SN 2017gmr: An Energetic Type II-P Supernova with Asymmetries

The full optical light curve can be seen in Figure 4, and the V-band light curve compared to other Type II SNe is shown in Figure 5. For reference, the r-band discovery magnitude is shown as an open hexagon, while the dotted line connects the pre-explosion upper-limit r-band magnitude 2 days prior in Figure 5. Overall the shape is that of a typical Type II SN with an extended plateau, albeit on the brighter end with a maximum M_V = −18.3 mag. The maximum occurs at ∼6 days after explosion for the U and B bands, ∼8 days for g and V, and closer to 10 days for r and i. This is consistent with the average rise times seen for the majority of Type II SNe (González-Gaitán et al. 2015; Rubin et al. 2016; Förster et al. 2018).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The light curves then remain at a relatively constant magnitude for the next 75 days until the fall off the plateau begins around day 85, with decline rates of 0.027, 0.011, and 0.003 mag day⁻¹ in B, V, and i, respectively. Using the method described in Valenti et al. (2016), we obtain the point at half of the fall at MJD 58,093.5 ± 0.4, or 95 days after our estimated explosion date. Between day 85 and day 105 the V-band light curve drops by 1.5 mag. This moderate post-plateau drop is on the lower end but consistent with other Type II-P SNe, particularly higher-luminosity events (Valenti et al. 2014).

The plateau length of SN 2017gmr is on the shorter side for comparable objects and has an average M_V = −17.8 mag (Figure 5), a value noticeably brighter than the norm (but similar to SN 2004et). According to Anderson et al. (2014), Faran et al. (2014), and Galbany et al. (2016), more luminous Type II-P SNe tend to exhibit shorter plateau durations, which coincides with the overall picture of SN 2017gmr. SN 2017gmr, SN 2013fs, SN 2004et, and SN 2008if all show similar luminosities and evolution over the first few days (Figure 5 inset) but then evolve to drastically different light-curve shapes. While SN 2017gmr and SN 2004et change very little over the first 3 months, SN 2013fs and SN 2008if show evolution more akin to Type II-L SNe, with a larger drop in luminosity over the first ∼75 days.

One rather intriguing feature seen in the early light curve of SN 2017gmr is the bump in luminosity that occurs a couple of days post-explosion, particularly in the bluest bands. In Figure 6 we show the ground-based U and B observations along with the Swift UV. From the U and B data we see a sharp rise over the first 2 days, then a drop of roughly 0.2 and 0.1 mag in U and B, respectively, and then a slow rise over the next few days back to the peak value. Unfortunately, no Swift data exist prior to day 2, so the light-curve behavior in the UV bands is unknown over the same time period. It is also possible that we are seeing undulations in the U and B light curves due to inhomogeneities in the CSM, particularly in some cases where the magnitude changes are larger than the uncertainties.

Zoom In Zoom Out Reset image size

Models recently produced by Moriya et al. (2018) do show this small bump in luminosity in the u and g bands with certain mass-loss and density configurations (see also Morozova et al. 2018). The key to creating this early bump is to have moderately dense CSM close to the progenitor. The Type II-P SN 2016X showed a similar bump in the Swift UV light curve over the first few days after explosion, although it did not seem to be present in the optical bands (Huang et al. 2018). Their explanation for the initial light-curve peak was a shock breakout cooling effect, but as we discuss in Section 4.8, we cannot fit this bump with standard shock-cooling models.

As we show in Figure 5, the radioactive tail of SN 2017gmr does not show the exponential decline of ⁵⁶Co decay of 0.98 mag 100 day⁻¹ (Woosley et al. 1989). While the B band declines around 0.9 mag 100 day⁻¹, V and i decline by 1.5 and 1.4 mag 100 day⁻¹, respectively. These values are more consistent with the mean value of 1.47 mag 100 day⁻¹ in the V band found by Anderson et al. (2014). By our last photometric observations around day 175, the V-band light curve is about 0.5 mag fainter than expected if we assume the canonical 0.98 mag 100 day⁻¹. The same behavior is seen in the bolometric light curve, as we discuss below. This deviation can be explained by incomplete gamma-ray trapping, a decrease in the energy input from shock interaction, as dust production in the ejecta, or some combination of the three.

Incomplete gamma-ray trapping has been documented in other Type II-P SNe. As mentioned above, Anderson et al. (2014) found that the mean decline rates at late times for a large sample of Type II-P SNe were roughly 0.5 mag 100 day⁻¹ larger in V band than the generally accepted value of 0.98 mag 100 day⁻¹. They also found that the more luminous the SN, the greater the deviation from the expected decay rate. Highly energetic explosions can also have large expansion velocities, which in turn leads to weaker trapping. Alternatively, if the distribution of ⁵⁶Ni is very asymmetric or mixed in the ejecta, the escape probability could be greater. If CSM interaction is occurring, it can also contribute to the luminosity at late times and would not follow the predicted rate of ⁵⁶Co decay. We will discuss these possible scenarios further in Section 6.

Multiple epochs of NIR data were obtained over the first 160 days of evolution. The NIR luminosity rose over the first 30–40 days after explosion (Figure 7). This was followed by a few weeks of nearly constant luminosity; then, starting around day 75, a steady decline begins in all filters and continues until our last observed epoch. We have plotted the NIR light curves of SN 2004et from Maguire et al. (2010) as a comparison, and they indicate that the NIR plateau is much shorter for SN 2017gmr than SN 2004et and that likely the late-time NIR luminosity is greater for SN 2017gmr as well.

