XMM-NEWTON OBSERVATIONS OF NGC 247: X-RAY POPULATION AND A SUPERSOFT ULTRALUMINOUS X-RAY SOURCE

Due to the soft spectrum of the ULX and the thin filter on the PN, the number of photons detected on the PN is about 3–4 times that on each MOS. Thus, we used only the PN data for spectral analysis. The energy spectrum was extracted from events in a 30''radius circular region around the source to avoid CCD gaps and with FLAG and PATTERN the same as those used in creating the image. Background was subtracted from events in a nearby circular region with a radius of 45'' on the same CCD and off the readout column of the ULX. To avoid spurious spectral structure caused by oversampling of the spectrum, the energy channels were grouped with a bin size of 1/5th of the local spectral resolution (in terms of full width at half-maximum) and to have at least 15 net counts per bin.

A single power-law subject to interstellar absorption is inadequate to fit the ULX spectrum, resulting in χ² = 198.0 for 44 degrees of freedom (dof). An absorbed multicolor disk (MCD) model provides a better fit with χ²/dof = 98.5/44, but is still inadequate, especially at energies above 1.4 keV where a hard excess is obvious, see the top panel in Figure 4. Therefore, we used a model with the sum of a power law and an MCD, which significantly improved the fit with χ²/dof = 63.4/42. To test the significance of the power-law tail, we simulated 10⁴ spectra based on the single MCD model. In none of the simulated spectra did addition of a power-law reduce the χ² by the amount seen in the real data. This suggests a chance probability less than 10⁻⁴. The F-test gives a consistent probability of 9.5 × 10⁻⁵ despite caveats about its validity in such a case (Protassov et al. 2002).

Zoom In Zoom Out Reset image size

Although the MCD plus power-law model improves the fit considerably, it still does not provide an adequate fit. As one can see from the middle panel of Figure 4, most residuals arise around 1 keV and appear like an absorption feature. Thus, we then tried two absorption components, both of which gave adequate fits. An absorption edge at 1.02 keV gave χ²/dof = 36.8/40 and a Gaussian absorption line with a centroid of 1.16 keV gave χ²/dof = 37.0/39. The residuals for both models are flat throughout the passband, see the bottom panel of Figure 4. The absorption equivalent width is 0.13–0.19 keV, depending on the model. Similarly, we tested the significance of the absorption component using simulations, and an edge was detected as significant in only 1 out of 10⁴ trails, indicative of a chance probability of ∼10⁻⁴.

The spectral parameters are listed in Table 3. Due to the poor fit of the MCD plus power-law model, we fixed the absorption column density at the best value, N_H = 3.2 × 10²¹ cm⁻², obtained from the best-fit model. Consistent parameters of the continuum components were obtained for all models. The unfolded spectrum from the MCD plus power-law model with absorption edge is plotted in Figure 5. The MCD is the dominant component of the spectrum, with a contribution of 96% of the total unabsorbed flux in the 0.3–10 keV band.

Zoom In Zoom Out Reset image size

Table 3. Spectral Parameters of NGC 247 ULX

Model	NH	Thermal		Power Law		Absorption			fX	LX	χ2/dof
		kT	R	Γ	NPL	E	τ	σ
$\bf a_1$	3.2a	0.117+0.003−0.003	1.34+0.12−0.13	1.6+0.6−0.7	0.5+0.5−0.3	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	1.75+0.18−0.15	2.92+0.16−0.16	88.0/43
$\bf a_2$	3.2+0.7−0.7	0.13+0.02−0.02	0.9+1.3−0.5	2.5+1.4−1.0	1.7+3.8−1.2	1.02+0.02−0.03	1.4+0.5−0.5	⋅⋅⋅	1.71+0.17−0.13	2.7+3.4−1.3	36.8/40
$\bf a_3$	3.5+0.9−0.8	0.123+0.022−0.012	1.3+1.2−0.4	2.0+1.1−1.0	0.9+1.3−0.6	1.16+0.05−0.03	0.26+0.08−0.08	0.08+0.04−0.03	1.74+0.18−0.14	3.4+4.0−1.6	37.0/39
$\bf b_1$	2.7a	0.103+0.003−0.002	1.35+0.07−0.11	1.6+0.6−0.7	0.6+0.5−0.3	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	1.76+0.18−0.15	1.74+0.09−0.09	81.5/43
$\bf b_2$	2.7+1.0−0.7	0.112+0.016−0.017	1.0+1.3−0.5	2.7+2.2−1.0	2.0+7.5−1.4	1.02+0.03−0.03	1.3+0.5−0.4	⋅⋅⋅	1.70+0.17−0.14	1.6+2.9−0.8	39.0/40
$\bf b_3$	2.9+0.9−0.7	0.109+0.017−0.013	1.2+1.2−0.5	2.1+1.2−0.9	1.0+1.4−0.6	1.15+0.04−0.03	0.23+0.07−0.08	0.08+0.04−0.03	1.74+0.18−0.14	1.9+1.7−0.9	38.9/39

