THE UNUSUAL SPECTRAL ENERGY DISTRIBUTION OF LBQS 0102–2713

The source spectra obtained from the ROR numbers 700121p-0 and 700121p-2 using the xselect command-line interface were grouped with the grppha command with group min 10. The merged data set was grouped with group min 30. The spectral fitting was performed using XSPEC version 12.5.0. The model components were (mo = phabs (zpo)). While the light curves do not show significant variations (Figure 1) the resulting photon indices are remarkably steep (Figure 2). LBQS 0102–2713 might be the quasar with the steepest soft X-ray photon index, detected in two individual and the merged observations, reported so far. For the merged data set we get Γ = (6.0 ± 1.3) and Γ = (5.8 ± 1.3) for the xselect and EXSAS software packages, respectively. The spectral fitting results from the xselect command line interface and the EXSAS software routines are listed in Table 2. The spectral parameters obtained from both software systems are consistent within the errors. As a security check we have used the NASA GSFC simulation software webpimms. The simulated spectra are consistent with the ROSAT power-law fits. This is an independent test, whether the calibration and response files available at the MPE site give the same results as the corresponding files produced by the NASA GSFC calibration team.

Table 2. Spectral Fitting Results for a Power-Law Model with Foreground Absorption of LBQS 0102–2713^a

	xselect			EXSAS
ROR number	Γ	NH	Norm	Γ	NH	Norm
700121p-0	$\rm 6.6\pm 2.5$	$\rm 6.2\pm 0.5$	$\rm 2.5\pm 2.9$	$\rm 5.5\pm 2.5$	$\rm 7.4\pm 1.5$	$\rm 4.2\pm 5.0$
700121p-2	$\rm 5.5\pm 2.5$	$\rm 3.7\pm 2.4$	$\rm 1.3\pm 1.8$	$\rm 6.5\pm 2.1$	$\rm 6.0\pm 4.7$	$\rm 4.5\pm 4.0$
Merge data set	$\rm 6.0\pm 1.3$	$\rm 4.8\pm 1.5$	$\rm 1.3\pm 1.6$	$\rm 5.8\pm 1.3$	$\rm 6.5\pm 4.4$	$\rm 1.5\pm 1.0$

Note. ^aThe N_H value is in units of $\rm 10^{20}\; cm^{-2}$ and the normalization is given in $\rm 10^{-4}\; photons\; cm^{-2}\; s^{-1}\; keV^{-1}$ at 1 keV.

Download table as:  ASCIITypeset image

The Galactic foreground absorption in the direction to LBQS 0102–2713 is very low with $N_H=1.2\,{\times}\, 10^{20} \rm cm^{-2}$. It is known that Γ and N_H are correlated in ROSAT PSPC fits, such that a larger N_H requires a steeper photon index to give a good fit. In Figure 3 we show the Γ–N_H contour plots for the merged data set and for the ROR number 700121p-2. For the ROR number 700121p-0 no reliable contour plot could be obtained. At a 99% confidence level, the photon index is about 3.5 at the Galactic N_H value. The photon index might indeed be flatter as Γ = (6.0 ± 1.3) and the object may be less special in X-rays than it appears from the simple power-law fits. However, the photon index obtained for a power-law fit with the Galactic N_H value fixed remains still steep for quasar soft X-ray SEDs (compared to Section 1). Longer observations with the present generation of X-ray telescopes are required to confirm whether the photon index of LBQS 0102–2713 is indeed extremely steep.

We note that a simulation of a power law plus an ionized absorber could also result into a flatter photon index. A simulated 50 ks XMM-Newton spectrum obtained from a power-law model with cold absorption plus an ionized absorber based on the grid25BIG_mt.fits model (C. Done 2008, private communication) gives a photon index of Γ = (4.4 ± 0.5) for a column density of the ionized absorber ranging between $\rm (1.1\hbox{--}1.5)\,{\times}\, 10^{22}\ cm^{-2}$ and an ionization parameter range between $\rm \xi = 20$ and 30. The parameter ranges explored range between $N_H=10^{21\hbox{--}24} \rm cm^{-2}$ and $\rm \xi =$ 10 to 1000. The flatter photon index occurs in a very narrow parameter range of the ionized absorber column density and the ionization value. When fitting the simulated data with the ROSAT PSPC response and the ROSAT exposure time, the photon index is $\rm (6.7\pm 1.8)$, consistent with the ROSAT spectral fitting results for a power-law model with the N_H value leaving free in the fit.

