The late time optical evolution of twelve core-collapse supernovae: detection of normal stellar winds Free

ABSTRACT

We analyse the late time evolution of 12 supernovae (SNe) occurring over the last ∼41 yr, including nine Type IIP/L, two IIb, and one Ib/c, using UBVR optical data from the Large Binocular Telescope (LBT) and difference imaging. We see late time (5–42 yr) emission from nine of the eleven Type II SNe (eight Type IIP/L, one IIb). We consider radioactive decay, circumstellar medium (CSM) interactions, pulsar/engine driven emission, dust echoes, and shock perturbed binary companions as possible sources of emission. The observed emission is most naturally explained as CSM interactions with the normal stellar winds of red supergiants with mass-loss rates in the range −7.9 ≲ log₁₀(M_⊙ yr⁻¹) ≲ −4.8. We also place constraints on the presence of any shock heated binary companion to the Type Ib/c SN 2012fh and provide progenitor photometry for the Type IIb SN 2011dh, the only one of the six SNe with pre-explosion LBT observations where the SN has faded sufficiently to allow the measurement. The results are consistent with measurements from pre-explosion Hubble Space Telescope images.

1 INTRODUCTION

For all massive stars (>8 M_⊙), the end stages of stellar evolution involve the ultimate collapse of the iron core. In most cases, this leads to an explosion and a luminous supernova (SN). These core-collapse supernovae (ccSNe) are categorized into subtypes based on their photometric and spectroscopic properties. To simplify the typing somewhat, Type IIP/L SN arises from red supergiants (RSGs), Type Ib/c SN comes from stars stripped of their envelopes, and Type IIn SN explodes into a dense circumstellar medium (CSM). In the standard picture of non-Type IIn explosions, SNe can support their near peak luminosities until the ejecta recombine and become optically thin, and then the luminosity in the nebular phase is from radioactive decay. In Type IIn SNe, there is a significant or dominant contribution to the emission from shock heating the dense CSM.

As part of a search for failed SNe first proposed by Kochanek et al. (2008b), we have been monitoring 27 nearby (<10 Mpc) galaxies to search for failed SNe. The survey also enables two unique studies of the stars which do explode. First, it can be used to study the pre-SN variability of the progenitor stars, on time-scales of ∼10 yr directly and for ∼100 yr for transiently formed dust (e.g. η Car; Humphreys & Davidson 1994), so far with null results almost down to the levels of the variability of local RSGs (e.g. Szczygieł et al. 2012; Kochanek et al. 2017b; Johnson, Kochanek & Adams 2018a). This is difficult to reconcile with claims that SN light curves require unusually high pre-SN mass-loss rates shortly before explosion [e.g. Morozova, Piro & Valenti (2018) and Wu & Fuller (2022) discuss many claimed examples]. Secondly, by waiting for the SN to fade, it is possible to obtain high precision UBVR photometry of the progenitor stars using image subtraction (e.g. Johnson, Kochanek & Adams 2017). Based on either SN 1987A (e.g. Seitenzahl, Timmes & Magkotsios 2014) or typical estimates of the radioactive yields of ccSNe, this should be feasible after ∼3–4 yr.

In the most recent search for new failed SN, Neustadt et al. (2021) noted that SN 2013am and SN 2013ej were still brighter than their progenitors a decade after explosion. This is illustrated in Fig. 1 for SN 2013ej and SN 20011dh, the one example showing the expected fading behaviour. Here, we are showing the difference between the most recent images and pre-SN images, so a ‘negative’ image of the progenitor appears once the SN becomes fainter than the progenitor. This has clearly happened for SN 2011dh and the observed signal matches the flux expected for the progenitor from pre-SN Hubble Space Telescope (HST) images (see below Maund et al. 2011; Van Dyk et al. 2011). Such a source clearly is not present for SN 2013ej, and this is difficult to reconcile with radioactive decay as an energy source. Other possible sources of late time emission are interaction with CSM (e.g. Chevalier 1982a), dust echoes (e.g. Chevalier 1986), pulsar/engine-driven emission (e.g. Kasen & Bildsten 2010), and shock-heated companion stars (e.g. Ogata, Hirai & Hijikawa 2021).

The puzzle of SN 2013am and SN 2013ej motivates this systematic examination of the late time emission from the 13 SNe listed in Table 1 and discussed in Appendix  A that have occurred in the LBT galaxies since 1980. Of these, seven exploded during the course of the survey and so we should be able to observe the vanishing of the progenitor. For the other six, we can only search for continuing emission. SN 2009dh is one of the seven in the LBT sample, but it is heavily obscured and suffers from poor image subtractions. We include it for completeness but do not consider it part of the main study. Most of these SN are well studied. Seven have existing reports of CSM interactions and pre-SN mass-loss constraints usually based on radio or X-ray observations, and five are reported to have either optical or infrared dust echoes. These earlier results are summarized in Appendix  A. The new finding here is that late time (decade(s)) optical emission comparable to the luminosity of the progenitor stars appears to be the norm rather than the exception. In Section 2, we describe the data, analysis methods, and the five potential sources of late time emission. We analyse the SNe in Section 3 and discuss the results in Section 4.

2 OBSERVATIONS AND MODELS

The data are obtained using the Large Binocular Camera (LBC; Giallongo et al. 2008) on the LBT (Hill, Green & Slagle 2006) in the U, B, V, and R bands with image subtraction done using ‘ISIS’ (Alard & Lupton 1998; Alard 2000). The details of the data processing are given in Gerke, Kochanek & Stanek (2015), Adams et al. (2017b), and Neustadt et al. (2021). For SNe with observations obtained pre-event, we built the reference images using the best observations before the explosion. With these reference images, difference imaging photometry will yield a ‘negative’ image of the progenitor as the SN fades completely.

Lower quality data points are flagged as those with seeing FWHM > 1.5 arcsec or an ISIS flux scaling factor of <0.8. This scaling factor is the counts per unit flux compared to the reference image and thus is an estimate of the transparency for each epoch relative to a flux reference epoch. For the calibrations, we follow the same methods as Gerke et al. (2015) and Adams et al. (2017b). Stars in the reference images are matched to the Sloan Digital Sky Survey (SDSS; Ahn et al. 2012) and the corresponding ugriz AB magnitudes are transformed into UBVR Vega magnitudes following Jordi, Grebel & Ammon (2006). Following Johnson et al. (2018a), we extract light curves for both the target and a grid of comparison points to characterize the local noise. An inner grid of four points is placed 7 pixels apart (∼1.6 arcsec given the plate scale of 0.23 arcsec pixel⁻¹). The outer grid is placed 15 pixels away or ∼3.5 arcsec. Figs 4–10 show the resulting difference image light curves grouped by SN type. The formal errors are shown but are generally small, which is why we look at the comparison points. We compute band luminosities (νL_ν) using the distances and extinctions in Table 1. The extinction corrections for SN 2002hh are enormous, nearly 10 mag in the U band.

We fit the late time (meaning well into the nebular phase) light curves as a linear function in luminosity (νL_ν for each band),

with t₀ being the mean age of the points being fit so that there are no significant covariances between the two parameters L_SN and β_SN. We do the same for the light curves of the comparison points. Tables 5 and 6 report L_SN and β_SN and their uncertainties for each of the SNe. From the comparison points, we report the mean values 〈L_i〉 and 〈β_i〉 and the dispersion of the values σ_L and σ_β. Cases like SN 2009hd where there is significant dispersion in the comparison sample are a strong indication that the values are dominated by systematics. As mentioned earlier, the image subtractions for SN 2009hd are not very clean, driving the measured values and the large scatter. We do not consider SN 2009hd further.

