We present the first result of a programme to search for large flux variations in the X-ray sources of the XMM Serendipitous Survey compared to previous ROSAT observations. An increase in X-ray flux by a factor >10 was discovered from the nucleus of the galaxy NGC 7589 on a time-scale of less than 5 yr. The 0.4–10 keV XMM spectrum can be approximated by a power law with photon index of 1.7–1.8, though it seems to flatten above 5 keV, suggesting a possible complex model, such as partial covering or disc reflection. A classification based on an analysis of its optical spectrum places NGC 7589 in the Seyfert region, but close to the Seyfert–LINER (low-ionization nuclear emission-line region) borderline on the active galactic nucleus (AGN) diagnostic diagrams. We classify NGC 7589 as either Seyfert 1.9 or LINER I, in the light of the detection of a broad Hα line, which makes NGC 7589 an AGN in the low-luminosity regime. We interpret the observed variability in terms of either changes in covering factor of absorbing gas in the AGN, or variability in the intrinsic X-ray luminosity. Should the latter be the case, the inferred Eddington accretion rate increased from the radiatively inefficient accretion-dominated regime to a value close to the putative critical value, at which a transition of the accretion mode is supposed to take place. This possibility presents a new prospect of studying accretion physics in the central black holes of external galaxies by direct observing changes of ‘spectral state’, as is common in stellar black hole binary systems.
In contrast to typical active galactic nuclei (AGN), low-luminosity AGN (LLAGN) as a class radiate at a power much lower than the Eddington luminosity, i.e. in terms of the bolometric/Eddington luminosity ratio (Eddington ratio), Lbol/LEdd < 0.01 (Ho 2004). As such, black holes in LLAGN are hypothesized to accrete via radiatively inefficient accretion flows (RIAF; see Quataert 2001 for a review), such as an advection-dominated accretion flow (ADAF; see Narayan, Mahadevan & Quartaert 1998 for a review), rather than an optically thick, geometrically thin standard accretion disc (thin disc hereafter; Shakura & Sunyaev 1973). In fact, the high end of their Lbol/LEdd distribution encompasses the critical value above which a RIAF is to be replaced with a standard thin disc. As a consequence, in LLAGN exhibiting violent variability of luminosity, a transition of the accretion mode is expected to take place once Lbol/LEdd crosses the critical value. This is where observations can be used to test black hole accretion theories. While such a transition of the accretion mode can, indeed, explain the observed changes of the ‘spectral states’ in X-ray binaries (e.g. Esin et al. 1997; Meyer, Liu & Meyer-Hofmeister 2000), the situation is not clear in massive extragalactic black hole systems. On galactic scales, the major difficulty in observations is the long time-scales of the X-ray variability, which is roughly scaled with the black hole mass. Timely detection of violent variability in AGN, especially in LLAGN, is of particular interest in this regard.
About 40 per cent of nearby galaxies exhibit LLAGN activity (Ho, Filippenko & Sargent 1997), the majority of which belong to a class known as low-ionization nuclear emission-line regions (LINER, Heckman 1980). Large-amplitude X-ray variability is sometimes seen in low-luminosity Seyferts (e.g. Nandra et al. 1997), whereas no similar behaviour is found in LINERs (e.g. Ptak et al. 1998; Komossa, Böhringer & Huchra 1999; Terashima et al. 2002).
In this Letter, we report the discovery of large-amplitude X-ray variability in the galaxy NGC 7589 at a redshift of 0.0298. This is one of the first results of a search of long-term highly variable X-ray sources (Yuan et al. in preparation; see also Yuan et al. 2002) of the XMM Serendipitous Survey (Watson et al. 2001). We selected candidates which varied by a factor of at least 10 for further follow-up studies in the X-ray and other wavebands. We used a luminosity distance of 123.3 Mpc calculated from the radial velocity of 8562 km s−1 (LEDA) relative to the cosmic microwave background by assuming H0= 71 km s−1 Mpc−1, ΩΛ= 0.73 and Ωm= 0.27.
