Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://hea-www.harvard.edu/~kim/pap/RedQ_1999.pdf Äàòà èçìåíåíèÿ: Mon Dec 6 20:20:39 2010 Äàòà èíäåêñèðîâàíèÿ: Mon Oct 1 23:48:53 2012 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: http astrokuban.info astrokuban

THE ASTROPHYSICAL JOURNAL, 516 : 9 õ 17, 1999 May 1
( 1999. The American Astronomical Society. All rights reserved. Printed in U.S.A.

RADIO-QUIET RED QUASARS DONG-WOO KIM1
Chungnam National University, Taejon, 305-764, South Korea

AND MARTIN ELVIS
Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, MA 02138 Received 1998 May 11 ; accepted 1998 December 8

ABSTRACT We have performed a successful targeted search for a population of red, radio-quiet, and probably absorbed quasars. Radio-quiet, optically red ROSAT PSPC X-ray sources brighter than 1 ] 10~13 ergs cm~2 s~1 were searched for red (O[E [2.0, O ¹ 20) counterparts in the Automated Plate-measuring Machine catalog of Palomar Sky Survey objects. Of 45 objects for which we obtained adequate followup optical spectroscopy, we have found seven red quasars, ïve with a \ [2. Their redshifts range opt from 0.06 to 0.31, and their luminosities are moderate, lying on the quasar/Seyfert boundary. These red quasars strengthen the case for a radio-quiet population that is the counterpart of the radio-loud red quasars found by Smith & Spinrad and Webster et al. Unidentiïed, fainter sources could increase the fraction of red quasars by up to a factor of 7. For the red quasars found here, the Ha/Hb ratios, optical slope, and X-ray colors all indicate that they are absorbed by A D 2 rather than having intrinsically red spectra. This amount of obscuration seems to hide D1%õ7%Vof quasars at a given observed ÿux or D3%õ20% when their ÿuxes are corrected to their intrinsic values. This size of population is consistent with earlier limits, with predicted values from Comastri et al., and is comparable with the rate found among radio-loud quasars. A large population of more heavily absorbed (A \ 5), fainter quasars equal V in size to the blue population could exist without violating existing upper limits, which is in accord with the Comastri et al. predictions. Subject headings : dust, extinction õ quasars : general õ X-rays : galaxies
1

. INTRODUCTION

Quasars are the canonical "" ultraviolet excess îî objects (Sandage 1965). Yet red quasars have been found in radioselected samples by Smith & Spinrad (1980) and Webster et al. (1995). Webster et al. (1995) proposed that a large fraction, perhaps 80%, of radio-loud quasars might have been hidden as red objects. Moreover, the currently favored explanations for the cosmic X-ray background invoke a population of heavily obscured active galactic nuclei (AGNs) 5 times more common than the unobscured population (Comastri et al. 1995). If the small number of known red quasars really is the "" tip of the iceberg îî of a large, even dominant, quasar population, then the consequences would be interesting : the overall AGN populationõand so the massive black hole populationõmay be 5 times larger than had been thought ; obscured quasars would be a long-lived evolutionary phase (see Sanders et al. 1988), or all quasars may be hidden along 80% of possible lines of sight ; and red quasars may contribute importantly to the cosmic X-ray background. The Webster et al. conclusion is widely disputed. Boyle & di Matteo (1995), Stickel et al. (1996), and Benn et al. (1998) all argue that any missing population must be smaller and perhaps insigniïcant, while Gunn & Shanks (1998) disagree. Here we present an X-rayõbased survey targeted explicitly at red AGNs to ïnd radio-quiet red quasars. Radio-loud red quasars are relatively easy to ïnd, since the radio emission is unaected by absorbing gas or dust and, in low-frequency surveys, usually comes from the large radio lobes that lie well outside any obscuring material in
1 Also at Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, MA 02138.