A blackbody fit to the merged data set does also result into an acceptable fit. The derived spectral parameters in the rest frame are: $N_H=(2.0\pm 2.3)\,{\times}\, 10^{20} \rm cm^{-2}$, kT = (0.13 ± 0.013) keV, and $n=(2.6\pm 2.8)\,{\times}\, 10^{-5}\ \rm photons\ cm^{-2}\ s^{-1}\; keV^{-1}$ at 1 keV. Due to the limited spectral resolution of ROSAT, we cannot disentangle between a power-law and a blackbody model. We compare the derived blackbody temperature with other published blackbody temperatures and find that they are in good agreement. Crummy et al. (2006) found values for the blackbody temperature for an AGN sample observed with XMM-Newton between 0.009 and 0.17 keV (their Table 2). Tanaka et al. (2005) have analyzed NLS1 spectra observed with XMM-Newton and found for a sample of 17 objects blackbody temperatures between 0.09 and 0.15 keV. Fiore et al. (1998) obtain blackbody temperatures of 0.16 keV for PG 1244+026 and NAB 0205+024, respectively based on ASCA observations. Puchnarewicz et al. (2001) obtain a value of 0.13 keV for the NLS1 galaxy RE J1034+396.

Morris et al. (1991) have published an optical spectrum of LBQS 0102–2713 in the wavelength range between 3200 and 7400 Å in the observers frame (compared to Figure 4). The Mg ii line at 2800 Å (rest frame) is clearly visible. Applying a Gaussian fit to this line we obtain a FWHM value of about 2200 km $\rm s^{-1}$ in the rest frame. This value is very close to the artificial border line between NLS1 galaxies and broad line Seyfert galaxies of 2000 km $\rm s^{-1}$ following the definition of Osterbrock & Pogge (1985). In addition Morris et al. (1991) first noted the strong UV Fe ii multiplet emission between about 2200 and 2500 Å in the rest frame. All this is typical for NLS1 galaxies and LBQS 0102–2713 can be considered as an X-ray bright NLS1 galaxy. This goes in line with the steep X-ray spectrum. In addition, a strong C iii line is found at 1909 Å rest frame and a strong emission feature at about 2075 Å which is usually interpreted as a bunch of Fe iii multiplets. The later feature is cataloged in the Vanden Berk et al. (2001) quasar composite paper and is certainly present in higher signal-to-noise composites.

Zoom In Zoom Out Reset image size

The observed 4450 Å flux density from the optical spectrum (compared to Figure 4) obtained in 1988 is about $f_{4450\;{\AA}}^{\rm obs.} = 3.0\,{\times}\, 10^{-16}\; \rm erg\ cm^{-2}\ s^{-1}\;{\AA}^{-1}$. Details of the spectroscopic measurements can be found in Section 2.3 of Morris et al. (1991). To convert the units from Å to Hz, we followed the relation from Hogg (2000) described in his Chapter 7. As the differential flux per unit log frequency or log wavelength is νf_ν = λf_λ (Hogg uses S instead of f) one gets f_ν = (λ²/c) ×  f_λ with f_λ = f^obs._4450 Å and λ = 4450 Å. The resulting observed flux density per Hz is $f_{4500 \;{\AA}}^{\rm obs.}= 2.0\,{\times}\, 10^{-27} \rm erg\ cm^{-2}\ s^{-1}\ Hz^{-1}$. To correct the monochromatic flux measurements for Galactic extinction, we used the Nandy et al. (1975) extinction law with R = A_V/E(B − V). Their Table 3 gives for $\rm (1/ \lambda (\mu \ m^{-1}))$ at 4450 Å an R value of 3.8. The E(B − V) value from the GALEX data obtained from NED is 0.02. According to Geminale & Popowski (2005), the relation between R and is (λ − V) = R((A_λ/A_V) − 1). Following their extinction curves shown in Figure 4, one obtains an value of about 0.5. The A_λ value at 4450 Å is then 0.076 and the A_V value using the relations given above is 0.066. The A_λ/A_V value is 1.15. The extinction-corrected flux and the uncorrected flux at 4450 Å are related by $f_{4450 \;{\AA}}^{\rm Ext.} = (10^{+4}\,{\times}\, A_{\lambda } / A_V)\,{\times}\, f_{4450 \;{\AA}}^{\rm obs.} = 2.6\,{\times}\, f_{4450 \;{\AA}}^{\rm obs.}= 5.3\,{\times}\, 10^{-27} \rm erg\ cm^{-2}\ s^{-1}\ Hz^{-1}$. The luminosity distance $D_L= 1.5\,{\times}\, 10^{28} \rm cm$. The resulting rest-frame UV luminosity density is $L_{2500\;{\AA}}= 1.6\,{\times}\, 10^{31}\; \rm erg\ s^{-1}\ Hz^{-1}$. The relation between l_uv and L_2500 Å is log L_2500 Å = l_uv = 31.21.

