Late-time Observations of the Type Ia Supernova SN 2014J with the Hubble Space Telescope Wide Field Camera 3

Over the years, several observational techniques have been developed to constrain the nature of the progenitor and explosion mechanism of Type Ia supernovae (SNe Ia; see the review by Maoz et al. 2014). These methods include, but are not limited to, reconstruction of the delay-time distribution through measurements of SN Ia rates (see Maoz & Graur 2017 for the latest summary), analysis of pre-explosion imaging (e.g., Li et al. 2011; Graur et al. 2014; Kelly et al. 2014; Graur & Woods 2018), and multi-wavelength observations (e.g., Horesh et al. 2012; Margutti et al. 2012, 2014; Chomiuk et al. 2016).

Recently, late-time observations of SNe Ia > 500 days past B-band maximum light (hereafter referred to simply as, e.g., >500 days) have been suggested as a new way to constrain SN Ia progenitor, explosion, and nebular physics. Several works have shown that, starting at ∼800 days, the decline of the optical light curves of SNe Ia begins to slow down. Until that point, the light curves are dominated by the radioactive decay of ⁵⁶Co. The slow-down in the decline rate of the light curves indicates that an additional source of energy is contributing to the heating of the SN ejecta.

Graur et al. (2016) showed that the light curve of SN 2012cg was consistent with a combination of ⁵⁶Co and ⁵⁷Co radioactive decays (as predicted by Seitenzahl et al. 2009). The same was shown by Shappee et al. (2017) for SN 2011fe and by Yang et al. (2018) for SN 2014J. According to Shappee et al. (2017), the late-time photometry of SN 2011fe was precise enough to rule out the single-degenerate progenitor scenario (Whelan & Iben 1973), but Dimitriadis et al. (2017) and Kerzendorf et al. (2017) showed that similar data could also be fit with models that assumed either atomic freeze-out (Fransson & Kozma 1993; Fransson & Jerkstrand 2015) or a changing fraction of positron trapping (which could be due, for example, to temporal variations in the magnetic field of the ejecta; Penney & Hoeflich 2014).

Graur et al. (2018a, 2018b) analyzed the late-time light curves of five SNe Ia and claimed a possible correlation between the decline rate of the late-time light curve and the intrinsic luminosity of the SNe. This was challenged by Jacobson-Galán et al. (2018); based on a single epoch of photometric measurements of SN 2013aa at ≈1500 days, they claimed that the ⁵⁷Co/⁵⁶Co mass ratio measured for SN 2013aa was significantly lower than predicted by the correlation.

The correlation claimed by Graur et al. (2018a, 2018b) suffers from two main sources of uncertainty: a statistical uncertainty due to the small number of SNe observed at >500 days and a systematic uncertainty due to the different assumptions that go into the construction of the pseudo-bolometric light curves on which the correlation is based. Here, I attempt to remove the second source of uncertainty by directly comparing observations of the SNe studied by Graur et al. (2018a, 2018b) in the same filters. This is made possible by new observations of SN 2014J.

The late-time light curve of SN 2014J was studied by Yang et al. (2018) using data taken with the Advanced Camera for Surveys (ACS) onboard the Hubble Space Telescope (HST). In Section 2, I present complementary data taken with the HST Wide Field Camera 3 (WFC3). I show that the new photometry is consistent with that of Yang et al. (2018) but, crucially, includes data in F438W and F555W , which allows a direct comparison with similar photometry of SNe 2011fe, 2012cg, 2015F, and ASASSN-14lp. In Section 3.1, I confirm the conclusion reached by Yang et al. (2018) that, although SN 2014J has spatially resolved light echoes, the photometry of the SN itself is not contaminated by an unresolved echo. In Section 3.2, I show that the correlation claimed by Graur et al. (2018a, 2018b) is still apparent when examining the observed—instead of pseudo-bolometric—light curves. I summarize my findings in Section 4.

SN 2014J was discovered by Fossey et al. (2014) on 2014 January 21.8 in the nearby (3.3 Mpc; Foley et al. 2014 ) starburst galaxy M82. With a SiFTO stretch value of 1.086 ± 0.010 (Conley et al. 2008; Childress et al. 2015), SN 2014J is a luminous but still normal SN Ia. According to Marion et al. (2015), the SN reached peak B-band light on 2014 February 1.7 (MJD 56689.74 ± 0.13); all phases in this work are measured relative to this date.

