NO X-RAYS FROM THE VERY NEARBY TYPE Ia SN 2014J: CONSTRAINTS ON ITS ENVIRONMENT

The Swift (Gehrels et al. 2004) X-Ray Telescope (XRT; Burrows et al. 2005) started observing SN 2014J on 2014 January 22, 10:13:52 UT (δt ∼ 8 days; PI: Brown, PI: Ofek). XRT data have been analyzed using HEASOFT (v6.15) and corresponding calibration files. Standard filtering and screening criteria have been applied. We find no evidence for a point-like X-ray source at the position of SN 2014J. Using observations acquired at δt < 20 days (total exposure time of 114 ks), the 3 σ count-rate limit is 2.4 × 10⁻³ counts s⁻¹ in the 0.3–10 keV energy range. The Galactic neutral hydrogen column density in the direction of SN 2014J is $\rm {NH}_{\rm {MW}}\lesssim 5.1\times 10^{20}\,\rm {cm^{-2}}$ (Kalberla et al. 2005).⁴ The estimate of the intrinsic hydrogen column (NH_int) is more uncertain. Observations of SN 2014J at optical wavelengths point to a large local extinction (see Section 3) corresponding to A_V = 1.7 ± 0.2 mag (Goobar et al. 2014). A more recent study by Amanullah et al. (2014) find consistent results around maximum light: A_V = 1.9 ± 0.1 mag. Assuming a Galactic dust-to-gas ratio, this would imply $\rm {NH}_{\rm {int}}\sim 4\times 10^{21}\,\rm {cm^{-2}}$ (Predehl & Schmitt 1995; Watson 2011). However, the low value inferred for the total-to-selective extinction R_V < 2 (Goobar et al. 2014; Amanullah et al. 2014) suggests the presence of smaller dust grains, more similar to the grain size distribution of the Small Magellanic Cloud (SMC). For an SMC-like dust-to-gas ratio the inferred hydrogen column is $\rm {NH}_{\rm {int}}\sim (7\hbox{--}15)\times 10^{21}\,\rm {cm^{-2}}$ (Martin et al. 1989; Welty et al. 2012). Gamma-ray burst (GRB) host galaxies sample instead the high end of the gas-to-dust ratio distribution (Schady et al. 2010). For a GRB-like environment the observed optical extinction would imply $\rm {NH}_{\rm {int}}\sim 10^{22}\,\rm {cm^{-2}}$. Based on these findings, in the following we calculate our flux limits and results for a fiducial range of intrinsic hydrogen column density values: $4\times 10^{21}\,\rm {cm^{-2}}<\rm {NH}_{\rm {int}}<10^{22}\,\rm {cm^{-2}}$. We adopt $\rm {NH}_{\rm {int}}\sim 7\times 10^{21}\,\rm {cm^{-2}}$ as "central value" in our calculations.⁵ It is worth noting that from the analysis of diffuse interstellar bands (DIBs) in M82, Welty et al. (2014) estimate a total hydrogen column density of ∼5 × 10²¹ cm⁻² in the direction of SN 2014J.

Based on these values, the unabsorbed XRT flux limit is $F_x^{3\sigma }<2.4\times 10^{-13}\,\rm {erg\,s^{-1}cm^{-2}}$ ($\rm {NH}_{\rm {int}}=7\times 10^{21}\,\rm {cm^{-2}}$), corresponding to $L_x^{3\sigma }<3.5\times 10^{38}\,\rm {erg\,s^{-1}}$. Restricting our analysis to 7 ≲ δt ≲ 15 days, when X-ray inverse Compton (IC) emission is expected to peak (Figure 2), we find a count-rate limit of 5.4 × 10⁻³ c s⁻¹ (exposure time of 27 ks, 0.3–10 keV), corresponding to $F_x^{3\sigma }<5.5\times 10^{-13}\,\rm {erg\,s^{-1}cm^{-2}}$ with $L_x^{3\sigma }<8.0\times 10^{38}\,\rm {erg\,s^{-1}}$.