The image subtraction light curves do not include the flux of the source in the reference image. For the SN with pre-SN images, this corresponds to the flux of the progenitor star. For the other SN, it is simply the mean flux of the SN over the images used to construct the reference image. We used aperture photometry to estimate the flux in the reference image, and the resulting luminosity is also reported in Tables 5 and 6. Particularly in the case of the progenitor stars, crowding generally makes the flux unmeasurable, so the luminosity estimates are dominated by systematic uncertainties. This is the reason why we focus on the subtracted light curves, which are little affected by crowding, separate from the reference flux.

Table 2 summarizes the cases where there are HST observations that can be used to explore this normalization. These are either pre-explosion observations for the SN with pre-SN LBT observations, or late time observations that (nearly) overlap the LBT observations. We converted the HST magnitudes to band luminosities using the HST zero points and the distances and extinctions from Table 1. These band luminosities are reported in Table 2 and are provided in the LBT light curves (Figs 4–10). The HST luminosities are absolute, while the LBT light curves are differential relative to the reference image. We are interested in basic normalizations, not modest filter differences, so we simply view the HST filter as representing the closest equivalent LBT filter (so, for example, V = F555W = F606W). While we are frequently worried about line emission, we do not think these details represent a serious concern given the wavelength ranges of the filters. Where the HST data overlaps the LBT light curve, it can be used to normalize the LBT data simply by shifting the LBT light curve to pass through the HST measurement. For example, this can be done trivially for SN 1993J. However, there was no way to do this uniformly, so the figures simply show the measured LBT differential luminosity evolution and the available HST data to provide the available information on absolute luminosities.

To characterize the late time photometric evolution of each of the SNe, we must consider the expected properties of the progenitor star and then the five possible sources of late time luminosity: radioactive decay, CSM interactions, dust echoes, neutron star spin down, and shock heated companions.

2.1 Progenitor stars

While we consider a broader range of SNe, the most interesting cases are the six where we can construct a reference image from pre-SN data. For a reference image built from pre-SN images, we will be left with an ‘inverse image’ of the progenitor once the SN has completely faded. SN 2011dh, shown in Fig. 1, is the one case out of these 6 where this appears to have occurred. We model its spectral energy distribution (SED) following the procedures of Adams & Kochanek (2015) using DUSTY (Ivezic & Elitzur 1997; Ivezic, Nenkova & Elitzur 1999; Elitzur & Ivezić 2001) so that we can consider self-obscuration and Solar metallicity stellar atmosphere models (Castelli & Kurucz 2003).

Fig. 2 shows the expected UBVR band luminosities along with the bolometric luminosity as a function of the initial progenitor mass M_ZAMS for the Solar metallicity PARSEC (Bressan et al. 2012; Marigo et al. 2013) isochrones. The Groh et al. (2013) progenitor luminosities are similar. The bolometric luminosity is roughly a power law in mass, with

In the PARSEC models, the stars are RSGs for M_ZAMS ≲ 30 M_⊙ so the stars are faintest in the U band and brightest in the R band. The mass-loss monotonically increases with mass, so there is a small transition region where the temperatures lead to the emission peaking in the optical, and then the stars are very hot, heavily stripped stars that are brightest in the U band and faintest in the R band. With binary interactions, the temperatures at a given mass can be very different, but the bolometric luminosity is still basically set by the helium core mass.

2.2 Radioactive decay

All SN will show emission due to radioactive decay, initially dominated by ⁵⁶Ni, then ⁵⁷Co and finally ⁴⁴Ti (e.g. Seitenzahl et al. 2014). A typical Type IIP SN synthesizes M_Ni < 0.1 M_⊙ (Sukhbold et al. 2016). The ⁵⁶Ni luminosity is

where τ₀ = 111.3 (Nadyozhin 1994). The time required for the luminosity to fade below a threshold luminosity, L, is

This neglects gamma-ray escape, which would lead to a more rapid decline in luminosity (Seitenzahl et al. 2014). For the typical Ni masses produced in a Type II-P SN of M_Ni < 0.1 M_⊙ (Sukhbold et al. 2016), the time-scale for the radioactive decay luminosity to be less than the progenitor luminosity is roughly 3–4 yr. For example, SN 1987A had a luminosity of ∼30 000 L_⊙ (∼3000 L_⊙) after 1000 (2000) d and M_Ni = 0.071 ± 0.003 M_⊙ (Seitenzahl et al. 2014). We will estimate the required M_Ni masses empirically by scaling the SN 1987A V-band light curve from Seitenzahl et al. (2014) to the LBT data. While we phrase the result in terms of M_Ni, the emission at these late times should be from ⁴⁴Ti.

2.3 CSM shock interactions

Following the SN, the expanding ejecta will shock heat any pre-existing CSM, exciting optical emission lines which contribute to the late time luminosity (e.g. Chevalier 1982a). For a spherically symmetric ρ ∝ 1/r² pre-SN wind, the scale of the optical shock luminosity is

where v_s is the shock speed, |$\dot{M}$| is the wind mass loss, v_w is the wind speed, and ϵ is the fraction of the luminosity emitted in the optical. For luminosities comparable to the progenitor or lower, the inertia of the swept up CSM is only modestly slowing the shock, so the ability of the shock to support late time emission is largely controlled by the efficiency ϵ of converting the shock energy into optical emission. The physics of ϵ is very complex, since it depends on the balance of emission mechanisms and the radiative efficiency (see e.g. Chevalier & Fransson 1994). For our standard results we will simply match the R-band luminosity (νL_ν) assuming ϵ = 0.1, v_s = 4000 km s⁻¹, and v_w = 10 km s⁻¹. We chose ϵ and v_s to match Chandra et al. (2022) but use v_w = 10 km s⁻¹ (instead of 45 km s⁻¹) to match the assumption of the RSG mass-loss models we consider later as well as most |$\dot{M}$| estimates from X-ray and radio studies. We focus on the R-band light curve since it contains the strong H α emission line. In our sample, CSM interactions have been invoked for SN 1993J (e.g. Suzuki et al. 1993; Leising et al. 1994; Suzuki & Nomoto 1995; Fransson et al. 1996), SN 2002hh (e.g. Chevalier et al. 2006), SN 2004dj (e.g. Chevalier et al. 2006, Chakraborti et al. 2012), SN 2011dh (e.g. Soderberg et al. 2012, Maeda et al. 2014), SN 2013ej (e.g. Mauerhan et al. 2017), and SN 2017eaw (e.g. Weil et al. 2020). These |$\dot{M}$| estimates are given in Table 4.

