NuSTAR Detection of a Hard X-Ray Source in the Supernova Remnant-molecular Cloud Interaction Site of IC 443

We extracted both the NuSTAR and the XMM-Newton spectra from a 30'' radius region centered on Src 1 (the 30'' cyan circle in Figure 3). Background spectra were selected from local surrounding regions on the same detector chip. We performed a joint spectral fit with NuSTAR and XMM-Newton data using XSPEC version 12.9.0 (Arnaud 1996).

To examine the SNR ejecta fragment model, we fit the 0.5–79 keV spectra with a model composed of a power law and an optically thin thermal plasma component MEKAL, both subject to foreground absorption, resulting in constant∗TBabs∗(MEKAL+powerlaw). It results in a satisfactory fit with ${\chi }_{\nu }^{2}$ = 0.9 for dof = 328 (left panel of Figure 4). All the model parameters for TBabs, MEKAL abd powerlaw are tied among the spectrum data sets, while the constant is frozen at 1 for the NuSTAR FPMA spectrum and left as a free parameter for all other spectra. The photon index of the power law is tightly constrained as Γ = 1.72 ± 0.08; the temperature of the thin plasma model is kT = 0.19 ± 0.04 keV; the best-fit absorption column density is N_H = (1.2 ± 0.4) × 10²² cm⁻². With this broadband spectrum, we derived a softer spectrum for Src 1 than previous results of ${\rm{\Gamma }}={1.48}_{-0.08}^{+0.20}$ (Table 5 in Bocchino & Bykov 2003) with improved error bars. The unabsorbed source flux is F_2–10 keV = (6.1 ± 0.1) × 10⁻¹³ erg cm⁻² s⁻¹ in 2–10 keV and F_10–10 keV = (8.0 ± 0.4) × 10⁻¹³ erg cm⁻² s⁻¹ in 10–40 keV. The correspondent luminosities in 2–10 keV and 10–40 keV are L_2–10 keV = (0.72 ± 0.02) × 10³² erg s⁻¹, and L_10–40 keV = (1.4 ± 0.1) × 10³² erg s⁻¹, assuming a distance of 1.5 kpc.

Zoom In Zoom Out Reset image size

To examine the PWN scenario, we also fit the spectra with the blackbody plus a cutoff power law (constant∗TBabs∗(bbody+cutoffpl)), which models the combination of thermal emission from the neutron star surface, and nonthermal emission from the PWN. This model results in an equally good fit with ${\chi }_{\nu }^{2}$ = 0.8 for dof = 331 (right panel of Figure 4). The best-fit absorption column density is N_H = (0.9 ± 0.2) × 10²² cm⁻². The blackbody temperature is derived to be kT = 0.11 ± 0.01 keV. The cutoff power law has a photon index of Γ = 1.4 ± 0.1, consistent with previous measurements of Γ = 1.48^+0.20_−0.08 using 1–10 keV data only, but with a cutoff energy at ${E}_{\mathrm{cut}}={18}_{-5}^{+10}$ keV. Since we only have one data bin above 20 keV with S/N > 3σ, the current data cannot unambiguously distinguish between a power law and a cutoff power law. However, we determined that the cutoff energy cannot be lower than 10 keV at the 90% confidence level. The model fitting results for both the SNR ejecta scenario and the PWN scenario are shown in Table 1.

Table 1.  Spectral Fitting Results for Src 1 with Joint NuSTAR and XMM Data

Parameter	Unit	MEKAL+power-law Modela	bbody+cutoffpl Modelb
NH	1022 cm−2	1.2 ± 0.4	0.9 ± 0.2
Γ	⋯	1.72 ± 0.08	1.4 ± 0.1
Ecut	keV	⋯	${18}_{-5}^{+10}$
normpl/cutoffpl	ph keV−1 cm−2 s−1	(1.6 ± 0.3) × 10−4	(1.2 ± 0.1) × 10−4
kT	keV	${0.19}_{-0.03}^{+0.04}$	0.11 ± 0.01
normMEKAL/BB	see note	${3}_{-2}^{+7}\times {10}^{-3}$	${1.7}_{-0.7}^{+1.9}\times {10}^{-5}$
F2–10 keV	10−13 erg cm−2 s−1	6.1 ± 0.1	6.1 ± 0.1
F10–40 keV	10−13 erg cm−2 s−1	8.0 ± 0.4	5.6 ± 0.4
${\chi }_{\nu }^{2}$ (dof)		0.9 (328)	0.8 (331)

