Overionized plasma in the supernova remnant Sagittarius A East anchored by XRISM observations

Figure LABEL:fig-image shows 6.45–6.8 keV (Fe He$\alpha$) Xtend and Resolve images extracted from the observation of Sgr A East. One can see that the Resolve observation of Sgr A East is highly affected by the bright transient source (low-mass X-ray binary), AX J$1745.6-2901$ (Hyodo et al., 2009), whose X-ray emission has a black-body continuum with absorption lines at Fe He$\alpha$ and Ly$\alpha$ and thus requires a careful treatment. Further assessment can be found later in this section. The Resolve spectrum extracted from the source region is shown in Figure LABEL:fig-spec, demonstrating that the Fe He$\alpha$ triplet is successfully separated for the first time.

Our primary interest is the Fe He$\alpha$ and Ly$\alpha$ line intensities. We perform plasma diagnostics based on these intensities measured with a semi-phenomenological model rather than simply relying on plasma models. This treatment is to minimize the potential bias due to the other spectral features including the high and uncertain background. The Resolve spectrum clearly exhibits evidence for an overionized plasma, i.e., enhanced Fe He$\alpha$ z/w and x/y ratios and coexistence of Fe ions with a broad range of charge states (H-like, He-like, and lower-ionized ones like Li-like and Be-like). We apply Lorentzian models with a Gaussian-like broadening to obtain the individual fluxes of the Fe He$\alpha$-z, y, x, w, Ly$\alpha$₁, and Ly$\alpha$₂ (six in total). To explain the other spectral features including weak lines and continua, we apply a two-temperature overionized plasma model (bvvrnei + bvvrnei in XSPEC) plus a power-law function (Ono et al., 2019) without the Fe He$\alpha$ and Ly$\alpha$ lines (see Suzuki et al. (2020) for the methodology).¹¹1The origin of the power-law component is thought to be either an ensemble of point sources or non-thermal emission associated with filaments in the remnant (Koyama et al., 2007), or many faint X-ray reflection nebulae (Ono et al., 2019). The Lorentzian widths (FWHM) are fixed to the individual natural widths. The Gaussian-like broadening (same for all the six lines in velocity dispersion) is tied to that of the bvvrnei model. All the parameters of the low-temperature component are fixed to the values in Ono et al. (2019) except for the emission measure, the ratio of which relative to the high-temperature component is fixed to the value in Ono et al. (2019). The contribution of the low-temperature component at Fe He$\alpha$ and Ly$\alpha$ is below 5% (Ono et al., 2019) and thus its uncertainty does not affect the discussion below. The power-law index is fixed to 1.0 (Ono et al., 2019). The electron temperature $kT_{\rm e}$, recombination timescale $\tau$, emission measure, velocity dispersion, and redshift of the high-temperature component, and power-law flux are treated as free parameters. The initial ionization temperature $kT_{\rm init}$ is fixed to 10 keV (Ono et al., 2019) because the model itself is insensitive to it without the six major lines. The interstellar absorption column, which is unimportant in the energy range we use, is fixed to $1.5\times 10^{23}$ cm^-2 (based on Ono et al. (2019), roughly consistent with Zhou et al. (2021)). As the input emission profile to generate ARFs, we use an Fe He$\alpha$ image obtained with Chandra.

The background of our observation is dominated by the Galactic center X-ray emission (GCXE; see Koyama (2018) for review) and AX J$1745.6-2901$. We determine the spectral model for AX J$1745.6-2901$ using an Xtend spectrum (ObsID: 300044010) extracted from a circle with a $1^{\prime}$ radius centered at (R.A., Dec.) = (266\fdg3985, $-29\fdg 0261$). The emission is approximated with a simple blackbody spectrum. We first ignore the absorption features at Fe He$\alpha$ and Ly$\alpha$ (Trueba et al., 2022), which are time-variable. The uncertainty due to this treatment is evaluated later in this section. When we apply this model to the Resolve spectra, we only allow the normalization to vary within $\pm\,50\%$ to account for uncertainties in the ray-tracing software when generating ARF, with the other parameters fixed. Meanwhile, we empirically model the spectrum extracted from the FoV of the nearby-sky observation (ObsID: 300045010) to obtain the GCXE spectral shape.²²2Details of the GCXE spectrum itself will be reported in a separate paper. Only the surface brightness is left as a free parameter, which is assumed to be the same for the source and background regions. The effect of the possible spatial dependence is evaluated later in this section (Muno et al., 2004; Chatzopoulos et al., 2015). As the emission profiles to be inputted to ARFs, we assume a point-like and flat sources for the AX J$1745.6-2901$ and GCXE, respectively.