As illustrated by the theoretical models of Dessart & Hillier (2022) or the late time spectra of SN 1993J from Matheson et al. (2000a), the optical CSM emission is dominated by emission lines, so the colour of the emission should distinguish it from continuum emission like dust echoes and shock heated secondaries. Fig. 3 shows the colours of their model with 1 M_⊙ of ejecta travelling at 10⁴ km s⁻¹, which provides a reasonable match to the Matheson et al. (2000a) spectra of SN 1993J 976 d after peak, as compared to the Solar metallicity ‘PARSEC’ (Bressan et al. 2012, Marigo et al. 2013) isochrones with ages of 10^6.6–10^7.2 yr. We also show the colours of the emission from the shock heated ring in SN 1987A using the fluxes from Larsson et al. (2019) over the time period from 3200 to 11 000 d after peak. In SN 1987A, the emission peaks nearly 22 yr after the explosion, with B, V, R luminosities of roughly ∼60, ∼10, and ∼160 L_⊙, respectively, that are orders of magnitude less than the luminosity scales of the progenitor stars shown in Fig. 2. This particular colour combination, B − V and V − R, appears to be a good combination for identifying systems dominated by line emission because the R band contains the strong H α emission, and the B band contains strong [O iii] emission while the H β/[O iii] emission lines in the V band are generally weaker. This should give CSM emission blue B − V colours and red V − R colours that are difficult for less line dominated spectra to mimic, as seen in Fig. 3.

2.4 Dust echos

When light from an SN encounters interstellar dust, it can be scattered to reach us with a light travel time delay. The detailed properties of light echos are dependent on multiple parameters such as dust location, geometry, and scattering optical depth (e.g. Chevalier 1986; Liu, Bregman & Seitzer 2003; Sugerman 2003; Rest et al. 2005; Rest et al. 2011). We can roughly characterize the luminosity as

where E_rad is the energy radiated in the photometric band, t_now is the time elapsed since peak, and τ_dust is the fraction of E_rad that is scattered to the observer. This model essentially assumes that the dust is in a shell a distance ct_now from the SN that is absorbing fraction τ of the radiated energy with light travel times then spreading the observed emission over time t_now. Dust echoes have (roughly) the emission weighted mean spectrum of the SN, weighted by the smoothly varying, direction-dependent scattering opacity of the dust. For our sample, dust echoes are reported for SN 1980K (Sugerman et al. 2012; Bevan, Barlow & Milisavljevic 2017), SN 1993J (Sugerman & Crotts 2002; Liu et al. 2003), SN 2002hh (Welch et al. 2007; Andrews, Smith & Mauerhan 2015), and SN 2003gd (Sugerman 2005). For these SNe, the observed light echos are generally bluer than the colour of the SN near peak.

We again make the estimates using the R-band data, in this case because it generally has smaller uncertainties. The data are collected using the LBC/Red camera for the R-band while cycling through UBV on the LBC/Blue camera, leading to shorter exposure times for the other filters. We estimate the total radiated energy E_rad using the Morozova et al. (2015) light-curve models normalized to the peak R-band peak luminosity given in Table 1. We then simply estimate τ_dust using equation (6). The basic physics of echoes requires the 1/t_now decay unless the effective optical depth is increasing with distance from the SN. Fig. 3 shows the mean, near-peak colours of SN 1993J, 2013am, 2013ej, 2016cok, and 2017eaw after correcting for extinction. The scatter in the colours is smaller than the symbol. We would expect dust echoes to be modestly bluer than these colours.

2.5 Engine-driven emissions

Another possibility for driving late time emissions is energy injection from a pulsar. If we assume dipole spin down, the luminosity is

for moment of inertia I = 10⁴⁵ g cm², spin period P = 10P₁₀ msec, and spin down time t_s = 10⁴t₄ yr. The spin down time is related to the magnetic field B = 10¹²B₁₂ gauss by

For the Crab pulsar (P = 0.033 s, t_s = 2600 yr), L = 1.1 × 10⁵ L_⊙. However, essentially none of this energy is converted into visible radiation at late times. For the optical magnitudes of the Crab from Sandberg & Sollerman (2009), a distance of 2 kpc (Kaplan et al. 2008) and an extinction of E(B − V) ≃ 0.4 mag (Green et al. 2019), the UBVR luminosities of the Crab are <10 L_⊙. Magnetar models (e.g. Kasen & Bildsten 2010), assume that the spin down energy is fully thermalized and that the amount radiated is regulated by the balance between the expansion and diffusion times. We are considering the emission at late times in the nebular phase, where the prediction of these models is that the luminous emissions are strongly suppressed and we would expect properties more like the Crab. We can use the mean luminosity and the linear decay of the light curve to place constraints on the period and period derivative. These then provide an estimate of the magnetic field strength required to power the late time emission.

2.6 Binary companion ejecta interaction

When the shock from a SN hits a binary companion, it heats and inflates its envelope, which leads to a period of enhanced luminosity (Wheeler, Lecar & McKee 1975; Fryxell & Arnett 1981; Marietta, Burrows & Fryxell 2000; Podsiadlowski 2003; Meng, Chen & Han 2007; Hirai, Sawai & Yamada 2014; Hirai, Podsiadlowski & Yamada 2018, Ogata et al. 2021). We use the numerically calibrated analytic model of Ogata et al. (2021) to constrain the companion mass (M_c), and the ratio of its radius to the orbital separation (R_c/a). The total energy injected into the companion from the outburst is

where |$p \simeq 8{\!-\!}10~{{\ \rm per\ cent}}$| is the energy injection efficiency, E_expl = 10⁵¹erg is the explosion energy, and

is the fractional solid angle subtended by the companion. The maximum luminosity of the companion is

where κ_fit is a fitting function from Ogata et al. (2021) for the average opacity at the bottom of the companion’s convective layer, G is the gravitational constant, c is the speed of light, and M_c is the companion mass. Thus, the maximum luminosity is determined by the companion mass. The time-scale over which the companion maintains this luminosity is

where α = 0.18. Since significant heating of a main sequence (MS) binary companion requires an orbit smaller than the RSG progenitors of Type II SNe, we consider this case only for the Type Ibc SN 2012fh.

3 DISCUSSION

Table 1 summarizes the SN we consider, and the Appendix  A has notes on the individual sources. After dropping SN 2009hd, there are 12 sources in total; 9 IIL/P, 2 IIb, and 1 Ib/c, of which 4, 1, and 1 have pre-SN LBT observations. Figs 4–10 show the LBT light curves of the SN grouped by type and with the epoch of explosion marked if it occurred after the start of the LBT project. Tables 5 and 6 give the results of the linear fits to the late time light curves of each SNe and its grid of comparison points where the fits extend from the epoch marked in the figure through the end of the light curve. We include fits for SN 2009hd even though we will not discuss it further.

The luminosities of the SNe in these light curves are all relative to the luminosity of the source in the reference image. For the SNe with pre-SN LBT data, the observed light curve is thus any present-day emission minus the luminosity of the progenitor. For the SN with only late time LBT data, it is the present-day emission minus the mean luminosity of the SN is the reference image flux. The aperture photometry luminosity of the SNe on the reference image is reported in Tables 5 and 6, but they should be generally regarded as limits due to the effects of crowding. Figs 4–10 include the HST luminosities from Table 2. The offset between the differential luminosity light curve and the HST luminosity is the luminosity in the reference image. Where the HST data are available, the actual light curve is simply the differential light curve shifted upwards in luminosity to pass through the HST data. What we see in general is that the necessary shifts are on the same scale as the changes in luminosity. The exception is SN 1980K, where the HST luminosity is roughly an order of magnitude larger. SN 1980K is also the only case where we can clearly see the SN in the LBT data at these late times.