Notes. The goodness of fit is evaluated by ${\chi }_{\nu }^{2}$ and the number of degrees of freedom (dof) is given in parentheses. The errors are 90% confidence. Unabsorbed Flux in 2–10 keV (F_{2–10 keV}) and 10–79 keV (F_{10–79 keV}) predicted by each model are reported here.

^aMEKAL+power-law model: tbabs∗(MEKAL+power-law). The MEKAL model is characterized by plasma temperature kT in keV, normalization (norm) of MEKAL defined as $\mathrm{norm}={10}^{-14}{[4\pi {D}_{A}{(1+z)}^{2}]}^{-1}\int {n}_{e}{n}_{{\rm{H}}}{dV}$, where D_A is the angular diameter distance to the source (cm), n_e and n_H are electron and H densities ^bbbody+cutoffpl model: tbabs∗(blackbody+cutoffpl). The bbody model is characterized by temperature kT in keV, and model norm defined as $\mathrm{norm}\,=$ ${L}_{39}^{2}/{D}_{10}^{2}$, where L₃₉ is the source luminosity in units of 10³⁹ erg s⁻¹, and D₁₀ is distance units of 10 kpc. The cutoffpl is characterized by photon index Γ, e-folding energy of exponential rolloff E_cut (in keV), and the model norm in ph keV⁻¹ cm⁻² s⁻¹ at 1 keV.

Download table as:  ASCIITypeset image

A previous study has pointed out that Src 1 shows no flux variation from 2000 to 2006 (Bykov et al. 2008). By comparing the 2–10 keV source flux in the 2015 NuSTAR observation with previous measurements by Chandra and XMM-Newton, we conclude that Src 1 has maintained the same flux level for 15 years from 2000 to 2015.

We also extracted NuSTAR and XMM-Newton spectra from a larger region with r = 80'' (the magenta circle in Figure 2), which contains all three sources (Src 1–3) and the extended emission features in between the three sources. The spectrum from the r = 80'' circle is softer compared to the r = 30'' circle centered on Src 1, resulting in Γ = 1.95 ± 0.05. The flux of the r = 30'' region makes up about 33% of the total flux of the r = 80'' region in 2–10 keV and ∼50% above 10 keV. This is consistent with previous results showing that there are thermal features in between sources 1–3, making the spectrum from a larger region softer.

Both Src 2 and Src 3 are not detectable in the 2015 NuSTAR observations, resulting in detection significance of 1σ and 2σ, respectively. The 3σ detection limit for Src 2 and Src 3 is F_2–10 keV ∼ 2 × 10⁻¹⁴ erg cm⁻² s⁻¹, which can be taken as the upper limit flux of these two sources. We reanalyzed the 2004 Chandra observation to obtain spectral parameters for Src 2 and Src 3. The regions used for source and background spectra extraction are shown in Figure 2.

The spectrum of Src 2 does not show any line feature. We thus first fit its spectrum with a simple absorbed power-law model TBabs∗powerlaw. The best fit gives an absorption column density of N_H = (1.0 ± 0.2) × 10²² cm⁻² and a photon index of ${\rm{\Gamma }}={2.4}_{-0.2}^{+0.3}$ with χ²/dof = 1.1/58. We also applied an absorbed thermal plasma model TBabs∗MEKAL to the Src 2 spectrum, which fits equally well. We fixed the abundance to solar, as it cannot be well constrained by the data. The thermal plasma model gives the best-fit temperature of ${kT}={2.8}_{-0.5}^{+0.6}$ keV. These results are consistent with the Bykov et al. (2008) analysis. However, we want to point out that the thermal plasma model cannot give an acceptable fit with the redshift fit to z = 0, even after the abundance parameter is set to free. The thermal model instead requires a redshift of z = 0.09 ± 0.01.