Zoom In Zoom Out Reset image size

The B − V color evolution of SN 2017gmr and a comparison to other SNe are shown in Figure 3. As in other Type II SNe, the color is initially blue and evolves rapidly toward the red as the large envelope of the RSG progenitor expands and cools, until it reaches the recombination phase and the rate slows (de Jaeger et al. 2018a). This continues over the duration of the optically thick plateau phase until a peak value of B − V = 1.5 mag. After day 100, once the exponential decline phase begins, the color gradually becomes bluer again.

The abundance of photometric data has allowed us to straightforwardly create a quasi-bolometric light curve using the routine superbol (Nicholl 2018). Following the description in Nicholl et al. (2016), the reddening and redshift-corrected photometry in each band were interpolated with the g band as reference and then converted to a spectral luminosity (L_λ). The bolometric luminosity was then computed from the integration of the SED for each epoch.

In Figure 8 we show the bolometric light curve produced from the observations (red) and those obtained with blackbody corrections (black), as well as the blackbody temperature (T_bb) and blackbody radius (R_bb) shown in orange and purple, respectively, in the bottom of Figure 8. The red light curve is pseudo-bolometric and is constructed by integrating under the filters from UV to IR. Swift UV coverage does not extend past ∼9 days, so a first-order polynomial is fit to the data and extended out to later epochs. As the contribution to the total bolometric luminosity falls quickly after the first few weeks, this does not add much uncertainty. The data have been corrected for an E(B − V) = 0.30 mag and adopting the distance modulus μ = 31.46 mag.

Zoom In Zoom Out Reset image size

As we mention above, the late-time light curve falls faster than expected for a fully trapped ⁵⁶Co decay of 0.98 mag 100 day⁻¹, with L_bol roughly 1 × 10⁴¹ erg s⁻¹ fainter that predicted on day 165. Integrating over the entire bolometric light curve gives a total radiated energy of 3.5 × 10⁴⁹ erg in the first 175 days.

With the advent of high-cadence transient searches in the past decade, several instances of pre-SN outbursts have been observed directly in the months to years before explosion (e.g., Fraser et al. 2013; Mauerhan et al. 2013; Ofek et al. 2013, 2014; Elias-Rosa et al. 2016; Tartaglia et al. 2016; Reguitti et al. 2019), although overall detectable outbursts are rare (Bilinski et al. 2015; Strotjohann et al. 2015). These outbursts are generally associated with SNe that have substantial circumstellar material as evidenced by their SN IIn-like behavior. However, many standard Type II-P/L SNe also show evidence for CSM material either as narrow emission lines in their early-time spectra (e.g., Khazov et al. 2016) or as early peaks in their light curves (Morozova et al. 2017). This CSM could have been deposited in the years or decades prior to explosion and could have been accompanied by faint pre-SN outbursts, as has recently been suggested in the gravity-wave-driven scenario of Shiode & Quataert (2014) and Fuller (2017), or in unsteady nuclear burning events or binary interaction (Smith & Arnett 2014).

The field of NGC 988 was observed by the DLT40 survey 56 times between 2015 January and 2017 September, just prior to the explosion of SN 2017gmr. During much of this time period, the DLT40 survey was coming online, with some prolonged down periods. No precursor outbursts were observed down to a typical limiting magnitude of r ∼ 19–19.5 mag (−12 > M_r > −12.5). We can therefore rule out bright eruptions like SN imposters or LBV eruptions with roughly M_r = −14 mag lasting several months, but not fainter or short-lived outbursts. This includes those LBV eruptions that have been found to have magnitudes of only M_r = −10 or −11 mag (Smith et al. 2011b).

Due to the well-sampled photometric data over the first few days after explosion in SN 2017gmr, we are able to model the early-time light curves using the prescriptions outlined in Sapir & Waxman (2017). To do this, we employed the code presented in Hosseinzadeh (2019) and described in Hosseinzadeh et al. (2018), which uses a Markov chain Monte Carlo (MCMC) routine to fit the light curve in each photometric band and outputs posterior probability distributions for physical parameters, such as the time of explosion, the temperature and luminosity 1 day after explosion, the time at which the envelope becomes transparent, and the progenitor radius. Data were only fit up to day 4 to still lie within the validity range described by Rubin & Gal-Yam (2017). The best fits to our data are shown in Figure 9.

Zoom In Zoom Out Reset image size

One day after explosion, the modeled temperature is (25.9 ± 0.1) × 10³ K (kK) with a luminosity of (2.9 ± 0.03) × 10⁴³ erg s⁻¹. From the Sapir & Waxman (2017) fits the estimated progenitor radius is 489 ± 22 R_⊙ (3.4 × 10¹³ cm), a value on the small end for an RSG, which theoretically can range in size from ∼100 to 1500 R_⊙ (Levesque 2017), but that is commensurate with observations of some Galactic RSGs (e.g., Wittkowski et al. 2017; Montargès et al. 2018). From these fits we also derive an explosion date of MJD 57,999.09 ±0.01 days. This value is further bolstered by our first observation obtained on MJD 58,000.27, or just over a day after the estimated explosion date, and our last nondetection on MJD 57,998.22. This is also consistent with the UV photometry obtained 2.5 days after discovery, which does not show a rise to peak that is seen in other bands (Figure 6).

