CONSTRAINTS ON THE PROGENITOR SYSTEM OF THE TYPE Ia SUPERNOVA 2014J FROM PRE-EXPLOSION HUBBLE SPACE TELESCOPE IMAGING

To locate the SN site in pre-explosion HST exposures with high precision, we acquired wide-field NIRC2 K-band AO images of the site of SN 2014J on 2014 January 25. The SN was used as the guide star to measure tip-tilt corrections. The FWHM natural seeing produced a PSF with 065 FWHM, while the AO-corrected images have a PSF FWHM of 01 near the SN position.

We first subtracted the median of a stack of bias exposures and then applied flat-field and distortion⁷ corrections to the NIRC2 images. Although exposures of SN 2014J were acquired for the minimum possible integration time, the number of counts registered by the several pixels closest to the peak of the PSF exceeded the range within which the detector has a linear response. Both to register the 18 AO exposures and then to measure the coordinates of the PSF of SN 2014J, we needed to determine the center of the PSF of the SN. We therefore fit a two-dimensional Gaussian model to the SN pixel intensity distribution in each of the 18 exposures and allowed the PSF center coordinates, FWHM, and normalization to vary.

Saturated pixels in the NIRC2 detector assume low data number (DN) values during readout. For each exposure, we identified (through visual inspection) and masked the one or two pixels whose low DN values were consistent with a saturated pixel. Our χ² goodness-of-fit statistic excludes pixels where the model value was in excess of the nonlinearity threshold. Centers of the fitted Gaussian models coincided with the positions of the most saturated pixel. The NIRC2 exposures show no evidence for bleeding from saturated pixels.

We then used the best-fitting SN coordinates to align the reduced images. The positions of the intensity peaks of many sources in M82 show variation with wavelength; thus, to mitigate the contribution of this astrometric noise and to maximize the number of common sources, we registered the coadded K-band image against archival HST F160W (H) images. To minimize the effect of any remaining astrometric distortion in the NIRC2 coadded image, we used sources only within the central 16'' × 16'' section of the NIRC2 coadded image when cross-registering with the HST F160W image. After extracting sources with SExtractor (Bertin & Arnouts 1996) and excluding extended objects with r₅₀ > 3 pixels, we used the tweakreg routine in the astrodrizzle package⁸ to fit for an astrometric solution from 68 common objects.

Figure 1 shows the astrometrically matched, coadded K-band exposure obtained with the NIRC2 AO system, along with the coadded HST F160W image. The position of the SN is α = 9^h55^m42.137(9)^s, δ = +69°40'25.40(5)'' (J2000.0) in the World Coordinate System (WCS) of the drizzled F160W image⁹ available from the Hubble Legacy Archive (HLA).¹⁰ The rms of the offsets between pairs of matched objects after applying the astrometric solution is 0057 in R.A. and 0055 in decl.

Zoom In Zoom Out Reset image size

Table 1 provides a list of the HST imaging data sets of the explosion site that we use to place limits on the flux of the progenitor. Except for the Advanced Camera for Surveys (ACS) exposures, the images within each data set were processed, drizzled, and coadded to a common pixel grid by the HLA team. The ACS coadded mosaics of drizzled images that we analyze are High Level Science Products made available by the observers who acquired the data (Proposal 10776; PI: M. Mountain).

As may be seen in the representative images in Figure 2, the local environment of SN 2014J exhibits strong surface brightness variations from both resolved and unresolved sources, as well as strong and varying extinction by dust. As in Maoz & Mannucci (2008), our approach for estimating nondetection limits is to add artificial point sources. We create a sequence of images, each with a superimposed artificial point source having FWHM appropriate to the HST instrument and bandpass filter. We create a pixelated Gaussian PSF and use the PyFITS package¹² to add the artificial source to the HST image. We place the PSF ∼1–2 pixels in a randomly determined direction away from the SN explosion site to avoid superimposing the PSF on any possible underlying counts from a progenitor, which would yield a biased upper limit. These offset positions are sufficiently modest that the local background is comparable to that at the SN site.