Src 2 has shown time variability since the year of 2000. Its 2–10 keV flux increased by a factor of ∼4 within five months in 2000, from ${F}_{\mathrm{2000,4}}={0.5}_{-0.2}^{+0.4}\times {10}^{-13}$ erg cm⁻² s⁻¹ in April to F_2000,9 = (1.8 ± 0.2) × 10⁻¹³ erg cm⁻² s⁻¹ in September. Then its flux decreased to ${F}_{2004}=(1.2\pm 0.1)\,\times {10}^{-13}$ erg cm⁻² s⁻¹ in 2004, and further decayed to F₂₀₀₆ = (0.6 ± 0.1) × 10⁻¹³ erg cm⁻² s⁻¹ in 2006 (Bykov et al. 2008). Its 2015 flux in 2–10 keV is constrained to ${F}_{2015}\leqslant 0.2\,\times {10}^{-13}$ erg cm⁻² s⁻¹ (3σ detection limit) by the NuSTAR observation, which suggests a flux decrease for an order of magnitude from 2000 to 2015. There is no evidence for significant spectral variation during this time range.

Similarly, we fit the Src 3 spectrum with both the power-law model and the thermal plasma model. The power-law model gives a marginally acceptable fit to the Src 3 spectrum with residuals between 3 and 4 keV (χ²/dof = 1.3/23). The fit can be improved by adding an emission line, resulting in TBabs∗(powerlaw+Gaussian) (χ²/dof = 1.1/22). The best fit gives a photon index of ${\rm{\Gamma }}={1.8}_{-0.3}^{+0.4}$, an emission line centroid energy of E = 3.7 ± 0.1 keV, and a slightly lower absorption of ${N}_{{\rm{H}}}=({0.7}_{-0.2}^{+0.3})\times {10}^{22}\,{\mathrm{cm}}^{-2}$ compared with Src 1 and Src 2. The 2–10 keV flux of Src 2 in 2004 is F_2–10 keV = (1.2 ± 0.1) × 10⁻¹³ erg cm⁻² s⁻¹. A thermal plasma model can also fit well to the Src 3 spectrum (${\chi }^{2}/\mathrm{dof}=1.1/22$), resulting in ${N}_{{\rm{H}}}=({0.6}_{-0.1}^{+0.2})\times {10}^{22}\,{\mathrm{cm}}^{-2}$ and ${kT}={10}_{-3}^{+10}$ keV, but requiring a high redshift of z = 0.8 ± 0.1, which implies an extragalactic origin for Src 3. Therefore, the line feature at ∼3.7 keV shall be taken with caution. While a redshift of z = 0 points to an Ar K line, a redshift of z = 0.8 is consistent with a shifted Fe K line instead.

Src 3 also shows time variability, but with a different trend compared to Src 2. The 2–10 keV flux of Src 3 increased from 2000 (F₂₀₀₀ = (0.3 ± 0.1) × 10⁻¹³ erg cm⁻² s⁻¹) to 2004 (F₂₀₀₄ = (0.6 ± 0.1) × 10⁻¹³ erg cm⁻² s⁻¹) and maintained at the same level until 2006 (F₂₀₀₆ = (0.63 ± 0.05) × 10⁻¹³ erg cm⁻² s⁻¹). However, this brightening trend does not seem to continue, as the Src 3 flux in 2015 cannot significantly exceed ${F}_{2015}=0.2\times {10}^{-13}$ erg cm⁻² s⁻¹, comparable to its flux level back in 2000.