The early spectra, shown in Figure 10, are typical for a young Type II-P SN, displaying a blue, mostly featureless continuum. Only strong interstellar Na I D absorption lines and a broad emission feature around 4600 Å (likely He ii λ4686) are seen. Neither the low-resolution FLOYDS spectrum nor the high-resolution Keck spectrum, taken within hours of discovery, shows signs of narrow high-ionization lines, other than narrow 55 km s⁻¹ Hα seen in the Keck HIRES echelle spectrum. This is different from other early-detected CCSNe which can show features of highly ionized nitrogen and carbon along with He and H. This is discussed further in Section 6.3.

Zoom In Zoom Out Reset image size

As the photosphere begins to cool, the continuum becomes redder and broad Balmer emission lines begin to appear with P Cygni absorption features. When Hα becomes pronounced a week after explosion, the peak appears blueshifted, centered at −5000 km s⁻¹. This is a common occurrence in Type II-P SNe, where the opaque hydrogen envelope preferentially obscures the redshifted, receding side of the line (Dessart & Hillier 2005a; Anderson et al. 2014). As the recombination front moves through the envelope, the red side becomes visible again and the emission-line peak becomes more symmetric.

Around a month after explosion, the SN is well into the plateau phase and the Ca ii IR triplet centered around 8600 Å emerges, along with a forest of metal lines blueward of 5000 Å (Figure 11). In particular, lines of Fe ii, including Fe ii λλ4924, 5018, and 5169, can be seen.

Zoom In Zoom Out Reset image size

By the end of the plateau phase other broad lines such as Ba ii λ6142, [Sc ii] λλ5527, 5658, and 6246 (blended with [O i]), and [O i] λλ6300, 6364 appear in the nebular spectra (Figure 12). Redward of Hα, strong [Ca ii] λλ7291, 7324 is seen, flanked on either side by He i λ7065, Fe ii λ7155, and O i λ7774. What appears to be K i λλ7665, 7699 is also detectable and distinct from O i by ∼day 120. The emergence of the He i λ7065 line starting around day 90 suggests the presence of a strong ionization source. Also of note is the strengthening of the Ca ii IR triplet, which has become almost as strong as Hα by day 165.

Zoom In Zoom Out Reset image size

Figure 13 shows the NIR spectral evolution from 2 to 149 days. Overall the spectra show a decrease in flux with increase in wavelength, typical of young CCSNe. The spectra from the first week are featureless (minus atmospheric absorption), but by day 13 some Paα emission begins to emerge. Over the next month Paβ and Brγ appear as the continuum flux decreases. Both He i 1.083 and Paγ are present, although slightly blended. As the SN drops from the plateau phase after 100 days, additional lines of O i, Si i, He i, and other weak hydrogen series are seen. The CO overtone between 2.3 and 2.5 μm is not present in our last two spectra as has been seen for other Type II SNe (Yuan et al. 2016; Rho et al. 2018; Sarangi et al. 2018; Tinyanont et al. 2019). This may help rule out dust formation, at least in the first 150 days.

Zoom In Zoom Out Reset image size

To help constrain the distance to SN 2017gmr, we have used the Expanding Photosphere Method (EPM; Kirshner & Kwan 1974), which relies on the relation between the photometric angular radius and the spectroscopic physical radius of the homologously expanding SN ejecta. Assuming that the outflow is radiating as a diluted blackbody, the observed SN magnitudes are fitted to a blackbody function multiplied by dilution factors, to derive the color temperature and the angular radius. Dilution factors based on atmosphere modeling of Type II SNe were adopted from Dessart & Hillier (2005b). Further, to eliminate the effect of filter response function ingrained in the observed broadband magnitudes, the response function is convolved with the blackbody model flux. The convolved function can be expressed in terms of the color temperature and the coefficient values taken from Hamuy et al. (2001). Following the same procedure undertaken in Dastidar et al. (2018), expansion velocities were calculated using the He i λ5876 and Fe ii λ5169 lines over the first 50 days of evolution.

The distance is derived from a linear fit to the data in the form of

$$\begin{eqnarray}&&t=D(\theta /{v}_{\mathrm{ph}})+{t}_{o},\end{eqnarray} \tag{ 1 }$$

where the slope is the distance and the y-intercept the date of explosion. This fit is shown in Figure 14. From this method we obtain an EPM distance of 18.6 ± 2.2 Mpc, a value consistent with the 19.1 Mpc used throughout the paper. It also indicates an explosion epoch of MJD 57,999.0 ± 1.9 days, which agrees well with the constrained explosion date discussed above.

