Fig. 1 shows the finding chart of RX J0134.2-4258 and the measured X-ray, optical, and radio positions of RX J0134.2-4258 are given in Table 1. The 2 (90% confidence) error circles of the RASS and pointed observations are marked. The optical position was derived from the US Naval Observatory scans of the ESO/SRC plates. The X-ray source corresponds to an optical counterpart that was identified as an AGN at z=0.237 (Grupe et al. 1998a). The error circle of the RASS observation was estimated by fitting a Gaussian profile to the data and averaging the positional uncertainties in RA and DEC. The position and error circle of the pointed observation was determined by using a maximum likelihood algorithm in the EXSAS software package (Zimmermann et al. 1998).
The 4.85 GHz radio position was obtained in the Parkes-MIT-NRAO (PMN) radio survey (Wright et al. 1994). The uncertainty on this position, which depends on the location and flux of the object (see Wright et al. 1994 for details), indicate that the optical position of RX J0134.2-4258 is nearly in RA from the cataloged radio source (Fig. 1). This fact, casting significant doubt on the possibility that the optical and radio sources were associated, prompted us to make a new pointed observation using the VLA. We used the task JMFIT on the new VLA data to fit a gaussian to the peak of the emission. The position and uncertainty (taken to be 1/3 the beam in each direction) are shown on Fig. 1 and listed in Table 1. We assert that the coincidence of the radio source and RX J0134.2-4258 is such to warrant a confident association of the optical/X-ray source with this radio object.
The RASS observation yielded 12029 counts, giving a count rate of 0.210.05 and an HR of -0.840.05 (Grupe et al. 1998a, 1999). A power law plus absorption model reveals that the intrinsic absorption in this source appears to be negligible and therefore the absorption was fixed at the Galactic value (; Dickey & Lockman 1990). The resulting energy spectral index is very steep: =6.92.9. Spectral fitting results are given in Table 2, and the result of this fit is shown in Fig. 2. A blackbody model was also fitted. The absorption column density approached zero when it was allowed to be free and therefore it was again fixed at the Galactic value. The resulting radiation temperature was 248 eV. We cannot distinguish between a power law and a black body model for the RASS data. The low signal to noise of the spectrum did not allow more sophisticated fits.
Table 2. Spectral fits to the RASS and pointed observation X-ray data of RX J0134.2-4258. "" is the column density given in units of , "Norm" is the normalization at 250 eV (rest frame) in , "" the soft energy spectral slope, "Break" the break energy in keV of the broken power law fit (rest frame), "" the hard energy spectral slope, "A" the black body integral in , and "" the radiation temperature in eV. The models used here are simple power laws (powl), broken power laws (brpl), Black body (bbdy), and a black body plus power law (bbpl).
The pointed ROSAT PSPC observation yielded 147643 counts corresponding to a count rate of 0.200.01 . This was nearly the same rate as was found in the RASS observation; however, the spectrum was much harder (HR=-0.120.01). A single power law fit to the total pointed observation spectrum with column density fixed to the Galactic value does not give a satisfactory description of the soft spectrum (Fig. 3, left panel). A single black body model does not give a good fit either. A broken power law gave a much better fit (Fig. 3, middle panel). However, to determine the uncertainties on the model parameters, it was necessary to fix the slope of the hard power law. We used two values: =1.0, the result of the free fit, and =0.8, the result from the ASCA spectra (see Sect. 3.2). Interestingly, for both cases, we found that the soft X-ray component had a similar slope as was found during the RASS observation. A black body plus a power law also fit the spectrum well (Fig. 3, right panel); in this case, all parameters except for the absorption, which was still fixed to the Galactic value, could be left free. These spectral fitting results are also summarized in Table 2.
We can infer the presence of a soft excess if a power law plus absorption model yields a column density less than the Galactic value. Fig. 4 displays the contours of power law fits to the RASS and pointed observation spectra with the absorption column free. The RASS spectra are consistent with a single power law model; however, the measured column density for the pointed observation is less than the Galactic column density, indicating the presence of a soft excess.