Zoom In Zoom Out Reset image size

We also analyzed the seven HST FOS spectra named as Y309010nT with n ranging from 2 to 8. The HST FOS spectra were obtained from the HST FOS proposal ID 6007: Comparison of the large scale structure in QSO absorbers and galaxies in the Galactic Poles, D. York, University of Chicago. The observed wavelength ranges between 1700 and 2300 Å. All seven exposures obtained in 1995 agree in their flux density values and the HST FOS absolute and relative calibration is very good (M. Rosa, ESO Garching 2009, private communication). In Figure 5, we show the spectrum for the science exposure number Y3090102T. The Lyα line is clearly visible and marked in the spectrum. The line is blended with a low-ionization line of N v at 1240 and 1242 Å in the rest frame. In addition, there is emission from the Lyβ line blended with the O vi line at 1032 and 1037 Å at their rest-frame wavelengths.

The observed 2300 Å flux density is about $f_{2300 \;{\AA}}^{\rm obs.} = 1.5 \,{\times}\, 10^{-15}\; \rm erg\ cm^{-2}\ s^{-1}\;{\AA}^{-1}$. The flux density at 2500 Å rest frame is f_2500 Å = f_1292 Å × (1292 Å/2500 Å)^−0.5 = 4.2 ×  10⁻²⁷ erg cm^-2 s⁻¹ Hz⁻¹. To extrapolate the corresponding rest-frame wavelength of 1292–2500 Å, we have used the α_o value of −0.5 of Richstone & Schmidt (1980). The extinction-corrected flux at 2500 Å is $\rm 1.4 \,{\times}\, 10^{-26}\ erg\ cm^{-2}\ s^{-1}\ Hz^{-1}$, about a factor of 3.3 larger compared to the non-extinction-corrected flux density. The slight discrepancy compared to the optical spectrum obtained by Morris et al. (1991) might be due to flux variability between the 1988 and 1995 observations or due to the extrapolation from 1292 to 2500 Å using a mean $\rm \alpha _o$ value of −0.5. The correction for Galactic extinction has been performed as described in the previous subsection. The R value at 2500 Å is 7.2 and the value is 7.2. The resulting A_λ/A_V value is 1.4. This results into a Galactic extinction correction with a factor of 3.3 and the extinction-corrected flux is $f_{2500 \;{\AA}}^{\rm Ext.} = 1.4\,{\times}\, 10^{-26}\; \rm erg\ cm^{-2}\ s^{-1}\ Hz^{-1}$. The resulting rest-frame luminosity density is $L_{2500\;{\AA}} = 4.0\,{\times}\, 10^{31}\; \rm erg\ s^{-1}\ Hz^{-1}$.

