Spitzer's Last Look at Extragalactic Explosions: Long-term Evolution of Interacting Supernovae

During our Spitzer/LASTCHANCE program (PID 14098; PI O. D. Fox), we observed 31 targets that met one of the following criteria: (i) high-profile, well-understood interacting and/or dust-forming SNe that were previously detected by Spitzer/IRAC and required another epoch to monitor their mid-IR evolution (based on Szalai et al. 2019a, and references therein), or (ii) young (≲1 yr) and relatively nearby (≲70 Mpc) SNe with known dense CSM (from optical spectra) or with the potential for dust formation in their ejecta (Type II-P SNe).

Our observations were designed to attain a signal-to-noise ratio (S/N) of 100 for a medium background according to the Spitzer/IRAC Sensitivity Performance Estimation Tool (SENS-PET). We aimed for this S/N so that any photometric uncertainties would be dominated by systematic effects associated with the background (host-galaxy) subtraction rather than insufficient integration time. This S/N also provides a buffer for detecting SNe that may have faded more quickly than predicted. Our observations used the medium-scale cycling dither pattern to provide good redundancy and capability for self-calibration.

Basic data on the SNe and their hosts—collected via the Open Supernova Catalog ¹⁵ (Guillochon et al. 2017), Simbad database ¹⁶ (Wenger et al. 2000), and NASA/IPAC Extragalactic Database ¹⁷ —are shown in Table 1. All of our LASTCHANCE targets were observed at one epoch—except SNe 2004dk and 2018fhw (two epochs) and 2017hcc (three epochs)—in both the 3.6 and 4.5 μm channels. We also checked the Spitzer Heritage Archive (SHA) ¹⁸ for pre-explosion Spitzer/IRAC images of the SN sites, to be used as templates (see next section).

We collected and analyzed IRAC post-basic calibrated data (PBCD). The scale of these images is 06 pixel⁻¹. Photometric analysis was carried out using the phot task of IRAF. ¹⁹ For isolated sources, we implemented aperture photometry on the PBCD frames using the phot task as a first step. We generally used an aperture radius of 2'' and a background annulus from 2'' to 6'' (2−2–6 configuration), and applied aperture corrections of 1.213, 1.234, 1.379, and 1.584 for the four IRAC channels (3.6, 4.5, 5.8, and 8.0 μm, respectively) as given in the IRAC Data Handbook; however, for a few bright sources extending to more pixels, we used the 3−3−7 configuration (aperture corrections: 1.124, 1.127, 1.143, and 1.234, respectively).

Identifying a point source within a host galaxy, where compact H ii regions and star clusters may also appear as pointlike sources in the images, can be difficult, especially if the target is faint or is superposed on a complex background. Therefore, when available, we also performed aperture photometry on template images (i.e., pre-explosion images, or very late-time images in which no point source appears at the position of the SN) and subtracted these on-site flux values from the SN fluxes. We call this technique "template-based background subtraction," or simply "background subtraction" hereafter.

Not all targets have templates. In these cases, the local background was estimated by measuring the nearby flux with apertures placed adjacent to the SN site in locations that qualitatively appear similar to the background underlying the SN (as was applied, e.g., by Fox et al. 2011). More complex modeling of the local background is beyond the scope of this paper.

In both background-subtracted and unsubtracted cases, we defined the source as a positive detection if (i) it showed epoch-to-epoch flux changes, and (ii) its flux was above the local background by at least 5 μJy and 15 μJy at 3.6 and 4.5 μm, respectively (according to point-source sensitivities in Table 2.10 of the IRAC Instrument Handbook ²⁰ ).

Following these steps, we find that 19 of 31 targets are detected (one object, SN 1995N, only at 4.5 μm). In three other cases (SNe 2011ft, 2017gas, and 2018fhw), there is a point source close to the coordinates of the SN; however, given the lack of template images, they cannot be confirmed. SN 2017gas is very close to the core, while SN 2011ft is a distant object that cannot be distinguished from its host (note that this latter site was also captured ∼6 yr earlier with Spitzer, but only at 3.6 μm, and there are no significant flux changes compared to this previously published value; see Szalai et al. 2019a). We discuss the case of SN 2018fhw later.

