Suzaku Observations of Fe K-shell Lines in the Supernova Remnant W51C and Hard X-ray Sources in the Proximity

Figure 1(a) shows an X-ray image obtained by Suzaku in the 4–8 keV band (color) overlaid by the 0.5–2 keV band image (yellow contours). Here the images of XIS0, 1, and 3 were merged. The NXB was subtracted and the vignetting was corrected. The outermost white line indicates the FOVs of the two observations. The hard X-ray objects previously reported, HX1, HX2 (CXO J192318.5$+$1403035), HX3east, and HX3west, were seen in the image. Soft X-ray emission would come from the SNR W51C. Figure 1(b) is the same as (a), but the yellow contours indicate the 1.4 GHz radio continuum (Stil et al., 2006; Beuther et al., 2016). The radio shell covers the whole FOVs. Since CR acceleration occurs in the shell, the 6.40 keV line should be searched in the whole FOVs.

We extracted a spectrum from the whole FOVs, using the FI and BI data. Here, the regions including the four hard X-ray objects (the circles and ellipses in figure 1) were excluded. We merged the FI (XIS0 and 3) spectra and fitted the FI and BI spectra simultaneously. We fitted the spectrum of the high-energy band above 4.5 keV in order to avoid contamination of a local soft X-ray emission from W51C (c.f. Hanabata et al. (2013); Sasaki et al. (2014); also see figure 1). We ignored the band above 10 keV for the FI spectrum and the band above 7.0 keV for the BI spectrum because the influence of the NXB is significant.

The fitted model was an absorbed thermal bremsstrahlung plus four Gaussians in addition to the cosmic X-ray background (CXB). We used phabs in XSPEC as the absorption model. The thermal bremsstrahlung is mainly due to the GRXE and the extended hard X-ray emission overlapping W51B (Hanabata et al., 2013), as well as a very small amount of thermal emission from W51C. The contribution of the thermal emission from W51C is only $\sim 2$% in the 4.5–10 keV band because of its low temperature ($\sim 0.5$ keV; Hanabata et al. (2013); Sasaki et al. (2014)). The four Gaussians represent the Fe K-shell lines, and the centroids were fixed to 6.40 keV and 7.05 keV (Fe\emissiontypeI K$\alpha$ and K$\beta$), 6.68 keV (Fe\emissiontypeXXV He$\alpha$), and 6.97 keV (Fe\emissiontypeXXVI Ly$\alpha$). The normalization of the Fe\emissiontypeI K$\beta$ was fixed to 0.125 times that of the Fe\emissiontypeI K$\alpha$ (Kaastra & Mewe, 1993; Smith et al., 2001). The parameters of the CXB were fixed to the values in Kushino et al. (2002).

The best-fit parameters are summarized in table 3.2. The intensities of the 6.40 keV and 6.68 keV lines were obtained to be $(1.4\pm 0.4)\times 10^{-8}$ and $(1.7\pm 0.4)\times 10^{-8}$ in units of photons s^-1 cm^-2 arcmin^-2, respectively. The spectrum with the best-fit model is presented in figure 2. In the case where the centroid of the 6.40 keV line was let free, its upper limit was obtained to be 6.430 keV at the 90% confidence level. Allowing the CXB flux to vary within the range of fluctuation in Kushino et al. (2002) does not affect the best-fit line intensity.