For our standard analyses, we use the image subtraction light curve if the linear fit estimate of the luminosity of the last epoch is positive. If the linear fit estimate of the luminosity of the last epoch is negative, we rescale the final luminosity to zero and use |$L_{\rm SN}^{\prime } = L_{\rm SN}-\beta _{\rm SN}(t_{\rm last}-t_0)$| where t_last the time of the last epoch. All of these latter systems are SNe without pre-explosion LBT images with the exception of SN 2011dh. This still means that our estimates of the ⁵⁶Ni mass, wind |$\dot{M}$|⁠, and dust optical depth τ will be underestimates, so we discuss the consequences of further increases in the luminosities due to the (remaining) uncorrected flux in the reference image. We will also provide estimates based on the slope β_SN. For radioactive decay we can make a direct estimate of the ⁵⁶Ni mass by matching the slope of the LBT light curve to that of SN 1987A. For the other two emission mechanisms we make a ‘standardized’ estimate using the luminosity β_SNΔt, the drop in luminosity over Δt = 10 yr. It is standardized in the sense that the time baselines for the most recent SN are much shorter than the full time baseline of the LBT data.

Of the nine Type IIP/L systems, eight show continued late time emission, particularly in the V and R bands. One of the Type IIb SN, SN 1993J, shows continued fading. Only the Type IIP SN 2005cs, the Type IIb SN 2011dh and the Type Ibc SN 2012fh do not show significant evidence for continued emission. For the systems with pre-LBT images, all but SN 2011dh and possibly SN 2012fh have a present-day luminosity in excess of the progenitor. In the next sections, we first discuss the progenitor of SN 2011dh, and then the potential sources of the continued emission for the other systems.

3.1 Progenitor constraints

As noted earlier, only SN 2011dh, a Type IIb SNe in NGC 5194, shows the originally expected fading. The SN was fainter than the progenitor within ∼600 d, as expected from a light curve dominated by radioactive decay. In a reverse time average of the subtracted light curves, we begin to see an upward trend at times earlier than +1400 d post explosion, so we adopt the reverse time average of the light curves after day  +2000 as the estimate of the progenitor flux. The principal limitation on our photometry is the calibration because the depths of the LBT data and the SDSS survey are poorly matched (particularly since SDSS only catalogues stars around the periphery of the galaxy). A shallower LBT image is needed to improve the calibration. None the less, the LBT magnitudes in Table 3 are very similar to those found by Van Dyk et al. (2011) and Maund et al. (2011) analysing the same HST F336W, F435W, F555W, F658N, and F814W images, roughly corresponding to the U, B, V, R, and I bands. That they are so similar despite the need for a better calibration of the LBT data is a demonstration that it should be straight forward to obtain very good photometry of SNe progenitors from the LBT data once they have faded. Fig. 11 provides an extinction corrected SED.

We fit the data assuming minimum photometric errors of 10 per cent, a fixed distance of 7.1 ± 1.2 Mpc to match Van Dyk et al. (2011) and Maund et al. (2011) and varying only the luminosity, temperature, and foreground extinction, with an extinction prior of E(B − V) = 0.05 ± 0.05 mag. This includes both the Galactic extinction and any contribution from the host. For the LBT data, the best fit has χ² = 21.0 for two degrees of freedom, with L_* = 10^{5.00 ± 0.04} L_⊙, T_* = 5294 ± 154 K and E(B − V) = 0.02 ± 0.02 mag. The poor fit is driven by an inability to find a model which is roughly flat in νL_ν for the B, V, and R bands and then drops rapidly enough to fit the U band.

We did not fit the HST narrow band H α data (F658N), and for the Van Dyk et al. (2011) photometry the best fit has χ² = 8.2 again for two degrees of freedom with L_* = 10^{4.98 ± 0.04} L_⊙, T_* = 6312 ± 357 K, and E(B − V) = 0.10 ± 0.04 mag. In their analysis, Van Dyk et al. (2011) adopted a temperature of T_* = 6000 K and a luminosity of L = 10^4.99 L_⊙ and a fixed (Galactic) extinction of E(B − V) = 0.04 mag. Finally, for the Maund et al. (2011) values we find χ² = 15.2, L_* = 10^{5.00 ± 0.03} L_⊙, T_* = 6317 ± 353 K and E(B − V) = 0.12 ± 0.02 mag. Maund et al. (2011) found L_* = 10^{4.92 ± 0.20} L_⊙ and T_* = 6000 ± 280 K. Both of the HST fits also struggle with the rapid drop down to the F336W point, but the impact on the χ² is smaller because of the larger uncertainty for HST F336W compared to the LBT U band. Using circumstellar dust instead of foreground extinction did not solve the problem of the poor fits to these data.

If we combine the LBT and Maund et al. (2011) data (again excluding the narrow band filter), we get χ² = 39.6 for 7 degrees of freedom, L_* = 10^{4.97 ± 0.04} L_⊙, T_* = 5601 ± 226 K, and E(B − V) = 0.04 ± 0.04. If we drop the U/F336W points, we find χ² = 2.1 for 5 degrees of freedom with L_* = 10^{4.99 ± 0.05} L_⊙, T_* = 6629 ± 324 K, and E(B − V) = 0.08 ± 0.04 mag. Fig. 11 shows these combined models. For the adopted distance estimate, there is an additional uncertainty in the luminosity of 0.15 dex from the uncertainties in the distance. This is likely already included in the Maund et al. (2011) luminosity uncertainty. All of these models agree on a progenitor luminosity very close to 10⁵ L_⊙ and Fig. 2 shows how these luminosity estimates compare to the progenitor models. The close match to the HST data emphasizes the point in Fig. 1 that faded progenitors should generally be trivially visible in the LBT data.

Maund (2019) argues for a late-time emission plateau at 2000–2500 d of roughly 10⁴ L_⊙ fading by approximately (− 330 ± 110) L_⊙ yr⁻¹, and argues for a dust echo since there are no indications of the presence of emission lines in their narrow-band images. Such a low level contribution would have little effect on this analysis since the estimated progenitor luminosity is an order of magnitude higher. Formally, we find somewhat steeper decay rates in Table 6.

3.2 Radioactive decay

As discussed in Section 2.2, we can estimate the required nickel mass by scaling the V band light curve of SN 1987A either to the observed light curve or the slope of the observed light curve. To set the basic scale, Table 4 gives the nickel mass corresponding to a V-band luminosity of 10³ L_⊙ at time t₀. The SN 1987A V-band light curve from Seitenzahl et al. (2014) extends to 4300 d. If t₀ < 4300 d, we match the LBT and SN 1987A luminosities at t₀. If t₀ > 4300 d but the time period used for the linear fit extends to times <4300 d (SN 2002hh, 2003gd), we use the linear fit estimate of the luminosity of the SN at 4300 d. If there is no overlap (SN 1980K, 1993J), we use the linear fit estimate for the earliest time included in the fits and the luminosity of 87A at 4300 d. This last case should underestimate the required mass. These mass estimates, given in Table 4, are impossibly large. Adding any additional luminosity (i.e. some multiple of the mass corresponding to 10³ L_⊙) will only exacerbate the problem.

