4.1. Flux transient behaviour in ROSAT observations
RX J0134.2-4258 was one of the softest AGN at the time of the RASS, but turned into an object with a rather flat X-ray spectrum by the time of its pointed observation. Its behaviour is different from that observed from other `transient' X-ray AGN. WPVS007 was practically `off ' in all follow-up ROSAT observations after the RASS. We explained this behaviour as a dramatic change in the comptonization parameters that shifted the Big Blue Bump emission out of the ROSAT PSPC range (Grupe et al. 1995b). In IC 3599 (and also NGC 5905) we observed a fading of the X-ray flux over several years. We (Grupe et al. 1995a, but see also Brandt at al. 1995; Bade et al 1996; Komossa & Bade 1999) came to the conclusion that these were X-ray outbursts that could have been caused by tidal disruption of a star by a central black hole.
RXJ0134.2-4258 can be considered to be a `transient' X-ray source as well. During the RASS, it was practically off at energies above 0.5 keV. While we usually find transient sources that were bright during the RASS, such as WPVS007 or IC 3599, RX J0134-4258 is an exception in that it became bright in its hard X-ray band after the RASS. What makes RX J0134.2-4258 so unusual is its dramatic spectral variability, both on long time scales (between the RASS and pointed observation) and on short time scales (during the pointed observation).
4.2. Precedents for similar behaviour
The results from our study of RX J0134.2-4258 are sufficiently unusual that we have searched the literature for examples of similar behaviour from other objects.
The short term spectral variability observed in RX J0134.2-4258 has some precedent. A hardness ratio analysis of the ROSAT lightcurves from Mrk 766 revealed that the spectrum also hardens when the flux increases. Leighly et al. (1996) first demonstrated that this behaviour was consistent with a non-varying soft excess component plus a power law with varying normalization. Those data have been recently re-examined, and the same results were found (Page et al. 2000).
The long term spectral variability may also have some precedents. Bedford et al. (1988) report the discovery and followup EXOSAT observations of the Seyfert 1.5 galaxy EXO 1128+691. They found the object clearly detected in the EXOSAT LE detector, which is sensitive to soft photons in the 0.05-2.0 keV band, but not detected by the ME detector (0.7-10 keV). The LE detector has no intrinsic energy resolution, but information from exposures using several filters show that the spectrum was very soft, with effective photon index . In contrast, a recent ASCA observation of EXO 1128+691 showed a strong detection in 2-10 keV X-rays (Leighly et al., in preparation). The spectral parameters from the ASCA observation predict 0.4 counts/s in the EXOSAT ME. While this count rate would be somewhat low for detailed spectral analysis (e.g. Turner & Pounds 1989), it should at least have been detected. Therefore, EXO 1128+691 may have undergone a similar spectral transition as experienced by RX J0134.2-4258. A difference is that during the EXOSAT observation, the LE flux was strongly variable, by a factor of 3 on time scales as short as 15 minutes; in contrast, during the pointed ROSAT observation of RX J0134.2-4258 we find that the soft component is constant, and the normalization of the hard component varies.
Another example of a possibly similar spectral transition is the one recently seen in the Narrow-line Seyfert 1 galaxy NGC 4051 (Guainazzi et al. 1998). A BeppoSAX observation of this object found the 2-10 keV flux to be a factor of 20 lower than the historic average. The remaining hard X-ray spectrum was very flat with a strong iron line, suggesting a Compton-reflection dominated spectrum, as is found in Compton-thick Seyfert 2 galaxies. Guainazzi et al. (1998) report that the soft X-ray spectrum was steep, but defer the details to another paper. While this also appears somewhat similar to the case of RX J0134.2-4258, a difference is that while the soft X-ray component is reported to be relatively strong compared with the rest of the spectrum, the EUVE light curve shows that it is much weaker than has been seen in previous observations by EUVE (J. Halpern 1999, P. comm.).
4.3. Possible origins of the spectral variability
4.3.1. Spectral variability as a change in the warm absorber
Komossa & Fink (1997) and Komossa & Meerschweinchen (2000) investigate whether the spectral variability in RX J0134.2-4258 can be explained by a `warm' ionized absorber. They noted that there is a deviation in the residuals of a power law fit to the pointed observation near 0.6keV (see Fig. 3) that could be interpreted as an absorption edge from OVII in the rest frame. In contrast, we found that a single power law model does not give a good fit and the spectrum is much better described using two component models in which case the residual disappears. This underlines the fact that with the spectral signal to noise and the low resolution of the PSPC data, a two component model and a warm absorber model cannot be distinguished. The warm absorber requires a very large column density () to explain the ultrasoft spectrum, and we cannot accommodate such heavy absorption in the ASCA spectra. Furthermore, we do not see a signature of this material in other wavebands. For example, Reynolds (1997) found that warm absorbers are generally associated with optical reddening in a sample of ASCA observations of Seyfert 1 galaxies; in contrast, RX J0134.2-4258 has one of the bluest optical spectra observed in the soft X-ray selected AGN sample. Furthermore, we find that for reasonable parameters, the cloud scenario cannot explain the rather similar spectral variability observed during the pointed observation, simply because for plausible source sizes and cloud velocities, the time scale of variability required is rather longer than observed.