Notes. XSPEC models $\bf a_1$: wabs * (diskbb + powerlaw); $\bf a_2$: wabs * edge * (diskbb + powerlaw); $\bf a_3$: wabs * gabs * (diskbb + powerlaw); $\bf b_1$: wabs * (bbodyrad + powerlaw); $\bf b_2$: wabs * edge * (bbodyrad + powerlaw); $\bf b_3$: wabs * gabs * (bbodyrad + powerlaw). N_H is the absorption column density in 10²¹ cm⁻². kT and R are temperature in keV and face-on radius in 10⁴ km of the thermal component, respectively; for the diskbb model, they correspond to the values at the innermost disk. Γ is the power-law photon index. N_PL is the power-law normalization at 1 keV in units of 10⁻⁵ photons keV⁻¹ cm⁻² s⁻¹. E and τ are energy in keV and optical depth of the local absorption component, respectively, and σ is the width of the Gaussian absorption in keV. f_X is the observed flux in 0.3–10 keV in units of 10⁻¹³ erg cm⁻² s⁻¹. L_X is the unabsorbed luminosity in 0.3–10 keV in units of 10³⁹ erg s⁻¹. Errors are quoted at 90% confidence level. ^aFixed at the value obtained from the best-fit model

Download table as:  ASCIITypeset image

We also replaced the MCD with a blackbody and obtained similar results as shown in Table 3. We note that the use of a more sophisticated model for interstellar absorption, such as tbabs, leads to almost identical results as obtained with wabs due to the moderate resolution and statistics.

We tried to fit the spectrum with an MCD plus Comptonization model, which is found to fit for ULX spectra with high statistics (Gladstone et al. 2009). This model provides an equally good fit as the MCD plus power-law model and results in almost identical parameters for the disk component. The disk temperature is 0.13 ± 0.02 keV and the inner disk radius is 0.9^+0.9_−0.5 × 10⁴ km assuming a face-on orientation. A similar edge is required in the model at 1.02^+0.02_−0.03 keV, with optical depth of 1.4^+0.5_−0.2. The Comptonization parameters are poorly constrained due to the low number of photons in the 2–10 keV band.

Light curves of the ULX were created using the PN events in the same region as for spectral analysis with FLAG = #XMMEA_EP and PATTERN ⩽ 4. Continuous intervals with low background were selected from within the GTIs. The 0.3–2 keV light curve shown in Figure 6 has a bin size of 200 s.

Zoom In Zoom Out Reset image size

For each continuous interval, power spectra were calculated from a light curve with a bin size equal to the CCD frame time, ∼73 ms, using 8192-point Fourier transforms. A final power spectrum was created by averaging all the individual power spectra with frequency channels geometrically binned by a factor of 1.2. Each point in the final spectrum is the mean of at least 33 individual data points, to assure validity in modeling the spectrum using minimal χ². The final power spectrum in the 0.3–2 keV band normalized following Leahy et al. (1983) is shown in Figure 6. A power-law plus a constant can adequately fit the power spectrum, with the constant describing the Poisson noise level of 1.995 ± 0.009 and the power law describing low-frequency noise with a slope of 1.22^+0.20_−0.16. The source count rate is 0.10 counts s⁻¹ in the 0.3–2 keV band and the background photon fraction is estimated to be 4.1%. The fractional root-mean-square (rms) power of the source is estimated to be 0.52 by integrating the power-law component over 0.001–0.1 Hz, or 0.31 extrapolating the power law to the frequency range 0.1–10 Hz. The power spectrum in the 2–10 keV band is consistent with that of Poisson noise, due to poor statistics.