Zoom In Zoom Out Reset image size

We have also measured the distance using the Standard Candle Method (SCM), which was first proposed by Hamuy & Pinto (2002) and later expanded on by other authors. SCM uses photometric magnitudes and expansion velocities at 50 days. For SN 2017gmr these values are m_V = 14.57 ± 0.04, m_R = 13.86 ± 0.02, m_I = 13.56 ± 0.03, and v_{Fe ii} = 5600 km s⁻¹. From these values we get SCM distances (in Mpc) of 16.10 (Hamuy 2005), 16.88 (Takáts & Vinkó 2006), 24.70 (Nugent et al. 2006), 14.38 (Poznanski et al. 2009), 10.24 (de Jaeger et al. 2017), and 13.31 (Gall et al. 2018). Except for Nugent et al. (2006), all other SCM distances are systematically lower than the EPM and kinematic distances. The same was found for SN 2017eaw and SN 2004et in Szalai et al. (2019) and could be due to CSM interaction or asymmetries. The SCM method relies on a correlation between the magnitude and expansion velocity at day 50, which could break down under these conditions.

To estimate the ⁵⁶Ni mass, we employ various methods from the literature, in particular those of Hamuy (2003), Jerkstrand et al. (2012), and Pejcha & Prieto (2015). These methods all rely on bolometric luminosities in the radioactive tail phase, so we use the constructed bolometric light curve discussed above (Figure 8). This results in measured ⁵⁶Ni masses of 0.130 ±0.026 M_⊙, 0.124 ± 0.026 M_⊙, and 0.090 ± 0.030 M_⊙, respectively, for the three techniques. In the Pejcha & Prieto (2015) calculation, we extrapolated the bolometric luminosity to day 200 and obtain an L_bol = (1.85 ± 0.9) × 10⁴¹ erg s⁻¹.

Other than SN 1992H, for which the actual ⁵⁶Ni mass could be as low as 0.06 M_⊙ depending on the distance used, and SN 1992am (Hamuy 2003), this is one of the highest ⁵⁶Ni masses reported for normal Type II SNe (Anderson 2019), higher if there is incomplete gamma photon trapping or if the SN is at a further distance than 19.6 Mpc, and lower if there is CSM interaction or if the SN is closer. According to Müller et al. (2017), less than 5% of Type II-P SNe have ⁵⁶Ni masses as large as 0.12 M_⊙. For comparison, other "normal" Type II-P SNe such as SN 1999em, SN 2003gd, and SN 2004dj each have ⁵⁶Ni masses of ∼0.02 M_⊙, or a full order of magnitude lower than estimated here (Elmhamdi et al. 2003b; Hendry et al. 2005; Vinkó et al. 2006).

We can also estimate the ⁵⁶Ni mass using a steepness factor S, where S = −dM/dt, a measure of the transition between the plateau and radioactive tail phases (Elmhamdi et al. 2003a). Generally an anticorrelation exists, where the steeper the transition, the lower the ⁵⁶Ni mass. Following Equation (7) in Singh et al. (2018), we measure a steepness factor S = 0.070 ±0.007 mag day⁻¹ , which corresponds to an estimated ⁵⁶Ni mass of ∼0.055 M_⊙. This is significantly smaller than the value obtained using the late-time bolometric luminosity and more consistent with other normal Type II-P SNe. This inconsistency could be due to the degree of mixed ⁵⁶Ni in the ejecta, since the same amount of ⁵⁶Ni will create a steeper decline if it is centrally located rather than mixed. The mixed ⁵⁶Ni will actually increase the radiative diffusion timescale, causing the transition to appear shallower.

In Figure 15 we show the evolution of the line velocities of both Hα and Fe ii λ5169 (shown as a function of radius over time). Hα falls from 15,000 km s⁻¹ near explosion to a relatively stable value of 7000–8000 km s⁻¹ during the radioactive tail. Fe ii λ5169, a more reliable measurement of photospheric velocity than Hα, settles to a late-time velocity of 3500 km s⁻¹. These expansion velocities are higher than average for Type II SNe and for Type II-P SNe in particular. In Figure 16 we show the comparison of SN 2017gmr optical spectra at various epochs with the well-studied SN 1999em and SN 2004et. At all epochs the line velocities of SN 2017gmr are faster than those of the other two.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

From Gutiérrez et al. (2017a), the mean velocities on day 53 for a sample of 122 Type II SNe (measured from the absorption minimum) are 6365 and 3537 km s⁻¹ for Hα and Fe ii λ5169, respectively. In comparison, SN 2017gmr has velocities on day 53 of 9330 and 5240 km s⁻¹ for Hα and Fe ii 5169, respectively. By day 115, the difference in the Fe ii λ5169 velocities has decreased, 2451 km s⁻¹ on average versus 3520 km s⁻¹ for SN 2017gmr, but Hα remains almost 2000 km s⁻¹ faster than the mean value of 5805 km s⁻¹.

The faster line velocities seem to correlate well with the high inferred ⁵⁶Ni mass and maximum luminosity of SN 2017gmr. Gutiérrez et al. (2017b) found a correlation between expansion velocities and ⁵⁶Ni mass that indicated that more energetic explosions (resulting in faster expansion velocities) created higher ⁵⁶Ni mass. When combined with previous conclusions of Hamuy & Pinto (2002), Hamuy (2003), and Pejcha & Prieto (2015), this suggests that the more energetic the explosion, the higher the luminosity, expansion velocity, and ⁵⁶Ni production. This may suggest that SN 2017gmr had an unusually energetic explosion, although low ejecta mass can also allow for high ejecta velocities.