We initiated deep X-ray follow up of SN 2014J with the Chandra X-ray Observatory on 2014 February 3, 20:10:39 UT (δt ∼ 20.4 days) under an approved Director's Discretionary Time proposal (PI: Margutti). Data have been reduced with the CIAO software package (version 4.6) and corresponding calibration files. Standard ACIS data filtering has been applied. The total exposure time of our observations is 47 ks. The observations mid-time corresponds to δt = 20.4 days since the explosion. No X-ray source is detected at the SN position (Figure 1) with a 3σ upper limit of 2.1 × 10⁻⁴ counts s⁻¹ in the 0.3–10 keV energy band, which implies an absorbed flux limit of F_x < 2.6 × 10⁻¹⁵ (erg s⁻¹cm⁻²) for an assumed F_ν∝ν⁻¹ spectrum. The count-rate limit is calculated as a 3σ fluctuation from the local background assuming Poisson statistics as appropriate in the regime of low-count statistics. The presence of diffuse soft X-ray emission from the host galaxy (Figure 1, left panel) prevents us from reaching deeper limits.

Zoom In Zoom Out Reset image size

The unabsorbed flux limit is $F_x^{3\sigma }<5.0\times 10^{-15}\,\rm {erg\,s^{-1}cm^{-2}}$ ($\rm {NH}_{\rm {int}}=7\times 10^{21}\,\rm {cm^{-2}}$, 0.3–10 keV energy band), with $(4.0<F_x^{3\sigma }<6.0)\times 10^{-15}\,\rm {erg\,s^{-1}cm^{-2}}$ for the range of NH_int values above. At the distance of 3.5 Mpc, the corresponding luminosity limit is $L_x^{3\sigma }<7.2\times 10^{36}\,\rm {erg\,s^{-1}}$, with $(5.7<L_x^{3\sigma }<8.7)\times 10^{36}\,\rm {erg\,s^{-1}}$.

WDs around the Chandrasekhar mass M_Ch accreting at a rate M_acc ≳ 3 × 10⁻⁷ M_☉ yr⁻¹ undergo steady hydrogen burning (e.g., Iben 1982; Nomoto 1982b; Prialnik & Kovetz 1995; Shen & Bildsten 2007). In this regime the WD accretes and retains matter. However, the system also loses material to the environment in three different ways: (1) wind from the donor star, (2) non-conservative mass transfer through RLOF, and (3) optically thick winds from the WD.

In symbiotic systems the wind from the giant companion star dominates the mass loss, with typical rates $\dot{M}=5\times 10^{-9}\;{\rm to}\;5\times 10^{-6}\,{M_{\odot }\,{\rm yr} ^{-1}}$ and wind velocities v_w < 100 km s⁻¹ (Seaquist & Taylor 1990; Patat et al. 2011; Chen et al. 2011). Our limit on the environment density around SN 2014J strongly argues against this progenitor scenario (Figure 4). From the analysis of the complete set of pre-explosion images (near UV to NIR) at the SN site acquired with the Hubble Space Telescope, Kelly et al. (2014) exclude SD progenitor systems with a bright red giant donor star. Consistent with these results, the mass-loss limit we derived from deep X-ray observations of the SN environment firmly and independently rules out this possibility and extends the conclusion to the entire class of red giant secondary stars.

Zoom In Zoom Out Reset image size

For systems harboring a main sequence, subgiant or helium star, the dominant source of mass loss to the surroundings is through non conservative mass transfer from the donor to the primary, accreting star. In this scenario the secondary star fills its Roche lobe and some of the transferred material is lost to the environment at the outer Lagrangian point: $\dot{M}_{\rm {trans}}= \dot{M}_{\rm {acc}}+\dot{M}_{\rm {lost}}$. We use the orbital velocity v ∼ a few 100 km s⁻¹ (or lower) as typical velocity of the ejected material. Observations of WDs in this regime indicate velocities up to ∼600 km s⁻¹ which we use in Figure 4 (e.g., Deufel et al. 1999). The efficiency of the accretion process ($\dot{M}_{\rm {acc}}/\dot{M}_{\rm {trans}}$) is poorly constrained. Our X-ray limit implies a very high efficiency of ≳ 99% (light-blue region in Figure 4) to avoid detectable signs of interaction with the lost material. A similar value was found for SN 2011fe using deep X-ray and radio observations (Margutti et al. 2012; Chomiuk et al. 2012). In this context Panagia et al. (2006) put a less constraining limit of >60%–80% on the efficiency of the same process using radio observations of a larger sample of 27 SNe Ia.