We re-analyzed the XMM-Newton observation of NGC 247 obtained in 2001 in a similar way. As the background is too high to find a quiescent level adaptively for the PN data, an additional requirement that the background count rate is less than 1.5 counts s⁻¹ for the PN is imposed to generate GTIs. The PN and the MOS spectra are fitted simultaneously with an MCD plus power-law model, resulting in χ²/dof = 53.4/49. The MCD component contributes about 88% of the unabsorbed luminosity in the 0.3–10 keV band, which is L_X = 8⁺¹³₋₄ × 10³⁹ erg s⁻¹, with a disk inner temperature of 0.15 ± 0.03 keV and a face-on inner radius of 1.2^+2.5_−0.7 × 10⁴ km. The photon index of the power-law component is poorly constrained, with a 90% upper limit of 3.8. The interstellar absorption has a column density of 3.6^+1.3_−1.0 × 10²¹ cm⁻². The best-fit parameters are consistent with those reported in Winter et al. (2006) within errors.

The XMM-Newton observation of NGC 247 obtained in 2009 December is, to date, the deepest X-ray observation of the galaxy and allows an X-ray source population study. As shown in Figure 2, sources inside the D25 ellipse of NGC 247 have a luminosity function well consistent with that of LMXBs. A population of X-ray sources dominated by LMXBs is reasonable given the low star formation rate of the galaxy.

For sources outside of the D25 ellipse of the galaxy, the log N(> S) versus log S distribution is roughly consistent with that expected for AGNs (Brandt & Hasinger 2005) (Figure 3) at fluxes above 2 × 10⁻¹⁴ erg cm⁻² s⁻¹. Below this flux, the distribution flattens significantly with decreasing flux. This may be caused by the decrease of sensitivity due to vignetting at large off-axis angles. Some of these sources have bright optical counterparts with an X-ray to optical flux ratio consistent with that of AGNs. Therefore, we conclude that most sources outside of the D25 ellipse of NGC 247 are likely background objects.

The observed X-ray spectrum of the ULX is completely dominated by a cool, thermal emission component with a temperature around 0.1 keV; a weak power-law tail with a photon index of about 2.5 is also evident but contributes only a small fraction of the flux. This spectral shape is in contrast with that of many other ULXs whose high energy component contributes the majority of the flux (more than 60%) and can be explained by an optically thick corona (Gladstone et al. 2009). In fact, Gladstone et al. (2009) fits can predict an intrinsically soft-dominated spectrum, but this is a model-dependent result, while the observed spectral shapes of their sources are always dominated by the emission from a hard tail. It is unlikely that the high energy component in NGC 247 ULX arises from an optically thick corona; otherwise, such a corona must have a tiny covering factor since most disk photons are unscattered.

The spectrum, however, is consistent with that of black hole binaries in the thermal state (McClintock & Remillard 2006; Remillard & McClintock 2006), and may hint at a connection in nature. An important observational characteristic of the thermal state is that the disk luminosity varies roughly with the fourth power of the disk inner temperature, equivalent to a constant disk inner radius. This source has shown a consistent disk inner radius derived from two XMM-Newton observations, though the constraint from the 2001 observation is loose. For a positive identification of the thermal state, one also needs timing information. This ULX displays red noise with a ν⁻¹ form above the white noise, which is consistent with that seen in the thermal state. However, the integrated rms power is significantly higher than that seen in the thermal state. For Galactic black holes, the total rms power in the 0.1–10 Hz band is usually less than 0.075 in the thermal state, and only in the hard state is the rms power found to be greater than 0.15 (Remillard & McClintock 2006). However, the X-ray spectrum of the ULX is much too soft to be classified as in the hard state.