There are other ways to create faster line velocities. If CSM interaction is occurring, it can excite Hα, Fe ii, and other lines at larger radii (and therefore higher velocities). This means that lines that would have otherwise already recombined in the outer, faster parts of the ejecta will be reionized and give the appearance of faster ejecta at later times. This can make it appear as if the ejecta velocities are changing slowly or stagnant, especially at later times. Faster expansion velocities can also arise from asymmetries in the explosion. Dessart & Hillier (2011) found that asphericities in the ejecta of an axially symmetric explosion can change the location of the P Cygni minimum with inclination as much as 30% in the photospheric phase. We explore the possibility of an asymmetric explosion in Section 6.4.

Narrow lines seen within the first few days of explosion can be useful to infer composition, velocity, and density of the CSM surrounding the SN progenitor (Gal-Yam et al. 2014). One of the most well-known objects displaying this phenomenon, SN 2013fs, showed narrow (∼100 km s⁻¹) lines of oxygen, helium, and nitrogen within the first few hours of explosion (Yaron et al. 2017; Bullivant et al. 2018). These high-excitation lines disappeared over the next 2 days, and eventually the spectra resembled those of a normal Type II SN. Similar behavior has been seen in SN 1983K (Niemela et al. 1985), SN 2006bp (Quimby et al. 2007), SN 2013cu (Gal-Yam et al. 2014), SN 1998S (Shivvers et al. 2015), PTF11iqb (Smith et al. 2015), SN 2016bkv (Hosseinzadeh et al. 2018), and SN 2014G (Terreran et al. 2016). Khazov et al. (2016) found that 14%–18% of their sample of Type II SNe showed signs of early narrow lines, which they conclude is a lower limit for the Type II SN population as a whole.

These narrow lines were interpreted as the flash ionization of a WR-like wind for SN 2013cu (Gal-Yam et al. 2014). Later interpretation suggested that it is instead possibly the ionization of the cool dense wind from an LBV/YHG progenitor (Groh et al. 2014), which is more consistent with a Type IIb SN progenitor. Furthermore, Smith et al. (2015) found that PTF11iqb had an RSG progenitor, and the early narrow lines were likely the result of shock ionization from CSM interaction. A similar conclusion about the progenitor of SN 1998S was also reached in Shivvers et al. (2015) and Mauerhan & Smith (2012). In other words, WR-like wind features (particularly of hydrogen-rich WNH type) can be seen in early spectra if there are enough high-energy photons to fully ionize the progenitor's cool dense wind.

SN 2017gmr was observed spectroscopically within hours after discovery, and likely within 1.5 days of shock breakout, yet the only narrow emission line seen was that of Hα (Figure 17), and only with the higher-resolution instruments. The Keck HIRES spectrum on day 1.5 (inset of Figure 17) shows a narrow Hα emission with a Gaussian FWHM velocity of ∼55 km s⁻¹. This is suggestive of an RSG wind (see Smith 2014). The spectral resolution of these data is ∼7 km s⁻¹, so the velocity of the ionized material is fully resolved. For reference, SN 1998S was observed with the same instrument 1.86 days after discovery and had a narrow component velocity of ∼40 km s⁻¹ (Shivvers et al. 2015, albeit with lines other than Hα also present). The day 2.3 HET spectrum also seems to show a narrow but weak Hα feature with a moderately higher intermediate-width FWHM velocity of ∼1000 km s⁻¹. The broadening of the line may be due to electron scattering in the CSM, and the narrow feature may be embedded within, but it has likely faded by this epoch. This feature is completely gone in the HET spectrum 3 days later; in its place is a broad, blueshifted Hα emission with an expansion velocity of 15,000 km s⁻¹.

Zoom In Zoom Out Reset image size

One other noticeable feature in the very early spectra is the broad emission around ∼4600 Å shown in detail in Figure 18. A similar broad bump was seen in SN 2006bp (Quimby et al. 2007) and SN 2013fs (Yaron et al. 2017; Bullivant et al. 2018) and was attributed to blueshifted He ii λ4686 formed from the SN ejecta beneath a CSM shell. Unfortunately, a chip gap from the HET spectrum of SN 2017gmr occurs just blueward of 4670 Å, but the overall shape is similar to that of the other SNe. SN 2006bp and SN 2013fs also showed narrow He ii λ4686 emission on the red edge of the broad 15,000 km s⁻¹ line, which is absent in SN 2017gmr.

Zoom In Zoom Out Reset image size

The lack of narrow high-ionization lines in the early spectra would seem to suggest that if nearby CSM was present, its density was too low to yield detectable emission. Alternatively, it could imply that the photons were not energetic enough to doubly ionize He in the CSM, even if SN 2017gmr likely had a very energetic explosion. If the CSM density was adequately high, this too could prevent narrow lines from forming, as it would self-absorb all of the high-energy photons. Another option would be that the narrow, high-ionization lines were present before our first spectrum at 1.5 days but were produced from asymmetric CSM, which was quickly enveloped by the spherically expanding SN ejecta (Smith et al. 2015). We will discuss this possibility further in the next section.

Another luminous Type II, SN 2016esw, was also caught within a day of explosion and showed no signs of high-ionization emission lines (de Jaeger et al. 2018b). The authors conclude that the progenitor of SN 2016esw was likely surrounded by low-density CSM some distance away from the surface of the star that eventually showed signs of interaction 2–3 weeks after explosion. Similarly, the Type II-P SN 2017eaw did not show early flash signatures (Van Dyk et al. 2019), except for possibly the ∼160 km s⁻¹ Hα line seen by Rui et al. (2019) on day 2.5. Unlike SN 2016esw, though, neither SN 2017eaw nor SN 2017gmr showed obvious signs of CSM interaction in the shape of the Hα emission line the first few weeks after explosion.