the host galaxy. An explicit, albeit small-scale, search for radio- and X-rayõloud but optically quiet quasars, which should include reddened quasars, was, however, not successful (Kollgaard et al. 1995). Radio-quiet red quasars are much harder to ïnd, although they might be expected to be much more common. In the normal unreddened population, radio-quiet quasars outnumber radio-loud quasars 10 to 1 (e.g., the Extended Medium Sensitivity Survey, Stocke et al. 1991 ; the Palomar Green catalog [PG], Kellerman et al. 1989). However, most optical quasar surveys are actively biased against ïnding red quasars. Since these surveys primarily search for UVbright objects (e.g., Markarian, Lipovetsky, Markarian, & Stepanian 1987 ; PG, Schmidt & Green 1983 ; the Large Bright Quasar Survey, Hewett, Foltz, & Chaee 1995 ; the Hamburg Quasar Survey, Engels et al. 1998), they are blind to red objects. As a result, optical bounds on how large a population of radio-quiet red quasars might exist are weak. X-ray selection provides a way of selecting red quasars efficiently : hard X-rays (2õ10 keV) penetrate even tens of magnitudes of optical extinction with minimal absorption. Even the lower energy band of ROSAT (0.5õ2.5 keV) is not strongly aected by optical extinction of up to D2 mag.2 Astrophysics might be against us, since although blue quasars are overwhelmingly X-ray loud (Avni & Tananbaum 1986), red ones might be intrinsically X-ray faint. Fortunately, however, radio-loud red quasars are known to be X-ray sources (Bregman et al. 1985 ; Elvis et al. 1994), so this is unlikely to be a problem.
2 For standard Milky Way composition and dust-to-gas ratio (Bohlin, Savage, & Drake 1978 ; Seaton 1979) the PSPC count rate is reduced by a factor of 1/e at A \ 1.7 (N \ 3 ] 1021 atoms cm~2) for a power-law V H photon index of 2.0, all at 0 redshift.

9


10

KIM & ELVIS

Vol. 516

Complete ÿux-limited X-ray surveys have found some red quasars or AGNs (Stocke et al. 1982 ; Kruper & Canizares 1989 ; Puchnarewicz et al. 1996, 1997). In particular the ROSAT International X-Ray/Optical Survey (RIXOS) survey (Puchnarewicz & Mason 1998) has identiïed a small sample of red quasars. As part of a more general search for minority populations of X-ray sources, we have used the ROSAT (Trumper 1983) archive of pointed PSPC (Pfeerman et al. 1987) data to design an efficient search strategy explicitly targeting red quasars. The ROSAT pointed archive provides us with a tenfold increase in the number of X-ray sources from before ROSAT (to about 70,000) and reaches more than 10 times fainter than the ROSAT All-Sky Survey (Voges et al. 1996). By carefully selecting a small number of interesting objects, we can make an efficient search for a radio-quiet population of red quasars. In this paper we ïnd a substantial population of radioquiet red AGNs on the quasar/Seyfert luminosity boundary.
2

. SAMPLE SELECTION

Most ROSAT sources at high Galactic latitude are blue, unobscured AGNs (e.g., Boyle, Wilkes, & Elvis 1997 ; Schmidt et al. 1997). We use this fact to efficiently select against such objects and so isolate any red AGN that may be in the ROSAT archive. As a compendium of the X-ray sources found by the ROSAT PSPC we have used the "" WGACAT îî catalog (which is named after its authors, White, Giommi, & Angelini 1995). This catalog was generated from ROSAT PSPC pointed observations using a sliding cell, detect algorithm. This method is sensitive to ïnding point sources but can also ïnd spurious sources where extended emission is present. WGACAT includes a quality ÿag that notes such dubious detections based on a visual inspection of the ïelds. We have only used X-ray sources with high detection quality in order to exclude the spurious sources. From the