The artificial point source in each successive image contains 0.1 mag additional flux than that in the previous image. We visually determine the limiting magnitude by identifying the first image for which the brightest pixel of the injected source has more counts than the neighboring local maxima produced by photon shot noise and background variation, and where pixels adjacent to the peak of the injected source have elevated counts consistent with a PSF.

To estimate the statistical significance of these visual detections for use in likelihood computations, we compute detection limits using a complementary technique. We calculate the rms of the background flux in a region without detected sources or a strong intensity gradient. We consider a 3σ detection to be one for which the source flux exceeds the background rms inside a 6 pixel aperture by a factor of three.

A circular aperture enclosing 80% of the flux of a point source is the default choice for most HST instruments when computing the signal-to-noise ratio (S/N) of a flux measurement using the Exposure Time Calculator.¹³ For the ACS Wide Field Camera (WFC), this corresponds to an area that includes ∼44 pixels (and corresponding background noise), and yields a limiting magnitude that is ∼0.7 mag brighter and less sensitive than the value we compute for a 6 pixel background aperture. From inspection of the artificial point sources, we find that visual detections, however, depend almost entirely on only the several pixels closest to the coordinates of the source. We note that the limiting magnitudes of HST images calculated by Li et al. (2011) in their analysis of the explosion site of SN 2011fe are consistent with estimates instead made using a circular aperture enclosing 80% of the flux.

As shown in Table 1, the visual magnitude limits are comparable to the 3σ limits estimated using the background statistics. We assign accordingly a 3σ significance to the visual limits for the purpose of computing likelihood functions.

To convert these limits on the apparent magnitudes of a progenitor system to constraints on a stellar source, we use the Pickles (1998) library to model the spectral energy distributions (SEDs) of candidate progenitor systems of the SN, as well as (following Li et al. 2011) blackbody spectra with temperatures of 35,000, 65,000, and 95,000 K. For each combination of HST filter and instrument in Table 1 and all spectroscopic templates, we compute synthetic magnitudes with and without extinction from dust. After removing E(B − V)_MW = 0.14 mag of foreground Milky Way (R_V = 3.1) extinction, Goobar et al. (2014) estimated that additional extinction of E(B − V)_SN = 1.22 ± 0.05 mag with R_V = 1.40 ± 0.15 can best reproduce the shape of a premaximum optical spectrum of SN 2014J. While we adopt the Goobar et al. (2014) extinction curve for our progenitor constraints, we additionally compute synthetic magnitudes instead with an R_V = 3.1 curve for M82, using the total A_V favored by Goobar et al. (2014). As we demonstrate in Section 5, we obtain similar progenitor constraints for these differing values of R_V.

Following Li et al. (2011), we first translate the upper limit on the flux for each filtered observation to a 2σ upper limit on the absolute magnitude M_V, using a distance modulus to M82 of 27.73 ± 0.02 mag (Jacobs et al. 2009). After repeating this for all observations, we identify the most constraining upper limit on M_V, and refer to this as the 2σ "1-frame" limit for each spectrum template. A combined constraint is estimated next by constructing a probability function that incorporates the upper magnitude limits in all filters. The total probability (as a function of M_V and the template spectrum) is the product of the probabilities of each filtered observation computed from the predicted model magnitude and the measured upper magnitude limit.

To calculate the combined progenitor system limit, we incrementally increase the absolute brightness M_V until a 2σ probability (p = 0.954) is reached. As in Li et al. (2011), the combined constraints are ∼0.2–0.8 mag deeper than "1-frame" constraints which rely on observations in a single filter. We would expect to detect, with 95% probability, such a 2σ source in a least one HST bandpass image, and in the coadded image in Figure 2.

Table 1 shows the detection upper limits for the coadded image of each data set, and Table 2 lists the corresponding constraint on M_J, in addition to M_V, computed for each Pickles (1998) and blackbody spectrum. For each spectrum, the photometric band providing the faintest individual absolute magnitude limit is also identified in Table 2.

In Figure 3, we plot the limits on the progenitor flux we estimate using the HST imaging. We also show measurements of the luminosity L_ν of RS Oph during its quiescent phase as a function of frequency ν, and the uncertainty arising from current constraints on its distance.