Here we also report the NuSTAR detection of two point sources in the FoV (obsID 30101058002). The point source to the south of Src 1 is outlined with a r = 10'' circle and listed as Ps 1 in Figure 1. Ps 1 was detected by XMM-Newton (named as XMMU J061801.5+222233), and also by Chandra (catalog source CXOGSG J061801.4+222228). Within the Chandra/ACIS spatial uncertainty, there is a point-like optical counterpart GAIA DR1 id. 3377008814910613632. This source is detected by NuSTAR at a significance level of 9σ with a total counts of 237 in 3–79 keV. We performed a joint Chandra and NuSTAR spectral analysis for this source. A simple absorbed power-law model can fit to the data very well, resulting in ${N}_{{\rm{H}}}={2.0}_{-0.7}^{+0.8}\times {10}^{21}\ {\mathrm{cm}}^{-2}$ and Γ = 2.0 ± 0.2 with ${\chi }_{\nu }^{2}=0.8$ for dof of 58. The observed flux is F_2–10 keV = (1.3 ± 0.6) × 10⁻¹³ erg cm⁻² s⁻¹ in 2–10 keV and F_10–40 keV = (1.4 ± 0.8) × 10⁻¹³ erg cm⁻² s⁻¹ in 10–40 keV. Alternatively, an absorbed bremsstrahlung model with ${kT}={5.8}_{-1.5}^{+1.6}$ keV and ${N}_{{\rm{H}}}={0.8}_{-0.5}^{+0.5}\times {10}^{21}\ {\mathrm{cm}}^{-2}$ can fit to the data equally well.

The other point source to the east of Src 1 is denoted as Ps 2 in Figure 1. This source is detected at an 8σ significance level with total counts of 173 in 3–79 keV. Its spectrum is well fit by either a power law with Γ = 1.9 ± 0.5, or a bremsstrahlung model with ${kT}={12}_{-6}^{+26}$ keV. The observed flux in 3–10 keV and 10–40 keV are F_3–10 keV = (8 ± 4) × 10⁻¹⁴ erg cm⁻² s⁻¹ and F_10–40 keV = (1.2 ± 0.7) × 10⁻¹³ erg cm⁻² s⁻¹.

There is already a known PWN (1SAX J0617.1+2221) in the IC 443 region, which has been assumed to be associated with IC 443. However, since there is an overlapping SNR G189.6+3.3 (Asaoka & Aschenbach 1994), the possibility of the existence of a second PWN in this complicated field cannot be ruled out. The Src 1 X-ray spectrum of a blackbody with kT ∼ 0.1 keV plus a power law (Γ ≈ 1.7) or a cutoff power law (Γ ≈ 1.4 with E_cut ∼ 18 keV) is consistent with the PWN scenario. The blackbody emission can be explained by thermal emission from the surface of a neutron star. The best-fit blackbody temperature kT = 0.11 ± 0.01 keV is consistent with those derived by Bykov et al. (2005, 2008), and corresponds to about 3 km radius. The nonthermal power-law emission (w/wo energy cutoff) can arise from the nebula surrounding the central compact object, or from the magnetosphere of the neutron star.

We performed a pulsation search using the NuSTAR events collected within a r = 50'' circle in different energy bands for Src 1. A pulsation signal is not detected at a 99% confidence level.

Since the PWN scenario cannot be confirmed due to the nondetection of the pulsation signal, we next explored the shocked molecular clump scenario proposed by Bykov et al. (2000). Bykov et al. (2000) investigated the nonthermal emission from a radiative SNR interacting with molecular clouds, which is similar to the case of IC 443. In the SNR–MC interaction site, the remnant drives both a forward shock into the dense molecular clump and a reverse shock into the radiative shell (Chevalier 1999). When the shock velocity is high enough, the hot shocked gas behind the shock can produce thermal X-ray emission with ISM/cloud abundance. This might explain the thermal plasma component from Src 1.

Nonthermal particles are accelerated at the MHD collisionless shock and then diffuse into the shocked clump and shell materials. These energetic particles can interact with the shocked materials and produce prominent nonthermal bremsstrahlung emission, producing sources bright in X-rays to MeV γ-rays in the SNR–MC interaction region. As long as the molecular clump density is much higher than that in the radiative shell, the emission from the interaction region is expected to be dominated by the shocked clump component. The spectrum of nonthermal electrons in a shocked clump is mainly shaped by the first-order Fermi acceleration. If large-scale turbulence driven by thermal and dynamical instability exists at the radiative shock, the second-order Fermi acceleration can also affect the resulting electron spectrum. At low energies ≲100 keV, the electrons suffer strong Coulomb loss, which is able to modify the electron spectrum. At high energies in the γ-ray window, the spectrum is expected to have an exponential-like cutoff. The maximum energy of the particles should be determined by either ion-neutral damping (O'C Drury et al. 1996; Ptuskin & Zirakashvili 2003) or limited particle acceleration time.

