RX J1301.9+2747: A HIGHLY VARIABLE SEYFERT GALAXY WITH EXTREMELY SOFT X-RAY EMISSION

J1302 was spectroscopically observed by the SDSS in 2007 March. Figure 1 shows the rest-frame spectrum for J1302 (black line), which is dominated by the starlight of the host galaxy. To subtract the starlight and the nuclear continuum, we followed the recipe described in detail in Dong et al. (2005). As seen in Figure 1, the galaxy starlight model (green line) gives a very good fit to the optical continuum (χ²/dof = 3648/3208). After the subtraction of stellar absorption lines, we fitted the emission-line spectrum, represented by a blue line, by using Gaussians to derive the parameters of the emission lines. [O iii] λλ4959, 5007, [N ii] λλ6548, 6583, Hα, and [S ii] λλ6717, 6731 emission lines are clearly detected with signal-to-nose ratio (S/N) > 5, while Hβ line is only weakly distinguishable with S/N ∼ 1.4. The right panel in Figure 1 displays the emission-line spectra, alongside the best-fit Gaussian models. There is no apparent broad component of Hα line, and a narrow Gaussian with a linewidth of FWHM ∼ 240 km s⁻¹ can provide a good fit. In order to verify the absence of the broad Hα line, we add an additional Gaussian to the narrow Hα, with the width fixed at 2000 km s⁻¹. The flux is allowed to vary in the fitting process. We found that the fitting was marginally improved by adding this component, and the S/N for the Hα broad component is only ∼0.5. This suggests that the broad Hα line, if there is one, is extremely weak in J1302. The ratios of the narrow lines [O iii] λ5007/Hβ > 4.8 (using 3σ upper limit of Hβ line flux), and [N ii] λ6583/Hα = 2.3, place J1302 into the Seyfert regime on the Baldwin–Phillips–Terlevich diagram of Kewley et al. (2006). The flux ratios of [S ii]/Hα and [O i]/Hα are 0.56 and 0.32, respectively, further strengthening the Seyfert nature of J1302 according to the line ratio diagnostic diagrams of Kewley et al. (2006).

Zoom In Zoom Out Reset image size

J1302 was observed by XMM-Newton EPIC cameras in 2000 December with a total exposure time of 29 ks. It was detected ∼7.3 arcmin away from the center of the field of view in the XMM-Newton imaging of the Coma cluster (ObsID 0124710801). The XMM-Newton data were reprocessed with the Science Analysis Software version 11.0.0, using the calibration files as of 2011 December. We used principally the PN data, which have much higher sensitivity, using the MOS data only to check for consistency. Spectra and light curves of the source were extracted from a circular region with a radius of 40'' centered at the source position for both PN and MOS cameras. Background spectra were made from source-free areas on the same chip using four circular regions identical to the source region. The epochs of high background events were examined and excluded using the light curves in the energy band above 12 keV.

The Chandra pointing observation of J1302 was taken in 2009 June for about 5 ks. The data were processed with CIAO (version 4.3) and CALDB (version 4.4.1), following the standard criteria. Figure 2 shows the Chandra X-ray contours of J1302, overlaid on the Hubble Space Telescope (HST) image in the B band. It is clear from the figure that the X-ray emission of J1302 is point-like. The center for the X-ray source is coincident with the optical nucleus of the galaxy (with a positional offset of ∼01). Given the subarcsecond spatial resolution of Chandra, we conclude that most of the X-ray emission, if not all, comes from the nuclear region of the galaxy, likely related to the AGN.

Zoom In Zoom Out Reset image size

Light curves from the XMM-Newton and Chandra observations are shown in Figure 3. The source exhibits large-amplitude count rate variations in both observations. It can be seen from the PN light curve that there is a giant flare with count rates rising by a factor of five times the average value, having a duration of ∼2 ks. The X-ray flux then declines to a relatively steady state. Such an X-ray flare is confirmed by the MOS light curves, in which a possible decline of another flare is also recorded at the beginning of the MOS observation. The time interval of the two flares is about 17 ks. Interestingly, a similar flare is seen in the Chandra light curve, with count rates increasing by a factor of seven within ∼1 ks. Similar amplitudes of flares in the XMM-Newton and Chandra observations which span ∼9 yr suggest that the flare behavior in J1302 seems persistent on a timescale of about a decade. The spectral variability during the flare will be explored in detail in Section 2.4.