No significant variability was detected from the RASS light curve; however, the statistics are poor enough that we cannot exclude variability with high confidence. The pointed observation consisted of 6 observation intervals (OBI); each OBI corresponds roughly to a 1 or 2 orbit exposure. Binning the data in each OBI showed clear variability in both the flux and hardness ratio (Fig. 5). The spectral variability is correlated with the flux variability in that the spectrum becomes harder when the count rate increases . Fig. 6 displays how the spectra changed. While the soft channels remain practically constant, the hard channels became fainter by a factor of almost 10. To investigate this further, we made spectral fits of OBIs 3 and 5 using a two component spectral model consisting of an absorbed black body and a power law (Table 2). These fits show that the radiation temperature of the spectra remains the same, and the biggest variation is seen in the normalization of the power law component. This means that the spectral variability can be consistently explained as originating from variability in the flux of the hard component.
RX J0134.2-4258 was not detected during the ROSAT HRI observation. An upper limit of 0.0011 cts was measured. Assuming the power law fit to the OBI 3 data, a count rate of 0.0028 cts was expected. We infer that during the one and a half hours of the observation the hard X-ray component was in its low state. The HRI is much less sensitive to lower energies than the PSPC; therefore the soft X-ray component was not detected.
RX J0134.2-4258 was clearly detected in our ASCA images. Fig. 8 shows the smoothed, combined GIS2 and GIS3 ASCA image in the hard (2.0-5.0 keV) band; RX J0134.2-4258 is the brightest source in the field, but several other sources are visible. The superior Point Spread Function (PSF) available from the ROSAT PSPC allowed us to identify several nearby X-ray emitting sources that could contaminate the ASCA spectrum (Fig. 7) and their positions are listed in Table 3. None of these objects have catalogued identifications.
Table 3. X-ray positions of the sources surrounding RX J0134.2-4258. The source number corresponds to the marked positions in Fig. 7.
The nominal source extraction regions, with radii 4 and 6 arcminutes for the SIS and GIS respectively, were used initially. The background spectra were estimated using source-free regions of the detectors. The source was not detected at high energies and thus the spectra were fitted in the energy bands 0.5-7.0 and 0.8-8.0 keV for the SIS and GIS respectively. The net count rates were 0.017, 0.016, 0.010 and 0.013 in SIS0, SIS1, GIS2 and GIS3, respectively. We fitted all four spectra separately, and found that, as usual, SIS1 yielded slightly flatter spectral parameters and higher absorption. However, the four detectors were consistent within 90% confidence and therefore all data from all detectors were used for spectral fitting.
Spectral fitting revealed evidence that the ASCA spectrum is contaminated by a nearby source. For the source extraction regions listed above, the GIS flux is about 40% larger than that of the SIS. While the cross calibration between detectors is not perfect, the differences should not be more than about 10% (this number is not known exactly because of changes in the SIS efficiency). The suspected contaminating source is #4 for the following reasons. First, it is about from RX J0134.2-4258, and therefore it falls within the GIS extraction region. Also, it can be seen as a faint extension toward the south in the ASCA image (Fig. 8); in contrast, source #3, which is also about from RX J0134.2-4258, is not clearly present in the ASCA image. Thirdly, the difference in fluxes between SIS and GIS increases as the source extraction region sizes are increased. This is expected because the GIS has a broader point spread function, but more importantly because the position of this contaminating source means that it falls off the chip completely for SIS1 and partially for SIS0. The photon index from a power law fit to all four detectors increases slightly as the source extraction region is decreased. This result suggests that the contaminating source may be somewhat harder than RX J0134.2-4258. This source is less than 6% as bright as RX J0134.2-4258 in the pointed ROSAT observation; however, the flux measurements listed above suggest that it contributes more than % to the ASCA spectrum. The difference may be due to a harder spectrum for the contaminating source (although analysis of the ROSAT spectrum does not clearly indicate this) or variability. A faint (magnitude ) counterpart is visible in the DSS image at the X-ray position; however, there is no cataloged identification. Because of the possible contamination, we list spectral fitting results from the regions listed above, as well as SIS alone and SIS+GIS with extraction regions in radius at which the contamination by a source distant should be minimal.