Since there is no Suzaku data that can be used as background regions in the vicinity of W51C, we estimated the intensities of Fe K-shell lines of the background, based on the GRXE model of previous studies. The distributions of the 6.40 keV and 6.68 keV lines accompanied by the GRXE were well studied by Uchiyama et al. (2013) and Yamauchi et al. (2016). The scale lengths along the Galactic plane of the 6.40 keV and 6.68 keV lines are $57^{\circ}\pm 50^{\circ}$ and $45^{\circ}\pm 10^{\circ}$, respectively, according to Uchiyama et al. (2013). As for the scale heights, Uchiyama et al. (2013) used the data mainly near the Galactic center, and the scale-height measurement was affected by the Galactic Bulge X-ray Emission (c.f. Koyama (2018)). The scale heights of the GRXE reported by Yamauchi et al. (2016) are more reliable because they measured a pure GRXE in the range of $|l|=10^{\circ}$–$30^{\circ}$. Therefore, we adopted the scale heights in Yamauchi et al. (2016), $\timeform{0D.50}\pm\timeform{0D.12}$ and $\timeform{1D.02}\pm\timeform{0D.12}$ for the 6.40 keV and 6.68 keV lines, respectively. Here, we considered the statistical errors of the scale length and scale height of each line obtained by Uchiyama et al. (2013) and Yamauchi et al. (2016). Uchiyama et al. (2013) calculated the differences between the GRXE model and the data for the individual FOVs of Suzaku and obtained their standard deviation to be 24%–30%. This uncertainty was also taken into account.

The line intensities at the central coordinate of the W51C regions ($l=\timeform{49D.11},b=\timeform{-0D.43}$) were estimated to be $(0.4\pm 0.3)\times 10^{-8}$ and $(2.0\pm 0.8)\times 10^{-8}$ in units of photons s^-1 cm^-2 arcmin^-2 for the 6.40 keV and 6.68 keV lines, respectively. The observed intensity of the 6.68 keV line was consistent with the model, but that of the 6.40 keV line is higher at the confidence level of 2.0$\sigma$. It indicates a hint of an enhancement of the 6.40 keV line emission in the W51C region.

For spectral analysis of the hard X-ray sources (HX1, HX2, HX3east, and HX3west), we also extracted spectra from the circle and ellipse regions in figure 1, respectively, and fitted the individual spectra with an absorbed thermal bremsstrahlung and four Gaussians as well as the CXB. The best-fit intensities of the 6.40 keV and 6.68 keV lines are shown in table 3.2. The line intensities are consistent with those of the W51C region within the error ranges. The only exception is the 6.68 keV line of HX1, whose intensity is higher than that of the W51C region at the significance level of $3.4\sigma$. According to Koo et al. (2002), HX1 is coincident with the compact H\emissiontypeII regions G48.9$-$0.3 and G49.0$-$0.3. In our hard X-ray image (figure 1), the emission peak of the HX1 region is very close to G49.0$-$0.3, and therefore the enhancement of the 6.68 keV line would come from G49.0$-$0.3.

To investigate the spatial distribution of the 6.40 keV line, we divided the W51C region into three regions: West, Center, and East as shown in figure 1. A spectrum of the high-energy band ($\geq 4.5$ keV) was extracted from each region excluding the circle and ellipse regions in figure 1. We fitted the spectra with an absorbed thermal bremsstrahlung plus four Gaussians and the CXB. The obtained line intensities are summarized in table 3.2. The intensities of the 6.40 keV line of the three regions are consistent within $1\sigma$ errors, whereas that of the 6.68 keV line is weakest in East, where is the farthest region from the Galactic plane among the three regions. The GRXE weakens with a distance from the Galactic plane (Uchiyama et al., 2013; Yamauchi et al., 2016). The weak intensity of the 6.68 keV line in East is consistent with that expected from the GRXE distribution.

The enhanced 6.68 keV line emission in G49.0$-$0.3 (HX1) implies the presence of high-temperature plasma. We fitted the G49.0$-$0.3 spectrum in the 2–10 keV band with an absorbed CIE plasma model (apec in XSPEC). Here, the spectrum of West was used as the background spectrum. The model explained well the G49.0$-$0.3 spectrum ($\chi^{2}/$d.o.f. = 58/52 = 1.12). The temperature and the abundance of the CIE plasma were obtained to be $kT=3.0^{+0.8}_{-0.7}$ keV and $0.5\pm 0.2$ solar, respectively. The flux of the 2–10 keV band was $3.7\times 10^{-13}$ erg s^-1 cm^-2. The spectrum is shown in figure 3 and the best-fit parameters are shown in table 3.4.