We can avoid the problem of the reference image flux by instead fitting the slope of the light curve, β_SN at time t₀. These results are also given in Table 4. For systems where there is no overlap with the SN 1987A light curves, we use the slope at the end of its light curve, which will again lead to an underestimate of the mass. We again find unreasonably high ⁵⁶Ni masses for the SNe with well-measured slopes. That late time emission cannot be powered by radioactivity is not surprising, but this analysis emphasizes the degree to which it is infeasible.

3.3 CSM interactions

Table 4 gives two estimates of |$\dot{M}$|⁠, both under the assumption that v_s = 4000 km s⁻¹, v_w = 10 km s⁻¹, and ϵ = 0.1 (equation 5). The first is based on the luminosity at time t₀ after shifting L_SN upwards to make the linearly interpolated luminosity at the last epoch to be zero if it is negative. The second is simply to use the change in luminosity β_SNΔt over Δt = 10 yr. We can also visualize the required mass-loss rates as shown in Fig. 12. The resulting estimates generally range from 10⁻⁷ to 10⁻⁵ M_⊙ yr⁻¹. Adding a luminosity of 10³, 3 × 10³, 10⁴, or 3 × 10⁴ L_⊙ for the flux in the reference image, corresponds to increasing the mass-loss rates by |$\dot{M} = 10^{-7.72}$|⁠, 10^−7.24, 10^−6.72, and 10^−6.24 M_⊙ yr⁻¹. We illustrate the cases of adding 3 × 10³ and 10⁴ L_⊙ in Fig. 12. If we increase the reference frame flux much more, we would start trivially seeing a source – for example, SN 1980K is a clear source in the R-band reference image with a luminosity of 1.7 × 10⁴ L_⊙, quite consistent with the 1.3 × 10⁴ L_⊙ found with HST just before the start of the LBT observations (see Table 2). Adding this to our standard estimate just using the difference light curve only increases |$\dot{M}$| from 10^−6.65 to 10^−6.26 M_⊙ yr⁻¹ – a quantitatively important change but one without any importance for our qualitative conclusions. Table 4 also reports the mass-loss rate corresponding to the estimated drop in luminosity β_SNΔt over Δt = 10 yr, which gives similar results. There is considerable freedom in the absolute scale of |$\dot{M}$| from choosing the parameters, particularly the shock velocity v_s and the radiative efficiency ϵ in equation (5).

Fig. 3 shows the V − R and B − V colours of SN 1987A (Larsson et al. 2019) and the model spectrum of Dessart & Hillier (2022), which is a good fit to late time spectra of SN 1993J. We chose these colours because the B and R bands contain strong hydrogen Balmer emission lines while the V band only has weaker lines in the Dessart & Hillier (2022) model spectra. These leads to red V − R colours and blue B − V colours that are difficult for a continuum emission source to mimic, as illustrated by the colours of the PARSEC stellar isochrones.

The inferred mass-loss rates are modest, ranging from ∼10⁻⁷ to ∼10⁻⁵ M_⊙ yr⁻¹. For comparison, Table 4 gives mass-loss estimates for these SNe from the literature. These are generally post-explosion estimates based on radio and X-ray observations, although there are exceptions (see the Appendix). None of the estimates are based on the late time optical emission, although late time CSM emission is well established for SN 1980K and 1993J (Milisavljevic et al. 2012), and SN 2004dj (Chakraborti et al. 2012; Nayana, Chandra & Ray 2018). Recently, Van Dyk et al. (2022) also found that the progenitors of SN 2013ej and SN 2017eaw are still brighter than their progenitors in the V band (as had already been noted by Neustadt et al. 2021), but not in the I band, in 2021 and 2020, respectively, and hypothesize that this may be due to ongoing CSM interactions. In both the Matheson et al. (2000a) spectra of SN 1993J and the theoretical spectra of Dessart & Hillier (2022), there are strong lines in the UBVR bands but not in the I band, consistent with these observations.

3.4 Dust echoes

Using the same procedures as in Section 3.3 and equation (6) yields the optical depth estimates given in Table 4. The required optical depths are not very large. We can recast equation (6) using E_rad = L_peakt_peak, where L_peak ∼ 10⁸ L_⊙ is the peak luminosity and t_peak ≃ 100 d is the duration of the plateau phase, so that τ ∼ (L_SN/L_peak)(t₀/t_peak). The luminosity has dropped by L_SN/L_peak ∼ 10⁴ or more, while the elapsed time is only t₀ ∼ 30t_peak, leading to required optical depths of τ ∼ 10^−3.5–10^−2.5. However, as illustrated in Fig. 13, the shapes of the light curves generally show phases with the optical depth increasing with time. We also see in Fig. 3 that the colours of the late time emission are different from the expectations for dust echoes. The dust echo colours should, essentially, be the colours of the SN shifted roughly parallel to the isochrones while the observed colours are generally shifted perpendicular to the isochrones. These facts appear to rule out dust echoes as a general explanation of the late time emission.

At first glance, these low required optical depths might seem to require significant dust echoes from essentially all the SNe because they are much lower than the extinction estimates in Table 1 and τ_R ≃ 2.3E(B − V). The missing element is that the dust echo at any given time is being produced by a very thin region of thickness ∼ct_peak ∼ 0.1 pc corresponding to the spatial extent of the luminosity transient. If the dust is uniformly distributed over a line-of-sight distance of 0.1–1 kpc, the optical depth scale relevant to producing an echo is fraction 10⁻³ to 10⁻⁴ of the total along the line of sight. Concentrating the dust in ‘sheets’ helps, but more at early times when the physical region producing the echo is small. This discussion assumes unresolved echoes where the surface brightness of the echo is not relevant to its detection.

The exception is SN 2002hh which is both heavily obscured (Table 1) and known to have a strong dust echo (e.g. Barlow et al. 2005; Pozzo et al. 2006; Welch et al. 2007; Otsuka et al. 2012). With the large corrections for extinction, SN 2002hh has the greatest late-time luminosity of any of the 12 systems by almost two orders of magnitude. While it is fading in the R band, and evolving little in the B and V bands, the U-band luminosity is clearly increasing. It is the only such system.

3.5 Pulsar engine-driven emission

If we use the engine model from equations (7) and 8 in Section 2.5 to fit the R-band luminosity and linear luminosity slope from Tables 5 and 6, we can estimate the required P and |$\dot{P}$|⁠. In Fig. 14, we show the results for SN‘2004dj, SN 2013am, SN 2013ej, SN 2016cok, and SN 2017eaw, along with known pulsars from Manchester et al. (2005). We can only make these estimates for the SN where the luminosity is decaying, since a positive slope would require the engine to be spinning up rather than down. We made these estimates assuming that all of the spin-down energy appears in the R band. If we assume it is only fraction ϵ of the spin-down energy, then the estimates move to longer periods, faster spin-down rates and higher implied magnetic fields since P ∝ ϵ², |$\dot{P} \propto \epsilon ^5$|⁠, and B ∝ ϵ^7/2. If we add 3 × 10³ or 10⁴ L_⊙ to the R-band luminosity for the flux in the reference image, the estimates shift to shorter periods, slower spin down rates, and weaker magnetic fields, but still require high magnetic fields.

If all the spin-down energy were radiated at R band, they would have properties roughly consistent with young magnetars and very different from the Crab. However, with the more realistic assumption that only a small fraction of the spin-down energy is radiated at R band, the estimates quickly shift to requiring magnetic field strengths that are unreasonable. Where there are X-ray observations discussed for the individual SNe, none are reported as originating from a pulsar wind nebula. We conclude that the late time emission is very unlikely to be due to a central engine.

