SUPERNOVA EJECTA IN THE YOUNGEST GALACTIC SUPERNOVA REMNANT G1.9+0.3

G1.9+0.3 is the youngest known Galactic supernova remnant (SNR), with an estimated supernova (SN) explosion date of ∼1900, most likely located near the Galactic center at a distance of 8.5 kpc (Reynolds et al. 2008, hereafter Paper I; Green et al. 2008, hereafter Paper II). Its spectrum is dominated by nonthermal continuum well described by synchrotron emission from a power-law electron distribution with an exponential cutoff (Paper I; Reynolds et al. 2009, hereafter Paper III). Our 237 ks Chandra observation in 2009 July showed conclusively thermal emission from ejecta from the north rim (near the radio, but not the X-ray, maximum), with strong line emission from intermediate-mass elements (IMEs) such as Si and S, as well as from Fe (Borkowski et al. 2010, hereafter Paper IV). In addition, we found evidence for emission from ⁴⁴Sc, produced by the radioactive decay of ⁴⁴Ti.

Only the outermost ejecta layers with free-expansion velocities in excess of 18,000 km s⁻¹ have been shocked so far in this very dynamically young SNR (Carlton et al. 2011, hereafter Paper V). The presence of Fe at such high velocities is surprising, since few SNe show such high-velocity Fe (or ⁵⁶Ni). The spatial distribution, abundances, and physical properties of ejecta are key for understanding the explosion mechanism in this likely Type Ia SNR. Chandra observed G1.9+0.3 for nearly 1 Ms in 2011, which clarified the ⁴⁴Sc detection (spectrum of the interior is shown in Borkowski et al. 2013) and provided detailed images and spectra. Here we report spatially resolved spectroscopy of SN ejecta, and interpret the results in the framework of an energetic and asymmetric Type Ia explosion.

Chandra observed G1.9+0.3 in 2011 May and July (see Table 1). The ACIS S3 chip was used in Very Faint mode. Data reduction and spectral extraction were done with CIAO v4.5 and CALDB v4.5.1. No significant particle flares were found. The total duration was 977 ks.

In order to examine spatial structures in spectral lines and continua, data cubes, 256³ in size, were extracted from merged event files. The spatial scale is one ACIS pixel (0492 × 0492), while spectral channels from 84 to 595 were binned by a factor of two (these correspond to the energy range of 1.2–8.7 keV, including almost all source photons). At the ACIS S3 spectral resolution, the binning introduces negligible spectral degradation above 1.2 keV.

X-ray spectra from various locations within the remnant were then extracted from each data set listed in Table 1, and then summed. The spectral responses were averaged using effective exposure times as weights. The background was modeled rather than subtracted, by fitting the spatially integrated spectrum of G1.9+0.3 together with a background spectrum. The X-ray analysis was performed with XSPEC v12.7.1 (Arnaud 1996), using C-statistics (unless mentioned otherwise) and APEC NEI v2.0 atomic data augmented with inner-shell processes (Badenes et al. 2006). We also updated inner-shell atomic data for all ions of Si, S, Ar, and Ca with more than three electrons present. Relevant collisional excitation cross sections and inner-shell ionization cross sections by electrons were calculated with the Los Alamos Atomic Physics Codes⁶ (Mann 1983; Clark et al. 1991). Line energies and branching ratios for radiative transitions are from Palmeri et al. (2008). Only strong resonance lines have been included.

Previous Chandra observations of G1.9+0.3 revealed thermal emission lines from ejecta in the radio-bright northern rim (Paper IV), but poor photon statistics precluded a search for spatial variations in composition. We searched for variations in the line emission within the 2011 May Chandra data cube. First, this cube was denoised using the spectro-spatial method of Krishnamurthy et al. (2010). Smoothed images at low, medium, and high energies were extracted from a more heavily denoised data cube containing all 2011 observations, and combined into a color image (Figure 1). Spectral variations are apparent, with the bright east (E) and west (W) lobes noticeably harder than the much fainter south (S) and north (N) rims, confirming azimuthal variations in the synchrotron spectra (Paper III).