Zoom In Zoom Out Reset image size

As both XMM-Newton and Chandra observations show peculiar flare behaviors, we attempt to quantify the spectral variability during flares by dividing the data into high and low flux intervals, using count rate thresholds of 0.35 counts s⁻¹ for XMM-PN and 0.08 counts s⁻¹ for Chandra. For simplicity, we classify the data above the count rate thresholds as belonging to the flare state, and those which fall below, to the quiescent state. The spectra data were grouped in the following manner: data from the XMM-Newton had at least 20 counts per bin to ensure the χ² statistics, and the Chandra data had at least 3 counts per bin and utilized the C-statistics which was adopted for minimization. Spectral fitting was performed using XSPEC (Version 12.6) and limited to the 0.2–2 keV range for XMM-Newton, since the emission is background dominated above that energy range. The Chandra data were fitted in the energy range between 0.3 and 3 keV. Throughout the model fittings, the Galactic column density was considered and fixed at $N_{\rm H}^{\rm Gal}=0.75\times 10^{20}$ cm⁻² (Kalberla et al. 2005).

Spectral variability is clearly present between the two states with the source being much harder in the flare state. To illustrate the differences in the spectral slope between the two states, we show the spectra together with an absorbed power-law model in Figure 4. While the fit is generally acceptable for the two states, as confirmed by the reduced χ² value (see Table 1), the photon indices obtained from a power-law fit are, however, extremely steep with Γ = $4.4^{+0.5}_{-0.4}$ for $N_{{\rm H}}=4.3^{+2.2}_{-1.8}\times 10^{20}$ cm⁻², and Γ = $7.1^{+0.9}_{-0.7}$ for $N_{{\rm H}}=3.6^{+1.9}_{-1.6}\times 10^{20}$ cm⁻², for the XMM flare and quiescent state, respectively. The photon indices belong to the steepest values obtained from AGN X-ray spectra. For comparison, the mean photon index in the 0.2–2.0 keV band is ∼2.9 for a sample of soft X-ray selected AGNs observed with the ROSAT RASS (Grupe et al. 2010). The result of this comparison indicates that the X-ray spectrum of J1302 is extremely soft compared to other AGNs. The absorption-corrected luminosity in the 0.5–2.0 keV range for this simple power-law model is 6.7 × 10⁴¹ and 2.8 × 10⁴⁰ erg s⁻¹ for the XMM flare and quiescent state, respectively.

Zoom In Zoom Out Reset image size

Table 1. Spectral Fitting Results for the XMM-Newton and Chandra Observation at Different States