The 2 keV rest-frame flux density is determined following Hogg (2000) and Weedman's (1986) Quasar Astronomy (Section 3.5, page 61). The photon index Γ is related to the energy index via $\rm \alpha _x = -\Gamma + 1$. For a photon index not equal to 2, the rest-frame 2 keV flux density is given by $f_{2\;{\rm keV}} = f(0.5\hbox{--}2.0)\,{\times}\, ((1+\alpha _x) / (1+z)^{\alpha _x}))\,{\times}\, ((\nu _{2\;{\rm keV}}^{\alpha _x} / (\nu _{2\;{\rm keV}}^{\alpha _x + 1} - \nu _{0.5\;{\rm keV}}^{\alpha _x + 1}))$. The unabsorbed ROSAT flux in the 0.5–2.0 keV energy band is $\rm 4.3\,{\times}\, 10^{-14}\ erg\ cm^{-2}\ s^{-1}$. As the photon index derived in the previous section was 6.0, we get for $\rm \alpha _x$ a value of −5. The corresponding 2 keV and 0.5 keV frequencies are $\rm 4.8\,\,{\times}\,\, 10^{17}\ Hz\ {\rm and}\ 1.2\,\,{\times}\,\, 10^{17}\ Hz$. This results into a flux density at 2 keV of $f_{\nu \ 2\;{\rm keV}} = 4.3\,{\times}\, 10^{-32}\; \rm erg\ cm^{-2}\ s^{-1}\ Hz^{-1}$. The 2 keV luminosity density is then $L_{2\;{\rm keV}} = 4.0 \,{\times}\, 10^{25}\; \rm erg\ s^{-1}\ Hz^{-1}$. We note that the extremely steep photon index has to be confirmed in future observations with the present generation of X-ray telescopes given the results from the contour plots shown in Figure 3. If the X-ray spectrum is flatter than obtained from the power-law fits, this would result in a less extreme α_ox value.

For a definition of α_ox = 0.384 ×  log(L_2 keV/L_2500 Å), one derives α_ox values of −2.3 and −2.2.

Zoom In Zoom Out Reset image size

Richards et al. (2006) show the mean quasar SEDs for optically luminous SDSS, all SDSS, and optically dim quasars (compared to Figure 11 of their paper). In Figure 6, we show an adopted version of the plot of Richards et al. (2006). The $\rm \nu L_{\nu }$ value at 2500 Å rest-frame wavelength is $\rm 4 \,{\times}\, 10^{47}\ erg\ s^{-1}$. This is about a factor of 10 more luminous compared to the mean of the optically most luminous SDSS quasars. However, there is a large dispersion in the optical M_B magnitudes for quasars. Brotherton et al. (2001) show the distribution of the M_B magnitudes for more than 11,000 quasars (their Figure 5). Their absolute magnitudes range between about −23^mag to −32^mag. LBQS 0102–2713 has a photometric IIIa-J magnitude of 17.52 mag as taken from NED at a wavelength of 4612 Å corresponding to 0.46 $\rm \mu$m. This close to the mean B wavelength of 0.40 $\rm \mu$m. The corresponding M_B value of LBQS 0102–2713 is −25.9. This might indicate that the scatter in the 2500 Å luminosities is also large and that LBQS 0102–2713 is less extreme in the UV than one would think at first glance when comparing the UV value to the mean of the optically most luminous quasars. The corresponding 2.0 keV rest-frame value of $\rm \nu L_{\nu }$ value is $\rm 1.4\,{\times}\, 10^{44}\ erg\ s^{-1}$ which is comparable to the optically dim SDSS quasars. The dispersion in the X-rays appears to be much less compared to the UV luminosities if one compares the X-ray luminosity of LBQS 0102–2713 to the X-ray luminosities as shown in the sample papers of Strateva et al. (2005), Vignali et al. (2003), or Gibson et al. (2008).