2 THE XMM X-RAY DATA OF NGC 7589
2.1 XMM observations
The XMM observations, data screening and source extraction are summarized in Table 1. It should be noted that the PN camera was not active during the first observation in the XMM orbit 272. The XMMscience analysis system (sas, v. 5.4) was used for data reduction. Source X-ray events were extracted from a circle of 32-arcsec radius. Background events were extracted from source-free regions using a concentric annulus of 40-/120-arcsec radii for the MOS detectors, and circles of 32-arcsec radius at the same CCD read-out column as the source position for the PN detector. In both observations, an X-ray source was detected at the position , coincident with the nucleus of NGC 7589 (Fig. 1). The source spatial profile in the 0.3–2keV band was found to be point-like. No time variability was found on time-scales shorter than the XMM observational intervals of about 12 ks. The galaxy was also observed with the XMM Optical Monitor (OM) with the UVW1 filter (2000–4000 Å) and was detected as an extended source.
2.2 The XMM X-ray spectra and fluxes
The MOS1 and MOS2 spectra were co-added to produce a single, combined spectrum. We used xspec (v.11.3) for spectral modelling. Spectral bins below 0.4 keV were excluded because of uncertain calibration below 0.35 keV (XMM-SOC-CAL-TN-0018). From orbit 361, we have both PN and MOS spectra; they were jointly fitted with independent normalizations. In comparisons of the two observations, the fitted spectral parameters agree with each other for most or all the parameters except the normalizations, which differ by a factor of ≃2.5. To improve photon statistics, we also performed fits to the two spectra from the different epochs with most of the parameter values tied together. The results are summarized in Table 2 and explained below.
Power-law models. A simple absorbed power law gives statistically acceptable fits to all the spectra; however, the fitted NH is lower than the H i column density along the line of sight NGalH= 4.06 × 1020 cm−2 (at the 92 per cent confidence level). The fits over the restricted 0.4–2 keV and 2–10 keV ranges yielded a photon index in the low-energy band Γlow-E= 1.72 ± 0.12 and one in the high-energy band Γhigh-E= 1.42+0.20−0.26, respectively, indicating a spectral flattening towards high energies. This can be seen from the residuals in Fig. 2, which compares the data to a fitted model using only the 0.4–2 keV spectra. In fact, a broken power-law model of Γlow-E= 1.75+0.10−0.16 and Γhigh-E= 1.03+0.40−0.47 improved the fit significantly, with Δχ2=−10 for two additional free parameters; the F-test gave a probability level for no improvement over the single power law fit as being <1 per cent. The fitted NH was also recovered to value consistent with NGalH.
Complex spectral models. Spectral flattening is often seen in AGN as a result of scattered X-rays, likely by an accretion disc, or by X-rays transmitted through optically thick gas. Accordingly, we fit disc-reflection and partial-covering models to the spectra. Both types of models, either cold or ionized, were found to provide excellent and statistically indistinguishable fits to all the data. The overall absorption NH values were in general consistent with NGalH. The disc inclination was not constrained and so was fixed at 30°. The disc-reflection models require a larger reflection/direct ratio (i.e. >2) than that from a medium subtending a 4π solid angle, implying that the primary continuum is partially obscured in this case. No significant iron Kα line was seen at around 6.4 keV; 90 per cent upper limits for the line equivalent width were found to be 550 and 110 eV for the first and second observations, respectively. Modelling the high-energy excess simply by an extremely broadened iron line (the discline model) gave no satisfactory fit.
X-ray fluxes and luminosities. The X-ray fluxes were calculated using the best-fitting power-law spectral model with NH=NGalH, i.e. Γ= 1.68 (1.73) for 0.5–10 (0.5–3) keV. In the 0.5–10 keV band, the Galactic absorption corrected fluxes are 9.1 × 10−13 and 3.7 × 10−13 erg cm−2 s−1 for orbits 277 and 361, respectively, which correspond to luminosities of 1.7 × 1042 and 0.7 × 1042 erg s−1, respectively. In consideration of the likely presence of the partial-covering or disc-reflection (assuming a flat disc subtending a 2π solid angle) component, the intrinsic X-ray luminosities are likely to be a factor of ∼2 and 3–5 higher, respectively.