Zoom In Zoom Out Reset image size

In our search for spectral lines within this cube, we used a continuous wavelet transform method to spatially localize ejecta emission. We use the package MassSpecWavelet developed by Du et al. (2006) for mass spectrometry data. We concentrated on Si, S, and Fe, the most abundant ejecta elements whose lines are prominent in young Type Ia SNRs, and which are also present in the large radio-bright N region of Paper IV. (We do not confirm the presence of Ca and Sc there, and a very weak Ar line is not statistically significant.) For the Fe Kα line, we considered all possible emission line candidates identified by MassSpecWavelet in a relatively broad (6.12–6.76 keV) energy range near 6.4 keV in order to allow for potentially large Doppler shifts. For Si and S Kα lines, our chosen energy ranges were 1.65–1.97 keV and 2.15–2.53 keV, respectively. In order to reduce the effects of noise, maps of estimated line strengths were examined for spatially contiguous regions of potential line emission. (In synchrotron-dominated SNRs where thermal and nonthermal emission is poorly correlated, this procedure could result in regions not obviously corresponding to spatial structures visible in broadband images.) For these regions, we then extracted and combined X-ray spectra from all 2011 observations. Here we discuss only selected regions with unambiguous line emission present, likely having highest density and shocked ejecta mass.

We identified four regions where either strong Fe Kα line and/or both Si and S Kα lines are present (see Figures 1 and 2). Three regions mostly overlap with the northern rim prominent in radio: inner and outer northern rims, and a smaller northeastern (NE) region. This confirms that most ejecta emission is associated with the radio-bright N rim. The fourth region in the west overlaps with the innermost X-ray and radio emission in this part of the remnant, demonstrating that ejecta emission is present elsewhere than just the northern rim. X-ray spectra are shown in Figure 3, with simple fits consisting of absorbed power-law continua and Gaussian lines (see Table 2 for fitted parameter values). All lines are broad (FWHM ∼15,000 km s⁻¹; we assumed the same Doppler width for the IMEs and Fe). Variations in X-ray spectra are apparent: Fe Kα is prominent in the northern rim regions, but not in the western rim where only Si and S lines are present. The outer northern rim exhibits strong Fe emission, but Si is weak and S is missing. Both Fe and the IMEs (Si + S) are present in the two remaining regions in the north, and are particularly strong in the inner northern rim.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Based on line identifications from the Gaussian fits, we fit vpshock models with variable abundances as given in Table 2. Here we focus on the inner northern rim which has the richest line emission. An absorbed vpshock fit to its spectrum gives, with 90% confidence intervals, a hydrogen column density N_H = 7.0(6.3, 7.8) × 10²² cm⁻², plasma temperature kT of 3.2(2.6, 3.9) keV, ionization age τ = 2.4(1.1, 4.0) × 10⁹ cm⁻³ s, and the following abundances relative to solar values of Grevesse & Sauval (1998): 3.0(1.5, 5.1) for Si, 4.1(2.8, 5.5) for S, and 7.4(4.1, 12) for Fe. Lines are Doppler-broadened (FWHM = 15,000 (9200, 21,000) km s⁻¹) and blueshifted with a velocity of 2700(600, 4900) km s⁻¹. Ejecta are blueshifted at a high confidence level of 98.0%. In this fit, we used Markov chain Monte Carlo (MCMC) methods (with uniform priors) instead of relying on C-statistics. Unlike the latter, estimated parameter values and their errors obtained with MCMC methods do not suffer systematic biases, and are more reliable. In particular, comparison with the FWHM value of 12,000 (8100, 18,000) km s⁻¹ for the inner N rim from Table 2 suggests that some (or all) line widths listed there slightly underestimate true line widths.