WGACAT catalog we have selected sources by the following criteria : 1. X-ray bright ( f [ 10~13 ergs cm~2 s~1) to allow X follow-up observations with other X-ray telescopes ; 2. well detected with a signal-to-noise ratio greater than 10 and a WGACAT quality ÿag, DQFLAG, greater than 5 ; 3. within r \ 18@ from the detector center to provide good positions ; 4. at high galactic latitude ( o b o [ 20¡) to minimize the fraction of Galactic stars (which are also red) ; 5. not within 2@ of the target position (at which point the source density reaches the background level) to select only random, serendipitous, sources ; 6. north of decl. \[18¡ in order to have two band measurements in the Automated Plate-measuring Machine (APM) catalog (McMahon & Irwin 1992), hence giving an archival optical color ; and 7. unidentiïed, with WGACAT class \9999 and no SIMBAD or NASA Extragalactic Database identiïcation.3 The ïrst six criteria selected 1624 sources. Since these sources were selected purely on their X-ray properties, they form a well-deïned sample from which to study the incidence of minority X-ray populations, including any radioquiet red quasars. Adding the requirement that a source be unidentiïed left 940 X-ray sources that could be examined for having red optical counterparts. We then searched the APM catalog of objects detected on the Palomar Sky Survey for optical counterparts to the unidentiïed X-ray sources. To ïnd counterparts, we used a search radius of 26A, which corresponds to about 95% conïdence for X-ray sources within 18@ of the PSPC detector center (Boyle et al. 1995a). Of these, 881 sources had APM catalog counterparts brighter than the limiting magnitudes O \ 21.5 and E \ 20. (The remaining "" blank îî ïelds are the
3 Although only "" unidentiïed îî sources were selected, one (1WGA J1118.0]4505 ; Table 1) turned out to be a known Seyfert 1 (Bade et al. 1995).

Fig. 1a

Fig. 1b

FIG. 1.õ(a) Optically identiïed EMSS sources in the a -(O[E) plane. Dierent sources are marked by dierent symbols. Our source-selection criteria ox (O[E [ 2, a \ 1.8) are seen as a lower right box. (b) Same as (a) but for our ROSAT samples. The a line shows how a slightly stricter criterion yields a ox ox larger fraction of red quasars.


No. 1, 1999

RADIO-QUIET RED QUASARS

11

subject of another study ; Elvis, Kim, & Nicastro 1999.) Among these 881, many had only O-band magnitudes (suggesting that they are blue), leaving 575 with O[E colors available. The combination of the ROSAT X-ray ÿux and the APM magnitudes allows us to create a rough classiïcation of the X-ray sources in our sample. The two Palomar O (blue) and E (red) magnitudes are close to Johnson B and Cousins R, respectively (Gregg et al. 1996). From the X-ray ÿux and the two optical magnitudes we can construct a two-color diagram of a -(O[E). a is deïned by the power-law index ox ox between 2 keV and 2500 A (Tananbaum et al. 1979 ; Stocke et al. 1991). This diagram allows us to select red objects and then reduce the Galactic stellar population among these red objects by selecting the X-rayõloud population. Here we make use of the observation that in stellar sources the X-ray ÿux for a given optical ÿux is much weaker than in AGNs and clusters (e.g., Maccacaro et al. 1988). Figure 1a shows Einstein Extended Medium Sensitivity Survey (EMSS) X-ray sources (Stocke et al. 1991) in the a -(O[E) plane. ox This plot clearly illustrates the distinction between Galactic stellar sources and extragalactic sources. Based on the EMSS source distribution, we divided the ROSAT -APM sources on the same plane (Fig. 1b). The O[E APM colors are not as accurate as the EMSS values, which are based on CCD photometry. As a result, the spread of observed colors is wider (Fig. 1b), and there will be some blue objects in the red zone and vice versa. To create our list of red quasar candidates, we ïrst excluded the 128 sources with a [ 1.8, because they are likely to be Galactic ox stars. Then we excluded another 360 sources with blue colors, O[E \2, because they are most likely just normal, blue, unobscured AGNs. This results in a ïnal sample of 87 X-ray sources deïned by the lower right corner of Figure 1b in the a -(O[E) plane. Our ox sample of 87 optically red X-rayõloud sources is a mere 0.1% of the D70,000 WGACAT sources. The fraction of X-ray sources that may be red quasars, however, is much larger : D5% of the initial X-ray selected sample, D15% of the unidentiïed sources with APM colors, and D20% of X-ray bright objects with APM colors that will primarily be AGNs. However, other classes of X-ray source than red quasars can inhabit this region of the a -(O[E) plane : for example, ox ïrst-ranked elliptical galaxies in distant clusters of galaxies. Optical spectroscopy is needed to ïnd red quasars. We have taken spectra for 51 of the 87 red quasar candidates, as is described in the next section.
3

