4.1. The hard X-ray properties of IRAS12393+3520
X-rays provide a strong evidence in favor of the existence of an AGN in IRAS12393+3520. We have detected a significant variability of the X-ray light curve. The minimum observed associated doubling/halving time is comprised in the range 30-75 ks. If the variability is intrinsic to the radiation emitted in the nuclear region of an AGN, usual light crossing arguments constrain the mass of the nuclear black hole to be , where is the typical size of the region emitting the bulk of the primary non-thermal continuum in unit of five Schwarzschild radii. The joint ROSAT/ASCA 0.1-10 keV spectrum has also a shape which closely resembles the one typically observed in radio-quiet AGN. Our data suggest that the solid angle subtended by the accretion disk is higher than due to a plane-parallel infinite slab. The upper limit on the EW of a fluorescent neutral iron line (600 eV) is not inconsistent with this scenario if solar abundances are assumed. However, much better quality of the data is required to confirm this hint.
The data require at 99% level of confidence the presence of an absorption edge, whose threshold energy is consistent with the K-shell photoionization energy of OVII . In at least 50% of Seyfert 1s the nuclear radiation is absorbed by substantially ionized matter (Reynolds 1997; George et al. 1998). In IRAS12393+3520 the properties of the absorbing matter are statistically poorly constrained and are somewhat dependent on the assumed continuum model. We conservatively derive a column density in the range 2.5- cm-2 [or -1.5 under standard assumptions], if the physical conditions required to sustain an oxygen state dominated by He-like are assumed. If the warm absorber contains dust in the Galactic dust-to-gas ratio, such a medium may account for the observed optical reddening and explain the missing H broad line, even in the presence of a strong broad H (which is actually detected only in some of the IRAS12393+3520 optical spectra, cf. Sect. 3). "Dusty" warm absorbers have already been invoked to explain the discrepancy between the amount of X-ray cold absorption and the optical reddening in a few Seyfert galaxies (Reynolds et al. 1997; Komossa & Fink 1998 and references therein), although Siebert et al. (1999) have recently shown that such a "simple" representation was not able to reproduce all the observed optical and X-rays data in the case of a well studied example, IRAS13349+2438.
The observed spectra can be formally accounted by a two-temperature optically thin plasma, in analogy with recent evidence on the X-ray emission of starburst galaxies (Della Ceca et al. 1998; Cappi et al. 1999). In this scenario, however, the observed short-term variability cannot be explained, unless it represents the diluted appearance of a much more variable underlying unresolved source. The lack of spectral variability during the ASCA observation points against the emerging of a further source in the spectrum in correspondence to the flux variation. The contribution of a power-law source in the 0.5-4 keV band is constrained to be lower than erg s-1. To reproduce the observed flux change, it should vary in seconds by an amount , where is the X-ray luminosity in units of erg s-1. This rules out the possible serendipitous detection of local X-ray binaries or superluminal sources; the latter ones exhibit indeed variability in flux by an amount up to two orders of magnitude, but have normally luminosities well within erg s-1. On the other hand, X-ray variability on such timescales is rather common in Seyfert galaxies (Nandra et al. 1997a and references therein). The timescales are too fast to be explained by mechanisms which do not invoke accretion on a supermassive black hole (e.g. the starburst model of Terlevich et al. 1992).
4.2. Comparison between X-ray and other wavelengths
An energetic argument points against the idea that hard X-rays are dominated by a nuclear starburst. No known starburst galaxy is so X-ray luminous (Ptak et al. 1999). The soft-X vs. 60µm and soft-X vs. UV luminosity ratios are and , more than one order of magnitude higher than typical values observed in star forming galaxies (SFG, Mas-Hesse et al. 1995). The observed X-ray luminosity erg s-1 is, on the other hand, not uncommon among Seyfert galaxies. In Fig. 7 the IRAS12393+3520 SED is shown, superimposed to the average "tracks" of the Mas-Hesse et al. (1995) sample for Seyfert 1s, Seyfert 2s, QSO and SFG. Caution must be employed when evaluating the SED in IRAS12393+3520, because it refers to non-simultaneous measurements. Optical spectroscopy suggests that the nuclear activity (as seen by the BLR), and therefore also the energy budget, might be variable with time. Given these caveats, one notes that IRAS12393+3520 follows apparently better a Seyfert 2 SED (or eventually a SFG one), being strongly under-luminous in UV and X-rays in comparison to the IR, if the normalization is done in the far-IR. These pieces of information are summarized in the IR, UV and X-ray color-color plot of Fig. 8, where IRAS12393+3520 is seen to lie in the region of Seyfert galaxies, and apparently type 2 rather than 1.