In the cases of the remaining nine targets, no point sources are detected at the position of the SN. Instead, some of these nondetections provided us with template images, which allowed us to carry out background subtraction on earlier archive data on these particular targets. In the cases of PTF11kx, SNHunt248, ASASSN-14dc, SN 2009ip, and SN 2012ca, background subtractions do not change the previously published flux values (see Szalai et al. 2019a, and references therein) beyond photometric uncertainties. Instead, the change in measured flux is more significant in the case of SN 2001em (221 ± 41 and 294 ± 36 μJy at 3.6 and 4.5 μm, respectively, versus 303 ± 41 and 349 ± 37 μJy given in Szalai et al. 2019a) located in a distant, nearly edge-on galaxy. We included updated fluxes and correponding absolute magnitudes of SN 2001em in Table 2.

Moreover, during the review of archival Spitzer data, we identified a variable mid-IR source at the position of the Type Ib SN 2003gk. While there are no published Spitzer data for this object (we missed it during our overview in Szalai et al. 2019a), we added it to our sample and analyzed it in the same way as the other targets (discussed in more detail below). In the case of the Type II-P SN 2017eaw, we also present here the previously unpublished photometry of late-time archival data.

In summary, we studied 32 objects during this program: 13 SNe IIn (11 positive detections, 1 unconfirmed detection, 1 nondetection), 7 SNe II-P (4 positive detections, 3 nondetections), 1 unclassified SN II (positive detection), 4 SNe Ib/c (2 positive detections, 1 unconfirmed detection, 1 nondetection), 3 SNe Ia-CSM (1 unconfirmed detection, 2 nondetections), and 4 intermediate-luminosity transients (2 positive detections, 2 nondetections). These detection rates—taking into account epochs and distances—are basically in agreement with the expectations based on earlier results (Tinyanont et al. 2016; Szalai et al. 2019a); see details later. For 12 of our targets, no Spitzer data have been published before (Type IIn PTF11iqb, ASASSN-14dc, SN 2015da, and SN 2017hcc; Type II-P SNe 2017aym, 2017ivu, 2018gj, 2018acj; SNe Ib 2003gk and 2004dk; and intermediate-luminosity transients AT2016jbu and SN 2017jfs).

In Figure 1, we present examples of new positive detections (SNe 2004dk, 2017hcc, and 2017jfs). Mid-IR fluxes and cases of nondetections are listed in Table 2. We also identify SNe where background subtraction was not possible. We note that in all these cases, the measured fluxes should be interpreted as upper limits given the complex underlying backgrounds. Flux uncertainties are generally based on photon statistics provided by phot; however, where background-subtraction photometry was carried out, an increase in the noise level by $\sqrt{2}$ is also taken into account.

Zoom In Zoom Out Reset image size

In Figure 2, we present 4.5 μm LCs of all SNe in our current sample and of some further well-known interacting SNe, along with their corresponding Spitzer archival data previously published by Szalai et al. (2019a, and references therein).

Zoom In Zoom Out Reset image size

The newest data originate primarily in this work, although some additional data from Jencson et al. (2019) and Tinyanont et al. (2019b) are also included (see details in Sections 3.1.3 and 3.1.5). We also present the 4.5 μm LCs of our SNe in Figure 3 for four subgroups: SNe IIn, SNe Ia-CSM, SNe Ib/Ic, and intermediate-luminosity transients.

Zoom In Zoom Out Reset image size

3.1.1. Type IIn SNe

3.1.2. Type Ia-CSM SNe

SNe Ia-CSM are comparably luminous to SNe IIn. This subclass is thought to arise from thermonuclear explosions surrounded by dense, H-rich shells of ambient CSM (producing SN IIn-like emission features in their late-time spectra; see, e.g., Silverman et al. 2013; Fox et al. 2015; Inserra et al. 2016). Few SNe Ia-CSM have been observed by Spitzer; Figures 2 and 4 show all of them. Analysis of this small sample suggests that this subclass, particularly in the mid-IR, is more homogeneous than SNe IIn (Graham et al. 2017; Szalai et al. 2019a). Only SN 2012ca is observed around the mid-IR "peak," reaching an absolute magnitude of almost −22 at 4.5 μm around 600 days after explosion, but the other objects in this subclass do seem to follow a qualitatively similar evolution. Apart from SN 2013dn, we also do not detect any of the SNe Ia-CSM between 1500 days and 3500 days after explosion.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We also discuss two nondetections of SN 2018fhw (ASASSN-18tb), the only new thermonuclear SN Ia in our sample. This object has unique optical properties that show some similarities but also striking differences to other SNe Ia-CSM (Kollmeier et al. 2019; Vallely et al. 2019). We observed the site of SN 2018fhw at two epochs (∼250 days and ∼450 days after explosion). While there is a pointlike source in all of the images, its position does not exactly match that of the SN (its center is ∼2'' from the position given by Vallely et al. (2019), see Figure 5; moreover, it does not show flux changes (within the uncertainties) between the two epochs in any channels. Given the lack of pre-explosion template images, we conclude for the moment that the source is not detected in our data. If we do calculate the absolute magnitudes of the detected mid-IR source, we obtain ∼−17 mag in both IRAC channels, which is a much lower value than that of other studied SNe Ia-CSM at these epochs.