3 The Long-Term X-Ray Variability
3.1 Fluxes from the ROSAT and Einstein data
NGC 7589 was within the field of view (10 arcmin off-axis) of a 5.4 ks pointing observation made by ROSAT with its Position Sensitive Proportional Counter (PSPC) detector (0.1–2.4 keV) on 1995 June 16–17. No source was detected at the position of NGC 7589. We derived a limit on the source flux in two ways. The first was a simple estimate using the count rate of the weakest detected source neighbouring NGC 7589. This led to a conservative upper limit on the source count rate of 2.7 × 10−3 count s−1. A more rigorous estimation was made using the PSPC X-ray image. Photon counts from the background and a possible source were extracted from a circle of 45-arcsec radius (enclosing 95 per cent of the power of the point spread function, hereafter PSF) at the position of NGC 7589, yielding a measured Cs+b= 17 count. Following Poisson statistics, the expected counts is distributed as (27) at the (one-tail) probability of 95 (99) per cent. The contribution of the background was estimated as Cb= 9.5 count, using the averaged local background estimated from an annulus of 90-/225-arcsec radii. The expected net number of source counts was estimated to be (17.5) count. After correction for the vignetting and extraction aperture, the source count rate was estimated to be ≤2.9(3.4) × 10−3 count s−1. Assuming the spectrum to be the same as that measured with XMM, we derived the 0.1–2.4 keV flux limit to be 5.9 (7.0) × 10−14 erg cm−2 s−1 at the 95 (99) per cent level, after correction for the Galactic absorption.
A less stringent flux limit of 5.6 × 10−13 erg cm−2 s−1 was set by the data extracted from the ROSAT All-Sky Survey performed in 1992. By searching the data archives of various X-ray satellites, we found that, in addition to ROSAT, NGC 7589 was also serendipitously observed by the Einstein Observatory with its Imaging Proportional Counter (IPC) detector at an off-axis angle of 10.8 arcmin for 2540 s in 1980 May. Similar estimation using a source extraction radius of 3 arcmin yielded a limit on the unabsorbed flux of 3.5 × 10−13 erg cm−2 s−1. Both limits are quoted at the 95 per cent level.
3.2 Variability in the X-rays and UV
For comparison of the fluxes among the different missions, we used the overlapping 0.5–2.4 keV energy band. The light curve in Fig. 3 reveals a dramatic increase in luminosity, by a factor of >10, occurring sometime between the ROSAT pointing and the XMM observation in orbit 272, on a time-scale ≲5 yr. Given the poor sampling, the flux seen by XMM may not represent the flux at the ‘highest state’, which means that the peak luminosity could be even higher. Therefore the actual variability amplitude must be >10. The flux decreased by a factor of 2.5 between the two XMM observations over a period of 5 months, whereas the overall 0.4–10 keV spectral shape appeared to remain unchanged.
The OM UV magnitude within a central 6-arcsec-radius aperture was measured to be 16.99 ± 0.06 and 17.03 ± 0.05 for orbits 272 and 361, respectively. However, the emission is dominated by the extended bulge and no significant point-like source is present; thus no meaningful constraint on variability could be obtained.
4 Optical Spectroscopic Classification
NGC 7589 was observed spectroscopically in the Sloan Digital Sky Survey (SDSS; York et al. 2000) on 2000 September 29, about 8 months prior to the high state caught by XMM. The spectrum was taken from the SDSS Data Release 2. The contribution of stellar light from the host galaxy was subtracted from the observed spectrum by modelling stellar absorption lines employing a range of galaxy spectral templates, using the method developed by Li et al. (2004) (A full account of the spectral analysis and estimation of the black hole mass for NGC 7589 is to be presented elsewhere.) Single-component Gaussians were fit to the emission lines. This provided satisfactory fits, except for Hα which required the presence of a second, broad component. The broad Hα line, which was blueshifted, had a width of FWHM ≈ 3428 km s−1, compared with the narrow lines of FWHM ≈ 228 km s−1.