We simultaneously fit the source and background spectra with the combined model, Sgr A East + GCXE + AX J$1745.6-2901$, taking into account the mutual contamination of each component between the two regions. We allow the normalization of the Sgr A East model for the background spectrum to vary within $\pm\,50\%$ relative to that for the source region to account for uncertainties in the ray-tracing software and thus in the ARF. As a result, we obtain an acceptable fit with a $C$-stat/d.o.f. $=2975.1/2589$ (Figure LABEL:fig-spec). The obtained parameters are summarized in Table 1. We successfully constrain the intensity ratios as Fe He$\alpha$-z/He$\alpha$-w $=1.39\pm 0.12$, Ly$\alpha$/He$\alpha$-w $=0.36\pm 0.07$. The best-fit Ly$\alpha$₁/Ly$\alpha$₂ intensity ratio is consistent with $\approx 2$, which is generally expected in thermal plasma. The constrained velocity dispersion, $\approx 109$ km s^-1, corresponds to an Fe ion temperature of $\approx 8$ keV if the broadening is purely due to thermal motion. One may expect to detect a strong radiative recombination edge of He-like Fe ions at $\approx 8.83$ keV as was the case for W49B (Ozawa et al., 2009). Although our model indeed reproduces a similar feature because of the spectral parameters similar to W49B (e.g., Yamaguchi et al. (2018)), the feature is much less noticeable because of its much lower flux than the contamination from the power-law component of Sgr A East, AX J$1745.6-2901$, and GCXE.

We evaluate the contribution of two processes which possibly alter the Fe He$\alpha$ forbidden-to-resonance ratio, i.e., the Fe He$\alpha$ resonance scattering and charge exchange (CX) X-ray emission. Details are described in Appendix A. The possible flux decrease of Fe He$\alpha$-w due to scattering is estimated to be $\approx 5\%$ at most. The possible contribution of the CX emission is also evaluated to be $<10\%$ to the Fe He$\alpha$-z/He$\alpha$-w and Ly$\alpha$/He$\alpha$-w ratios. Thus, both will not affect the plasma parameters significantly.

To account for systematic uncertainties in the background estimation, we consider uncertainties associated with the GCXE flux and Fe-K absorption features of AX J$1745.6-2901$. Based on the Fe He$\alpha$ flux distribution over a $17^{\prime}\times 17^{\prime}$ region around Sgr A^∗ (Muno et al., 2004) and the projected stellar mass ratio of the source and background regions (Chatzopoulos et al., 2015), we take into account the possible spatial dependence of GCXE by scaling the best-fit GCXE flux of the source region by $+40\%$ and search for the best-fit parameters again. The Fe He$\alpha$ and Ly$\alpha$ fluxes are slightly changed but within $\approx 10\%$, both smaller than the statistical errors and are insignificant. We evaluate the effect of the absorption features of AX J$1745.6-2901$ by replacing the model with a one with the K-shell absorption lines of ionized Fe determined with the Xtend spectrum and confirmed with the Resolve spectrum extracted from westmost pixels (dashed gray line in Figure LABEL:fig-spec)³³3Details of the spectral features of AX J$1745.6-2901$ will be reported in a separate paper.. The resultant line fluxes of Sgr A East are slightly increased, by $\approx 18\%$ at Fe Ly$\alpha$, yielding insignificant changes in the plasma parameters.