3.1.3. Stripped-envelope SNe

3.1.4. Type II-P SNe

3.1.5. Intermediate-luminosity Interacting Transients

Figure 3 and 4 also show [3.6] – [4.5] μm color curves in every case when data in both channels are available. Mid-IR color evolution can be, with some limitations, a good indication of temporal changes of dust temperature. In the first few hundreds of days after explosion, when hot (T ≳ 1000 K) components of gas and dust tend to dominate the ejecta and/or its environment, the IRAC measurements probe the peak of the dust blackbody emission and provide the best constraints on the physical parameters. Such measurements have been illustrated with several SNe II-P (see, e.g., Sugerman et al. 2006; Meikle et al. 2007; Kotak et al. 2009; Andrews et al. 2010, 2011b; Szalai et al. 2011; Szalai & Vinkó 2013) and several interacting SNe (SN 2005ip—Fox et al. 2010, 2020; Stritzinger et al. 2012; SN 2006jd—Stritzinger et al. 2012; SN 2010jl—Andrews et al. 2011a; Fransson et al. 2014; Sarangi et al. 2018; Bevan et al. 2020; SN 2013L—Andrews et al. 2017; Taddia et al. 2020; SN 2014C—Tinyanont et al. 2019b; SN 2015da—Tartaglia et al. 2020).

Since most of the data presented in this paper were obtained during the Warm Spitzer mission, we are limited to the 3.6 and 4.5 μm data. Longer-wavelength data would be necessary to probe (much) colder dust, which has been observed in SN 1987A (Dwek et al. 2010; Matsuura et al. 2011; Indebetouw et al. 2014) and in young Galactic SN remnants (see, e.g., Williams & Temim 2017, for a review). A limited number of 5.8–24.0 μm measurements were obtained by Spitzer during its cryogenic phase on a few extragalactic SNe; from these data, the presence of cold dust can also be inferred (even within a few years after explosion, see Szalai et al. 2019a, for a review). Unfortunately, the assumed (significant) cold dust reservoirs cannot be efficiently surveyed; however, above ∼200 K, we will be able to get a much more detailed picture of the dust content of SNe from the expected data from the forthcoming James Webb Space Telescope (JWST).

Even with just two data points, we can derive dust temperatures and radii of the dust-containing regions via fitting blackbodies (or simple dust models) to two-point SEDs (see, e.g., Fox et al. 2011, 2013; Szalai et al. 2019a). First, it is important to determine whether any line emission contributes to the 3.6 and 4.5 μm measurements. As shown in some SNe II-P (e.g., Kotak et al. 2005; Szalai et al. 2011; Szalai & Vinkó 2013), additional flux from the 1–0 vibrational band of CO at 4.65 μm may arise in the first ∼500 days, strongly affecting the 4.5 μm LCs and leading to values of [3.6] – [4.5] ≈ 1.5–3 mag in this period (see, e.g., SNe 2004et and 2017eaw in Figure 4).

The colors of SNe IIn have a relatively small scatter (∼0.5–0.7 mag), indicating the long-time presence of hot dust above ∼1000 K. This is in agreement with the results of Fransson et al. (2014) and Sarangi et al. (2018) on SN 2010jl and of Kokubo et al. (2019) on KISS15s. The color of SN 2010jl turns continuously redder after ∼1000 days, probably indicating the decreasing dust temperature, which is also observed in SN 2007rt and SN 2013cj (at least up to their last observed epochs). In contrast, the colors of SN 2005ip and SN 2006jd indicate nearly constant (or, at least, slowly changing) dust temperatures out to ∼4500 days, in agreement with the results of Fox et al. (2010, 2013, 2020) and Stritzinger et al. (2012). SN 2010mc seems to follow a similar trend to SN 2005ip.