The emission-line ratios place NGC 7589 in the Seyfert region of the diagnostic diagrams of Veilleux & Osterbrock (1987), close to the borderline with the LINERs (see Fig. 4). The same conclusion also holds if we use the classical definition of LINERs (Heckman 1980) which only makes use of oxygen emission-line ratios. For LINERs, [O ii]λ3727/[O iii]λ5007 > 1 and [O i]λ6300/[O iii]λ5007 > 1/3 (and lower for ‘[O i]-weak’ LINERS), whereas NGC 7589 shows [O ii]λ3727/[O iii]5007λ≈ 0.5 and [O i]λ6300/[O iii]λ5007 = 0.1. This places NGC 7589 in a region populated by Seyferts (see fig. 1b of Filippenko & Terlevich 1992) but, again, close to the borderline with LINERs. We note that the measured line ratios are insensitive to the galaxy spectrum subtraction – the same result holds even when the unsubtracted spectrum is used. The detection of a broad component in Hα reveals a type I AGN and leads to the classification of NGC 7589 as either a Seyfert 1.9 or possibly a LINER I.
5.1 Classification of NGC 7589
On the classification diagrams based on emission-line ratios, the distribution of the LLAGN in the Ho (2004) sample has a continuous and smooth transition between Seyferts and LINERs. This makes the conventional, ‘operational’ dividing lines somewhat arbitrary and physically insignificant. NGC 7589 falls within the area in which galaxies are classified as Seyferts; however, it is too close to the border region to allow an ultimate classification as either a Seyfert or LINER. A similar conclusion holds when the emission-line width is considered (e.g. Ho, Filippenko & Sargent 2003). We therefore treat NGC 7589 as a member of the important ‘borderline objects’, which probe the physics in transition from Seyfert to LINER. Optically, NGC 7589 is a giant, bulge-dominated low-surface-brightness galaxy (Pickering et al. 1997).
The high-amplitude X-ray variability detected is exceptional, given the transition nature of NGC 7589. While there is a trend in Seyferts showing enhanced X-ray variability with decreasing luminosity (e.g. Nandra et al. 1997), the trend breaks at low luminosities; in particular, LINERs do not vary much in X-rays (e.g. Ptak et al. 1998; Komossa et al. 1999; Terashima et al. 2002). Those that do rarely exceed a variability of a factor of two. The high-amplitude variation in the X-ray luminosity of the order of magnitude of 1042 erg s−1 on time-scales of a few years suggests black hole accretion as the ultimate source of the power in NGC 7589. This is in line with the AGN nature of NGC 7589 indicated by the broad Hα line.
We may ask whether the XMM‘high state’ or the ROSAT‘low state’ is more representative of the average X-ray luminosity. We assume that the narrow-line luminosities in the SDSS spectrum represents the average state of NGC 7589. By making use of the LHα∼LX correlation for LLAGN (Halderson et al. 2001), a luminosity of LX∼ 1 × 1041 erg s−1 (∼1.5 × 1041) is predicted in the 0.1–2.4 keV (0.5–10 keV) band, assuming Γ= 1.7. This value is significantly below the high-state luminosity, and is comparable to the low-state upper limit. This suggests that probably NGC 7589 normally behaves like a LLAGN, and underwent an outburst that was caught by XMM. At the high state, LX reached the lower end of the typical Seyfert luminosity range (1042−44 erg s−1), and the very upper end of the LINER range. The power-law model is again consistent with both Seyfert and LINER types, whereas the partial covering/reflection model would favour a Seyfert.
5.2 The X-ray variability
In general, X-ray variability in AGN may have two distinct origins: variations in intrinsic radiation and variations in absorption/obscuration. Below we discuss briefly our results in this context. For NGC 7589, the mass of the central black hole is of the order of 107 M⊙, as found in our ongoing follow-up work (Yuan et al. in preparation).
Variable absorption. Evidence for variable absorption has been found in some Seyfert galaxies, especially in Type 2 and intermediate-type, on time-scales of months to years (e.g. Risaliti, Elvis & Nicastro 2002) through variation of NH and/or the covering factor (e.g. NGC 3227: Komossa et al. 2001, NGC 3516: Guainazzi, Marshall & Parmar 2001). The absorbers are postulated to be gas clouds in the ‘broad-line region’ (BLR) producing relatively fast variations, and a parsec-scale ‘torus’ producing relatively long-term variations. In particular, NGC 7589 might be an analogue to the LLAGN in M 51, in which a suspected large-amplitude X-ray variation could be well explained by placing a thick (NH > 1024) absorber on and off the line of sight (Fukazawa et al. 2001).