A self-consistent model for the nonthermal emission from a shocked clump requires a proper treatment for the shock interaction between a remnant and molecular clouds (Chevalier 1999), plus a good recipe for particle injection and acceleration in an MHD collisionless shock (Caprioli et al. 2010, and reference therein), which is beyond the scope of this paper. For simplification, we instead assume that the nonthermal electrons follow a power-law spectrum in momentum based on first-order Fermi acceleration, resulting in $N(E){dE}\,\propto {({E}^{2}+2{{Em}}_{e}{c}^{2})}^{-\tfrac{s+1}{2}}\times (E+{m}_{e}{c}^{2}){dE}$, where E is electron kinetic energy and s is electron spectral index. The normalization can be determined from the energy density of electrons with energies above 10 keV, w_e(>10 keV). Here we assume that the CR pressure is 10% of the ram pressure, and that the electron to proton number ratio is K_ep = 0.01 at momentum p = 1 GeV c⁻¹, and ${w}_{e}(\gt 10\,\mathrm{keV})\approx 7.0\,\times \,{10}^{-10}{n}_{3}{u}_{s,2}^{2}\,\mathrm{erg}\,{\mathrm{cm}}^{-3}$, where n₃ is the number density of clumps in 10³ cm⁻³ and u_s,2 is the shock velocity in 100 km s⁻¹. At low energies, the power-law spectrum is modified by substantial Coulomb loss (Vink 2008). The second-order Fermi acceleration is neglected here, which can overcome the Coulomb loss at low energies. As a result, we leave the Coulomb break as a free parameter in the data fitting. We calculated the bremsstrahlung spectrum using analytic cross sections given in Haug (1997), with e − e bremsstrahlung in the X-ray band neglected. With the above settings, the NuSTAR data can be fit well with a nonthermal bremsstrahlung with s = 2.0 without invoking Coulomb loss (see the blue dashed line in Figure 5), giving a shocked clump mass of M_c ≈ 6.5 M_⊙. In the case of s = 2.2, we found that the data requires Coulomb loss with a break at E_Cou ≈ 15 keV (black solid line in Figure 5), giving a molecular clump mass of M_c ≈ 5.0 M_⊙. The treatment here for the bremsstrahlung emission from a shocked clump is similar as that discussed in Uchiyama et al. (2002) and also agrees with the simulation results in Bykov et al. (2000) with a more complicated setup.

Zoom In Zoom Out Reset image size

Fast-moving X-ray knots were first discovered in Vela SNR with ROSAT by Aschenbach et al. (1995). Head-tail structure of apparent size of 84 × 41 (0.6 × 0.3 pc at a distance of 250 pc) was revealed with high-resolution Chandra observation (Miyata et al. 2001). Because of the relative high abundance in supernova nucleosynthesis products, the observed fast-moving knots are interpreted as isolated supernova ejecta fragments, which are interacting with surrounding media and composed of heavy elements.

X-ray line and continuum emission from supernova ejecta fragments interacting with dense molecular clumps have been discussed for the situation in Vela SNR (Bykov 2002) and IC 443 (Bykov et al. 2008). It is believed that nonthermal particles are accelerated at the MHD collisonless shock and then diffuse into the shocked materials. The X-ray continuum emission of bremsstrahlung nature is produced in a similar way as in shocked molecular clumps (see Section 5.2). The X-ray line emission, a signature of SNR ejecta, is mainly due to K-shell ionization induced by energetic particles in the cold metallic part of ejecta fragments. The resulting X-ray emission is most prominent when the ejecta fragments are colliding with a dense molecular clump. Therefore, SNR ejecta fragments and shocked molecular clumps are likely to exist next to each other.