The NE rim spectrum is similar to the inner N rim spectrum, with best-fit values of N_H = 7.3(6.5, 9.1) × 10²² cm⁻², kT = 3.5(2.2, 4.5) keV, τ = 1.0(0, 4.8) × 10⁹ cm⁻³ s, Si = S =2.6(1.2, 4.6), Fe =2.6(0.8, 5.0). Lines might be redshifted by 4200(− 1200, 18,000) km s⁻¹ and broad (FWHM = 19,000 (7600, 43,000) km s⁻¹). While the fitted temperatures and ionization ages for the N and NE rims are similar, the relative velocity inferred from the fits is 6900 km s⁻¹, and such high velocity is supported by the large line broadenings required for both regions. We can thereby infer that the systematic line shifts for various regions seen in Table 2 are mostly caused by bulk motions, though ionization effects may still play a role.

In the vpshock fit to the inner N rim spectrum, the X-ray continuum is produced by free–free emission on protons and α particles, and we set the abundances of Ar and Ca equal to S. Strongly oversolar abundances of Si, S, and Fe confirm that emission lines are produced by SN ejecta. Mass emission measures (EMs) are n_eM_Si(S) = 4.7(4.5) × 10⁻⁴ cm⁻³ M_☉ for the IMEs, and n_eM_Fe = 2.1 × 10⁻³ cm⁻³ M_☉ for Fe. The total mass EM is equal to 0.22 cm⁻³ M_☉, and with n_e ≈ 2τ/t_SNR = 1.5 cm⁻³, the inferred shocked mass is 0.14 M_☉, several times larger than the spatially integrated shocked ejecta mass of 0.033 M_☉ (derived from the overall remnant's dynamics, and assuming a standard Type Ia explosion; Paper V; see also Section 4). This discrepancy suggests that the assumption that the continuum is thermal is incorrect; given a nonthermal continuum, it is likely that pure heavy-element ejecta, without any H or He, produce the lines seen in Figure 3. This makes Si, S, and Fe EMs somewhat uncertain, particularly for Fe if plasma temperatures are significantly higher than 3.2 keV. As an example, a fit to the inner N rim with a power-law model and a plane shock model with an assumed temperature of 10 keV and without any H or He gives EMs of n_eM_Si(S) = 4.6(2.6) × 10⁻⁴ cm⁻³ M_☉ for the IMEs, and 2.4 × 10⁻⁴ cm⁻³ M_☉ for Fe. A similar fit to the outer N rim spectrum gives 1.7 × 10⁻⁴ cm⁻³ M_☉ for Si and 2.1 × 10⁻⁴ cm⁻³M_☉ for Fe. In the W rim, n_eM_Si(S) = 8.9(2.6) × 10⁻⁴ cm⁻³ M_☉, with τ = 0.14(0, 1.4) × 10⁹ cm⁻³ s and velocity of −1200(− 3200, 3100) km s⁻¹. Since the upper bound of 1.4 × 10⁹ cm⁻³ s on τ is less than our best estimate of ionization age in the inner N rim, ejecta densities likely decrease from N to W, and then continue decreasing south where ejecta emission becomes much fainter.

The presence of shocked heavy-element ejecta in G1.9+0.3 is surprising, since all ejecta that have been shocked so far were expelled by the SN with velocities exceeding ∼18,000 km s⁻¹ (Paper V). In particular, Fe (or ⁵⁶Ni) is not commonly found at such high velocities in SN spectra, except for unusually energetic Type Ia SNe and rare core-collapse (CC) hypernovae. In G1.9+0.3, Fe is prominent only in the northern rim, not in the western rim where only Si and S lines have been detected, implying a strongly asymmetric SN explosion. There is no evidence for any jet-like features expected in jet-triggered energetic CC SNe, and other considerations (Paper I) also favor a Type Ia origin, so in the following discussion we focus on Type Ia explosions.

Studies of high-velocity Fe-group elements in Type Ia SN spectra are rare since they require frequent acquisition of optical and UV spectra soon after the SN explosion. The most detailed study so far is of a normal albeit bright Type Ia SN, SN 2010jn (Hachinger et al. 2013), with a broad light curve indicating substantial amounts of ⁵⁶Ni. In their preferred SN model, Hachinger et al. (2013) found good agreement with observations only when they empirically adjusted abundances in the delayed detonation model WS15DD3 of Iwamoto et al. (1999). Fe-group elements are present even at velocities exceeding 30,000 km s⁻¹; at 20,000 km s⁻¹, the mass fraction of ⁵⁶Ni and decay products is ∼5%, with an additional ∼2% of directly synthesized Fe present. No published Type Ia model has enough high-velocity Fe-group elements to account for their presence in SN 2010jn.