spectrograph slit to obtain spectra of both at once. If this the ïrst spectra did not ïnd an AGN, we then observed the next faintest, if present. Since the density of (blue) AGNs at B \ 21 is only D0.005 per error circle (e.g., Zitelli et al. 1992), we expect onlya1in4 chance of AGN coincidence in the 51 spectra, so stopping once an AGN is found will not produce a signiïcant number of false identiïcations. We performed optical spectroscopy with the Multiple Mirror Telescope (MMT) on 1997 March 13õ15, with the Fred Lawrence Whipple Observatory (FLWO) 60A telescope on 1996 November 16õ17 and 1997 February 12õ13, and with the Cerro Tololo Inter-American Observatory (CTIO) 60A telescope on 1997 February 3õ5. We used longslit apertures of 2@@õ3@@ ] 180@@ and gratings with 300 gpm. The spectral resolutions are 6 and 9 A for the MMT and 60A telescopes, respectively. Wavelength coverage is about 3500õ8000 A. We took bias, dome ÿat, and twilight sky frames each night, and the corresponding corrections (bias subtraction, ÿat ïelding, and illumination correction) were applied separately to each night of data. At least two standard stars were observed each night for spectrophotometric calibration. The observing conditions were not photometric, except for the CTIO run, so the absolute calibration is subject to a signiïcant uncertainty. However, relative intensities (such as a line intensity ratio and an optical power-law index) are accurate within 20%, as is conïrmed by multiple observations of the same source. Six sources are of undetermined nature because they are too faint and so gave spectra of too poor a signal-to-noise ratio. 3.2. Classiïcation of Spectra Of the 45 sources observed at a good signal-to-noise ratio, we have identiïed seven red quasars (Table 1). The results for these seven red quasars are presented in this paper. (The full data set will appear elsewhere.) The red quasars are mixed in with 18 stars and a small number of normal blue quasars, narrow emission line galaxies, and elliptical galaxies (Table 1). The elliptical galaxies are likely to be brightest cluster galaxies. We will report on these separately. For the remaining nine X-ray sources, the red optical candidate within the error circle turned out to be a star (mostly late type), but it is not likely that these red stars are the counterparts, because their a values are too large ox for a star (see above ; Maccacaro et al. 1988). The remaining optical candidates are not red, and hence we stopped making further observations. These nine sources and the three blue quasars measure the blue contamination of the sample and should not be considered as part of the list of red X-ray counterparts. The optical spectra of the seven red quasars are shown in Figures 2a and 2b. Broad lines of Ha, Hb, and Mg II are clearly seen in the spectra as well as bright narrow lines (e.g., [O III] j5007), making the AGN character of the objects unambiguous. In Table 2 we tabulate source position, redshift, optical magnitude and color, X-ray ÿux and X-ray

. OBSERVATIONS

3.1. Optical Spectroscopy A typical X-ray error circle contains just 1õ2 optical objects in the APM catalog. Since we have selected against blue counterparts, we began by observing the brightest red counterpart. If two objects were present, we aligned the

TABLE 1 SUMMARY OF OPTICAL IDENTIFICATIONS OF RED X-RAY SOURCE COUNTERPARTS Measurement 1.8 [a [ 1.6 (O º 19) ...... ox a \ 1.6 (O \ 19) ............ ox Total ........................ Total 22 29 51 Red Quasars 1 6 7 Too Faint 0 6 6 NLXG 0 2 2 Elliptical Galaxies 1 4 5 M Stars 17 1 18 Other Stars 1 0 1 Blue Quasars 0 3 3 Not Red 2 7 9


12

KIM & ELVIS

Vol. 516

Fig. 2a

Fig. 2b

FIG. 2.õ(a) Observed spectra of the ïve radio-quiet red quasars (O[E [ 2 mag, a [ 2.0). Broad lines such as Ha, Hb, and Mg II are clearly seen in the opt spectra as well as bright narrow lines (e.g., [O III] j5007). Those strong lines are marked in the ïgures. Steep continuum slopes and very weak Hb line strengths indicate a signiïcant amount of dust extinction (A [ 2 mag), in contrast to (b), two quasar spectra with a \ 2.0 and relatively strong Hb lines. V opt