Zoom In Zoom Out Reset image size

In Figure 7, we have plotted the available multiwavelength data in units of $\rm \nu f_{\nu }$ versus the wavelength and compare these data points to the mean quasar SED for radio-loud objects from Richards et al. (2006, their Figure 10). All data points are extinction corrected. For the ROSAT data, we show only the lowest and highest energy data point. While at 2 keV the object is X-ray weak compared to the mean quasar SED, at the softest energies LBQS 0102–2713 is a very powerful emitter and the $\rm \nu f_{\nu }$ value even exceeds the 2500 Å UV value. The HST FOS data points are about a factor of 2–5 larger compared to the mean quasar SED. Here we note again that the scatter in the individual UV luminosities compared to the mean might be larger as pointed out in the previous subsection and that therefore the object might be less extreme in its UV luminosity. The optical data points and the K_S magnitude follow the mean quasar SED. The main result from this plot is the discrepancy between the X-ray and HST FOS data points as these bands are discontinuous. The standard picture assumes that the rest-frame UV photons are emitted from the accretion disk and that the hard X-ray photons are arising from the accretion disk corona. From Figure 7, it is obvious that there is a discrepancy between the UV and the X-ray photons. The optical data from Morris et al. (1991) are also offset with respect to the HST FOS data points. This is most probably due to a significant contribution from the host galaxy and the rest-frame optical photons are not expected to be related to the accretion disk. We have cross-checked the discontinuity between the UV and X-ray data points. The GALEX NUV wavelength is 2315.7 Å (compared to Morrissey et al. 2007, their Table 1). The NUV GALEX calibrated flux from NED is 282.2 $\rm \mu$Jy. This converts into flux densities of $\rm 2.8\,{\times}\, 10^{-27}\ erg\ cm^{-2}\ s^{-1}\ Hz^{-1}$ or $\rm 1.6\,{\times}\, 10^{-15}\ erg\ cm^{-2}\ s^{-1}\ \AA^{-1}$. The GALEX NUV flux density value is consistent with the flux density shown in the HST FOS spectrum of Figure 5.

Leighly et al. (2007) favor an intrinsically X-ray weakness model for the quasar PHL 1811. The 2 keV X-ray emission is intrinsically weak rather than absorbed. In this case, one expects weak low ionization of semiforbidden UV lines. For the blend of the $\rm Ly{\alpha }$ and N v lines, an EW value of 15 eV is obtained. No EW values are given for the blend of Lyβ and O vi. The authors argue that this points to an intrinsically weak X-ray model. The argumentation for an intrinsically X-ray weakness model is based on individual lines such as Na i D, Ca ii H and K, and C iv. The rest-frame EW values for the blend of Lyβ and the O vi lines and the $\rm Ly{\alpha }$ and the N v lines shown in Figure 5 are about 12 and 50 Å, respectively. In the following, we compare these values with quasar composite spectra. Brotherton et al. (2001) give for the Lyβ plus O vi lines an EW value of 11 Å in the rest frame. For the blend of $\rm Ly{\alpha }$ and N v, the EW value is 87 Å. The corresponding values reported by Vanden Berk et al. (2001) are 9 and 94 Å, respectively. Zheng et al. (1997) found values of 16 and 102 Å, respectively. The EW values obtained for LBQS 0102–2713 are typical to the mean quasar composite values and the source appears not to be intrinsically X-ray weak.

There are other models in the literature which are speculative or do have a low probability to explain the X-ray weakness.

Gibson et al. (2008) present in their Figure 2 a decreasing trend of α_ox with increasing 2500 Å UV luminosity density. This is a strong observational constraint as an increased UV luminosity density results into steeper α_ox values. In the case of PHL 1811 and LBQS 0102–2713, we indeed see lower 2 keV flux densities compared to other quasars. If one compares the UV and X-ray luminosity densities of LBQS 0102–2713 which are log l_UV = 31.20 for the Morris et al. (1991) spectrum and log l_UV = 31.60 obtained from the HST FOS spectra and an X-ray luminosity density of log l_X = 25.60 obtained from ROSAT, with Figure 8 of Vignali et al. (2003) the X-ray weakness becomes immediately apparent. The reason for the so-called global X-ray Baldwin effect is presently not known. It is speculated that a patchy or disrupted accretion disk corona is related to the UV luminosity density (Gibson et al. 2008). However, there is no direct observational proof for this model.

Zoom In Zoom Out Reset image size

One could speculate that UV and X-ray variability of the source might account for the low α_ox value. Assuming that LBQS 0102–2713 was in 1992 in a low X-ray flux state, then a factor of about 10 at 2 keV is required to obtain the canonical α_ox value of −1.8 for radio-loud quasar as shown in Figure 11 of Richards et al. (2006). A factor of 10 in three years with respect to the optical observations appears unlikely as only a few AGNs are known to exhibit flux variability of a factor of 10 or more. In addition, variability seems unlikely given Hook et al. (1994) who find only weak optical variability with σ = 0.17.