XMM-PN Flare State
wabs*modela	NH	Γ	kT	τb	χ2/dof	F0.5–2 keVc	L0.5–2 keVc
	(1020 cm−2)		(eV)			(10−14 erg s−1 cm−2)	(1040 erg s−1)
powl	$4.3^{+2.2}_{-1.8}$	$4.4^{+0.5}_{-0.4}$			40.4/28	40	67
bbody	0.75(fixed)d		$99^{+6}_{-5}$		55.5/29	38	54
diskbb	0.75(fixed)		$134^{+9}_{-9}$		43.7/29	39	55
compTT	$4.3^{+2.2}_{-1.8}$			$0.16^{+0.07}_{-0.05}$	40.3/28	40	67
XMM-PN Quiescent State
powl	$3.6^{+1.9}_{-1.6}$	$7.1^{+0.9}_{-0.7}$			27.9/40	1.6	2.8
bbody+powl	0.75(fixed)	$4.3^{+1.6}_{-1.9}$	$43^{+6}_{-3}$		26.3/39	1.7	2.5
diskbb+powl	0.75(fixed)	$3.7^{+2.0}_{-2.0}$	$52^{+5}_{-5}$		25.9/39	1.7	2.6
compTT	$2.9^{+1.4}_{-1.3}$			0.016(<0.03)	27.7/40	1.6	2.7
Chandra Flare State
wabs*model	NH	Γ	kT	τ	C/dof	F0.5–2 keV	L0.5–2 keV
	(1020 cm−2)		(eV)			(10−14 erg s−1 cm−2)	(1040 erg s−1)
powl	3.7(<11)	$3.2^{+1.1}_{-0.6}$			22.5/22	34	51
bbody	0.75(fixed)		$195^{+24}_{-20}$		27.6/23	35	47
diskbb	0.75(fixed)		$275^{+51}_{-38}$		22.5/23	35	47
compTT	3(<20)			$0.44^{+0.28}_{-0.28}$	22.5/22	34	50
Chandra Quiescent State
powl	0.75(fixed)	$4.5^{+0.6}_{-0.6}$			16.5/14	3.7	5.3
bbody+powl	0.75(fixed)	$3.5^{+0.8}_{-1.0}$	$29^{+19}_{-16}$		8.0/12	3.2	4.5
diskbb+powl	0.75(fixed)	$3.5^{+0.8}_{-1.0}$	$32^{+25}_{-18}$		8.0/12	3.2	4.4
compTT+powl	0.75(fixed)	2.8(<3.9)	$2.8^{+1.1}_{-3.1}$	0.01(<0.09)e	8.1/12	3.1	4.3

Notes. ^aSpectral model (as given in XSPEC) multiplied by a neutral absorption (wabs) with column density N_H. wabs is the photo-electric absorption model using Wisconsin–Morrison & McCammon (1983) cross-sections. ^bPlasma optical depth in the compTT model. We fixed the plasma temperature at 20 keV to obtain the constraints on the scattering optical depth, and the seed photons were assumed to follow Wien's law with a temperature of 22 eV (see Section 2.4 for details). ^cF_{0.5–2 keV} is the observed 0.5–2 keV flux in units of 10⁻¹⁴ erg cm⁻² s⁻¹. L_{0.5–2 keV} is the unabsorbed luminosity in the energy range of 0.5–2 keV, in units of 10⁴⁰ erg s⁻¹. ^dThe column density was fixed to Galactic value $N_{\rm H}^{{\rm Gal}}$ if the fitting yields an $N_{\rm H}<N_{\rm H}^{{\rm Gal}}$. ^ePegged at the minimum value allowed in XSPEC.

Download table as:  ASCIITypeset image

In order to further investigate the spectral variability in J1302, we then attempted to fit the spectra with a blackbody (bbody in XSPEC) or Multiple Color Disk model (MCD, diskbb in XSPEC), and a thermally Comptonized disk model (compTT in XSPEC; Titarchuk 1994); both of these alternative models have been used to fit the spectra of Galactic BHBs (e.g., Done et al. 2007), and the soft X-ray excess emission in AGNs (e.g., Porquet et al. 2004; Patrick et al. 2012). The diskbb model integrates over the surface of the accretion disk to form a multicolor blackbody spectrum, and compTT is an analytic model that self-consistently calculates the spectrum produced by the Comptonization of soft seed photons in a hot corona above the accretion disk. The physical parameters of the compTT model are: the soft photon temperature (kT₀), the temperature of the Comptonizing electrons (kT_e), and the plasma scattering optical depth (τ). For our fitting with the compTT model, a disk geometry was assumed for the Comptonizing region, and the seed photons were assumed to follow Wien's law with a temperature of 22 eV (the expected disk temperature in Section 3.2). Because the temperature and optical depth of the Comptonizing plasma are strongly coupled (both are equally involved in shaping the spectrum) and thus cannot be constrained simultaneously, we fixed the plasma temperature at 20 keV and obtained constraints on the optical depth.¹ The single Comptonized model yields consistent fitting results with the previous simple power-law model for the spectrum at both states. The Compton optical depth is $\tau =0.16^{+0.07}_{-0.05}$ and τ < 0.03 for the flare and quiescent state, respectively. For the spectrum in the XMM-Newton flare state, a multicolor-disk blackbody gives an equivalent fit, which is statistically better than the simple blackbody emission.