For the ROSAT low state, taking the best-fitting cold partial covering model for the XMM spectrum at the highest state (Table 2) and increasing artificially the covering factor up to ≳0.95, the soft X-ray flux was then suppressed below the ROSAT detection limit. For the flux decrease between the two XMM observations, we considered the joint fit with the partial covering model (Table 2) by assuming the continuum levels to be the same and tying them together, and we still obtained an acceptable fit, though slightly worse than that given in Table 2; this yielded a much larger value of NH and the covering factor (8.0 × 1023 cm−2 and 0.72) at the epoch of orbit 361 than those of orbit 272 (1.9 × 1023 cm−2 and 0.29). Both the time-scales of <5 yr and of 5 months are consistent with the motion of gas clouds (≈3400 km s−1) in the BLR for a ∼107 M⊙ black hole.
Variable intrinsic emission: For the ROSAT low state, the Eddington ratio was estimated conservatively to be Lbol/LEdd < several times 10−4, much lower than the critical accretion rate (see Section 1). This value indicates that the accretion proceeded via a RIAF. A RIAF has been postulated to be present in some LLAGN (e.g. Fabian & Rees 1995; Lasota et al. 1996; Quataert et al. 1999; Ptak et al. 2004). In particular, the prototype of such a class, the LINER NGC 4258 (Lasota et al. 1996), has values of Mbh and LX similar to those we found for NGC 7589. A thin disc would be truncated and only exist at large radii (e.g. Quataert et al. 1999; Lu & Wang 2000; Meyer & Meyer-Hofmeister 2002). The broad iron line and reflection hump, which are characteristic of typical Seyferts and believed to be evidence for a thin disc, are not expected to be seen in a RIAF. These different accretion modes also predict distinct spectral shapes in the broad-band spectral energy distribution (SED) and possibly in the X-ray spectra. Therefore a good X-ray spectrum and/or SED in a low state is essential to test the RIAF hypothesis.
In the XMM high state, the identification of the accretion mode is not clear. The value of Lbol/LEdd was estimated to be a few per cent, which is immediately close to the critical value. Thus the accretion flow was possibly in a transition state between a RIAF and a thin disc. Had the flux ever reached a peak much higher than the XMM high state sometime between the observations, a transition of the accretion mode might have taken place, according to the current accretion theories. The observed time-scale of ≲5 yr is consistent with the mass diffusion time-scale on which a thin disc drifted inwards down to several Schwarzschild radii RS from a transition radius of up to ∼ hundreds RS. However, the quality of the obtained X-ray spectra does not allow us to distinguish between a dominating thin disc (extending down to a few RS) and a disc truncated at ∼ hundreds RS, though the probable spectral flattening – if this is indeed due to reflection of the X-rays – would favour the former. Deep X-ray observations for NGC 7589 and for more objects of this kind are essential for a serious confrontation of current accretion theories.
Large X-ray variability events appear to be rare. The first results from our programme, which searched over 386 XMM fields previously covered with ROSAT pointed observations, yielded only one, or possibily two, detections. The total sky area covered (66 deg2) contains, statistically, ∼26 galaxies brighter than 15 magB and ∼180 galaxies brighter than 16.4 magB (cf. 15.2 magB for NGC 7589). A further extensive search, as well as the measurement of the variability time-scale, is needed in order to be able to estimate the event rate.
We thank Cheng Li and Tinggui Wang for helping with the subtraction of the stellar spectrum, and Christopher Brindle for help in verifying the XMM OM results. WY thanks Matteo Guainazzi for useful discussion. Funding for the creation and distribution of the SDSS Archive has been provided by the Alfred P. Sloan Foundation, the Participating Institutions, the National Aeronautics and Space Administration, the National Science Foundation, the US Department of Energy, the Japanese Monbukagakusho, and the Max Planck Society. This research has made use of the NASA/IPAC Extragalactic Database (NED).