We compare the determined line intensities with the plasma models in AtomDB to constrain the plasma state. Figure LABEL:fig-diag (a) indicates that the CIE plasma model can not explain the observed high forbidden-to-resonance intensity ratio with a reasonable range of electron temperature, where a sufficient amount of both He- and H-like Fe atoms are present. Figure LABEL:fig-diag (b) shows that the overionized plasma model can explain the observed line intensity ratios with sufficiently high initial ionization temperatures $kT_{\rm init}$. Figure LABEL:fig-diag (c) and (d) quantify the constrained ranges of the electron temperature, present and initial ionization temperatures, and recombination timescale based on the Fe He$\alpha$-z/He$\alpha$-w and Ly$\alpha$/He$\alpha$-w intensity ratios. The electron temperature $kT_{\rm e}$ and present ionization temperature $kT_{\rm z}$⁴⁴4CIE temperature equivalent to the ionization state which explains the observed line ratios (e.g., Masai (1994)). are estimated to be $kT_{\rm e}=1.6\pm 0.2$ keV and $kT_{\rm z}=4.7\pm 0.4$ keV, respectively. The initial ionization temperature is thus constrained to be $kT_{\rm init}>4$ keV. Note that no $kT_{\rm init}$ values can explain the observations when $\tau>8\times 10^{11}$ s cm^-3, as such timescales lead to much lower ionization states than those observed.

The electron temperature and recombination timescale directly constrained from the Fe He$\alpha$ and Ly$\alpha$ line ratios are consistent with those shown in Table 1. This fact means that the plasma state required from the Fe He$\alpha$ and Ly$\alpha$ lines can explain other spectral features including the satellite lines as well. Thus, we apply a model without replacing the Fe He$\alpha$ and Ly$\alpha$ lines with Lorentzian functions for Sgr A East and indeed obtain a very similar fit, with $C$-stat/d.o.f. = 2976.1/2595.

The derived recombination timescale, $\tau<8\times 10^{11}$ s cm^-3, can be used to evaluate the age of Sgr A East. If we assume that the Fe-K emitting plasma is dominated by ejecta, which is reasonable given its center-filled distribution and over abundance (Koyama et al., 2007; Ono et al., 2019), we can derive the plasma density: typical CSM + ejecta masses of $\gtrsim 1$ solar with the radius of the Fe He$\alpha$ emission traced with Chandra, 1.6 pc yield a plasma density $n_{\rm e}\gtrsim{1}$ cm^-3 for both solar-abundance or pure-Fe plasma, which are similar to the estimates from the plasma emission measures in previous studies (Maeda et al., 2002; Sakano et al., 2004; Koyama et al., 2007; Ono et al., 2019) and this work. The age of the remnant should be larger than the elapsed time after the overionization occurred, $\tau/n_{\rm e}\approx 2500~{}{\rm yr}\,(\tau/8\times 10^{11}~{}{\rm s}~{}{\rm cm}^{-3})(n_{\rm e}/10~{}{\rm cm}^{-3})^{-1}$. This will be more meaningful once some additional constraints are obtained.

The velocity dispersion of $\approx 109$ km s^-1 is converted to an upper-limit expansion velocity of $\approx 200$ km s^-1 (XRISM Collaboration, 2024). If we assume that the Fe-K emitting plasma is dominated by ejecta, expansion velocities will be kept high, such as $\gtrsim 1000$ km s^-1, even after shock-heating (e.g., Sato & Hughes (2017); XRISM Collaboration (2024)). Thus, the small expansion velocities inferred in Sgr A East will require certain processes to suppress the expansion. The inferred high ambient density is qualitatively consistent in this sense. Alternatively, if the progenitor wind produced dense circumstellar material (CSM), it would naturally cause a suppression of the expansion. The low Fe ion temperature $<8$ keV, along with the high electron temperature $\approx 1.6$ keV, generally requires a long thermal relaxation timescale $\gtrsim 10^{11}$ s cm^-3 and a moderate shock velocity $\lesssim 2000$ km s^-1 in the unshocked ejecta frame, given non-equilibrium of the temperatures of different elements immediately behind shocks (XRISM Collaboration, 2024; Ohshiro et al., 2024). We note that this may not hold for Sgr A East because the evolutionary path may have been unusual, e.g., the remnant may have experienced a quick equilibration behind the shock due to dense CSM as we discuss in Section 4.2.

The upper limit of the observed redshift of Fe He$\alpha$ is converted to a line-of-sight velocity of $\approx 20$ km s^-1. This supports the idea that Sgr A East is very close to Sgr A^∗: line-of-sight component of the Galactic rotation velocity, $<50$–150 km s^-1 away from the solar system at a distance of 1–10 pc from Sgr A^∗ (Lugten et al., 1986), may be observed as $<20$ km s^-1 after applying the standard corrections for the line-of-sight velocities of the solar system ($\approx 10$ km s^-1) and Earth ($\approx 28$ km s^-1). This estimate is also similar to the velocities of molecular clouds in the vicinity ($\sim 50$ km s^-1: e.g., Tsuboi et al. (1999, 2011)).