We can use the results of Hachinger et al. (2013) to estimate the mass and EM of shocked ejecta in G1.9+0.3. The density structure in the WS15DD3 model can be approximated well by an exponential ejecta model of Dwarkadas & Chevalier (1998) with the explosion energy E₅₁ (in units of 10⁵¹ erg) of 1.43 and the ejecta mass equal to the Chandrasekhar mass. The ejecta mass M_ej(v) for free-expansion velocities larger than v is equal to $M_{{\rm ej}}/M_{{\rm ej}}^{{\rm tot}} = 0.5 [(v/v_e)^2 + 2 v/v_e + 2] \exp (-v/v_e)$, where $M_{{\rm ej}}^{{\rm tot}}$ is the total ejecta mass and v_e is the exponential velocity scale (ρ_ej∝exp (− v/v_e)/t³). With E₅₁ = 1.43 and $M_{{\rm ej}}^{{\rm tot}} = 1.38\, M_\odot$, v_e = 2940 km s⁻¹ instead of v_e = 2440 km s⁻¹ (for assumed E₅₁ = 1) in Paper V. By repeating their analysis for this more energetic SN model, we arrive at v/v_e = 6.4, v = 19,000 km s⁻¹, and ρ_ej = 0.12 amu cm⁻³ for ejecta entering the reverse shock in G1.9+0.3 at the present time. The total shocked ejecta mass is equal to 0.064 M_☉, nearly twice as large as 0.033 M_☉ in the model of Paper V with v/v_e = 7.3. We estimated EMs at EM_Si ∼ 6 × 10⁻⁴ cm⁻³ M_☉, EM_S ∼ 1 × 10⁻⁴ cm⁻³M_☉, and ${\rm EM_{^{56}Ni+Fe}} \sim 3 \times 10^{-4}$ cm⁻³ M_☉, using abundances of 15%, 3%, and 7% (by mass; Hachinger et al. 2013), respectively. These spatially integrated EMs are comparable to the EMs we find in the inner northern rim (for a temperature of 10 keV), perhaps suggesting that ejecta there are several times denser than in the WS15DD3 model.

Most Type Ia SN models lack the high-velocity Fe-group elements seen in SN 2010jn and G1.9+0.3. Meakin et al. (2009) report the presence of substantial (>10% by mass) amounts of ⁵⁶Ni up to velocities of 18,500 km s⁻¹ in their gravitationally confined detonation models. High-velocity (up to 18,000 km s⁻¹) ⁵⁶Ni is present in only one of a dozen three-dimensional (3D) delayed-detonation models of Seitenzahl et al. (2013). In this model (N3), deflagrations were initiated in only three central locations within the white dwarf (WD), producing a strongly asymmetric and energetic explosion that synthesized nearly 0.7 M_☉ of ⁵⁶Ni. All explosions with a small number of ignition kernels resulted in energetic explosions, but fast-moving (v ≳ 12,000 km s⁻¹) ⁵⁶Ni is present only in models N3 and N5. In both models, part of a deflagration plume had already risen to the surface of the WD when the first delayed detonation occurred. The outermost deflagration ashes present within this plume were then accelerated to high velocities by the SN shock wave. G1.9+0.3 might have been produced by such an explosion. The strongly asymmetric distribution of Fe in G1.9+0.3 supports this hypothesis, since it is in qualitative agreement with the asymmetric distribution of ⁵⁶Ni in the N3 model. (The absence of strong azimuthal variations in expansion rate (Paper V) disfavors an external density gradient as an alternative explanation.)