We found the enhancement of the 6.68 keV line in the compact H\emissiontypeII region G49.0$-$0.3. In H\emissiontypeII regions, stellar winds experience a shock transition, and the shocked hot winds fill the shell region to emit thermal X-rays. The shock temperature is expressed as follows:

where $m_{\rm H}$ is the mass of a hydrogen atom, $\mu=0.62$ is the mean molecular weight, and $v_{w}$ is the stellar wind velocity (e.g. Ezoe et al. (2006)). Since the observed temperature of the G49.0$-$0.3 region is $kT=3.0^{+0.8}_{-0.7}$ keV, the wind velocity is estimated to be $v_{w}\sim 1500$ km s^-1. The most massive star in G49.0$-$0.3 is an O9 star (Kim et al., 2007), and a typical wind velocity of O stars is 1000–3000 km s^-1 (Prinja, 1990). The observed abundance was $0.5\pm 0.3$ solar, which is consistent with the values in the other H\emissiontypeII regions (e.g., Hyodo et al. (2008)). Therefore, the enhancement of the 6.68 keV line would be due to the high-temperature plasma associated with the H\emissiontypeII region.

Hanabata et al. (2013) reported the extended hard X-ray emission overlapping W51B (called “Reg.3”) although the G49.0$-$0.3 region was excluded in their analysis. They found that the spectrum of the hard X-ray emission was represented by either a high-temperature thermal plasma model or a non-thermal emission, but they did not distinguish which model is better. “Reg.3” coincides with a part of West and Center regions. Our analysis found no enhancement of the 6.68 keV line compared to the GRXE in both regions, which supports the non-thermal scenario.

We found the enhancement of the 6.40 keV line at the significance level of $2.0\sigma$ although it is not a robust result. Here we discuss possible origins of the 6.4 keV line enhancement, assuming that it is true.

One possible origin of the enhancement is the ionizing thermal plasma of the SNR. However, we conclude that the 6.40 keV line is not likely of the plasma origin. First, because the thermal plasma of W51C is in the ionizing state and its temperature is $kT\sim 0.6$–$0.7$ keV (Hanabata et al., 2013; Sasaki et al., 2014), at which a detectable Fe K-shell line cannot be emitted. Second because, as mentioned in section 3.1, the upper limit of the centroid of the observed 6.40 keV line is 6.430 keV, which corresponds to the ionization state of Ne-like Fe and the ionization time-scale of $\sim 200$ yr according to the NEI code in XSPEC. This is two orders of magnitudes younger than the age of W51C (Koo et al., 1995). Here we assume the electron number density of plasma of 1 cm^-3.

The 6.40 keV line can be generated by X-ray photoionization. The Fe K-shell ionization energy is 7.1 keV, and therefore soft X-ray sources such as W51C contribute little to the 6.40 keV emission. Candidates are the four hard X-ray sources, HX1 (G49.0$-$0.3), HX2, HX3west, and HX3east. The flux of the 6.4 keV line is calculated as

