The Discovery of Soft X-ray Loud Broad Absorption Line Quasars
aa r X i v : . [ a s t r o - ph ] J a n The Discovery of Soft X-ray Loud Broad Absorption Line Quasars
Kajal K. Ghosh andBrian Punsly ABSTRACT
It is been known for more than a decade that BALQSOs (broad absorption linequasars) are highly attenuated in the X-ray regime compared to other quasars,especially in the soft band ( < A ) flux density ratios that are higher than typical nonBAL radioquiet quasars. Our sample of 3 sources includes one LoBALQSO (low ionizationBALQSO) which are generally considered to be the most highly attenuated inthe X-rays. The three QSOs are the only known BALQSOs that have X-rayobservations that are consistent with no intrinsic soft X-ray absorption. Theexistence of a large X-ray luminosity and the hard ionizing continuum that itpresents to potential UV absorption gas is in conflict with the ionization statesthat are conducive to line driving forces within BAL winds (especially for theLoBALs). Subject headings: (galaxies:) quasars: absorption lines — galaxies: jets — (galax-ies:) quasars: general — accretion, accretion disks — black hole physics
1. Introduction
One of the biggest challenges to our understanding of broad absorption line quasars(BALQSOs, hereafter) is how lithium like species (those species producing the resonant UVabsorption lines) can form within an out-flowing wind in the presence of the hard ionizing Universities Space Research Association, NASA Marshall Space Flight Center, VP62, Huntsville, AL,USA N H & cm − and most BALQSOs have absorption columnsof 10 cm − ≤ N H ≤ cm − (Punsly 2006). As a qualifier, it is possible that the modestradio jet seen in some BALQSOs can also contribute to the X-ray flux and some of thesesources might not appear X-ray suppressed due to a secondary source of X-rays from therelativistic jet ∼
10 - 100 pc from the central quasar (Brotherton et al. 2005). Thus, wemake a distinction between radio quiet (RQBALQSOs), log
R <
1, and radio loud BALQSOs(RLBALQSOs), log
R >
1, where R is the k-corrected 5 GHz to 2500˚ A flux density ratio(the RQBALQSOs considered here have log R < N H & cm − have fairly typical hard X-ray to UVflux density ratios for radio quiet quasars, but are highly attenuated in the soft X-rays(Aldcroft and Green 2003; Shemmer et al 2006; Murray et al 1995). To this date, all knownRQBALQSOs are highly absorbed in the soft X-rays. The lone exception is the X-rayspectrum of IRAS 07598+059 that is consistent with no absorption. However, the X-ray fluxrelative to the UV is suppressed by a factor of more than 20 compared to a typical radioquiet quasar, thus it is believed that the central quasar X-ray source is completely obscuredand we are seeing a secondary weak source, likely the central source seen in reflection off anelectron scattering mirror (Imanishi and Terashima 2004).We have searched the ROSAT database to look for evidence of soft X-ray loud BALQ-SOs. Previously, the only BALQSO ROSAT detection in the literature is 1245-067 and it ishighly absorbed, N H ≈ cm − (Green and Mathur 1996). We have found approximately40 likely detections of BALQSOs by ROSAT and are preparing a catalog of these. However,most of these either are not necessarily strong compared to the UV flux or have poor photonstatistics, so it is difficult to determine the X-ray luminosity with much certainty. In sortingthrough the catalog, we have segregated three AGN that are the most convincing examplesof RQBALQSOs that are soft X-ray loud (defined by soft X-ray (0.3 keV) to UV (3000˚ A )flux density ratios that are higher than typical nonBAL radio quiet quasars), one of whichhas low ionization UV absorption lines (LoBALQSO). This is even more surprising since thethe LoBALQSOs require the maximal amount of X-ray screening, otherwise the gas will beover-ionized for the formation of Li-like low ionization species (Murray et al 1995). TheseBALQSOs could either be intrinsically exceptional objects or objects that are fortuitouslyconfigured to reveal some unforseen physical features of the QSO central engine. 3 –
2. The X-ray and UV Data