LBQS 0102–2713 appears to be very similar to BAL quasars which show strong UV- and weak 2 keV X-ray emission (compared to Gallagher et al. 2006). Their α_ox values range between −1.65 and −2.48. The majority of the objects have values smaller than −2.0. They authors argue that the X-ray weakness of BAL quasars is due to neutral intrinsic absorption with column densities between about $\rm (0.1\hbox{--}10)\,{\times}\, 10^{23}\ cm^{-2}$. As more soft X-ray photons are expected for a simple neutral absorber, the authors argue that the absorption is more complex. Partial covering or ionized absorbers can account for this observational fact. However, LBQS 0102–2713 do not show such high values for the neutral absorber compared to BAL quasars. The foreground absorption is about $\rm 6\,{\times}\, 10^{20}\ cm^{-2}$. This is at least a factor of about 20 lower compared to BAL quasars. LBQS 0102–2713 is an object which exhibits an unusual combination of UV brightness, X-ray weakness and no significant absorption by neutral matter along the line of sight. In addition, there are no significant indications that the object is intrinsically X-ray weak, in contrast to the argumentation for PHL 1811 by Leighly et al. (2007). This parameter combination is new and needs to be explained. With the present available X-ray data, we are limited in providing self-consistent models but would like to present these new results to the community.

The object might exhibit the steepest photon index reported from a quasar so far. For a simple power-law fit with neutral absorption left free in the fit we get photon indices of Γ = (5.8 ± 1.3) or(6.0 ± 1.3) by using the xselect command line interface and the EXSAS software package, respectively. However, if the foreground N_H value is set to the Galactic value of $\rm 2\,{\times}\, 10^{20}\ cm^{-2}$ (Dickey & Lockman 1990) the photon index is about 3.5 (compared to the Γ–N_H contour plots shown in Figure 3). The photon index still remains steep compared to previous studies listed in Section 1. In addition, Puchnarewicz et al. (1995) applied a broken power-law fit to RE J2248-511 with a photon index of $\rm 4.13^{+5.85}_{-0.60}$ up to the break energy of 0.26 keV. The Galactic N_H value was fixed to $\rm 1.4\,{\times}\, 10^{20}\ cm^{-2}$. For the object RE J1034+396 Puchnarewicz et al. (1995) obtain for their broken power-law fit a value of Γ = 4.45^+0.25_−0.32 up to a break energy of 0.41 keV. The N_H value was fixed to the Galactic value of $\rm 1.5\,{\times}\, 10^{20}\ cm^{-2}$. We note that the low N_H values, although fixed to their Galactic values, result in similar steep photon indices as obtained for LBQS 0102–2713 with the N_H value fixed in the fit.

There are open problems which cannot be solved given the present set of multiwavelength data. First, we found that the UV emission is discontinuous to the X-ray emission, e.g., the UV photons which are expected to arise from the accretion disk appear not to be correlated with the X-ray photons. Second, if the steep soft photon index of $\rm (6.0\pm 1.3)$ will be confirmed by new X-ray observations this steepness needs to be explained. Based on the contour plots shown in Figure 3, the object might not be as extreme as one would expect from the simple power-law fits. Third, the UV luminosity density is about a factor of 10 more luminous compared to the mean of the most luminous SDSS quasars. However, as pointed out there is a large spread in the M_B magnitudes for quasars and the UV brightness might also show a large scatter compared to the mean value at a certain frequency. Finally, if the 2 keV X-ray weakness is confirmed by other observations, the combination of UV brightness, X-ray weakness and the absence of significant absorption by neutral matter results into an unusual combination observational parameters which cannot be explained with the available set of multiwavelength data. Finally, what is the spectral complexity in the soft and hard X-ray band unresolvable for ROSAT? We know from observations with the present generation of X-ray telescopes that new spectral components can often be added to the fit and that the real spectrum might be more complex. We expect to see this in the soft band with higher signal-to-noise observations. In addition, observations above 2.4 keV are required to better constrain physical models for the unusual observational parameters detected in LBQS 0102–2713.