colors, and a , as well as osets between the optical and ox X-ray positions. In Figure 1b each class of identiïed sources is plotted in the a -(O[E) plane. The distribution of these sources can be ox compared with the EMSS sources in Figure 1a, conïrming that most X-ray sources with low a are indeed ox quasars or galaxy clusters. The selection technique ïnds 7/51 red quasars, i.e., 8% efficiency. A slightly stricter criterion, a \ 1.6 (instead of 1.8), would have selected red ox

quasars more efficiently (Table 2 ; Fig. 1b) : only one M star (instead of 19 stars) would have been present, with only one red quasar lost, i.e., 6/29, a little over 20%. Of the initial 87, 71% (62) remain when a \ 1.6 is the boundary. ox 3.3. Observed Optical Properties of the Red Quasars We measured the optical continuum slopes by ïtting a power law to the continuum spectra after excluding the strong emission lines (Table 3). Although all the sources

TABLE 2 BASIC PROPERTIES OF RED QUASARS Name Right Anglea Declinationa Osetb opt 11.3 5.0 4.3 7.0 4.0 (a z \ [2.0) 0.0647 0.3120 0.1514 0.2748 0.1467 opt [ [2.0) 0.1829 0.1151 2.76 15.75 1.290 0.926 1.295 1.263 18.27 16.33 2.72 2.44 1.515 1.522 7.74 2.42 1.62 1.65 1.52 1.459 0.860 0.959 0.859 0.535 0.750 1.711 1.512 1.417 1.040 16.38 20.55 19.16 21.38 19.33 2.44 2.35 2.08 2.28 2.33 1.636 1.180 1.466 1.114 1.451 fc X a soft d a hard e O O[E a ox

J2255.5]0536 J1234.3]2614 J1218.1]2956 J0909.7]4302 J1143.6]5521

...... ...... ...... ...... ......

22 12 12 09 11

55 34 18 09 43

31.0 21.8 07.1 43.6 35.5

05 26 29 43 55

36 13 55 02 20

01 28 21 47 21

[ 0.9 [a J1051.4]3358 ...... J1142.6]4624 ...... a b c d e 10 51 28.3 11 42 41.2 33 58 04 46 24 21 8.0 3.1

Optical coordinate in Equinox J2000. Dierence of X-ray and optical positions in arcseconds. In units of 10~13 ergs~1 cm~2. X-ray spectral index in 0.1õ 0.8 keV. X-ray spectral index in 0.8õ2.0 keV.


No. 1, 1999

RADIO-QUIET RED QUASARS
TABLE 3 LINE, CONTINUUM PROPERTIES, AND LUMINOSITIES OF RED QUASARS ADD D4000 ? Ha Name a opt a FWHM (a J2255.5]0536 J1234.3]2614 J1218.1]2956 J0909.7]4302 J1143.6]5521 ...... ...... ...... ...... ...... [2.39 [2.64 [2.36 [2.43 [2.11 5962 7184e 5011 2178 3223 opt fc \ [2.0) 29.8 5.1e 5.2 2.3 5.1 0.54 2.70e 0.56 0.91 0.51 [21.6 [21.1 [20.8 [20.0 [20.5 [24.1 [23.5 [22.9 [22.3 [22.9 1.02 8.19 1.21 4.29 1.07 Ld M(O) M(E) L (X) 43 b

13

J1051.4]3358 ...... J1142.6]4624 ...... a b c d e f Optical In units In units In units Mg II. Hb.

[0.93 [1.41

( [ 0.9 [a [ [2.0) opt 1797f 7.1f 7.10f 3104 81.3 4.90

[22.1 [23.0

[24.8 [25.4

3.07 6.75

spectral index,f P laopt . l of 1043 ergs s~1. of 10~15 ergs cm~2 s~1. of 1043 ergs s~1.