A search for optical counterparts in the Sloan Digital Sky Survey Data Release 4 ofROSAT PSPC sources from the White-Giommi-Angelini Catalog, White et al. (1994), wasperformed by Suchkov et al (2006). They found 1744 tentative identifications of ROSATsources with optical spectra typical of AGN. Our in depth inspection of the SDSS spec-tra of these 1744 AGNs has identified approximately 44 AGNs as BALQSOs. This sec-tion describes the data for the three candidate BALQSOs that are soft X-ray loud, SDSSJ023219.52+002106.8, SDSS J101949.75+450256.0 and SDSS J132229.66+141007.3 (corre-sponding to the WGA counterparts 1WGA J0232.3+0021, 1WGA J1019.7+4502 and 1WGAJ1322.4+1410, respectively). Hereafter, the SDSS designations will be abbreviated as 0232,1019 and 1322.Event files for the ROSAT PSPC observations of 0232, 1019 and 1322 were retrievedfrom the ROSAT archive. We use the locally-developed software tool, LEXTRCT, for theextraction of source and background light curves and spectra (Tennant 2006). Good timeinterval filtering was applied to all the event files. We extracted source counts from circlesof 60 ′′ radius around each source and corresponding backgrounds from nearby source-freeregions. Background regions were around 3 to 5 times larger than the source regions. X-raylight curves were binned into 1000 s bins. X-ray spectra of all three sources were binned sothat there were at least 30 counts per fitting bin. Spectral redistribution matrices and ancil-lary response files were retrieved from the ROSAT archive. XSPEC version 11.3 was used tofit the 0.2 – 2.4 keV energy spectra with an absorbed ( phabs ) powerlaw ( powerlaw ) model.Confidence contours in the Γ (photon index)– N H plane were generated for the models. Theconfidence contours indicate that N H < × cm − with >
99% confidence. The photonstatistics are not good enough to segregate out the small intrinsic absorption column densityfrom the Galactic value of N H . Thus, N H in of our favored model fit (column 6 of Table 1) isfixed at the Galactic value. We also fitted these spectra with a redshifted, absorbed, powerlawmodel. Best fitting model parameters are similar to the previous values. Table 1 describesthe details of our fits to the data. The first two columns are the source name and redshift.The next column is X-ray luminosity at the rest frame of the QSOs in the 2-8 keV energyband. Γ is tabulated in columns 4 and 5 based on χ and Cash statistics, respectively. The“goodness of fit” appears in column 7. In the following columns, we define two measures ofthe X-ray flux density strength relative to the UV: a UV to soft X-ray spectral index definedin terms of the flux density, F ν as α os ≡ .
537 log[ F ν (3000˚ A ) /F ν (0 . α ox ≡ .
384 log[ F ν (2500˚ A ) /F ν (2 . N H plane, the intrinsic X-ray luminosity is even larger thanindicated in column 3 (i.e., α os and α ox are even flatter than in columns (8) and (9)). Thus, 4 –Fig. 1.— The UV spectra from SDSS (from left to right, 0232, 1019, 1322), the best fitcontinuum is in magenta. 5 –our claims that these sources are soft X-ray loud are conservative. Deeper X-ray observationsare required to improve the accuracy of our spectral fits.The SDSS spectra of the 3 soft X-ray loud RQBALQSOs in the regions of the BALs arepresented in Figure 1. The spectra were retrieved from the SDSS database and were analyzedusing the IRAF software. First, the spectrum was de-reddened using the Galactic extinctioncurve, Schlegel et al. (1998), then the wavelength scale was transformed from the observed tothe source frame. The spectra were fitted in XSPEC with a powerlaw plus multiple gaussianmodel, including fits for both emission lines and absorption lines (Arnaud 1996). All themodel parameters were kept free. The best fit to the SDSS data was determined using χ minimization. If an emission bump around 2500 ˚ A is present then the continuum fit in theCIV (1550 ˚ A ) region was extrapolated to longer wavelength. The BALnicity indices quotedin Table 1 are a consequence of the method of spectral fitting described above and othermethods might produce different results. However, the exact BALnicity index is not criticalto this discussion, the spectra in Figure 1 clearly show BALs and that is the essential pointof relevance here. In columns 10 and 11 of Table 1, we compute the BALnicity indices (BIs)for the high ionization CIV trough and the low ionization Al III trough, respectively, per themethods of Weymann et al. (1991); BI > IRAF is the Image Reduction and Analysis Facility, written and supported by the IRAF programminggroup at the National Optical Astronomy Observatories (NOAO) in Tucson, Arizona.