where $\epsilon$ is the Fe fluorescence yield, $\Omega$ is a solid angle with which the molecular clouds are seen from the source, $N_{\rm H}$ is the hydrogen column density of the molecular clouds, $Z_{\rm Fe}$ is the Fe abundance, $\sigma_{\rm Fe}(E)$ is the cross-section of the photoionization for Fe atoms, and $A(E)=K(E/1~{}{\rm keV})^{-\Gamma}$ is the X-ray spectrum of the irradiating X-ray source with the photon index $\Gamma$ and normalization $K$. Here, we used the upper limits of the hydrogen column densities $N_{\rm H}$ of the shocked CO and H \emissiontypeI of $N_{\rm H}\geq 4.0\times 10^{16}$ cm^-2 and $N_{\rm H}\geq 1.8\times 10^{21}$ cm^-2, respectively (Koo & Moon, 1997a, b). Also, we adopted $\epsilon=0.34$ (Bambynek et al., 1972), $Z_{\rm Fe}=3\times 10^{-5}$ (Lodders, 2003), and $\sigma_{\rm Fe}(E)=6\times 10^{-18}(E/1~{}{\rm keV})^{-2.6}$ cm² (Henke et al., 1982). The photon index $\Gamma$ of the four hard X-ray sources ranges 1.8–3.2 (Sasaki et al. (2014), this work) and the flux of the enhanced 6.40 keV line we obtained in this work is $F_{\rm 6.4~{}keV}=5.0\times 10^{-6}$ photons s^-1 cm^-2. Substituting these values into equation (2), we obtained the normalization $K$ of $A(E)$ to be $(0.1$–$3)(4\pi/\Omega)(1.8\times 10^{21}~{}{\rm cm}^{-2}/N_{\rm H})$ photons s^-1 cm^-2 keV^-1. Then, integrating $A(E)$, we estimated the required X-ray flux of the irradiating source to be $F_{X}=(0.07$–$1)\times 10^{-8}(4\pi/\Omega)(1.8\times 10^{21}~{}{\rm cm}^{-2}/N_{\rm H})$ erg s^-1 cm^-2 in the 0.3–10 keV band, which is at least two orders of magnitude higher than the typical flux of the hard X-ray sources; $\sim 10^{-13}$ erg s^-1 cm^-2 for HX1 (this work) and $\sim 10^{-12}$ erg s^-1 cm^-2 for the other sources (Sasaki et al., 2014) regardless of the solid angle. It is unlikely that the 6.40 keV line is of the photoionizaiton origin.

The most probable origin of the enhancement of the 6.40 keV line would be the interaction between the low-energy CRs and molecular clouds. In fact, W51C is thought to be a CR accelerator; it is one of the brightest gamma-ray SNRs in the Galaxy (Ajello et al., 2020), and the high ionization rate was measured (Ceccarelli et al., 2011). Previous observations have shown that there are shocked CO clouds at the boundary between Center and West, and the centroids of the gamma-ray emissions and the position of the high ionization rate are consistent with Center (figure 1).

Makino et al. (2019) demonstrated that both the 6.40 keV line and gamma-ray emissions produced by the proton-proton interaction can be explained by a CR escaping model for SNRs interacting with molecular clouds. Assuming the 6.40 keV line in the W51C region is of the proton origin, we applied this model to W51C and found that the 6.40 keV line intensity calculated from the information of gamma-ray spectrum (Aleksić et al., 2012) and molecular clouds (Carpenter & Sanders, 1998) is $0.3\times 10^{-8}$ photons s^-1 cm^-2 arcmin^-2, which is smaller than the observed value. However, this may reflect that low-energy CRs are being further accelerated behind the shock of the SNR through the interaction between the shock and the clouds (Inoue et al., 2012).

If the 6.40 keV line emission originates from low-energy protons interacting with molecular clouds, we can expect a positive correlation between the luminosities of the 6.40 keV line and gamma-rays. Besides W51C, there are several SNRs where both the 6.40 keV line and gamma-ray emission have been observed: 3C391, Kes 79, N132D, W28, W44, G323.7$-$0.1, and IC 443 (Sato et al. (2014, 2016); Bamba et al. (2018); Nobukawa et al. (2018); Saji et al. (2018); Nobukawa et al. (2019); Ajello et al. (2020)). It would be valuable to investigate the relation between the 6.40 keV line and 0.1–100 GeV gamma-ray luminosities. As shown in figure 4, we found a hint of a positive correlation between the two observables (the correlation coefficient is $r\sim 0.3$) although the sample size is limited.