3.6 Binary shock interaction

We only consider the Type Ib/c SN 2012fh for this case because the progenitors of all the other SNe should be giants, where it is impossible to significantly heat a main-sequence companion as they subtend such a small solid angle. The companion star to a stripped SN should generally be a cooler star and dominate the optical emission even before being heated, so we assume that we can neglect the effect of the progenitor luminosity on the subtracted light curves. Under these assumptions, we estimate that a companion can have increased in luminosity by no more than L_max ≃ 10⁵ L_⊙ (equation 11). Based on Fig. 10, SN 2012fh faded within ∼440 d after peak luminosity, which we use as our limit on the inflation time-scale τ_infl (equation 12). Fig. 15 translates these limits into the allowed companion mass M_c and radius relative to the binary semimajor axis R_c/a.

Using solar metallicity ‘PARSEC’ isochrone with log₁₀(Age)  = 6.7, for which no M_c < 20 M_⊙ companions have evolved, we also show curves with constant semimajor axes of a = 50, 100, and 150 R_⊙ given the stellar radii R_c of the models. At least for these models, the binary would have to be wider separation than 50–100R_⊙ at the time of the explosion. If the progenitor of SN 2012fh was stripped through binary interactions, then the mass–radius relations of the single-star ‘PARSEC’ isochrones will be incorrect at some level.

4 CONCLUSIONS

Motivated by the discovery that SN 2013am and SN 2013ej were still optically brighter than their progenitors a decade after explosion (Neustadt et al. 2021), we systematically investigated the evolution over the last 14 yr of the 12 ccSNe that occurred in the LBT search for failed SNe (Kochanek et al. 2008a; Gerke et al. 2015; Adams et al. 2017a, b; Basinger et al. 2021; Neustadt et al. 2021) galaxies from 1980 onwards. We used difference imaging techniques to look for continued evolution in their UBVR luminosities. Difference imaging has the advantage of largely eliminating the problem of crowding in these ground based data, but some quantitative conclusions depend on the difficult to estimate flux of the targets in the reference image. As part of the survey, we analysed the SED of the progenitor of SN 2011dh from the LBT data and find results consistent with the results from pre-SN HST data (Maund et al. 2011; Van Dyk et al. 2011). This analysis also demonstrates that the LBT easily has the sensitivity to probe emissions at the level of |$\sim 10~{{\ \rm per\ cent}}$| of the progenitor luminosity or better (see Fig. 11). We also considered shock heated binary companions to the Type Ibc SN 2012fh and set limits on the companion mass M_c and the ratio of the companion radius to the semimajor axis R_c/a (see Fig. 15).

We focus these conclusions on the interesting finding that of the 11 Type II SNe in the sample, only two (the Type IIP SN 2005cs and the Type IIb SN 2011dh), do not show continued, evolving SN emission 5–42 yr post-explosion. This includes the two oldest systems studied, the Type IIL SN 1980K and the Type IIb SN 1993J. The continued emission is coincidentally on the scale of the luminosity of the progenitors (∼10⁴ L_⊙). For the six of these SNe that occurred after the LBT survey started, we know this is true of five (the exception is SN 2011dh) because we constructed the reference image from pre-SN images so that the luminosities in the subtracted light curves are relative to the luminosity of the progenitor. Apparently, SN like SN 1987A (e.g. Woosley 1988; Fransson et al. 2007; Seitenzahl et al. 2014) or SN 2011dh which rapidly fade to luminosities well below those of their progenitors are the exceptions, not the norm.

For completeness, we considered radioactivity, CSM interactions, dust echoes and magnetar/engines as possible drivers. Radioactivity requires impossible masses and dust echoes generally require too much dust. Neutron star spin-down can produce the necessary luminosity but only for unreasonable magnetic fields. Theoretically, central engines are not believed to produce significant optical emission at these phases (e.g. Kasen & Bildsten 2010), which is certainly true of the Crab pulsar today. The time evolution of the luminosity and the colours of the emission are also generally inconsistent with dust echoes. Not surprisingly, since CSM emission has been observed at earlier phases for many of these sources (see Table 4 and Appendix  A), the only logical possibility is continuing CSM emission aside from SN 2022hh where it is probably due to a very strong, continuing dust echo.

Mass-loss is a crucial component of RSG evolution (e.g. Reimers 1975; de Jager, Nieuwenhuijzen & van der Hucht 1988; Nieuwenhuijzen & de Jager 1990; van Loon et al. 2005; Mauron & Josselin 2011; Beasor, Davies & Smith 2021). If we use the progenitor luminosity scaling from equation (2), we can express the Nieuwenhuijzen & de Jager (1990) mass-loss rate as

and the van Loon et al. (2005) mass-loss rate as

With the temperature fixed at 3500 K, the mass-loss rates for 8 M_⊙ < M < 20 M_⊙ span |$-5.9 \lt \log \dot{M} \lt -4.8$| and |$-5.6 \lt \log \dot{M} \lt -4.8$|⁠. These estimates are somewhat higher than our estimates in Table 4. While there are significant quantitative uncertainties in these estimates, particularly through the assumed shock velocity (v_s = 4000 km s⁻¹) and the radiative efficiency (ϵ = 0.1) it would be difficult to significantly raise them given the lack of large numbers of X-ray bright Type II SNe (see Dwarkadas 2014). There are also arguments that these models for RSG mass-loss rates are too high (e.g. Beasor et al. 2021).

None the less, if we use these mass-loss rates to predict the CSM luminosity (equation 5) relative to the progenitor luminosity (equation 2), we find that

and

for the two mass-loss prescriptions. Since the luminosity of an RSG progenitor in a particular band is less than L_prog, it is clear that CSM emission from normal RSG winds can easily exceed the band luminosity of the progenitor provided the efficiency of converting the shock luminosity into optical emission is ϵ ≳ 0.01. The interesting question then becomes the duration of the emission, since we generically find that the emissions are fading.

One possibility, requiring theoretical study, is the physics of the efficiency ϵ. While the shock luminosity for a fixed shock velocity is constant for a 1/r² wind density, it would still be fairly natural for ϵ to evolve since most emission mechanisms scale as the square of the density rather than being proportional to the density like the shock luminosity. The fading could also be due to having wind density profiles that are generically steeper than the 1/r² density profile of a constant |$\dot{M}$| wind (see e.g. Dwarkadas & Gruszko 2012).

The missing element for making these conclusions more quantitative is to obtain absolute calibrations. Waiting for the SN to fade to black seems to require challenging levels of patience given that SN 1980K seems to still be going strong. On the other hand, SN 1993J does appear to be close to fading away. Fortunately, single orbit HST observations can provide the necessary absolute flux levels for two to three filters depending on the desired level of control over systematic problems (cosmic rays, bad pixels, etc.). For two filters, the preferred choices are probably R and B given the properties of observed (Matheson et al. 2000a) and theoretical (Dessart & Hillier 2022) spectra of CSM interactions, with adding V if a third filter is included.