If the ULX is in the thermal state, then the source would be an interesting candidate to be an intermediate-mass black hole. The accretion disk in the thermal state is believed to extend all the way to the innermost stable circular orbit (ISCO) around the central black hole. The ISCO radius depends only on the black hole mass and spin, as R_ISCO = 6GM/c² for a non-spinning black hole or six times smaller for an extremely spinning object, where G is the gravitational constant, M is the black hole mass, and c is the light speed. Assuming a hardening correction of 1.7 and that the maximum temperature occurs at 2.4 times the ISCO radius (Makishima et al. 2000), the ISCO radius can be expressed as R_ISCO ≈ 1.2R_in, where R_in is the apparent inner radius derived directly from the normalization of the MCD component. Adopting the parameters from the best-fitted model, MCD plus power law with absorption edge, and assuming a face-on disk, we have R_in = (0.5–2.2) × 10⁴ km in the 90% confidence region, corresponding to the ISCO radius of a 600–3000 M_☉ Schwarzschild black hole, or a more massive black hole if spinning or the disk is tilted. The observed luminosity is about a few percent of the Eddington limit of such a black hole, which is consistent with Eddington ratios typically found in the thermal state. Similar results can be obtained with the Gaussian absorption line.

Due to the lack of a characteristic feature in the power spectrum, we are unable to constrain the black hole mass via timing. The rms–mass relation derived for AGNs (Zhou et al. 2010) may have issues in extrapolation to lower masses. The faintness in the 2–10 keV band of the source prevents us from implementing the means proposed by González-Martín et al. (2011) to estimate the black hole mass.

A few other ULXs show a dominant soft, thermal emission component and are classified as supersoft ULXs: M101 ULX-1 (Pence et al. 2001; Kong & Di Stefano 2005; Mukai et al. 2005), Antennae X-13 (Fabbiano et al. 2003), M81 ULS1 (Swartz et al. 2002; Liu 2008), and NGC 4631 X1 (Carpano et al. 2007; Soria & Ghosh 2009). These sources also show extreme variability: the observed fluxes vary by up to a factor of a few tens, and the intrinsic luminosity changes may be as large as factors of 10²–10³ (Kong et al. 2004; Mukai et al. 2005; Liu 2008). Short-term variability has been detected in almost all of these sources at timescales down to a few ks (Swartz et al. 2002; Mukai et al. 2005; Carpano et al. 2007; Liu 2008). M101 ULX-1 also showed a power spectrum with a ν⁻¹ form at frequencies below 10⁻³ Hz (Mukai et al. 2005) during its outburst in 2000 March with variability more pronounced at high energies (0.8–2 keV) than in the lower band (0.2–0.8 keV) (Mukai et al. 2003). For a direct comparison, we calculated the rms power of NGC 247 ULX in the same energy bands and found that the fractional rms is 0.80 in the 0.8–2 keV band and 0.44 in the 0.2–0.8 keV band. Similar to NGC 247 ULX, NGC 4631 X1 also exhibits an absorption edge at 1.01 keV with an optical depth of 2.0 in its high state (Soria & Ghosh 2009). These similarities suggest that NGC 247 ULX may be a new member of the supersoft, ultraluminous family.

The other soft ULXs seem to keep a constant temperature during huge changes in luminosity, indicative of a dramatic change of the emitting surface area of the same order. Fabbiano et al. (2003) have pointed out that this would lead to difficulty in interpreting the emission as due to accretion power release within a few Schwarzschild radii of the black hole. For NGC 247 ULX, the four ROSAT and two XMM-Newton observations have detected a change by a factor of about three of the observed flux in 0.1–2 keV, which is significantly smaller than others. Therefore, the thermal state interpretation has not been ruled out for NGC 247 ULX; new, multiple observations are the key to revealing its nature.