For this work, I used WFC3 images of SN 2014J in the broad-band filters F438W, F555W, and F814W taken under HST programs GO-13626 (PI: Lawrence), GO-14146 (PI: Lawrence), GO-14700 (PI: Sugerman), and SNAP-15166 (PI: Filippenko). These programs imaged SN 2014J between 2014 March 11 and 2017 December 17. In this work I concentrate on the late-time light curve of SN 2014J, and so only present data starting at 2015 February 2, when the SN was ∼365 days old.

I used the AstroDrizzle task included in the DrizzlePac Python package³ (Hack et al. 2012) to align the HST FLC images and remove cosmic rays and bad pixels. Next, I aligned the images using the IRAF⁴ routines xregister and wcscopy (Tody 1986, 1993). Finally, I created difference images of the SN by subtracting the last epoch in each filter from all earlier epochs. The resultant difference images are shown in Figure 1. A visual inspection does not detect the SN in the template images or in the final difference image in each filter.

Zoom In Zoom Out Reset image size

SN 2014J is known to have spatially resolved light echoes (Crotts 2015; Yang et al. 2017), which are apparent in the difference images in Figure 1. To ensure that the light echoes did not contaminate the photometry of the SN, I performed aperture photometry with a 04-diameter aperture, and used a sky annulus with a radius larger than the farthest reach of the resolved light echoes. I applied the latest aperture corrections, as described by Deustua et al. (2016).⁵ The resultant photometry, in Vega mags, is presented in Table 1 and Figure 2. In epochs where the SN is undetected, I measured 3σ upper limits by estimating the magnitude at which the SN would have a signal-to-noise ratio (S/N) of 3.

Zoom In Zoom Out Reset image size

Unlike the clean F438W and F555W difference images, the F814W difference images shown in Figure 1 are speckled with other point sources aside from SN 2014J. A close look at some of these point sources shows that they appear, with different brightnesses, in multiple epochs. Thus, these point sources are likely to be variable stars that, while visible in F814W, are extincted in F438W and F555W .

The location of SN 2014J is known to suffer from high host-galaxy extinction. Marion et al. (2015), for example, measure a host-galaxy E(B − V) = 1.23 ± 0.06 mag and R_V = 1.46. Based on these values, and assuming a Cardelli et al. (1989) extinction law, throughout this work I use host-galaxy extinction values of A_F438W = 3.2, A_F555W = 1.9, and A_F814W = 0.7 mag. These values are consistent with the B, V, and I extinctions measured by Foley et al. (2014) from the early-phase light curve of SN 2014J. For line-of-sight Galactic extinction, I rely on the Schlafly & Finkbeiner (2011) measurements of A_F438W = 0.6, A_F555W = 0.5, and A_F814W = 0.2 mag.

As shown in Figure 2, the F555W photometry measured here is consistent with the F606W photometry measured by Yang et al. (2018). This is expected, as both filters cover roughly the same spectral range. On the other hand, the F438W photometry is systematically fainter by ≈1 mag, which is probably due to the fact that the ACS F475W filter is roughly 2.4 times wider than the WFC3 F438W filter and so encompasses a prominent spectral feature at ∼5400 Å (likely a combination of [Fe II] and [Fe III]; Graham et al. 2015) which is left out of the F438W filter.

In this section, I use the photometry measured in Section 2 to show that SN 2014J is not contaminated by an unresolved light echo (Section 3.1) and that an analysis of the late-time F438W and F555W light curves of SNe Ia is still consistent with the Graur et al. (2018a, 2018b) claim that more luminous SNe Ia have flatter light curves at late times (Section 3.2).

3.1. Light Echoes

To date, several SNe Ia have been shown to exhibit light echoes, which are produced when the light of the SN is reflected off nearby dust sheets. Unresolved light echoes will contaminate the photometry of the SN and cause its light curve to flatten out at late epochs (Schmidt et al. 1994; Sparks et al. 1999; Cappellaro et al. 2001; Quinn et al. 2006; Wang et al. 2008, 2018; Drozdov et al. 2015 ).