The SN shock also eventually reaches the edge of the wind. In the Weaver (1976) self-similar solution for a wind expanding into a constant density medium, the contact discontinuity between the two fluids lies at

where n is the ambient density and t is the time since the start of the wind. Except for time, the parameter dependence is weak, so the time scale for the SN shock to reach the contact discontinuity is unfortunately long (1300 yr for v_s = 4000 km s⁻¹). Moreover, the RSG wind may be expanding into a lower density bubble formed by a faster main sequence wind, leading to a still more distant contact discontinuity (e.g. Dwarkadas 2005). The wind termination shock can lie at a significantly smaller radius than the contact discontinuity, but centuries are not much of a gain over millennia.

Even if it seems unlikely that we will observe these boundaries of the RSG wind, it is still of interest to both continue to observe the evolution of these SNe and to extend the study to the small numbers of still older historical SNe in the LBT galaxies. Moreover, the estimated time scale to reach the edge of the wind is just an estimate—in SN 1987A, the shock started to interact with the wind boundary after only 20 yr (e.g. Larsson et al. 2019). There is also the possibility that the progenitor had a transient high mass-loss period 100s to 1000s of years pre-explosion. Such events have been inferred from SN light curves (e.g. Pastorello et al. 2007; Ofek et al. 2013; Milisavljevic et al. 2015; Margutti et al. 2017; Mauerhan et al. 2018; Chandra et al. 2020). For example, SN 2002hh is postulated to have a massive (∼0.1 M_⊙) dusty shell at a distance of 0.03–0.3 pc (e.g. Barlow et al. 2005). Like Andrews et al. (2015), we see no evidence for interactions with the postulated shell, which leads to a minimum distance to the shell of 0.08(v_s/4000 km s⁻¹) pc that begins to strongly constrain these models. And, since they are all fading, it will eventually be possible to do the progenitor photometry of those SNe with pre-explosion LBT imaging – just after far more time than expected.

ACKNOWLEDGEMENTS

We thank Krzysztof Stanek for his assistance with LBT calibrations and troubleshooting. We also thank Todd Thompson for valuable discussions. JN and CSK are supported by NSF grants AST-1814440 and AST-1908570. The LBT is an international collaboration among institutions in the United States, Italy, and Germany. LBT Corporation partners are: The University of Arizona on behalf of the Arizona university system; Istituto Nazionale di Astrofisica, Italy; LBT Beteiligungsgesellschaft, Germany, representing the Max-Planck Society, the Astrophysical Institute Potsdam, and Heidelberg University; The Ohio State University, and The Research Corporation, on behalf of The University of Notre Dame, University of Minnesota and University of Virginia.

DATA AVAILABILITY

The data underlying this article will be shared on reasonable request to the corresponding author.

References

APPENDIX A: INDIVIDUAL SNE

Here, we summarize the individual SNe. Table 1 lists the SN and gives the adopted distances, Galactic and host extinctions. Table 2 lists HST observations that can be used to provide absolute calibrations of the LBT light curves, and Table 4 gives previous estimates of mass-loss rates |$\dot{M}$|⁠.

SN 1980K

The Type II-L SN 1980K was discovered in NGC 6946 (Wild & Barbon 1980) on 1980 October 28 reaching a maximum of V = 11.4 mag on November 3 1980 (Barbon et al. 1982) and has been extensively studied far into the nebular phase (e.g. Chevalier 1982b; Dwek 1983; Montes et al. 1998; Fesen & Becker 1990; Leibundgut et al. 1991; Chevalier & Fransson 1994; Milisavljevic et al. 2012). Based on the radio emission, Chevalier (1982b) estimated a rough mass loss rate of |$\dot{M} \sim 10^{-5}M_\odot$| year⁻¹ for v_w = 10 km s⁻¹, and Montes et al. (1998) argue for a break to a region with a lower wind density due to a sudden drop in the radio fluxes around 1995. Dwek (1983) argued for a dust echo from a shell formed at the wind/ISM boundary with a detailed model in Sugerman et al. (2012) placing the shell ∼15 pc from the SN. They find weak evidence for optical fading between 2005 and 2008 and strong evidence for an infrared excess. The source they identify as the SN is clearly visible in the first LBT images, taken 4 months after the Sugerman et al. (2012) HST observations in January 2008. The LBT observations of NGC 6946 now span 13 yr with approximately 57 epochs per filter.

SN 1993J

SN 1993J in M81 is the prototype of the Type IIb SNe class (Schmidt et al. 1993; Suzuki et al. 1993; Filippenko, Matheson & Ho 1993; Nomoto et al. 1993; Filippenko, Matheson & Barth 1994). It was discovered on 1993 March 28 (Ripero et al. 1993). Using ground based observations obtained prior to the SN, the progenitor of 1993J was identified as K0 RSG (Aldering, Humphreys & Richmond 1994), and its eventual disappearance has been observed (Maund & Smartt 2009). The progenitor was not variable by more than 0.2 mag (Cohen, Darling & Porter 1995). There is extensive evidence for CSM emission from late time spectra (e.g. Matheson et al. 2000b, c), X-ray emission (e.g. Suzuki et al. 1993; Leising et al. 1994; Suzuki & Nomoto 1995; Fransson et al. 1996; Kundu et al. 2019), and radio emission (e.g. van Dyk et al. 1994; Kundu et al. 2019). VLBI observations monitored the expansion of the remnant for almost a decade (Bietenholz, Bartel & Rupen 2003), tracking the expansion to 19 000 au (0.1 pc) and implying average expansion velocities of ∼10⁴ km s⁻¹. The Kundu et al. (2019) models assume a mass-loss rate of |$\dot{M}=4 \times 10^{-5}\, {\rm M}_\odot$| yr⁻¹ before 3000 d. The LBT observations of M81 now span ∼14 yr with approximately 32 epochs per filter. For SN 1993J, the HST observations of Fox et al. (2014) provide an absolute calibration of the V and R band light curves (Table 2).

SN 2002hh

SN 2002hh is a Type IIP SN first discovered on 2002 October 31.1 (Li 2002) by LOSS in NGC 6946. Barlow et al. (2005) argue for a massive (∼0.1 M_⊙) dusty shell at a distance of ∼0.04 pc, and the dust echoes are discussed further in Pozzo et al. (2006), Welch et al. (2007), and Otsuka et al. (2012). Andrews et al. (2015) spectroscopically searched for any evidence of shock interactions with such a dusty shell in 2013, 10.5 yr after explosion, finding none. The spectra appear similar to the SN near peak, suggesting that the emission is dominated by a dust echo. We can use the HST observations from Otsuka et al. (2012) to normalize some of the LBT bands (Table 2). This galaxy has been observed by the LBT since mid-2008 with a total of 57 epochs in each filter.

SN 2003gd

SN 2003gd is a Type IIP SN discovered in NGC 628 (M 74) on 2003 June 12.82 (Evans & McNaught 2003) with a general study in Hendry et al. (2005). Archival HST and Gemini data identified its progenitor, an RSG star with an estimated mass of |$8^{+4}_{-2}M_\odot$| (Smartt et al. 2004, Hendry et al. 2005) and the progenitor is observed to have disappeared (Maund & Smartt 2009, Maund, Reilly & Mattila 2014). Two years after explosion it had a faint, resolved (by HST) dust echo with |$\sim 1 {{\ \rm per\ cent}}$| of the optical flux of the SN (Sugerman 2005). This galaxy has been observed for a total of 32 epochs.