SNe Ia-CSM and interacting SNe Ib/c basically show a nearly constant or slowly changing mid-IR color evolution within their observational ranges. This is in agreement with the results of previous studies (e.g., Fox & Filippenko 2013; Graham et al. 2017; Tinyanont et al. 2019b), in which the authors revealed slow evolution of dust temperatures in some of these SNe. The only exception is SN 2012ca, which shows a more intense reddening (cooling) in the first few hundred days.

The colors of interacting intermediate-luminosity transients, just like their LCs, exhibit large heterogeneity. The mid-IR color evolution of most of these objects has been described in detail by Jencson et al. (2019). Regarding our new targets, AT2016jbu and SN 2017jfs have been observed at only one epoch, while SN Hunt248—captured in total four times—shows a nearly constant, relatively blue color between 136 and 360 days.

Beyond the mid-IR, signs of ongoing circumstellar interaction can be most efficiently detected and traced in the form of radio, X-ray, or Hα emission. Fox et al. (2011) proposed a scenario whereby pre-shocked dust shells are radiatively heated by ongoing shock interaction at the inner radii, but the exact relationship between the different wavelengths may not be straightforward. Long-term multiwavelength monitoring, however, is quite limited; see recent reviews by Chevalier & Fransson (2017) and Chandra (2018), as well as (for example) Weiler et al. (2002), Dwarkadas & Gruszko (2012), and Vinkó et al. (2017) regarding radio, X-ray, and Hα data, respectively.

Figures 6 and 7, and Table 4, show an overview of long-term multiwavelength (radio, X-ray, Hα, near-IR, and mid-IR) LC evolution of many SNe in our sample (as well as Figure 8, zooming into the early-time evolution of SNe 2010jl and 2015da; see discussion below). While there are still only a few objects with well-sampled multichannel data sets, some conclusions can be drawn.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

First, from the limited data available, near-IR and mid-IR LCs appear to have qualitatively similar evolution, particularly in the K band, which suggests that the dominant source of the detected flux at these wavelengths is the same—thermal continuum emission and not line emission. The availability of instruments with access to wavelengths longer than ∼2 μm is expected to be limited for high-cadence follow-up observations of transient objects in the near-term post-Spitzer era. Long-term near-IR (K-band) observations can be useful for monitoring the late-time evolution of the hottest dust when long-wavelength observations are unavailable. This could be particularly useful during the upcoming era of the Nancy Grace Roman Space Telescope (NGRST), which will be usable for similar follow-up observations of SNe (its longest-wavelength filter, F184, covers 1.68–2.00 μm, falling between the traditional H and K bands). We note, however, that the fundamental vibrational overtones of CO may contribute to not only 4.5 μm but also K-band fluxes of certain SNe (see, e.g., Jencson et al. 2017, 2019).

Furthermore, the average brightness levels of the objects seem to correlate with each other in every wavelength range (except radio). Dwarkadas & Gruszko (2012) and Ross & Dwarkadas (2017) published a compilation of X-ray LCs of different types of SNe, and their comparative figures show similarities to our mid-IR LC compilation (Figure 2) from the viewpoint that (some) SNe IIn represent the brightest sources, while observed luminosities of other SNe scatter over a wide range. In the near-IR and Hα, one can also see similar trends. Nevertheless, it is important to note that, at this time, lack of data imposes strong limitations on a more detailed comparative analysis, especially in the cases of non-SN IIn interacting SNe.

A further purpose of our current study has been to find any direct correlation between long-term X-ray/Hα and mid-IR LC evolution, which has been poorly studied to date in the literature. While comparative analyses of long-term X-ray, radio, Hα, optical, and near-IR SN LCs were published in several cases (see, e.g., Stritzinger et al. 2012; Fransson et al. 2014; Chandra et al. 2015), mid-IR LCs have not usually been (directly) involved in these comparisons.

A famous exception is the very nearby SN 1987A, for which the most complete long-term multiwavelength mid-IR/X-ray data sets were published (Bouchet et al. 2006; Dwek et al. 2010; Arendt et al. 2016). Only in this case can both very well-sampled mid-IR (from 3.6 to 24 μm) and X-ray evolution of an SN be followed for several thousand days, unfolding the ongoing interaction of the SN blast wave with a pre-existing dusty equatorial ring. In the listed papers, the IR-to-X-ray flux ratio (IRX) has been applied for investigating the process of gas–grain collisions and cooling of the shocked gas. As a shock sweeps though a medium with pre-existing dust, the IRX (at these mid-IR wavelengths) would be expected to decline as collisionally heated dust cools or is destroyed by sputtering. An epoch of dust formation in a CDS could act to increase the IRX, as could radiative processes (heating of the dust), in which case there is no direct physical interpretation of the IRX. However, we note that the level of mid-IR activity measured in SN 1987A is far below the detection limit of Spitzer for the more distant extragalactic SNe, maybe because the CSM forms a ring rather than a shell in SN 1987A (however, even these low fluxes could be available for JWST).