The spectrum at the XMM-Newton quiescent state, however, shows an excess of emission at energies above ∼0.7 keV when fitted with a thermal model (bbody or diskbb). The addition of a power law to the model improves the fit with very high statistical significance (χ² decreased by 13.9 for two extra parameters, at a 99.98% level according to F-test). The power-law component contributes ∼15% of the total luminosity in the 0.3–2 keV band. In this case, we obtain an effective blackbody temperature of $kT_{\rm BB}=43^{+6}_{-3}$ eV, comparable to the seed photon temperature assumed in the compTT model. Although having large uncertainties, the additional power-law component is relatively steep with photon index $\Gamma =4.3^{+1.6}_{-1.9}$, and it is close to what is observed in the flare state. The spectral fitting results for the PN data, alongside the observed flux and intrinsic luminosity in the 0.5–2 keV range for each model, are shown in Table 1. Note that we used the same models to fit the MOS data and found that the results agree well with the PN data.

The Chandra spectra were fitted with the same models used in the XMM-Newton observation and the results are listed in Table 1. During the first run we found that the photon indices derived from the simplest power-law model are systematically flatter than the values for the XMM-Newton data. The spectrum during the flare can be well fitted by a power-law model (C/dof = 22.5/22). On the other hand, a power-law model is not sufficient to fit the data at the Chandra quiescent state. The addition of a soft thermal component with respect to the power-law model improves the fit significantly (C value decreases by ∼9 for two extra parameters, corresponding to a significance level of 98.7%). The resulting best-fit photon index is flatter, with $\Gamma =3.5^{+0.8}_{-1.0}$. Similarly, we obtain an effective disk temperature, when fitted with a blackbody, of $kT_{{\rm BB}}=29^{+19}_{-16}$ eV. This value is slightly lower than the blackbody temperature for the XMM-Newton quiescent state, but the parameter is loosely constrained due to the poor statistics of the Chandra data. The power-law component in this case contributes ∼40% of the total luminosity in the 0.3–2 keV, indicating a possible change of the power-law or the blackbody emission between Chandra and XMM-Newton observations.

Although a Comptonized model as opposed to a power-law model also provides a good fit for the Chandra flare state spectrum, it is not sufficient to fit the data at the quiescent state. The addition of an extra hard power-law component is needed at a significance level of 98.7% according to F-test. The best-fitted optical depth τ for the compTT model is not significantly different from that obtained with the XMM-Newton data. The unabsorbed luminosity in the 0.5–2 keV band, based on the best-fit power-law model and the diskbb+power-law model for the Chandra flare and quiescent state, is 5.1 × 10⁴¹ and 4.4 × 10⁴⁰ erg s⁻¹, respectively.

X-ray observation with Chandra, which has superb spatial resolution of ∼05, revealed the presence of an AGN in J1302: the center of the bright unresolved X-ray emission coinciding with the optical point-like nucleus of the galaxy (with a position offset of ∼01). Both the Chandra and XMM-Newton observations show that it has Seyfert-like X-ray luminosity of ∼10⁴¹ erg s⁻¹ in the energy range of 0.5–2 keV (∼10⁴² erg s⁻¹ in the 0.2–2 keV band), and rapid X-ray variability down to a timescale ∼1 ks. Additionally, the optical spectrum of J1302 displays Seyfert-like narrow emission-line ratios. All these observational facts point to the presence of an AGN (Seyfert nucleus) in this galaxy. This can be further supported by the point-like appearance of a nuclear source from HST imaging observations with ∼01 resolution (Caldwell et al. 1999).

Based on ROSAT X-ray observations, Dewangan et al. (2000) argued for the presence of an AGN in J1302, a view that is consistent with ours on the basis of the new observations. However, based on their observed optical spectrum, Dewangan et al. conclude that the galaxy nucleus is more like a LINER. In this paper, we have carefully modeled the host galaxy's starlight, especially the stellar absorption features, and subtracted them from the new SDSS spectrum, enabling us to accurately measure the weak AGN emission lines in J1302 (e.g., Dong et al. 2005), thus more resolutely confirming the Seyfert nature based on the line ratio diagnostics.