Table 1: X-ray Spectral Models and Balnicity Indices of Soft X-ray Loud BAL Quasars
QSO z L x Γ Γ N H χ /dof α os α ox BI CIV BI AlIII Type a χ Cash b (km/s) (km/s)0232 2.04 0 .
99 2 .
61 2 .
55 2.73 2.2/6 1 . +0 . − . . +0 . − . ± . ± . ± . .
27 2 .
89 2 .
62 1.07 0.9/4 1 . +0 . − . . +0 . − . ± . ± . ± . . .
62 3 .
03 1.76 3.9/6 1 . +0 . − . . +0 . − . ± . ± . ± . a X-ray luminosity derived from an absorbed powerlaw model in the QSO rest frame from 2.0 -8.0 keV, inunits of 10 ergs/s b Neutral hydrogen absorption column used in the model in units of 10 cm −
3. Source Identification
Since these objects are so exceptional and the ROSAT error circles are so large, it is im-portant to verify the identifications with the SDSS AGN by Suchkov et al (2006). In Figure2, we present SDSS “finding charts” for the 3 ROSAT detections. All of the SDSS sourceswithin the 90% confidence source error radius are labeled. Verification of the identificationsin Suchkov et al (2006) is predicated on the individual probabilities that each source withinthe confidence source error radius is not the ROSAT source: the last column in Table 2. Inorder to arrive at these probabilities, the table first lists the source classification (column 2)for each of the sources in the three ROSAT confidence error radii in Figure 2. The sourceclassifications in Table 2 are based on SDSS photometry using the results of Newberg et al(1999); Strateva et al (2001) and spectra when available. The classifications are amended bythe SDSS values of m B and the VLA/FIRST radio flux densities in the next two columns.This allows a comparison of each putative ROSAT source with the appropriate ROSAT“standard candle” that is identified in column 5. Using the source classification and the hy-pothesis that each source is the ROSAT detection, each source can be compared individuallyto the appropriate “standard candle” X-ray properties in columns 6 - 8 in order to assesswhether it is a viable identification. The standard candle values are listed on top and theputative source values are listed below in bold and in parenthesis to clearly distinguish thetwo. For example, consider the F-star that is source 1 in the field of 1WGAJ0232.3+0021 inFigure 2 (row 1 of Table 2). Given the value of M V associated with its stellar classificationand m V from SDSS photometry, one can determine the distance to the star. If one assumesthat this is the ROSAT source then one can compute the intrinsic X-ray luminosity in theROSAT band, log( L X ) = 34 . ± .
20 which is much larger than any of the ROSAT F starsof similar M V in Suchkov et al (2003), log( L X ) = 29 . ± .
50. If the Suchkov et al (2003)data fits a log normal distribution then we can rule out the F-star as the ROSAT sourcewith > . L X ) ofROSAT F-stars beyond 3 σ , does not obey Gaussian statistics (i.e., there are extra outliersin the tail), the F-star can not be ruled out categorically. The M-stars are treated in asimilar manner using the “standard candle” log( L X ) values for M-stars of matched M V fromMarino et al. (2000).Similarly, there is a standard candle for each field galaxy in Figure 3 that is determinedby the galaxy type, the 1.4 GHz flux density and m B . The X-ray fluxes of radio quietgalaxy standard candles are given by Refreiger et al. (1998). These standard candle fluxesare compared to the observed ROSAT fluxes in the 1WGA fields in column 6. The ROSATfluxes in the 1WGA fields are three orders of magnitude too large to be consistent radioquiet galaxy counterparts. The quasar soft X-ray properties have been studied in Laor et al.(1997) by means of α os . The quasar values of α os are compared to the Laor et al. (1997) 7 –Table 2: Probability of False Identification of the Sources in the 1WGA Fields Source Type m B Putative Source ) False ID a Flux b Log L x c α os d source 1 FV7-9 16 ± . < .
99 ROSAT F stars, ... 29 . ± . > . . < M V < . (34 . ± . )source 2 spiral 20 . ± . < .
99 radio quiet 4 . ± . > . (6592) source 3 quasar e . ± . < .