4.3.2. Spectral variability as absence and recovery of a corona
One possible scenario to explain the longterm spectral variability observed in RX J0134.2-4258 is the absence followed by the recovery of the accretion disk corona. A corona is necessary to produce the hard X-rays if the accretion disk is optically thick and geometrically thin. A popular version of the disk-corona model holds that the accretion power is dissipated primarily in the corona, and the disk merely reprocesses part of it (Haardt & Maraschi 1993; Svensson & Zdziarski 1994). However, it has recently been shown that in some Narrow-line Seyfert 1 galaxies, there is far too much power in the soft excess component, believed to be emitted from the disk, to be driven by reprocessing of the hard X-ray component (e.g. Pounds et al. 1995). Thus, the disk emission must be primary, the corona will not be necessary as the primary agent for dissipation of the accretion energy, and therefore it could conceivably be absent. It is possible that RX J0134.2-4258 had no corona during the RASS observation, and it developed one before the pointed observation. The corona regeneration time scale must be fairly short for this scenario to be viable. If the corona is produced by reconnection of buoyant magnetic flux tubes that have been built up to equipartition by convective dynamos in the disk, the relevant time scales are the convection time scales plus reconnection time scales (e.g. Haardt et al. 1994). These authors find a loop variability time scale could be as short as minutes for a object and therefore the corona could conceivably regenerate within the time between the RASS and pointed ROSAT observations. The question of what could cause an object to lose or regrow a corona is a very interesting one that we will not speculate on. On the other hand, it is not necessarily clear that this mechanism could explain the spectral variability observed on short time scales during the pointed observation.
4.3.3. Spectral variability as a strengthening of the radio component
Another possible explanation for the spectral variability observed in RX J0134.2-4258 is suggested by the fact that it is radio-loud, and possibly variable, and the fact that during the ASCA observation the photon index was flat. Radio-loud quasars are well known to have flatter X-ray spectral indices than radio-quiet ones, and they are also known to be brighter X-ray sources (e.g. Lawson & Turner 1997). A widely accepted explanation for this fact is that the X-ray emission includes a additional harder component associated with the radio jet. Therefore, it is possible that the increase in the normalization of the hard X-ray component is caused by a strengthening of the radio jet component. We note that compared with the GINGA sample of quasars, RX J0134.2-4258 is about a factor of 10 deficient in X-ray luminosity (see Lawson & Turner 1997, their Fig. 8). However, taking into account the scatter in the correlation and variability, it is not completely implausible that the hard X-rays are being dominated by a component associated with a radio jet.
This scenario may have a precedent in the case of 3C 273. A set of ASCA observations of this object reveal that a soft excess and iron line appears, and the spectrum becomes steeper when it is fainter (Cappi et al. 1998). It was proposed by Cappi et al. (1998) and also by Haardt et al. (1998) that 3C 273 in its bright X-ray state is dominated by the jet component, and in its faint state, the Seyfert nucleus is revealed. On the other hand, it must be noted that ASCA observations of the other two radio-loud NLS1s revealed canonical NLS1s spectra, i.e. a steep hard X-ray spectrum and a weak soft excess (Siebert et al. 1999; Leighly 1999a).
The existence of radio-loud NLS1s may be somewhat surprising, given the results of Ulvestad et al. (1995) on their small sample of Seyferts, but also because NLS1s, and especially RX J0134.2-4258, are known to be strong emitters of Fe II emission. It has long been considered that Fe II emission is generally weak in radio-loud objects, and [O III] is correspondingly strong. Boroson & Green (1992) in their study of PG quasars found that one of the clearest differences between radio-loud and radio-quiet objects is in their Fe II emission, although steep radio spectrum objects may have strong Fe II (also Joly 1991; Bergeron & Kunth 1984). However, the situation could be similar to that for broad-absorption line quasars (BALQSOs). Early on it was found that BALQSOs are almost never radio-loud (Stocke et al. 1992). Later it was found that many BALQSOs can be described as radio-moderate; they tend to lie at the upper edge of the R distribution for radio quiet quasars (Francis et al. 1993). More recently it has been proposed that radio-moderate or radio-intermediate objects are ones that are in fact not radio-loud but rather have enhanced radio emission because they have a weak radio jet which is beamed in our direction (Falcke et al. 1996). Such an interpretation may be attractive for some NLS1s because of their observed rapid and possibly coherent X-ray variability. However, NLS1s do not appear to be universally radio-intermediate. RX J0134.2-4258 is drawn from a soft X-ray selected sample of AGN (e.g. Grupe 1996; Grupe et al. 1999). A search of the NRAO VLA Sky Survey (NVSS; Condon et al. 1998) reveals that most of the northern ones are very weak radio emitters.
© European Southern Observatory (ESO) 2000
Online publication: March 28, 2000