When the SN reappeared from behind the Sun in 2018 July, we obtained one high-resolution echelle spectrum with MIKE on Magellan/Clay on day 312. The late-time analysis on SN 2017gmr is beyond the scope of this paper and will be discussed in depth in an upcoming paper, but due to the implications for the early-time evolution, we are including the Hα and [O i] lines here. In Figure 19 we show in red the day 312 spectrum compared to the other moderate-resolution MMT spectra. Instead of a single broad line, Hα clearly shows three intermediate-width peaks. The same is seen in the [O i] doublet, although the right peak of the λ6300 line is stronger than the red peak of Hα due to the overlap of the blue peak from the λ6364 line of the doublet. Signs of this asymmetry can even be seen in the day 154 spectrum.

Zoom In Zoom Out Reset image size

In Figure 20 we show that the multipeaked Hα can be fit with three Lorentzians, one centered at 0 km s⁻¹ (6563 Å) and blue and red peaks at roughly ±1700 km s⁻¹ (±35 Å). The same velocities are seen in [O i], for both the λλ6300 and 6363 lines, but the doublet nature of the line makes it appear distinctly different. The red peak of λ6300 would fall at ∼6335 Å, while the blue peak of λ6364 would fall at ∼6330 Å, making the red peak of λ6300 seem as bright as the blue peak, and swamping the emission at the center of the line.

Zoom In Zoom Out Reset image size

Double-peaked emission lines in SN spectra are often interpreted as ejecta interacting with asymmetric CSM, most commonly in a disk or torus (Hoffman et al. 2008; Mauerhan et al. 2014; Smith et al. 2015; Andrews et al. 2017). In this scenario, the underlying broad component traces emission from the free expansion of the SN ejecta, while the intermediate components are formed in the post-shock region between the forward and reverse shocks created as the ejecta crashes into the CSM. When the fast-moving SN ejecta collides with the slow-moving CSM, depending on the density of surrounding material, the CSM can be accelerated from speeds of 10–100 km s⁻¹ up to thousands of kilometers per second. The red and blue peaks therefore are the result of the ejecta accelerating the CSM material radially outward from the explosion. In the case of SN 2017gmr the CSM was likely accelerated from a normal RSG wind speed of ∼55 km s⁻¹ to the observed intermediate feature speed of ∼1700 km s⁻¹. Examples of other Type II SNe at somewhat similar phases to SN 2017gmr showing multipeaked Hα are shown in Figure 20.

The fact that we do not see narrow emission lines does not necessarily discount the possibility of SN 2017gmr being a partially CSM interaction powered event. CSM interaction can be inferred based on the intermediate-width line shapes and velocities. As explained in Smith et al. (2015), Smith (2017), and Andrews & Smith (2018), a disk-like geometry in the CSM may allow the CSM interaction to be hidden below the photosphere after the disk is enveloped by the fast SN ejecta. If the region of CSM interaction is happening below the ejecta photosphere and the CSM is sufficiently dense, it can be hidden for long periods of time because the sustained CSM interaction luminosity itself keeps the surrounding SN ejecta ionized and optically thick. This could help explain the extended high luminosity of SN 2017gmr. All that is required is that the disk or torus of material has a limited radial extent (i.e., ≤100 au) so that it can be overrun early by the SN photosphere. Only when the photosphere recedes internal to the CSM location (which has been pushed outward to 1700 km s⁻¹ owing to the Doppler acceleration) will the intermediate-width lines be revealed.

If we assume that high-ionization lines were observable prior to our 1.5-day spectrum, we can use the expansion velocity of Hα (15,000 km s⁻¹) to infer that the outer edge of the CSM must be closer than 1.8 × 10¹⁴ cm (or ∼2500 R_⊙). This is roughly the same radius inferred for SN 2013cu (Gal-Yam et al. 2014) and PTF11iqb (Smith et al. 2015).

The other possibility is that the multiple peaks seen in the emission lines could come from asymmetries in the ejecta (Maeda et al. 2008; Taubenberger et al. 2009). This was the scenario presented for SN 2003jd (Mazzali et al. 2005), SN 2004dj (Chugai et al. 2005, shown in Figure 20), SN 2010jp (Smith et al. 2012a), and SN 2016X (Huang et al. 2018; Bose et al. 2019). In a forthcoming paper Nagao et al. (2019) find that there is strong polarization in SN 2017gmr indicative of an aspherical explosion. Nonuniformity of ⁵⁶Ni could cause uneven ionization and excitation in the ejecta and produce multipeaked emission lines. SN 2004dj showed strong Hα asymmetry immediately after the plateau phase ended, during the epoch of increased polarization (Leonard et al. 2006). As we show in Figure 19, distinct multiple peaks are not present until sometime between 154 and 312 days, or a significant time period after the end of the plateau. Also of note is that there is a component at rest velocity at late times in SN 2017gmr that would have to come from some spherically distributed radioactive material.