Here we discuss how the overionized plasma in Sgr A East was produced. We briefly examine three candidate cases which are widely considered as the origin of the overionization in supernova remnants, rapid cooling of the plasma via (a) thermal conduction with cold material (Kawasaki et al., 2002) or (b) adiabatic expansion from dense CSM (Itoh & Masai, 1989), and (c) enhanced ionization due to irradiation of charged particles or photons (e.g., Ono et al. (2019)). In the cases (a) and (b), the ionization temperature $kT_{\rm z}$ before a transition to overionization states is similar to or lower than the electron temperature $kT_{\rm e}$ assuming an ordinary evolution. A rapid cooling of electrons causes a transition and realizes $kT_{\rm z}>kT_{\rm e}$. In the case (c), the electron temperature is kept the same before and after the transition, and only the ionization state is quickly enhanced and realizes $kT_{\rm z}>kT_{\rm e}$.

If we assume that the thermal conduction (a) was the origin of the overionization in Sgr A East, the electron temperature before the transition should have been $>4$ keV. It is hard to realize such high electron temperatures in supernova remnants in general. In fact, electron temperatures this high have never been observed in remnants, although may be possible theoretically in such a dense environment, where the plasma would quickly reach CIE. The thermal conduction scenario would not naturally explain such high initial temperatures due to the expected ordinary evolution before the transition, and thus is less favorable.

The adiabatic expansion scenario (b) may satisfy the high initial ionization temperature because it assumes the existence of dense CSM, developing a strong reverse shock at an early stage ($\sim 100$ yr; Itoh & Masai (1989)). The resultant higher shock velocity and higher density than without dense CSM in combination quickly achieve a high ionization state. Indeed, X-ray observations performed within $\sim 100$ days of several supernovae exploding in dense CSM suggested the presence of plasma with high electron temperatures $>10$ keV (e.g., Chandra et al. (2024)). Since the CSM breakout should occur within a few 100 yr, the recombination timescale should be similar to or larger than the “age times present plasma density”. Thus, if the age of the remnant is smaller than $\approx 25000~{}{\rm yr}\,(n_{\rm e}/1~{}{\rm cm}^{-3})^{-1}$, this scenario can explain the recombination timescale measured in our spectroscopy. Better constraints on the age is desired to test this scenario more precisely.

Realizing sufficiently high ionization states to explain the observed overionized plasma by enhanced ionization (c) will be hard in general in the case of collisional ionization with charged particles (Sawada et al., 2024). Another possibility is photoionization by a luminous X-ray source. Having such a high luminosity source in the vicinity of a remnant is rare and the photoionization scenario has never been considered as a plausible origin for overionized remnants (e.g., Kawasaki et al. (2002)). In the case of Sgr A East, however, intense photoionization may be possible because of the presence of Sgr A^∗ in the immediate vicinity. As already suggested by Ono et al. (2019), a luminosity $L\gtrsim{10^{40}}$ erg s${}^{-1}\,(n_{\rm e}/1~{}\text{cm}^{-3})(R/1~{}\text{pc})^{2}$, with the $R$ being the distance from Sgr A*, is required to realize an ionization state equivalent to the CIE with 5 keV ($\log\xi\approx 3$). This lies in between those of the hypothetical outbursts of Sgr A^∗, $L\sim 10^{39}$ erg s^-1 a few 100 yr ago (Ryu et al., 2009, 2013) and $L\sim 10^{44}$ erg s^-1 $\sim 10^{5\text{--}6}$ yr ago (Su et al., 2010; Nakashima et al., 2013). Thus, if a past outburst of Sgr A^∗ with $L\gtrsim{10^{40}}$ erg s^-1 occurred $\lesssim 2500~{}{\rm yr}\,(n_{\rm e}/10~{}{\rm cm}^{-3})^{-1}$ ago, it would have initiated the overionization of Sgr A East. We note that, we would expect overionization of other elements in the swept-up interstellar medium as well in this scenario. A work on the broad-band spectrum, which is important in this sense, will be reported separately.