3.2.1. ASCA spectral fitting results
The spectra were first fitted with a power law plus Galactic absorption, and a good fit was obtained (Fig. 9). No features are apparent in the spectra. Iron lines are frequently found in the ASCA spectra of Seyfert galaxies, although there appears to be some dependence on luminosity (e.g. Nandra et al. 1997). Addition of a narrow () iron line does not improve the fit; the upper limit is given in Table 4. Because the upper limit on the equivalent width is consistent with those measured from iron lines detected in other Seyfert galaxies as luminous as RX J0134.2-4258 ( W corresponding to a 2-10 keV flux of ) we conclude that the statistics are too poor to recognize the iron line if present. Evidence for ionized or "warm" absorption is also frequently found in the ASCA spectra of Seyfert galaxies (e.g. Reynolds 1997). Addition of absorption edges at 0.74 and 0.87 keV in the rest frame, relevant for absorption by O VII and O VIII respectively, did not improve the fit significantly, except in the case where the extraction region was ( for 1 degree of freedom). However, the observed energy of this edge for z=0.237 is 0.6 keV, very near the edge of the bandpass in the SIS in the observer's frame. Since it is not detected with the other extraction regions, the reality of this feature is suspect. The upper limits (Table 4) encompass typical optical depths found in Seyferts and NLS1s (e.g. Leighly 1999a) and we conclude that the poor statistics prevent detection of a typical warm absorber. However, a warm absorber with very large optical depth () is clearly ruled out.
The power law energy spectral index in RX J0134.2-4258, measured to be 0.8, has the distinction of being rather flat for an NLS1. In a study of 25 spectra from 23 NLS1s, Leighly (1999a) found an average index of 1.19 with an intrinsic dispersion of 0.3. Therefore, such a flat index is not unknown. However, the other cases of best fit indices occurred in Mrk 766 and NGC 4051, both of which have somewhat peculiar properties for NLS1s, or in objects with complex spectra and very weak hard tails (PHL 1092, PG 1404+226, and Mrk 507). Steeper hard spectra were found in other objects. Leighly (1999a) also found that many NLS1s that have little evidence for absorption have evidence for a soft excess. There is a little evidence for a weak soft excess in RX J0134.2-4258 in that that best fit value above spectral slope above 2 keV, measured to be 0.6, is flatter than the best fit index over the whole energy band. However, the indices are consistent within their statistical error, and addition of a soft excess component does not improve the fit significantly. It is interesting to note, for reasons that will become apparent below, that the spectral index is quite close to the average found from radio-loud quasars observed by GINGA (; Lawson & Turner 1997).
Because it appears that the ROSAT and ASCA spectral fitting results may be consistent except for a normalization factor arising from variability between the two observations, we proceed to make a joint fit (Fig. 9, Table 4). A power law with Galactic absorption provides an acceptable but not a good fit ( for 245 d.o.f.) and there are residuals at low and high energies. Addition of a black body improves the fit greatly ( for two additional degrees of freedom). This joint fit shows that the power law component normalization is a factor of about 3.5 larger in the ROSAT pointed observation than in the ASCA observation. Again, no additional absorption, warm absorber as modeled by O VII and O VIII edges or iron line are required. It has to be noted that in one simultaneous ROSAT and ASCA observation of NGC 5548, ROSAT data systematically appear steeper than ASCA data (Iwasawa et al. 1999) and therefore there is some doubt whether the cross calibration between the two missions is good enough to warrant a simultaneous fit. However, in this case, we see no evidence for a cross calibration problem.
3.2.2. X-ray variability during ASCA observation
We observed RX J0134.2-4258 to vary during the observation. The light curve from the SIS detectors binned by orbit is shown in Fig. 10; a steady decline by a factor of about 2 was observed. We calculated the fractional amplitude of variability, also called the excess variance, defined as the measured variance minus the measurement error and normalized by the square of the mean (e.g. Nandra et al. 1997; Leighly, 1999b). This parameter is a useful one to quantify variability in ASCA observations because it can be shown that under a set of justifiable assumptions, it is related to the inverse of the variability time scale (Leighly 1999b). For a light curve with 128 second binning, the excess variance is . Fig. 11, adapted from Leighly (1999b), shows that this excess variance is high compared with broad-line Seyfert galaxies with similar luminosities but consistent with Narrow-line Seyfert 1 galaxies. Therefore, although the hard X-ray spectrum appears to be rather flat for an NLS1, the variability properties are unremarkable.
Fig. 12 displays the optical spectrum of RX J0134.2-4258 obtained in September/October 1995. The spectrum is characteristic of a typical Narrow-line Seyfert 1 galaxy. The broad permitted blend of FeII emission commonly present in NLS1s severely contaminates the spectrum, making an accurate measurement of the flux and width, in particular of the H and O[III] lines difficult. In order to account for this, we employ the FeII subtraction method introduced by Boroson & Green (1992) and now commonly used (e.g. Grupe et al. 1999; Leighly 1999a). The FeII emission line spectrum from a high signal-to-noise optical spectrum of the prototype strong Fe emitter I Zw 1 is first convolved with a Gaussian and then scaled until the width and intensity of the lines approximately match those seen in the RX J0134.2-4258 spectrum. This best-fitting FeII model was then subtracted and the remaining emission lines examined.