At high mass transfer rates, optically thick winds developing at the WD surface are expected to self-regulate the mass accretion to a critical value $\dot{M}_{\rm {acc}}\sim 7\times 10^{-7}\,{M_{\odot }\,{\rm yr}^{-1}}$, the exact value depending on the hydrogen mass fraction and WD mass (Hachisu et al. 1999; Han & Podsiadlowski 2004; Shen & Bildsten 2007). Our analysis argues against this progenitor scenario, even in the case of fast outflows with v_w ∼ a few 1000 km s⁻¹ (dark blue region in Figure 4).

Finally, thermonuclear burning of hydrogen rich material on the surface of a WD is expected to generate super-soft X-ray emission, which should be thus detectable in pre-explosion images at the SN site. Consistent with our findings above, Nielsen et al. (2014) found no evidence for a super-soft X-ray source at the position of SN 2014J using deep Chandra X-ray observations spanning the time range 1 ≲ t ≲ 14 yr before explosion. These observations allowed Nielsen et al. (2014) to rule out SD systems with high effective temperature of the super-soft emission kT_eff > 80 eV.

Non-steady mass enrichment of the progenitor surroundings in the years preceding the SN can create a low-density cavity around the explosion, which would naturally explain our X-ray null detection. The evacuated region around the progenitor system can either be (1) the consequence of the cessation of mass loss from the companion star or (2) the result of repeated nova shell ejections by the WD.

Recurrent novae. A WD near the Chandrasekhar mass (M ⩾ 1.3 M_☉) accreting at $\dot{M}_{\rm {acc}}\sim (0.1\hbox{--}3)\times 10^{-7}\,{M_{\odot }\,{\rm yr} ^{-1}}$ experiences repeated nova explosions as a result of unsteady hydrogen burning on its surface (e.g., Iben 1982; Starrfield et al. 1985; Livio & Truran 1992; Yaron et al. 2005). Shell ejections associated with the nova outbursts evacuate a region around the progenitor system by sweeping up the wind from the companion star. The result is a complex CSM structure, shaped by the fast nova shells and by the slower wind from the donor star. This scenario has been invoked to explain the evidence for interaction in the spectra of some Type Ia SNe (e.g., SN 2002ic; Wood-Vasey & Sokoloski 2006). The main outcome of this process is the presence of regions with very low density, reaching n_CSM ∼ 10⁻³–1 cm⁻³, in the proximity of the explosion center even in the case of very powerful winds from the donor star (Figure 5).

Zoom In Zoom Out Reset image size

The immediate SN environment critically depends on the time of the SN explosion with respect to the time of the last nova shell ejection, as illustrated by Figure 5. Figure 5 shows the results from the simulations by Dimitriadis et al. (2014) for a symbiotic progenitor system consisting of a WD plus red giant star losing material with rate $\dot{M} =10^{-6}\,{M_{\odot }\,{\rm yr} ^{-1}}$ and wind velocity v_w = 10 km s⁻¹. The nova recurrence timescale in this simulations is Δt = 100 yr.

From Figure 5 it is clear that our observations obtained at δt ∼ 20 days do not probe the range of densities associated with nova cavities.⁷ However, if the progenitor of SN 2014J experienced recurrent nova outbursts, our X-ray null detection implies that (1) a nova shell must have been able to clear out the environment at R ∼ 10¹⁶ cm and (2) the wind from the donor star has not yet been able to refill this volume by the time of the SN explosion.

The size of the cavity cleared out by a nova shell depends on the shell dynamics (e.g., Moore & Bildsten 2012). For typical shell ejection velocities v_sh = 1000–4000 km s⁻¹ and mass M_sh = 10⁻⁷–10⁻⁵ M_☉ (Yaron et al. 2005) expanding into a medium enriched by the secondary star mass loss with a rate $\dot{M}=10^{-7}\hbox{--}10^{-6} \,{{M}_{\odot }\ {\rm yr}^{-1}}$, the shell rapidly evolves from free expansion to the Sedov–Taylor phase on a timescale of t_ST ∼ a few days. It then transitions to the snowplow phase at t_SP ∼ a few months. Adopting v_sh = 4000 km s⁻¹, t_ST = 2 days, t_SP = 2 months as typical parameters as indicated by observations of the recurrent nova RS Oph (e.g., Bode & Kopal 1987; Mason et al. 1987; Sokoloski et al. 2006), a nova shell would reach R ∼ 10¹⁶ cm on a timescale of ∼40 yr. For a wind environment R∝t^2/3 during the Sedov–Taylor phase and R∝t^1/2 during the snowplow phase. The slower wind from the donor star would need Δt_refill ∼ 300 yr to refill this region with new material (for v_w = 10 km s⁻¹).