To test whether the photometry of SN 2014J is contaminated by an unresolved light echo, I repeat the exercise done by Graur et al. (2018a, 2018b) and compare, in Figure 3, the late-time B − V colors of SN 2014J to those of SN 2011fe (Shappee et al. 2017), which is known to be free of light echoes (Graham et al. 2015; Shappee et al. 2017). I also compare these colors to those of the two SNe at peak (as measured by Munari et al. 2013 and Foley et al. 2014 for SNe 2011fe and 2014J, respectively). All magnitudes have been converted from the WFC3 filter sets to the Johnson–Cousins BVI filters and corrected for Galactic and host-galaxy extinctions. The filter-set conversions were calculated by comparing synthetic photometry of a late-time spectrum of SN 2011fe at 981 days (see Graham et al. 2015; Graur et al. 2018b).

Zoom In Zoom Out Reset image size

The B − V colors of both SNe are redder at late times than they were at peak. SN 2014J also has a V − I color of −1.2 ± 0.2 mag at 988.4 days, which is bluer than the peak V − I color of ∼−0.3 mag. This behavior, where the SN is bluer in V − I but redder in B − V, is consistent with the late-time colors of SNe 2011fe, 2015F, and ASASSN-14lp (Shappee et al. 2017; Graur et al. 2018a, 2018b). This is inconsistent with a light echo, which would scatter the SN light to bluer wavelengths and make it appear bluer than it was at peak in all colors. Thus, even though SN 2014J has spatially resolved light echoes, it appears that the aperture photometry conducted here is not contaminated by an unresolved echo, consistent with the findings of Yang et al. (2018).

3.2. The Decline Rate of the Late-time Light Curve

Previous papers on the late-time light curves of SNe Ia used multi-band observations to construct a "pseudo-bolometric" light curve in the optical wavelength range of ∼3500–10000 Å. Briefly, in each epoch, a spectrum of the SN is morphed to fit the observed photometry, which is first corrected for Galactic and host-galaxy extinction. A pseudo-bolometric flux is then measured by integrating the morphed spectrum over the observed wavelength range. Finally, the pseudo-bolometric flux is converted to a luminosity, taking into account the distance measured to either the SN or its host galaxy. This technique, though instructive, might introduce systematic uncertainties for several reasons.

• 1.  
  Only SN 2011fe has spectra taken 500–1000 days after the explosion (Graham et al. 2015; Taubenberger et al. 2015). The pseudo-bolometric light curves of SNe 2015F, ASASSN-14lp, and 2014J were also constructed using these spectra (Graur et al. 2018a, 2018b; Yang et al. 2018).
• 2.  
  Those spectra only sample SN 2011fe at 593, 981, and 1034 days. According to Graham et al. (2015), SN 2011fe shows little spectral evolution between the 593 and 981 day spectra. It is tacitly assumed that this is true of other SNe Ia as well.
• 3.  
  Some SNe did not have full multi-band coverage throughout the phase range in which the pseudo-bolometric light curve was constructed, and extrapolations were used to fill in the gaps (e.g., SN 2015F and ASASSN-14lp).
• 4.  
  Although Galactic line-of-sight extinction toward the SNe is well measured, there is still an ongoing debate as to the extinction law (parameterized by the total-to-selective extinction ratio, R_V) at the location of SNe Ia (e.g., Folatelli et al. 2010; Chotard et al. 2011; Scolnic et al. 2014; Amanullah et al. 2015).
• 5.  
  The measurement uncertainties of the distances to the SNe are not propagated into the final luminosities. This would impact the scaling of the resultant pseudo-bolometric light curve, which could bring into question the interesting claim that the late-time light curves first converge at ∼600 days before diverging at different decline rates (Graur et al. 2018a; Jacobson-Galán et al. 2018).

Here, I remove these systematic uncertainties by foregoing the construction of a pseudo-bolometric light curve for SN 2014J. Instead, in Figure 4 I compare the F438W and F555W photometry of SN 2014J measured in Section 2 to those of SNe 2011fe, 2012cg, 2015F, and ASASSN-14lp in the same filters. The F555W comparison is of most value, since previous works have shown that SNe Ia are most luminous in this filter. Moreover, there is some evidence that the F555W flux should trace the bolometric light curve. For example, Milne et al. (2001) showed that between 50 and 600 days, a constant fraction of the luminosity of SNe Ia is emitted in the V band, which is comparable to F555W .

Zoom In Zoom Out Reset image size

The photometry of the different SNe in Figure 4 is not corrected for Galactic and host-galaxy extinction. Such corrections, with their accompanying systematic uncertainties, would only impact the scaling of the measurements, which is not important for the analysis below.