We measured line widths of FWHM=900100 and 670200 for H and [OIII]5007, respectively. The ratio of the FeII to H emission is the strongest among all objects in the entire sample of soft X-ray selected AGN ( = 12.3, see Grupe at al. 1999). As we noted in Grupe et al. 1998a, the object has a very blue spectrum ( = -0.1). We were not able to detect any [OIII]5007 emission in the spectra taken two years earlier; however, the [OIII] is apparent in the better signal-to-noise spectrum presented here (after FeII subtraction). Otherwise, the results are the same as obtained from our other spectra (Grupe et al. 1999).
RX J0134.2-4258 was detected in the Parkes-MIT-NRAO (PMN) survey (Wright et al. 1994) with a flux density of 559 mJy at 4.85 GHz. Our new VLA observation yielded a flux of mJy at 8.4 GHz and the source is completely unresolved in our map. The two radio fluxes allow us to derive a radio spectral index of -1.4, assuming no variability between the two observations. This radio spectral index is fairly steep compared to core-dominated sources, but not unheard of. Applying the criterion of Kellermann et al. (1989), who use R=10 as the dividing line between radio-loud and radio-quiet quasars 2, the source is radio-loud, since R=71. The radio flux corresponds to a luminosity of log (P) = 25.3 W Hz-1. Thus, also using the luminosity criteria (e.g. Joly 1991; Miller et al. 1993), RX J0134.2-4258 should be classified as a radio-loud AGN. We note, however, that if the steep spectrum is real and not a consequence of variability, the radio flux is highly frequency dependent.
Narrow-line Seyfert 1 galaxies are generally considered to be weak radio sources. Ulvestad et al. (1995) found that the luminosities among 15 NLS1s observed with the VLA were were generally less than log P=22.5 W Hz-1. Only two other NLS1s are known to be radio-loud. One is PKS 0558-504 (e.g. Remillard et al. 1991) and the other is RGB J0044+193 (Siebert et al. 1999). Both of these objects are fairly faint radio sources, with R and 27, respectively. Thus, RX J0134.2-4258 has the relatively strongest radio emission of an NLS1 so far discovered.
It is possible that RX J0134.2-4258 is a variable radio source. It is not a catalogued member of the Parkes 2700 MHz survey catalog. The flux limit of this catalog depends on the location in the sky, but we note that there are catalogued objects within 10 degrees with fluxes as low as 50 mJy, although this appears to be the lower limit of the catalog. The PMN flux at 4.85 GHz predicts a flux of 125 mJy at 2700 MHz, assuming ; therefore, it should have been detected by the Parkes survey unless it is variable. Objects with steep spectra are generally not variable, but since the observations defining the index were not simultaneous, the steep index could in fact be a consequence of variability. Evidence for variability in the radio has also been found in one of the other radio-loud NLS1s, RGB J0044+193 (Siebert et al. 1999). On the other hand, many NLS1s observed simultaneously at more than one radio frequency show steep rather than flat spectra (Moran et al. in prep.).
3.5. Spectral energy distribution
Fig. 13 displays the spectral energy distribution of our source. The object was also detected by IRAS at 60 µm and we display upper limits at 12, 25, and 100 µm (Grupe et al. 1998a). The ROSAT X-ray data in the plot are displayed as the black body plus power law model fit to the RASS and pointed observation spectra. For both cases, the radiation temperature and the hard X-ray spectral slope were fixed (see Table 2).
We used the nonsimultaneous HST (2000) and X-ray spectra to compute , defined to be the point-to-point slope between 2500Å and 2 keV (Wilkes et al. 1994). The pointed ROSAT observation, the ASCA observation and the RASS observation yielded values of of 1.47, 1.68, and 2.00 respectively. Based on the regressions obtained by Wilkes et al. (1994), the predicted values of are 1.61 and 1.43 for radio-quiet and radio-loud AGN of this optical luminosity, respectively. The values obtained for RX J0134.2-4258 lie between these predicted indices, and in two cases are lower than the radio-quiet index. Therefore, although it is impossible to account for the effect of variability, the X-ray emission in RX J0134.2-4258 does not appear to be as strong as is found in other radio-loud AGN with similar optical luminosities.
© European Southern Observatory (ESO) 2000
Online publication: March 28, 2000