Several pieces of evidence point, however, against an identification of IRAS12393+3520 as a Seyfert 2. First, there is no evidence of the high excitation typical of Sey2's: the [OIII] over line ratio is difficult to measure accurately, as [OIII] is faint and is hidden in the corresponding stellar absorption, but is not larger than three. In addition, the [OII]/[OIII] ratio is much larger than one, and it is the combination of these line ratios, with the [NII]/ one, which lead to the LINER claim for this object. This is not compatible with a Seyfert 2 identification. Moreover, a high excitation is not seen either when the broad component is absent, so that the variations cannot be described as a transition between a Seyfert 1 and a Seyfert 2 phase, as seen in some other cases (e.g. Mkn 993, Tran et al. 1992; Mkn 6, Khachikian & Weedman 1971; Mkn 1018, Cohen et al. 1986). Second, the X-ray intrinsic neutral absorption column density is negligible. No model where the nuclear power-law is strongly absorbed yields comparably good fits to the ROSAT/ASCA spectrum as the simple models of Table 2, even if one allows the covering fraction of the absorbing matter to be lower than 1. One might suppose that the X-ray spectrum is totally dominated by scattered radiation from an otherwise invisible nucleus, as observed in several "reflection-dominated" Seyfert 2s (Turner et al. 1997; Matt et al. 1997; Guainazzi et al. 1999). The observed variability, however, introduces severe constraints on this possibility. Assuming a typical variability timescale of s, light crossing arguments imply a lower limit on the electron numerical density cm-3. If the absorption occurs in the same medium, an optical depth to scattering of the order of one implies an optical path cm. Such a low value would imply that the scattering/absorption occurs in the proximity of the galactic nucleus, hardly compatible with the idea that the matter hiding the nuclear region is located at distances 1-100 pc, in the shape of a, more or less homogeneous, azimuthally symmetric structure (Antonucci & Miller 1985; Maiolino & Rielke 1995; Greenhill et al. 1996). Finally, the appearance sometimes of a broad component is a definite sign in favor of a Seyfert 1 classification.
One clue to explain the observed SED is to assume that the AGN is a weak Seyfert 1, which is over-luminous in IR in comparison to the objects of the same class of comparable X-ray luminosity. This interpretation can be naturally linked to the puzzling absorption pattern emerging from the optical/UV spectroscopy. Very broad Ly and H lines have been observed in (not simultaneous) observations of the core of IRAS12393+3520 (M96). All these evidences suggest that the geometrical distribution of the circumnuclear neutral absorbing matter is far more complex than the simple equatorially symmetric pattern, which is assumed in the 0-th order Seyfert unification theories (Antonucci & Miller 1985; Antonucci 1993). The Narrow Line Region (NLR) could be absorbed by dusty, optically thick matter distributed on spatial scale of a few hundreds of parsec, which does not (not always?) intercept the line of sight towards the nuclear environment. This dust might simultaneously be responsible for the IR emission through reprocessing of the incoming nuclear radiation during particularly active phases (or when the line of sight between the nucleus and the dust is unobscured). In the same framework, the under-luminosity in and the relative over-luminosity in IR in comparison to the directly observed X-ray of nuclear origin can be simultaneously explained. A similar scenario had been originally suggested by Maiolino & Rieke (1995), who proposed that "intermediate" Seyfert galaxies are seen through a 100 pc-scale torus coplanar with the plane of the galaxies. This idea has received a further support after the results of a recent HST slew survey, which lead to the discovery that the distribution of matter in the nuclear environment of radio-quiet, nearby AGN is indeed highly patchy, with dusty lanes protruding from several kilo-parsecs to a few hundreds of parsecs towards the center (Malkan et al. 1998). Recently, Maiolino et al. (1999) pointed out that barred galaxies (like IRAS12393+3520) tend to exhibit the highest values of IR to X-ray luminosity ratio (see, however, a different point of view in Regan & Mulchaey 1999). This might suggest that stellar bars are very efficient in driving gas to the circumnuclear region and therefore to provide high amount of warm (AGN-heated) dust in the nuclear environment, which would reprocess the nuclear high-energy continuum.
The bulk of the IR radiation could alternatively be produced by an intense star formation episode, occurring on spatial scales of the order or larger than the NLR. The SED in the IR alone is typical of starburst galaxies and does not satisfy the various criteria defined to select AGN in IRAS data (de Grijp et al. 1985; Désert & Dennefeld 1988). Also, the standard IR/radio parameter q, discussed by Condon at al. (1995), has a value of 2.85 for IRAS12393+3520, showing that the AGN, if present, is not dominating the IR emission and/or that the object is radio-weak. Finally, the observed infrared luminosity ( erg s-1; M96) is almost two orders of magnitude higher than expected from the AGN 1-10 keV luminosity ( erg s-1), if the IRAS12393+3520 AGN SED has a ratio typical for quasars with erg s-1 (4.5; Elvis et al. 1994).
While both explanations for the IR emission are plausible, the energetic argument just discussed favors the hypothesis that the IR emission is dominated by star-formation. However, even in this case, an additional (nuclear) source of high-energy radiation is required to explain the hard X-rays. As shown in Fig. 8, IRAS12393+3520 exhibits indeed a much higher X-ray to IR or UV luminosity flux than typically observed in SFG galaxies. It is likely to be provided by a separate component, which is straightforward to identify, in the light of the ASCA/ROSAT results, with an active nucleus. On the other hand, a spatially and physically structured and time varying absorber is also required to yield the observed different absorptions towards the nucleus, of the BLR and of the NLR lines. The dusty environment of a star forming region could provide only the last. IRAS12393+3520 is therefore best described by a central, weak, AGN, with a BLR partly obscured by a structured absorber, and a well absorbed NLR, mixed with a region of intense star formation of perhaps larger extension.
© European Southern Observatory (ESO) 2000
Online publication: March 17, 2000