As a simple test of the Graur et al. (2018a, 2018b) correlation, I perform a linear fit to the photometry of each SN to measure the decline rates of their light curves. The resultant rates are presented in Table 2 and shown in the insets of Figure 4 as a function of the stretch, s, of the SNe. The latter is correlated with the intrinsic luminosity of the SNe through the Phillips (1993) relation.

Table 2.  Light Curve Decline Rates

Supernova	Stretch	ΔF438W	χ2/DOF	ΔF555W	χ2/DOF	ΔF555W	χ2/DOF
		mag/(100 days)		mag/(100 days)		mag/(100 days)
		500 < t < 1000 days		500 < t < 1000 days		800 < t < 1300 days
SN 2015F	0.906 ± 0.005	1.05 ± 0.07	0.02/1	0.77 ± 0.02	26/7	1.07 ± 0.06	11/3
SN 2011fe	0.969 ± 0.010	0.81 ± 0.01	260/3	0.67 ± 0.01	17/3	0.62 ± 0.01	13/4
SN 2012cg	1.063 ± 0.011	⋯	⋯	0.66 ± 0.03	⋯	⋯	⋯
SN 2014J	1.086 ± 0.010	0.76 ± 0.02	11/2	0.62 ± 0.01	2/2	0.49 ± 0.07	0.1/1
ASASSN-14lp	1.150 ± 0.050	0.42 ± 0.10	0.03/1	0.39 ± 0.05	8/3	0.39 ± 0.05	8/3

Download table as:  ASCIITypeset image

Ideally, the decline rate should be measured at t > 800 days, when previous works have shown that the late-time light curve begins to appreciably deviate from domination by pure ⁵⁶Co radioactive decay. This is possible in F555W, but not in F438W, where there are fewer measurements. For the latter filter, I measure the decline rate at 500 < t < 1000 days. The decline rates in F555W are measured in two phase ranges: 800 < t < 1300 days, shown in Figure 4, and 500 < t < 1000 days, to facilitate a comparison with F438W .

SN 2011fe only has HST observations starting at ≈1100 days, but Shappee et al. (2017) also provide earlier photometry in B and V. Because the latter measurements are consistent with the HST measurements taken at similar times (as shown in Figure 4), I include them in my analysis. For SN 2012cg, which only has two F555W measurements, I measure the decline rate between those two visits and use it as an estimate of the decline rate at 500 < t < 1000 days.

The light curve decline rates in both filters show the same trend as the pseudo-bolometric light curves in previous works, i.e., more luminous SNe Ia (with larger stretch values) exhibit flatter late-time light curves. However, as also noted by previous works, a larger SN sample is required to substantiate this trend.

I have presented new HST photometry of SN 2014J when it was ∼360–1300 days old. These observations, taken with WFC3, complement the HST ACS observations of this SN presented by Yang et al. (2018).

I have used these data to verify that, although SN 2014J exhibits spatially resolved light echoes, the photometry of the SN itself is not contaminated by an unresolved echo.

Whereas previous works used the photometry of each SN to construct a pseudo-bolometric light curve, in this work I conducted a direct comparison of the F438W and F555W light curves of SNe 2011fe, 2012cg, 2014J, 2015F, and ASASSN-14lp. This removes any systematic uncertainties introduced by the construction of the pseudo-bolometric light curve, as well as the use of extinction corrections and distance measurements.

I have measured the light curve decline rates of the five SNe Ia mentioned above at 500 < t < 1300 days in both F438W and F555W and have shown that these are consistent with the correlation claimed by Graur et al. (2018a, 2018b): more luminous SNe Ia (with higher stretch values) have slower-declining late-time light curves. However, although in this work I have minimized the sources of systematic uncertainty that might affect this claim, the main source of uncertainty—the small sample size—remains. A larger sample of late-time SNe Ia, observed with the same filters analyzed here, is required to definitively test the Graur et al. (2018a, 2018b) correlation.

I thank the anonymous referee for helpful comments and suggestions. O.G. was supported by NASA through HST-GO–14611 and HST-GO-15415. This work is based on data obtained with the NASA/ESA Hubble Space Telescope, all of which was obtained from MAST. Support for MAST for non-HST data is provided by the NASA Office of Space Science via grant NNX09AF08G and by other grants and contracts. This research has made use of NASA's Astrophysics Data System and the NASA/IPAC Extragalactic Database (NED) which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with NASA.

Software: DrizzlePac (Hack et al. 2012), IRAF (Tody 1986, 1993).