The lack of detectable broad permitted lines prevents us from estimating the central BH mass of J1302 using the conventional linewidth–luminosity–mass scaling relation. In the SDSS spectral fitting, we found that the strongest narrow line, [O iii] λ5007, is marginally resolved with Gaussian σ = 58 ± 9 km s⁻¹ (after correcting for the instrument resolution). Using the width of the [O iii] λ5007 line as a proxy for the stellar velocity dispersion of the host galaxy, we obtained a BH mass of M_BH = 8 × 10⁵ M_☉ with an intrinsic scatter of 0.5 dex (e.g., Xiao et al. 2011).

The bolometric luminosity for J1302 can be estimated from the optical continuum luminosity. We retrieved the high resolution HST/WFPC2 images of J1302 in the B (F450W filter) and I (F814W filter) passbands from the HST archive. The HST observations (data set U39D0301M–U39D0304M) were made in 1997 July, with two 600 s exposures in the B band and two 400 s exposures in the I band. The images were processed using the standard HST pipeline routines in IRAF/STSDAS.² We then performed two-dimensional profile decompositions of this galaxy with the code GALFIT (version 3.0; Peng et al. 2010). Our model consists of an exponential disk component, a bulge component, and an unresolved central point source for the nuclear AGN emission. In our GALFIT modeling, the point-spread function was generated by the $\tt {Tiny\,Tim}$ software (Krist 1995). Note that the bulge for this galaxy displays a box/peanut shape, which is commonly seen in edge-on barred disk galaxies; consequently we added a boxiness parameter to the bulge profile in GALFIT. The model generally matches the data well ($\chi ^2_\nu \sim 1.02$). The inferred flux for the central point source in the B band is M_B = −15.8, corresponding to a nuclear luminosity of νL_νB = 8 × 10⁴¹ erg s⁻¹. For comparison, the B-band bulge luminosity of this galaxy derived from the GALFIT decomposition is L_{B, bulge} ∼ 6 × 10⁴² erg s⁻¹. If indeed the nuclear emission comes from an AGN,³ we estimate the bolometric luminosity to be 1 × 10⁴³ erg s⁻¹ using the B-band luminosity for the central point source by adopting a bolometric correction of 13 (Marconi et al. 2004). For a BH mass of 8 × 10⁵ M_☉, the accretion rate in Eddington units is L/L_EDD ∼ 0.1, which suggests J1302 is accreting at high Eddington ratio.

One of the remarkable features of J1302 is the extreme softness of the X-ray spectra. The best-fitted power-law index for the spectrum in the XMM-Newton quiescent state (Γ ∼ 7) is one of the steepest soft X-ray photon indices among AGNs (e.g., Grupe et al. 1995; Boller et al. 2011). Understanding the origin of the ultrasoft X-ray emission will help to pin down the nature of the source. Some AGNs such as NLS1s can be very soft (e.g., Boller et al. 1996; Middleton et al. 2007), showing a strong soft X-ray excess over an underlying power-law component. The strength of the soft excess can be quantitatively described as the ratio of flux at 0.5 keV to the power-law extrapolation of the fitting to the spectrum above 2 keV (Middleton et al. 2007). Since no significant hard X-ray emission above ∼2 keV was detected in the XMM observation of J1302, we estimated a 90% confidence upper limit on the count rates in 2–7 keV, ∼6.7 × 10⁻⁴ counts s⁻¹, and converted it to an upper limit on the extrapolated flux at 0.5 keV. XSPEC simulations of models with power-law Γ = 2, 2.5, and 3 show that the lower limits on the ratios defined above are 37, 15, and 6.1, respectively. Note that the ratios for a sample of NLS1s (Middleton et al. 2007) are usually found to be less than 10. Thus, the soft emission relative to that at the hard X-rays in J1302 is extremely strong compared to other AGNs.