An alternative explanation of the soft, thermal emission is that it arises from a photosphere that has a radius of 10⁹ cm, inferred from the blackbody component. The shortest timescale of variation above white noise is about 10² s (Figure 6), which places a constraint that the emitting area must be smaller than 10¹² cm, compatible with the size obtained from the blackbody parameters. This photosphere is unlikely the surface of a white dwarf, as the observed luminosity and temperature are too high to be explained by steady nuclear burning on a white dwarf surface (Starrfield et al. 2004). Plus, presence of the power-law component with a 0.3–10 keV luminosity of 1.2 × 10³⁸ erg s⁻¹ is unexpected in such a case. It has been suggested that supercritical accretion could drive massive outflows above the disk and form an optically thick photosphere, with a temperature expected to be about 0.1 keV, which could shield the inner disk emission when viewed at a high inclination angle (Ohsuga et al. 2005; Poutanen et al. 2007). In this picture, the supersoft ULXs and normal ULXs are the same objects with different viewing angles. This interpretation may be in accord with the absorption feature observed at ∼1 keV, which is likely due to L shell transitions of highly ionized iron. Besides this ULX and NGC 4631 X1, less similar absorption features near this energy have been detected in the ULX NGC 1365 X1 (Soria et al. 2007) and in the quasar PDS 456 (Reeves et al. 2003), and are explained as a spectral signature of a massive outflow. The absence of absorption from low-Z elements like oxygen could be a consequence of high ionization.

We conclude that the X-ray properties of the NGC 247 ULX are unlike any popular emission state known in black hole binaries. The source exhibits a spectrum like the thermal state, but shows variability much stronger than seen in the thermal state. A possible strong absorption feature is detected. This source and other supersoft ULXs may exhibit a new accretion state that is different from those seen in Galactic black hole binaries and the majority of ULXs.

A variety of potential thermal emission components have been suggested to be present in the energy spectrum of ULXs, including the soft excess (0.1–0.4 keV) found in the majority of ULXs (Feng & Kaaret 2005; Stobbart et al. 2006), the high-temperature (∼keV) ULXs (Makishima et al. 2000), and supersoft ULXs. The properties of these thermal components are summarized and compared in Table 4, with NGC 247 ULX and M82 X41.4+60 in the thermal state highlighted. Among them, a disk interpretation is best established for M82 X41.4+60 because its thermal state is consistent with those of stellar-mass X-ray binaries in terms of timing and spectral properties and spectral evolution (Feng & Kaaret 2010). For the others, definitive evidence is lacking for positive determination of their nature, except that the disk explanation is ruled out for other supersoft ULXs other than NGC 247 ULX. Based on the timing information, it appears unlikely that the soft excess in "normal" ULXs and the soft (dominant) spectral component in supersoft ULXs have the same physical origin.

PHL 6625 was identified as a radio-quiet background QSO at z = 0.38 based on a Mg ii emission line (Margon et al. 1985). A single power-law model with interstellar absorption provides an adequate fit to its X-ray spectrum. The parameters listed in Table 2 are obtained from fitting in the 0.3–10 keV energy range, and the hydrogen column density is very close to the Galactic value at N_H = 2 × 10²⁰ cm⁻², indicative of negligible absorption within NGC 247. However, if we include data down to 0.2 keV, two more energy bins, then the best-fit absorption is N_H = 5.3^+1.2_−1.0 × 10²⁰ cm⁻². This is caused by the low flux between 0.2 and 0.3 keV and is consistent with the results obtained from ROSAT observations, in which a total absorption column density N_H = 5.4^+2.4_−2.1 × 10²⁰ cm⁻² was derived from fitting in the 0.1–2 keV band, and the excess absorption mainly takes effect at energies below 0.3 keV (Elvis et al. 1997). Those authors claimed a positive detection of absorbing materials in NGC 247 assuming a Galactic N_H = 1.4 × 10²⁰ cm⁻² based on 21 cm observations of neutral hydrogen. Here, we confirm this conclusion with tighter constraints, even assuming a higher Galactic absorption. Therefore, with future high-resolution spectroscopic observations, this QSO could be used to study the interstellar medium within NGC 247.

For the 2009 data, the power-law photon index at the rest frame is Γ = 2.10^+0.13_−0.09. We extracted spectra for this object from the 2001 XMM-Newton observation and found a harder spectrum with Γ = 1.64 ± 0.17 at a flux level of (3.7 ± 0.6) × 10⁻¹³ erg cm⁻² s⁻¹ in 0.3–10 keV similar to that in 2009. The fluxes observed with XMM-Newton are about 10 times lower than the ROSAT flux of (5.9 ± 0.7) × 10⁻¹² erg cm⁻² s⁻¹ in the 0.2–2.4 keV band (Elvis et al. 1997). No short-term X-ray variability has been detected for this object.