Zoom In Zoom Out Reset image size

The luminosity of RS Oph is comparable to our upper detection limit ($R_V^{\rm M82} = 1.4$; $A_V^{\rm M82} = 1.7$ mag), and the bright white dwarf primary in RS Oph contributes significantly in bluer optical bandpasses, unlike the primary of T CrB. Therefore, we create a near-UV to near-IR model of the spectrum of RS Oph to account for the contribution of the white dwarf, and to show clearly the measurements that contribute to our constraints.

Anupama & Mikołajewska (1999) performed a decomposition of an optical spectrum of RS Oph in its quiescent phase and find an A2 to A4 stellar spectrum for the hot component, and an M0 to M2 III spectrum for the giant companion. We fit for the linear combination of the corresponding Pickles (1998) stellar spectra that best matches the broadband magnitudes of RS Oph during quiescence. Li et al. (2011) use an effective temperature of 3750 to 3650 K for the giant companion typical of M0 to M2 giant stars (Straizys & Kuriliene 1981) in their SN 2011fe progenitor analysis.

Evans et al. (1988) compiled J-, H-, and K-band magnitudes of RS Oph obtained during quiescence from 1971 through 1982 by Swings & Allen (1972), Feast & Glass (1974), Szkody (1977), Sherrington & Jameson (1983), and Kenyon & Gallagher (1983). These measurements show modest variation of ±0.2 mag, and we use the median values (J ≈ 7.67, H ≈ 6.92, K ≈ 6.62 mag) as representative values during the quiescent phase. Magnitudes of RS Oph measured by Two Micron All Sky Survey (Skrutskie et al. 2006) in 1999 (J = 7.63 ± 0.02, H = 6.85 ± 0.04, K = 6.50 ± 0.01 mag; see also Darnley et al. 2012), as well as values synthesized from a near-IR spectrum (Rushton et al. 2010; Skopal 2014), are consistent with the magnitudes in quiescence assembled by Evans et al. (1988). We use the optical BVRI photometry of RS Oph during quiescence measured by Skopal (2014) to construct our model. The Skopal (2014) optical measurements are, on average, fainter than those presented by Zamanov et al. (2010) for several epochs. While Schaefer (2010) combine optical and IR colors of RS Oph taken on two dates to infer a possibly fainter flux in the IR, the combined values they present do not appear to be representative of the system in quiescence.

To estimate the luminosity of RS Oph, we use E(B − V) = 0.73 mag (Snijders 1987), which is consistent with the value derived by Anupama & Mikołajewska (1999), for an R_V = 3.1 Cardelli et al. (1989) Milky Way extinction curve. Estimates for the distance to RS Oph include ∼1.4 kpc from modeling of the 2006 outburst (e.g., Rupen et al. 2008), ∼1.6 kpc from radio observations of expanding material in the 1985 (Hjellming et al. 1986) and 2006 (Sokoloski et al. 2006) outbursts, and ∼3.1 kpc based on the assumption that the secondary fills the entire Roche lobe (Livio et al. 1986; Barry et al. 2008).

With a V-band total magnitude of 11.5 in quiescence, we estimate a range in absolute magnitude of −1.49 ⩽ M_V ⩽ −3.22 for distances of 1.4–3.1 kpc. Using our decomposition of the spectrum of RS Oph in quiescence, we estimate that the giant star should be ∼0.65 mag fainter in V than the total magnitude of the binary system.

As can be seen in Figure 2, the SN 2014J coordinates that the AO analysis favors are at a greater angular distance from the nearest resolved point source than the Tendulkar et al. (2014) position. This suggests that the star at a smallest angular offset is not a mass donor to the white dwarf progenitor of SN 2014J. Our coordinates also position SN 2014J closer to the center of an apparent dust cloud whose silhouette is visible in optical HST images.

Figure 4 shows the derived constraints on the position of the progenitor system of SN 2014J in the Hertzsprung–Russell (H-R) diagram. The upper luminosity limit excludes the bright extent of the region, marked in pale red, occupied by red giant stars. The faint boundary of the red giant region corresponds approximately to the least luminous red giants in the Hipparcos catalog (Perryman et al. 1997), and we truncate the red giant region at an upper effective temperature of 5000 K. Stars more luminous than the plotted M_V ≈ 3.8 mag bright boundary may instead be classified as red supergiant stars.