The origin of the soft excess is still unclear. The spectrum at the XMM quiescent state can be fitted well by a Comptonized model and a blackbody plus power-law emission equally. Interestingly, the fitted blackbody temperature ($kT_{\rm BB}=43^{+6}_{-3}$ eV) is much lower than the canonical values of ∼0.1–0.2 keV found for AGNs (Crummy et al. 2006). Standard accretion disk models (Shakura & Sunyaev 1973) give a maximum effective temperature of the accreting material $kT_{\rm max}\sim 11.5(\dot{m}/M_8)^{1/4}$ eV, where $\dot{m}$ is the mass accretion rate in Eddington units and M₈ = M_BH/10⁸ M_☉. Using M_BH = 8 × 10⁵ M_☉ and $\dot{m}\sim 0.1$ for J1302, we obtained kT_max ∼ 22 eV, which is comparable to the fitted blackbody temperature. Note that, although with larger uncertainties, the lower fitted temperature for the Chandra quiescent state data ($kT_{\rm BB}=29^{+19}_{-16}$ eV) is more compatible with the predicted maximum disk temperature. Therefore, the ultrasoft X-ray emission in J1302 may be connected with the direct thermal emission from the accretion disk. Another constraint on the X-ray spectra due to thermal disk emission can come from the optical/UV data. The optical B-band flux of the nuclear point source from the HST observation is, however, much higher than the extrapolated MCD flux in the optical (M_B ∼ −12.7), about one order of magnitude difference. This difference cannot be explained by the contamination from nuclear star clusters, as they have typical absolute I-band magnitudes between −14 and −10 (Böker et al. 2002), much lower than the observed I-band flux of J1302 nucleus (M_I ∼ −16.8, obtained with the same GALFIT decomposition of the HST/WFPC2 image as detailed in Section 3.1). To further investigate this difference we need a more realistic disk spectral modeling, and fit to broader band data, which is beyond the scope of this paper.

Another unusual feature of J1302 is that it clearly shows two distinct states in the X-ray: a flare (or eruptive) state and a quiescent state. The amplitudes of the flare in the Chandra and XMM-Newton light curves look very similar, with the count rates increased by a factor of 5–7 within ∼1–2 ks. Both the XMM-Newton observation in 2000 and the Chandra observation in 2009 detected rapid flares, suggesting that the flare itself appears repetitive and occurs very frequently in the object. In fact, ROSAT PSPC observations also found that the object is highly variable and demonstrates a rapid flare event in the light curve lasting ∼1.3 ks (Dewangan et al. 2000). The rapid energetic flare in J1302 is fairly rare among Seyfert galaxies and quasars, although some extreme variations have also been found in objects such as IRAS 13224-3809 (Boller et al. 1997), PKS 0558-504 (Wang et al. 2001), and I Zw 1 (Gallo et al. 2004). As discussed by Wang et al. (2001), a magnetic coronal model, in which electrons in the corona are continuously heated by magnetic reconnection, can produce a rapid energetic X-ray flare. A realistic physical model to explain a rapid energetic flare in AGNs and the comparison with Galactic BHBs are beyond the scope of this work and will be presented elsewhere.

The spectral variability is clearly present between the two flux states with the source being much harder in the flare state. This behavior is in marked contrast to what is commonly seen in AGNs and BHBs (e.g., Markowitz & Edelson 2004; Done et al. 2007). Clearly, the spectral variability cannot be explained by any simple spectral model. One possibility is that the spectrum in the quiescent state is represented by a relatively stable thermal emission from the accretion disk, and the spectral variability is caused by the flux variations of a second additional component such as strongly Comptonized power-law emission. This perhaps explains why the photon index for the power-law emission in the XMM-Newton flare state (which dominates the spectrum) is consistent with the additional power-law component in the quiescent state, whose strength is relatively weak. In this case, the underlying power-law emission is still steep (Γ ∼ 4) compared to other AGNs, the nature of which remains understood. Note that a similar spectral change has also been seen in the Chandra observation of J1302, though the spectra statistics for the data is low. Longer X-ray observations are required to examine the presence of persistent flares and to investigate the origin of spectral variability during flare periods. This can in turn shed new light on the characteristics of corona and accretion flows around BHs.