Source 28. The source is hard, and its spectrum can be described by a power law with Γ = 1.27, but the residuals vary systematically with energy. Adding an MCD component improves the residuals and results in χ²/dof = 31.7/54. The disk has a temperature at the inner radius of 0.35^+0.14_−0.10 keV with a bolometric luminosity of 2 × 10³⁷ erg s⁻¹ assuming a face-on geometry, and the power-law component has a photon index Γ = 0.6^+0.2_−0.3. The disk component has a fractional luminosity of 11% in the 0.3–10 keV band. If this source is a black hole binary, it is possibly in the hard state.

Source 34. The source sits at the north end of the NGC 247 disk and was bright during the 2001 XMM-Newton observation with Γ = 2.3^+1.5_−0.9, N_H = 1.9^+2.4_−0.8 × 10²¹ cm⁻², and f_X = 2.0 × 10⁻¹³ erg cm⁻² s⁻¹ in the 0.3–10 keV band, or L_X = 4.6 × 10³⁸ erg s⁻¹ corrected for absorption. However, it became faint in the 2009 XMM-Newton observation, with a flux of 3.4 × 10⁻¹⁴ erg cm⁻² s⁻¹ in the 0.3–10 keV band, six times dimmer than in 2001. It is possibly source 3 in Fabbiano et al. (1992) from Einstein observations and source 1 in Read et al. (1997) from ROSAT observations. The ROSAT observations revealed a luminosity of 1.7 × 10³⁸ erg s⁻¹ corrected for Galactic absorption only (at 3.4 Mpc) in the 0.1–2 keV band or 3 × 10³⁸ erg s⁻¹ corrected for all absorption. The luminosity in the same band from the 2001 XMM-Newton observation is about 6 × 10³⁸ erg s⁻¹.

There is no bright optical counterpart found at the X-ray position. However, the source is associated with a possible star-forming region, seen in the ultraviolet image from the 2001 XMM-Newton observation, with diffuse emission and star clusters nearby. Therefore, the source is likely an X-ray binary in NGC 247 and is perhaps an accreting black hole.

Source 36. This source could be source 3 in Read et al. (1997). The best-fitted column density is consistent with the Galactic value. Its flux estimated from the 2001 XMM-Newton observation is consistent with that in 2009. In the X-ray image, the source seems to show extended features and may consists of multiple objects, which can only be resolved with future Chandra observations.

Source 61. Its X-ray spectrum cannot be adequately fitted by a single power-law model, with χ²/dof = 80.9/60; adding a Gaussian absorption line significantly improves the fit with χ²/dof = 58.1/57. The absorption line has a central energy of 0.9 ± 0.1 keV and a width of σ = 0.22^+0.28_−0.16 keV with an optical depth of 0.29^+0.58_−0.19. The equivalent width is calculated to be around 0.30 keV. Its observed flux in 2001 was about twice of that in 2009. There is a bright optical source with B = 18.83 and R = 18.58 within 2'' of the X-ray position, and the X-ray to optical flux ratio is estimated to be −0.3. As discussed above, this source is likely a background AGN. Due to the low absorption column density, the absorption line is likely produced intrinsic to the source. Similar absorption has been seen in AGNs (Blustin et al. 2002).

Source 13. This is the hardest source detected with XMM-Newton. It is obvious in the hard band (2–10 keV) but undetected in the soft band (0.3–2 keV), consistent with N_H > 5 × 10²² cm⁻² for a power-law spectrum with Γ = 2. The source lies about 7' west of the nucleus of NGC 247 and is likely a highly absorbed background AGN.

Sources 41, 50, 59, and 72. These four sources are the softest besides the ULX and are only detected in the soft band (0.3–2 keV) with a hardness ratio around −0.9, which is consistent with a spectrum with a blackbody temperature <0.4 keV or a power-law photon index Γ > 2.5, assuming Galactic absorption. Sources 41 and 72 are outside of the D25 ellipse of NGC 247, but sources 50 and 59 are inside. Their X-ray luminosities are about 10³⁶–10³⁷ erg s⁻¹. These sources do not appear to have bright optical counterparts. They could be canonical supersoft sources that are powered by nuclear burning on the surface of white dwarfs or high redshift AGNs.