In general, it is difficult to disentangle the two mechanisms. The low polarization at early times is explained by Nagao et al. (2019) as the hydrogen envelope hiding a highly asymmetric helium core that is only observable when the optical depth decreases. We suggest that it could also be explained partially (or in full) by the spherical symmetry of the hydrogen envelope erasing the polarization signatures of deeply embedded asymmetric CSM interaction. The deviation from ⁵⁶Co decay in the late-time light curve can be due to incomplete γ photon trapping caused by a nonspherical ejecta, or it could be due to a decrease in the shock interaction. Whatever the mechanism, the emission-line shapes emerging during the nebular phase indicate a deviation from spherical symmetry, whether it be from asymmetric stellar ejecta or shock interaction with a disk or torus of CSM.

As we briefly mention above, the optical and bolometric light curves decline faster than 0.98 mag 100 day⁻¹ attributed to ⁵⁶Co decay. The fast decline could indicate the halting of shock interaction as a primary energy source, or that there is incomplete trapping of gamma-rays as we discuss in Section 6.1. It could also be due in all, or part, to dust formation in the ejecta.

Along with a decrease in optical luminosity from the growth of dust grains, we can also expect to see a blueshifted asymmetry in the optical emission lines since dust in the ejecta would attenuate the receding red side of the SN more than the blue. First detected in SN 1987A (Lucy et al. 1989), evidence for dust formation has been seen in many CCSNe, including SN 2003gd (Sugerman et al. 2006), SN 2004et (Kotak et al. 2009), SN 2005ip (Smith et al. 2009; Fox et al. 2010; Stritzinger et al. 2012; Bevan et al. 2019), SN 2006jd (Stritzinger et al. 2012), SN 2007od (Andrews et al. 2010; Inserra et al. 2011), SN 2010jl (Smith et al. 2012b; Gall et al. 2014), and one of the clearest cases, SN 2006jc (Mattila et al. 2008; Smith et al. 2008). In conjunction with the emission-line asymmetry and a decrease in the optical light curve, a corresponding increase in the IR luminosity is often observed as new dust grains form in the ejecta.

It is unlikely that dust has formed in SN 2017gmr by ∼150 days for a few reasons. First, a blackbody fit to the optical and NIR spectroscopy and photometry around day 150 indicates T_bb = 6800 K, a temperature much too high for grain condensation. Second, the bolometric light curve also shows a deviation from expected ⁵⁶Co decay. If dust formation was occurring, the light curves in individual bands will change, but the total bolometric light curve would be unchanged. Finally, as we mention above, the NIR spectroscopy during the early nebular phase fail to reveal the first overtone of CO (Figure 13). Normally the detection of CO heralds the formation of dust (Gerardy et al. 2000; Sarangi & Cherchneff 2013). Therefore, the blue-peaked hydrogen emission profiles and the fast decline in L_bol are likely due to other physical characteristics of the SN such as asymmetries and CSM interaction, not dust formation. This does not discount the possibility that in later epochs we may begin to see signatures of dust condensation in the ejecta.

Photometric data for SN 2017gmr was obtained from a variety of telescopes (see Section 2), resulting in an extremely high cadence optical light curve (Figure 4), as well as an early-time Swift UV+optical light curve (Figure 6). We briefly describe the instrumentation and data reduction techniques here, although if a telescope+instrument combination is not specifically mentioned, the data were reduced in a "standard" way, including image detrending (bias subtraction and flat-fielding), cosmic-ray removal, point-spread function (PSF) or aperture photometry, along with flux calibration performed against standard catalogs (e.g., Landolt standard stars or the Sloan Digital Sky Survey [SDSS]). The full ground-based optical data set is presented in Table 1, while the Swift data are presented in Table 2.

First, continued monitoring of SN 2017gmr was done by the DLT40 survey's two discovery telescopes, the PROMPT5 0.4 m telescope at Cerro Tololo Inter-American Observatory and the PROMPT-MO 0.4 m telescope at Meckering Observatory in Australia, operated by the Skynet telescope network (Reichart et al. 2005). The PROMPT5 telescope has no filter ("Open"), while the PROMPT-MO telescope has a broadband "Clear" filter, both of which we calibrate to the SDSS r band (see Tartaglia et al. 2018, for further reduction details).

Las Cumbres Observatory UBVgri-band data were obtained with the Sinistro cameras on the 1 m telescopes, through the Global Supernova Project. Using lcogtsnpipe (Valenti et al. 2016), a PyRAF-based photometric reduction pipeline, PSF fitting was performed. UBV-band data were calibrated to Vega magnitudes (Stetson 2000) using standard fields observed on the same night by the same telescope. Finally, gri-band data were calibrated to AB magnitudes using the SDSS (SDSS Collaboration et al. 2017). Because the Las Cumbres data are the most comprehensive, and there are differences across the instrument/filter pairs, all other data sets were shifted by small amounts to match the Las Cumbres magnitudes in Figure 4. These values ranged between 0.05 and 0.15 mag. The nonshifted values are all included in Table 1.

Optical photometry in the BVRI bands was obtained at the 60/90 cm Schmidt telescopes at Konkoly Observatory; see Vinkó et al. (2012) for a description of the instrumentation and data reduction techniques. Further, BVRI photometry was obtained with the Super-LOTIS (Livermore Optical Transient Imaging System; Williams et al. 2008) 0.6 m telescope at Kitt Peak National Observatory; these data were reduced in a manner similar to that described in Kilpatrick et al. (2016), and PSF photometry using standard IRAF procedures was then done on the resultant images.

