the environment, as well as for the progenitor (Section 4.1). All of these objects appear to be point-like, based on their Dolphot photometric properties, discussed above. We show in Figure 10 a CMD displaying the objects in the environment, corrected by two different assumptions of the total visual extinction, A_V = 1 and 2 mag, and adjusted by the assumed distance (the uncertainty in the distance is included in the uncertainties in absolute magnitude). A Cardelli et al. (1989) reddening law was assumed here with R_V = 3.1. We note, of course, that this is a simplification, since, over this size scale, the extinction could be variable. The photometry for the progenitor candidate is included in the figure but corrected for reddening by the same amounts as the overall environment, not as we have done in Section 4.3.

Zoom In Zoom Out Reset image size

One can see a number of stellar objects in the SN 2017ein environment. One interesting inference to point out is that although the progenitor candidate may be among the bluest objects in the environment, it is not necessarily the most luminous. We have compared the photometry for these objects with the PARSEC-COLIBRI model isochrones from Marigo et al. (2017) and MIST isochrones from Choi et al. (2016), both at a subsolar metallicity Z = 0.01. Note that these models are all based on single-star evolutionary tracks. We can approximately fit these populations, using both sets of models, with a mix of three different ages (4, 7, and 13 Myr), if the extinction is A_V ≈ 1 mag. The objects are poorly fit if A_V are closer to 2 mag, particularly for the objects with (F555W–F814W)₀ < 0 mag, which are likely massive main-sequence stars. We therefore consider this assumed extinction to be an upper limit on the actual extinction. Assuming the lower extinction, the youngest objects are generally quite blue. Three evolved cool supergiants appear consistent with the 13 Myr isochrone, with one or two additional objects consistent with the blue loop of the PARSEC-COLIBRI track at that age.

The corresponding turnoff masses for the PARSEC-COLIBRI isochrones are ∼57, 28, and 16 M_⊙. These masses are quite similar for the MIST isochrones, at ∼60, 26, and 15 M_⊙. That an SN has occurred recently in this environment would imply that it is a member of the younger, more massive population. The inferred initial masses are consistent with those of the high-mass BPASS binary models that appear to agree with the locus of the progenitor candidate in the CMD. However, if the assumption of coevality breaks down (e.g., dispersal through random motions), then it is possible that the SN progenitor was a member of one of the older, less massive populations, consistent with the expectation of a more moderate-mass binary progenitor. Furthermore, Maund (2017) showed that, within 100 pc in the environments of SNe II-P, a mixture of populations at different ages could be colocated, one of those populations being related to a given SN II-P progenitor. A similar situation might be in effect for the SN 2017ein environment.