99 radio quiet ... ... 1 . ± . > . z = 0 .
768 unobscured quasar (0 . + . − . ) source 4 BALQSO 19 . ± . < .
99 radio quiet ... ... 1 . ± .
26 0 . (1 . + . − . )1WGAJ1019.7+4502 source 1 M3.5 - M4.5 20 . ± . < .
98 ROSAT M3.5 ... 27 . ± .
05 ... 0 . (31 . ± . )source 2 spiral 22 . ± . < .
98 radio quiet 0 . ± . > . (3217) source 3 M4.5 - M5.5 17 . ± . < .
98 ROSAT M4.5 ... 26 . ± .
61 ... > . (29 . ± . )source 4 spiral 21 . ± . < .
98 radio quiet 2 . ± . > . (3217) source 5 elliptical 22 . ± . < .
98 radio quiet 3 . ± . > . (3217) source 6 BALQSO 19 . ± . < .
98 radio quiet ... ... 1 . ± .
26 0 . (1 . + . − . )1WGAJ1322.4+1410 source 1 FV8-9 19 . ± . < .
91 ROSAT F stars, ... 29 . ± . > . . < M V < . (34 . ± . )source 2 FV6-7 20 . ± . < .
91 ROSAT F stars, ... 29 . ± . > . . < M V < . (34 . ± . )source 3 spiral 21 . ± . < .
91 radio quiet 3 . ± . > . (12046) source 4 elliptical 22 . ± . < .
91 radio quiet 3 . ± . > . (12046) source 5 BALQSO 18 . ± . < .
91 radio quiet ... ... 1 . ± .
26 0 . (1 . + . − . ) a Upper limit on flux density from VLA/FIRST b X-ray flux in the observers frame 0.2 - 2.4 keV (10 − ergs/s cm − ) c Putative source luminosity is computed from χ fit to ROSAT data in figure 1 d α os is defined in Laor et al. (1997) and the characteristic quasar range of values is from that paper e This is not an obscured (red quasar) based on SDSS colors and 2MASS upper limit on flux density that yields, R − K < − .
8. The photometricredshift is based on SDSS colors < .
4. Discussion
In this Letter, we describe 3 BALQSOs that have soft X-ray to UV ratios that arelarger than typical quasars (see the α os values in Table 2). Hence, the designation as softX-ray loud. However, the α ox values are not exceptional for quasars due to the steep X-rayspectral indices (Strateva et al 2005). These are also the first known BALQSOs with X-rayabsorption consistent with pure Galactic absorption. The tentative discovery of soft X-rayloud BALQSOs is completely unexpected based on theoretical treatments of BALQSOs.The conundrum posed by this new class of AGN is how can a BAL region coexist witha powerful source of X-rays, since even a modest flux of X-rays will over-ionize the gasmaking it impossible to form Li-like atoms Murray et al (1995). On a speculative note,a relativistic X-ray jet beamed towards earth and away from the BAL gas is a possibility(Punsly 1999a,b). It is now known that there are BALQSOs with relativistic jets beamedtoward earth and a jet similar to the one in Figure 1 of Ghosh and Punsly (2007) that isradio weak, but X-ray strong would conform to the properties of the BALQSOs in Tables 1and 2. There are extragalactic jets that have X-ray and radio properties in-line with thesessources. Extreme high frequency peaked BL Lac objects such as PG1553+113, at z ≈ & ergs/s, a steep X-ray spectrum andwould have a radio flux at 1.4 GHz of about 0.1 - 0.5 mJy in agreement with the propertiesof the BALQSOs presented here (Osterman et al 2006). REFERENCES
Aldcroft, T.L., Green, P.J., 2003 ApJ
Astronomical Data Analysis Software and Systems V , eds.Jacoby G. and Barnes J., ASP Conf. Series volume 101, p17Brotherton, M. et al 2005, AJ textbf130 2006 9 –Ghosh, K. and Punsly, B. 2007 ApJL , 177.Murray, N. et al 1995, ApJ , 23White N.E., Giommi P. & Angelini L. 1994,http://wgacat.gsfc.nasa.gov/wgacat/wgacat.html