The ejecta emission in G1.9+0.3 appears stronger than expected in the heuristic SN 2010jn model of Hachinger et al. (2013). One weakness of that model is the assumption of spherical symmetry. The SN explosion could have been strongly asymmetric, as in the N3 model or in the gravitationally confined detonation models of Meakin et al. (2009) where the density profiles are much steeper in the direction of the off-center detonation site than on the other side of the WD, leading to an order of magnitude variation in density of high-velocity ejecta. This could explain the lack of perfect agreement between predictions based on the heuristic model of Hachinger et al. (2013) for SN 2010jn and Chandra observations of G1.9+0.3. The heavy-element ejecta in the northern and western rims of G1.9+0.3 were probably several times denser than in the spherically symmetric WS15DD3 model, perhaps because of their location opposite to the off-center detonation site.

The coincidence of the radio maximum with a region of substantial Fe suggests the possibility of a contribution to the radio flux from positrons from the decay of ⁵⁶Co injected into the shock-acceleration process with initial energies of ∼1 MeV. 20% of ⁵⁶Co decays produce a positron (Bé et al. 2006); the ∼10⁻³ M_☉ of initial ⁵⁶Ni we infer from the inner N rim will eventually produce a total of 4 × 10⁵¹ positrons. However, those injected at early times will lose energy in the dense ejecta due to ionization losses; once thermalized, they will rapidly annihilate. Martin et al. (2010) estimate the slowing-down time for positrons emitted at time t as $t_{\rm sd} \sim 10^{-3} E_{51}^{3/2}/M_{\rm ej}^{5/2} t^3$ yr, with M_ej in M_☉. Then positrons emitted earlier than t_esc given by t_sd = t will annihilate. For E₅₁ = 1 and M_ej = 1.4 M_☉, t_esc ∼ 50 yr. Given the 111 days mean life of ⁵⁶Co, an insignificant fraction of positrons is expected to survive to late times, unless they are accelerated first. The observed radio flux from the inner N rim is about 150 mJy at 1.4 GHz; assuming a spectral index of 0.62 (Paper III) and a power-law distribution of electrons down to a minimum Lorentz factor of 10, standard synchrotron formulae give

$$\begin{equation} N_e = 2.7 \times 10^{50} (B/100\ \mu {\rm G})^{-1.62}\ {\rm electrons,} \end{equation}$$

where nonlinear shock amplification would be required to reach magnetic field strengths of the order of 100 μG. For this value, all positrons emitted after about 300 days would need to survive and be accelerated. This seems exceptionally unlikely. We conclude that decay positrons cannot account for the radio maximum in the N rim; further radio observations under way may cast light on this issue.

G1.9+0.3 offers us a unique view of the outermost, strongly asymmetric Fe-rich ejecta of a likely Type Ia SNR. Their spatial distribution should allow for distinguishing among various Type Ia explosion models. Energetic delayed-detonation models are promising candidates as they may account for both the observed asymmetry and the presence of high-velocity Fe. The delayed-detonation N3 model of Seitenzahl et al. (2013) produces asymmetric ejecta rich in Fe-group elements, although not at high enough free-expansion velocities. High-velocity radioactive ⁵⁶Ni affects early-time Type Ia SN light curves (Piro 2012; Piro & Nakar 2013, 2012). Its highly asymmetric spatial distribution, as observed in G1.9+0.3 and obtained in recent 3D Type Ia simulations such as the N3 model, should be taken into account when interpreting light curves and spectra of Type Ia SNe. Future 3D hydrodynamical simulations of asymmetric Type Ia explosions, including additional delayed-detonation models, followed by modeling of their collision with the ambient interstellar medium, would be most helpful for understanding the most recent known SN explosion in our Galaxy. Such strongly asymmetric explosions are expected to produce azimuthal variations in expansion rate that can be measured with future Chandra and Very Large Array observations. In a broader context, studies of this and other young Type Ia SNRs that show large-scale asymmetries in their Fe-rich ejecta distribution (e.g., SN 1006 and Kepler's SNR; Yamaguchi et al. 2008; Uchida et al. 2013; Burkey et al. 2013) are invaluable for understanding Type Ia explosions and their mysterious progenitors.

This work was supported by NASA through Chandra General Observer Program grants SAO G01-12098A and B.