Zoom In Zoom Out Reset image size

With the exception of values for RS Oph (see Section 4.3), we plot the ranges of effective temperature and M_V values collected by Li et al. (2011) for Galactic candidate systems. Our luminosity limits are comparable to our faintest estimates for the luminosity of RS Oph in quiescence, including the uncertainty in its distance in the Galaxy. The upper M_V line intersects the M_V–T area corresponding to the less luminous Galactic symbiotic system T CrB (Hachisu & Kato 2001).

U Sco is a recurrent nova and supersoft X-ray source with a large white dwarf mass of 1.55 ± 0.24 M_☉ (Thoroughgood et al. 2001) and a subgiant companion. This candidate progenitor system is substantially too faint for detection in our archival images, as was the case for SN 2011fe.

A candidate He-rich single-degenerate progenitor system is the He nova V445 Puppis (V445 Pup). We use the Woudt et al. (2009) estimate for M_V based on a parallax distance measurement of 8.2 ± 0.5 kpc from the expansion of a bipolar shell after an eruption in year 2000. The upper luminosity limits for SN 2014J extend across the region of the H-R diagram corresponding to V445 Pup. Figure 4 shows the theoretical limits on the He channel computed by Li et al. (2011) using Liu et al. (2010) models and bolometric corrections from Torres (2010). Our measured limits are fainter than the predicted luminosities of He-star-channel progenitors with comparatively cool temperatures (T ≲ 35,000 K).

For comparison, we plot the luminosity limits for the progenitors of SN 2011fe (Li et al. 2011) and SN 2006dd (Maoz & Mannucci 2008); the latter provides a constraint typical of SNe Ia other than SN 2011fe. For companions with very low effective temperature (T < 3000 K), limits on the progenitor SN 2014J are significantly fainter than those for SN 2011fe, because the SN 2014J archival images extend to the near-IR.

Now referring to the stellar evolutionary tracks off the main sequence computed by Lejeune & Schaerer (2001), we find that a mass donor with effective temperature cooler than 4000 K would need to have a mass less than 2 M_☉. The plotted evolutionary tracks are models with abundance (Z = 0.02) consistent with the metallicity of M82 starburst regions (Förster Schreiber et al. 2001). The derived luminosity limits are significantly brighter than a hypothetical binary system of two white dwarfs that does not experience a long-lived merger phase (Shen et al. 2012), or a white-dwarf/main-sequence binary (e.g., Wheeler 2012).

Comparison among the bright red ($R_V^{\rm M82} = 3.1$ mag; $A_V^{\rm M82} = 1.7$ mag; Goobar et al. 2014), the dashed pale red ($R_V^{\rm M82} = 3.1$ mag; $A_V^{\rm M82} = 1.7$ mag), and the dashed pale blue ($R_V^{\rm M82} = 2$ mag; $A_V^{\rm M82} = 2.5$ mag) lines shows that the progenitor constraints exhibit only modest change within the range of dust parameters reported from analyses of the SN 2014J light curve and spectra (e.g., Goobar et al. 2014; Patat et al. 2014).

While the total integration times of pre-explosion HST broadband imaging of the site of SN 2014J are almost all greater than that of the site of SN 2011fe, and M82 (d ≈ 3.5; Jacobs et al. 2009) is at a smaller distance than M101 (d ≈ 6.4 Mpc; Shappee & Stanek 2011), the upper flux limits at optical wavelengths that we estimate for the progenitor of SN 2014J are less constraining than those for the progenitor of SN 2011fe (Li et al. 2011). The brighter optical constraints on the progenitor of SN 2014J arise from the high extinction inferred along the line of sight to SN 2014J, and from the bright and varying local background in M82. In contrast, the spectra and colors of SN 2011fe were consistent with a lack of extinction and it occurred at a location with low surface brightness. In the ACS WFC F814W FLT images of the explosion sites, we measure an average of ∼0.13 counts s⁻¹ pixel⁻¹ close to the explosion site of SN 2011fe in M101, and ∼5.3 counts s⁻¹ pixel⁻¹ close to Type Ia SN 2014J in M82.