These results indicate that if the recurrent nova scenario applies to SN 2014J, then (1) at least one nova shell has been ejected at t ⩾ 40 yr before the SN and (2) the recurrence timescale of nova outbursts is shorter than 300 yr (consistent with the timescales observed in recurrent novae, the shortest known time between outbursts being ∼1 yr at the time of writing; Tang et al. 2014). In this respect it is worth mentioning that a complementary search for nova outbursts in archival optical images revealed no evidence for nova explosions at the location of SN 2014J in the ∼1500 days before the SN (Goobar et al. 2014). No evidence for interaction has been found for SN 2014J at δt ⩽ 20 days since the explosion, either in the form of light-curve re-brightenings, Hα emission, or time–variable Na D features (Goobar et al. 2014; Zheng et al. 2014; Tsvetkov et al. 2014), consistent with the expansion in a very clean environment. The same conclusion is supported by the deep limits on the radio synchrotron emission by Chomiuk et al. (2014) and Chandler & Marvil (2014).

Cessation of mass loss. Our observations at δt ∼ 20 days cannot exclude the presence of material in the environment at larger distances R ≳ 10¹⁶ cm and are thus consistent with any scenario that predicts cessation of mass loss from the progenitor system at t ⩾ 300 × (v_w/10 km s⁻¹)⁻¹ yr before the terminal explosion ("delayed explosion models" in Figure 4). This condition is naturally satisfied by (1) spin-up/spin-down models (Di Stefano et al. 2011; Justham 2011) and (2) core-degenerate (CD) scenarios for Type Ia SNe (where the WD merges with the core of an asymptotic giant branch, AGB, star; Ilkov & Soker 2012).

In both cases rapid rotation might stabilize the WD against explosion, allowing the WD to grow above M_Ch. The actual delay time τ between completion of mass transfer and explosion is uncertain and depends on the physical mechanism that regulates the WD spin-down (i.e., reduced accretion, ceased mass transfer, angular momentum transfer, magnetic breaking, gravitational waves). A wide range of values seems to be allowed by theory (τ ≲ 10⁵ yr to τ > 10⁹ yr, e.g., Lindblom et al. 1999; Yoon & Langer 2005; Ilkov & Soker 2012; Hachisu et al. 2012). Here it is worth noting that τ ⩾ 10⁵ yr is enough for the circumbinary material to become diffuse and reach ISM-like density values at the explosion site, consistent with our findings. Additionally, for τ ≳ 10⁸–10⁹ yr, even if the donors started the mass loss as giant, sub-giant, or main-sequence stars, by the time of the explosion they have exhausted most of their envelopes, their remaining envelopes have shrunk inside their Roche lobes and/or they are likely to have already evolved into compact objects (e.g., He or C/O WDs) or low-mass stars (e.g., Di Stefano et al. 2011; Justham 2011; Hachisu et al. 2012). All these effects would suppress any prominent signature of the mass-donor companion both before (i.e., in pre-explosion images) and after the explosion (i.e., anything that originates from the interaction of the SN shock with the medium), naturally accounting for our null detections.

In double-degenerate models the merger of two WDs leads to the final explosion. The general expectation is that of a "clean" environment at R ≳ 10¹³–10¹⁴ cm with ISM-like density.⁸ No stellar progenitor is expected to be detectable at the distance of SN 2014J in near UV to NIR pre-explosion images. Additionally, since the WDs were unlikely to burn hydrogen just prior to explosion, no X-ray source is expected to be found in pre-explosion images either. Our deep X-ray non-detection is consistent with this scenario, but can hardly be considered a confirmation of this progenitor channel. We also note that our results are consistent with the predictions of the CD scenario, where the actual merger involves a WD and the core of an asymptotic giant branch star, as suggested in the case of SN 2011fe (Soker et al. 2014).

WD–WD mergers are, however, expected to enrich the ISM in a number of ways including (1) tidal tail ejections (Raskin & Kasen 2013), (2) mass outflows during the rapid accretion phase before the merger (Guillochon et al. 2010; Dan et al. 2011), (3) winds emanating from the disk during the viscous evolution (Ji et al. 2013), and (4) shell ejections (Shen et al. 2013; see also Soker et al. 2013). Each of these mechanisms predicts the presence of relevant mass at different distances from the explosion center. Below we discuss the predictions of these scenarios at R ∼ 10¹⁶ cm which is relevant for the Chandra null detection of SN 2014J.