Data in the BVugriz bands were taken with the IO:O imager on the Liverpool telescope and were reduced using the standard IO:O pipeline; aperture photometry was performed using custom python scripts and pyraf. The data were shifted +0.17 mag in B to match the Las Cumbres data. Data from the 1.30 m DFOT and 2.01 HCT telescopes were reduced as described in Dastidar et al. (2019) performing PSF fitting photometry using DAOPHOT II (Stetson 1987). Instrumental magnitudes were converted to standard magnitudes using a set of local standard stars and observations of either Landolt standard or SDSS fields.

The Swift UVOT analysis uses the pipeline of the Swift Optical/Ultraviolet Supernova Archive⁵² (SOUSA; Brown et al. 2014). The method is based on that of Brown et al. (2009), including subtraction of the host galaxy count rates, and uses the revised UV zero-points and time-dependent sensitivity from Breeveld et al. (2011). For SN 2017gmr we do not have template images to subtract the underlying galaxy flux. In this case, however, the largest contributor to the background is scattered/reflected light from the nearby bright star evident in Figure 1. The reported UVOT magnitudes use a background position that to the eye approximated the brightness of the galaxy and halo at the SN position. The errors have been conservatively increased to match the range of magnitudes measured with a variety of halo-free and bright halo regions. The full Swift data set is presented in Table 2 and is plotted in Figure 6.

Raw NIR data from NOTCam were reduced using the NOTCam Quicklook reduction package, and PSF photometry was then performed using standard IRAF procedures. The REM telescope is equipped with an optical and an IR camera, which observes simultaneously the same field, thanks to a dichroic placed before the telescope focal plane. IR images were corrected for dark current and flat-fielded, and subsequently median-stacked to obtain a background frame for each filter. The background-subtracted images were geometrically aligned and then stacked to obtain a final image for each filter, and the background in the locations of SN 2017gmr was modeled with a low-order polynomial surface and subtracted. The flux of the SN and the local sequence were measured through PSF fitting. For both instruments, photometric calibration was done using Two Micron All Sky Survey stars in the field. The resulting data set can be found in Table 3.

A high-cadence spectral sequence of SN 2017gmr was taken with low-, medium-, and high-resolution instrumentation throughout the rise, plateau, and fall from plateau of the SN. A log of these observations can be found in Table 4. These spectra were reduced using standard techniques, including bias subtraction, flat-fielding, cosmic-ray rejection, local sky subtraction, and extraction of one-dimensional spectra. Most observations had the slit aligned along the parallactic angle to minimize differential light losses. Flux calibration was done with standard-star observations, and most spectra were rescaled to match the photometric light curve at a given epoch. We discuss some details of the spectroscopic reductions below, but if a particular telescope+instrument combination is not mentioned, it was reduced in a standard way as described above.

Las Cumbres optical spectra were taken with the FLOYDS spectrographs mounted on the 2 m Faulkes Telescope North and South at Haleakala, USA, and Siding Spring, Australia, respectively, through the Global Supernova Project. A 2'' slit was placed on the target at the parallactic angle. One-dimensional spectra were extracted, reduced, and calibrated following standard procedures using the FLOYDS pipeline (Valenti et al. 2014). HIRES spectra were reduced using the MAuna Kea Echelle Extraction (MAKEE) data reduction package⁵³ (written by T. Barlow). MIKE spectra were reduced using the latest version of the MIKE pipeline⁵⁴ (written by D. Kelson).

A sequence of NIR spectra of SN 2017gmr were also taken and are logged in Table 5. All NIR spectra were taken using a classical ABBA technique, dithering the object along the slit in order to facilitate good sky subtraction. Further, the slit was oriented along the parallactic angle to minimize slit losses due to atmospheric differential refraction (Filippenko 1982). In all cases, an A0V star was observed either before or after the science observations in order to correct for telluric absorption and flux-calibrate the data, following the prescriptions of Vacca et al. (2003).

Gemini/GNIRS data were taken in cross-dispersed mode with the 0675 slit, yielding continuous wavelength coverage from 0.8 to 2.5 μm and an R ∼ 1000. These data were reduced with the XDGNIRS pipeline provided by Gemini Observatory, as described in Sand et al. (2016) and Hsiao et al. (2019).

IRTF spectra were taken in SXD mode and the 05 slit, yielding wavelength coverage from ∼0.8 to 2.4 μm and R ∼ 1200. These data were reduced using the publicly available Spextool software package (Cushing et al. 2004), as described in Hsiao et al. (2019).

Two NIR spectra were taken with the Son OF ISAAC (SOFI) spectrograph mounted on the NTT telescope (Moorwood et al. 1998), using both the Blue and Red grisms, giving a broad wavelength coverage of 0.9–2.4 μm. The SOFI spectra were taken as part of the ePESSTO program and were reduced as described in Smartt et al. (2015).

Finally, a single FIRE spectrum was taken using the high throughput prism mode with a 06 slit, giving continuous wavelength coverage from 0.8 to 2.5 μm and a resolution of R ∼ 500 in the J band. The spectrum was reduced with the purpose-built firehose pipeline (Simcoe et al. 2013) as described in detail in Hsiao et al. (2019).