Another example is the Type IIn SN 2005ip, which has also been monitored for quite a long time (∼1000–5000 days; Fox et al. 2020). The mid-IR and (soft) X-ray (Katsuda et al. 2014; Smith et al. 2017) LCs seem to follow similarly decreasing trends, in agreement with the scenario that the primary source of the observed mid-IR flux is warm dust radiatively heated by energetic photons emerging from CSM interaction, and that the shock may be finally reaching the outer extent of the dense CSM shell. The multiwavelength evolution, including Hα, shows a shape that resembles that of an unabsorbed X-ray LC (see Smith et al. 2017, and also Figure 6). SN 2006jd, another SN IIn, shows a quite similar behavior to that of SN 2005ip at various wavelengths; however, there is a lack of data after ∼2000 days (except in the mid-IR).

Based on the results of the first extended Spitzer survey of interacting SNe, Fox et al. (2013) also presented figures summarizing parallel mid-IR, X-ray, and optical (R-band) evolution of several SNe IIn; however, most of those LCs consist of only very few points.

In order to map similar correlations in our sample, we plotted the parallel evolution of all six SNe where both mid-IR and X-ray fluxes are available, and also calculated the IRX ratios for these objects (see Figures 9 and 10). Unfortunately, the amount and sampling of these data allow us to make only a preliminary analysis. Based on that, it seems that in SNe 2004dk, 2005ip, and 2006jd (expected to have radiatively heated pre-existing dust in their environments), all wavelengths trace each other (as proposed by, e.g., Fox et al. 2011).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

While in most cases, comparative analysis of mid-IR and X-ray data was possible at epochs after ∼1000 days, it would be very interesting to see similar analyses at earlier phases. The only object with well-sampled mid-IR and X-ray data before ∼1000 days is the Type IIn SN 2010jl (see Figures 6 and 8–10). While SN 2010jl exhibits near-IR and mid-IR rebrightening and a quite long plateau after ∼400 days, it is preceded by an X-ray/Hα peak somewhere between 200 and 300 days. As previously noted (e.g., Fransson et al. 2014; Gall et al. 2014; Sarangi et al. 2018; Bevan et al. 2020), an IR excess before ∼400 days may be the consequence of heating of pre-existing dust, while mid-IR flux at a later time may originate either from new dust in the CDS and/or the ejecta or from CSM dust. Sarangi et al. (2018) presented a comparative analysis of mid-IR and X-ray LCs, and we complete that comparison here with the addition of an Hα LC (adopted from Fransson et al. 2014).

If we take into account general similarities between mid-IR and near-IR and between X-ray and Hα (see above), we can also find some further cases where early-time multiwavelength evolution can be well traced. Two other SNe IIn, SN 2015da (Tartaglia et al. 2020) and KISS15s (Kokubo et al. 2019), also show mid-IR LC evolution similar to that of SN 2010jl in the first ∼1000 days (up to the end of their data sets), probably going through similar circumstellar and dust-formation processes. In the case of SN 2015da, this is also strengthened by the parallel evolution of Hα and mid-IR luminosities (Figure 8 (right panel), adapted from Tartaglia et al. 2020).

In SNe IIn 2005ip and 2006jd, one can see long plateaus in both the near-IR and X-ray/Hα ranges during the first several hundred days, in accordance with their later-time LCs also tracing each other (as described above).

There are also well-sampled near-IR/mid-IR and Hα LCs of the Type II-P SNe 2004et and 2017eaw and of the Type II-P/L SN 2013ej (see Figure 7). These LCs trace each other quite well, during both their declining and their small rebrightening phases (the latter ones can be seen at ∼800–1000 days). This can be explained well by CDS dust formation described in Section 3.1.4.

As a conclusion of this part, our preliminary studies show that comparative analysis of long-term multiwavelength (especially X-ray/Hα versus mid-IR) data sets is an essential (and maybe the only really useful) tool to differentiate between the existing dust forming and heating scenarios and thus to get a complete picture of the physical processes going on in the circumstellar environments of interacting SNe. We plan to carry out a more detailed investigation of this topic, expecting high-quality data in the near future (e.g., from approved JWST and associated programs).