SN 2004dj

The Type IIP SN 2004dj was discovered on 2004 July 31.76 in NGC 2403 (Nakano et al. 2004) with early observations in Vinkó et al. (2006), Tsvetkov et al. (2008), and Zhang et al. 2009 with extensions to ∼1000 d in Vinkó et al. (2009). Van Dyk, Li & Filippenko (2003) and Maund & Smartt (2005) identify two progenitor candidates, and Maíz-Apellániz et al. (2004) and Vinkó et al. (2009) estimate a mass of ∼15 M_⊙ based on the age of the associated star cluster, Sandage 96. There is evidence of CSM emission from early optical spectroscopy, late-time X-ray, and radio observations (Chugai, Chevalier & Utrobin 2007; Chakraborti et al. 2012; Nayana et al. 2018). Mid-IR Spitzer observations, late time optical photometry, and HST data from days ∼106 to −1393 after explosion provide evidence for dust formation in the ejecta (Szalai et al. 2011, Meikle et al. 2011). There are 52 LBT epochs for NGC 2403.

SN 2005cs

SN 2005cs is a Type IIP SN in NGC 5194 (M 51) that was discovered on 2005 July 28.9 (Kloehr et al. 2005). There are general observations of SN 2005cs in Pastorello et al. (2006), Brown et al. (2007) and Pastorello et al. (2009). Based on archival HST and Gemini images, the progenitor was identified as a K3 RSG, with an approximate mass of (8 ± 2)M_⊙ (Maund, Smartt & Danziger 2005; Li et al. 2006). Early UV, optical, and a non-detection in the X-rays provide an upper limit of |$\dot{M} \lesssim 10^{-5}\,{\rm M}_\odot$| yr⁻¹ on the pre-SN mass-loss rate (Brown et al. 2007). There are 31 LBT epochs for NGC 5194.

SN 2009hd

SN 2009hd occurred in NGC 3627 (M66) on 2009 July 2.69 (Monard 2009) and was classified as a Type IIL (Elias-Rosa et al. 2011). This SN is located in a region with very high host extinction (see Table 1). Elias-Rosa et al. (2011) reports an upper limit on the luminosity of the progenitor of L_bol ≲ 10^5.04 L_⊙ which we can use as an upper limit on the contributions of the progenitor to the reference image. The image subtraction residuals near the SN are unusually large, so while we have 27 LBT epochs, four of which were taken pre-outburst, the quantitative results are not very useful, particularly when combined with the high extinction.

SN 2011dh

SN 2011dh, was discovered in NGC 5914 (M51) on 2011 May 31.89 (Griga et al. 2011; Silverman, Filippenko & Cenko 2011) and classified as a Type IIb SN. There are general studies of SN 2011dh in Arcavi et al. (2011), Sahu, Anupama & Chakradhari (2013), Ergon et al. (2014), and Ergon et al. (2015). The progenitor was yellow super-giant (Maund et al. 2011; Van Dyk et al. 2011) with properties discussed in Section 3.1 and it appears to have been slightly variable (Szczygieł et al. 2012). Folatelli et al. (2014) argue that blue star detected in deep Near-UV HST images is a binary companion to the progenitor, but this is disputed by Maund et al. (2015). X-ray (Maeda et al. 2014) and radio (Soderberg et al. 2012) observations found evidence for CSM interactions, with a mass-loss rate of order |$\dot{M} \sim 3 \times 10^{-6}\, {\rm M}_\odot$| yr⁻¹ for v_w = 20 km s⁻¹. VLBI observations monitored the expansion of the remnant to a radius of (7.4 ± 0.3) × 10¹⁶ cm implying an average expansion velocity of ∼19 000 km s⁻¹ (de Witt et al. 2016). We have five observations before the SN and 26 afterwards, for a total of 31 LBT epochs.

SN 2012fh

This system was discovered on 2012 October 18.856 (Nakano et al. 2012) in NGC 3344, and was classified as a Type Ic by Tomasella et al. (2012) and Takaki et al. (2012). As in Johnson et al. (2017), we will use a more generic Type Ibc classification because the spectra were collected long after peak. Photometry for 2012fh is available in Shivvers et al. (2019) and Zheng et al. (2022). Johnson et al. (2017) place limits on the luminosity of the progenitor, finding that it likely had to be a lower mass star stripped by binary interactions. We have 10 LBT epochs prior to the SN and 16 afterwards for a total of 26 epochs This SN is the only candidate for which a binary companion analysis is carried out.

SN 2013am

This Type II-P SN was discovered on 2013 March 21.638 in NGC 3623 by Nakano et al. (2013) and classified as a young Type II SN (Benetti et al. 2013). There are a general studies of it in Zhang et al. (2014) and Tomasella et al. (2018), additional light curve data in Galbany et al. (2016) and de Jaeger et al. (2019) and additional spectra in Silverman et al. (2017). The spectrum in Silverman et al. (2017) at 257 d appears to be developing the box-shaped line profiles typical of CSM interactions. We have 8 pre-SN epochs of LBT data, one just 5 d before the discovery, and 29 epochs in total.

SN 2013ej

This SN was discovered on 2013 July 24.83 by Lee et al. (2013) in NGC 628 (M74) and classified as a young Type II (Kim et al. 2013). It is generally discussed in Valenti et al. (2014), Huang et al. (2015), Huang et al. (2015), Dhungana et al. (2016), and Yuan et al. (2016) where the latter argues that it should be classified as a Type IIL rather than a Type IIP. Fraser et al. (2014) argue that a red source in archival HST images might be the progenitor. Early optical observations (Bose et al. 2015), X-ray observations (Chakraborti et al. 2016), and spectropolarimetry (Mauerhan et al. 2017) all support the presence of CSM interactions. We have 13 pre-SN epochs of LBT data and 35 in total.

ASASSN-16fq/SN 2016cok

ASAS-SN (Shappee et al. 2014; Kochanek et al. 2017a) discovered the Type II-P SN ASASSN-16fq (SN 2016cok) in NGC 3627 (M66) on 2016 May 28.30 Bock et al. (2016). There is photometry of this SN in de Jaeger et al. (2019). Kochanek et al. (2017b) estimate that the progenitor was a 8−12M_⊙ RSG with a luminosity of 10^4.5–10^4.9 and that it had no significant pre-SN variability. The SN was not detected in 23 ksec of Swift observations, but there is no discussion of the implications (Bock et al. 2016). We have 16 LBT epochs of pre-SN LBT data, and 26 in total.

SN 2017eaw

SN 2017eaw is a Type II-P discovered on 2017 May 14.24 (Dong & Stanek 2017; Wiggins 2017) in NGC 6946. There are general studies of the SN and its progenitor in Kilpatrick & Foley (2018), Van Dyk et al. (2019), and Szalai et al. (2019). There is additional photometry in Buta & Keel (2019) and evidence for dust formation after the explosion (Rho et al. 2018, Tinyanont et al. 2019). Late time spectra show clear evidence of CSM interactions (Weil et al. 2020). The progenitor luminosity is roughly 10^4.9 L_⊙ with a dusty wind or shell (Kilpatrick & Foley 2018). The 40 pre-SN LBT observations have no evidence for optical variability from the progenitor over its last roughly decade (Johnson, Kochanek & Adams 2018b), which essentially rules out suggestions for late pre-SN outbursts (Kilpatrick & Foley 2018, Rui et al. 2019). We have 40 pre-SN LBT epochs, and a total of 54.