Tidal tail ejection. Some 10⁻⁴–10⁻² M_☉ of material is expected to be tidally stripped and ejected with a typical velocity v_ej ∼ 2000 km s⁻¹ just prior to the WD–WD coalescence (Raskin & Kasen 2013). The resulting structure of the CSM and the distance R_ej of the overdensity region created by the mass ejection from the explosion center critically depend on the delay time between the tidal tail ejection and the final explosion Δt_ej. In particular, the ejected material would require Δt_ej ∼ 10⁸ s to reach the distance of interest R_ej ∼ 10¹⁶ cm.⁹ Raskin & Kasen (2013) calculate that the tidal tail density at this distance would correspond to an effective mass-loss rate $\dot{M}\sim 10^{-2}\hbox{--}10^{-5}\,{M_{\odot }\,{\rm yr}^{-1}}$ which is ruled out by our observations. Our results thus argue against WD–WD mergers with delay times of the order of 1–10 yr, but allow for short delay times Δt_ej < 10⁶ s (including systems that detonate on the dynamical timescale Δt_ej ∼ 10²–10³ s or on the viscous timescale of the remnant disk Δt_ej ∼ 10⁴–10⁸ s). Long lag times Δt_ej ≫ 10⁸ s are also consistent with our findings since the interacting material would be located at larger distances.

Mass outflows during rapid accretion. Guillochon et al. (2010) and Dan et al. (2011) suggested that a phase of mass transfer and rapid accretion (with accretion rates reaching 10⁻⁵–10⁻³ M_☉ s⁻¹) might precede the actual WD–WD merger. As for SD systems, mass can be lost at the Lagrangian point, thus shaping the environment that the SN shock will probe later on. According to the simulations by Dan et al. (2011), by the time of the merger the mass of the unbound material is M_ej  ∼  10⁻²–10⁻³ M_☉ and could in principle produce observable signatures as soon as the SN shock and ejecta expand and interact with this material. The dynamics of the ejected material depends on v_ej, M_ej, and n_ISM. The location of the ejected material R_ej at the time of the SN explosion critically depends on Δt_ej. Both Δt_ej and v_ej are at the moment poorly constrained. For v_ej ∼ a few 1000 km s⁻¹ (i.e., comparable to the escape velocity), our deep density limit at R  ∼  10¹⁶ cm implies that the system did not eject substantial material in the 2–3 yr before the SN explosion.

Disk winds. WD–WD mergers that fail to promptly detonate have been recently suggested to produce a rapidly rotating, magnetized WD merger surrounded by a hot thick accretion disk before producing a Type Ia SN (Ji et al. 2013). A fraction of the disk mass is lost through magnetically driven winds. Ji et al. (2013) calculate that for near-equal mass WD, M_ej ∼ 10⁻³ M_☉ is gravitationally unbound and ejected with v_ej ∼ 2600 km s⁻¹. As for the other scenarios, a critical parameter is the interval of time between the mass ejection and the SN explosion Δt_ej (assuming that a SN occurs). The low density of SN 2014J at R ∼ 10¹⁶ cm suggests Δt_ej > a few yr.

Shell ejection. In the case of He + C/O WD progenitors in the context of the "double detonation" model (Livne 1990), Shen et al. (2013) have recently proposed that hydrogen-rich material can be ejected by the binary system in multiple ejection episodes hundreds to thousands of years prior to the merger (see their Figure 3). In their simulations Shen et al. (2013) find that a total of M_ej = (3–6) × 10⁻⁵ M_☉ of material can be transferred to the environment and it is ejected with initial velocity v_ej ∼ 1500 km s⁻¹ (i.e., comparable with the He WD circular velocity). In close analogy to nova shells, the ejected material interacts with the local ISM and shapes the local environment of the explosion (Shen et al. 2013, their Figure 4). Both the delay time between the onset of the mass-transfer and the disruption of the He WD and the time of the last ejection critically depend on the evolutionary history of the He WD, with the "older" He WD in the simulations of Shen et al. (2013) giving origin to mass ejections ∼200 yr before the terminal explosion. Our finding of a very low density environment at R ∼ 10¹⁶ cm constrains the most recent mass ejection in SN 2014J (if any) to have happened at t > 3 yr before the explosion.