Substructures of Src 1 revealed by Chandra show a few bright clumps (Src 1a, Src 1b etc., see Figure 2) embedded in an extended diffuse emission. The Src 1a Chandra spectrum can be best fit with a power-law with ${\rm{\Gamma }}={1.5}_{-0.4}^{+0.5}$ plus a thermal plasma model with ${kT}={0.2}_{-0.1}^{+0.6}$ keV (Bykov et al. 2008). The spectrum of the fainter subclump Src 1b can be described by a power-law with Γ ∼ 2 with large error bars. The observed IR/X-ray morphology can possibly be explained by the interaction between IC 443 and clumpy molecular clouds, which involves multiple shocks with different velocities (Bykov et al. 2008). The shocks are likely to be driven by the fast-moving ejecta fragments, since a region next to Src 1a and 1b shows evidence for possible Si K-shell line. Considering the complicated morphology of Src 1, it is likely that it is composed of both SNR ejecta (or a PWN) and shocked molecular clumps.

The other two isolated X-ray sources in the same SNR–cloud interaction site, Src 2 and Src 3, are not detectable in the 2015 NuSTAR data, resulting in 1σ and 2σ detection, respectively. Src 2 showed a brightening in X-rays by a factor of four within five months in 2004, and then its X-ray flux kept decreasing from 2004 to 2015. The 2015 X-ray flux of Src 2 is ∼10% of its highest X-ray flux in record. The 2–10 keV X-ray spectrum of Src 2 is featureless and can be equally fit with either a power law with Γ ∼ 2.4, or a thermal plasma with kT ∼ 2.8 keV and z ∼ 0.1. Its time variability and X-ray spectral shape are consistent with either an SNR ejecta fragment, or a shocked molecular clump illuminated by an SNR shock front in 2004. Alternatively, Src 2 could be a background galaxy.

Src 3 demonstrates different X-ray characteristics in both time variability and spectrum. The X-ray flux of Src 3 increased from 2000 to 2006, but dropped back to its previous X-ray flux in 2015. Its X-ray spectrum in 2–10 keV is well fit with a power law with ${\rm{\Gamma }}={1.8}_{-0.3}^{+0.4}$ plus an emission line at E = 3.7 ± 0.1 keV. Based on the detection of the line emission, Src 3 can be best interpreted as an SNR ejecta fragment if it is associated with IC 443. Its X-ray spectrum can also be fit with a thermal plasma model with ${kT}={10}_{-3}^{+10}$ keV with a redshift of z = 0.8, in which case the 3.7 keV line shall be a shifted Fe K line. Therefore, an extragalactic origin, such as an AGN, is an alternative scenario for Src 3. Future Chandra monitoring of long-term variabilities of Src 2 and Src 3 will help to clarify their source nature.

The SNR ejecta/shocked clump interpretation of hard X-ray sources within SNR–MC interaction region can potentially explain a large population of hard X-ray sources in dense environments of star-forming regions, such as the Galactic center. In the NuSTAR Galactic center survey, we detected a few hard X-ray sources in the SNR–MC interaction sites (Zhang et al. 2014; Nynka et al. 2015). Among them, G359.97-0.038 is most likely arising from SNR–MC interaction.

G359.97-0.038 is located at the interaction site of the Galactic center SNR Sgr A East and a giant molecular cloud M-0.02-0.07 (a.k.a. the 50 km s⁻¹ cloud), strongly indicated by the presence of OH 1720 MHz maser (Pihlström & Sjouwerman 2006). It shows a similar hard continuum spectrum (Γ = 1.3 ± 0.3) as Src 1 in the IC 443 system, with no significant line features. Therefore, as discussed in Sections 5.2 and 5.3, G359.97-0.038 can be a shocked molecular clump. While NuSTAR detected a few X-ray sources in the vicinity of Sgr A East, Chandra revealed many more isolated X-ray sources in the interaction site of the Sgr A East and M-0.02-0.07 cloud system. In future works, we will apply the SNR ejecta/shocked clump models developed here on the unidentified X-ray sources around Sgr A East. Considering the rich environment for star-forming regions at the Galactic center, a large population of sources with hard X-ray spectra could bear the same source nature.