were selected based on a red O[E color, in some cases the observed optical continuum shape is relatively ÿat. This is because of both line emission contributing to the blue and/or red bands and uncertainties on the O and E magnitudes, particularly when the object is faint (M. Irwin 1996, private communication). The power-law index (F D laopt ) l ranges from [0.9 to [2.6. A steep optical continuum, a \ [2, is found in ïve of the seven red quasars, while opt even the remaining two, intermediate red quasars are redder ([1.5 \a \ [1) than is found for UV-excessõselected opt quasars ([0.2 ^ 0.8 ; Neugebauer et al. 1987). To illustrate the spectral dierences between these two groups, we display the spectra separately in Figures 2a (steep) and 2b (intermediate). In addition to the dierence in continuum shape, these two groups also dier in their Hb line strengths (see Figs. 2a and 2b). The group with the steep optical continuum have only weak Hb lines or no detection, whereas the group with a relatively ÿat continuum have stronger Hb lines. Since the ratio of Ha to Hb is sensitive to optical extinction, this suggests more reddening in the steep slope group than the intermediate slope group (Table 3), which is in accord with the optical continuum slopes. To quantify this eect for those quasars with no detected Hb line, we estimated its upper limit using a simple method that assumes a box proïle with a base equal to 3000 km s~1 (the mean FWHM of detected Hb lines) and a height equal to 3 times the ÿuctuation noise on the continuum. This is a conservative measurement, because the peak of a Gaussian proïle would be more easily detected than the ÿat top of a box proïle, particularly when the line width is considerably larger than the spectral resolution. Monte Carlo simulations using a Gaussian line proïle assuming Poisson statistics show that the box proïle overestimates the upper limit by up to 50% for the adopted line width, while it reproduces consistent results when the line width is comparable with the spectral resolution. For the two objects whose optical spectra do not cover the Ha line, we have instead used the [O III]/Hb ratio as a measure of relative Hb strength. For all ïve quasars with a \ [2, the Ha/Hb ratios are greater than 5, while theopt III]/Hb[ 0.8. The line [O

ratios of the two remaining quasars are smaller (Table 2), which is consistent with less reddening in intermediateslope objects. None of the characteristic galaxian stellar absorption features4 are seen in our spectra. Most strikingly, no 4000 A break is seen in any of the ïve red quasars for which our spectra cover that region, including all of the steep-slope group. Typical values of D(4000)5 are 1õ1.2, as is expected from the measured optical slopes. These compare with values of 2 ^ 0.2 for normal E and S0 galaxies Dressler & Shectman (1987). Hence any starlight continuum contribution to the red quasar continuum must be minor. 3.4. X-Ray Colors of the Red Quasars To determine the rough X-ray spectral properties of the seven red quasars, we ïrst double-checked in the PSPC images that the sources were cleanly separated from any confusing sources, then measured their X-ray hardness (HR \ H/M) and softness (SR \ S/M) ratios based on the count rates in the standard ROSAT PSPC bands : soft, S (0.1õ 0.4 keV) ; medium, M (0.4õ 0.86 keV) ; and hard, H (0.87õ2 keV). These ratios are then converted to eective X-ray spectral indices, a and a (Table 2), to correct for the variable soft hard Galactic line-of-sight absorption and the energy-dependent point-spread function. (These are not physical slopes but should be considered analogous to U[B and B[V colors ; see Fiore et al. 1998 for a detailed discussion of the estimation and usage of eective X-ray spectral indices.) Due to the low signal-to-noise ratio of X-ray data, individual spectral indices are not reliable. However, the locus of their colors forms a useful indicator of global X-ray properties. We compare the colors of the red quasars with those of normal radio-quiet quasars in Figure 3. The large ïlled symbols are the red quasars reported here, while the cloud of small dots represents radio-quiet quasars from the sample of Fiore et al. (1998). On average, the red quasars have smaller a than a , indicating a cuto spectral soft hard shape. [The line pairs around the periphery of the ïgure
4 CH G-band j4304, Mg I j5175, Ca]Fe j5269, Na jj5890, 5896. 5 D(4000) \ F (4050õ 4250 A)/F (3750õ3950 A), Dressler & Shectman l l (1987).


14

KIM & ELVIS

Vol. 516

FIG. 3.õEective soft and hard ROSAT PSPC X-ray spectral indices (see text) of the red quasars (large ïlled symbols) compared with normal radio-quiet quasars (small dots ; Fiore et al. 1998). Steepõoptical spectrum (a\ [2) red quasars are shown as circles, and intermediateõoptical spectrum ( [ 1.5 \a\ [1) red quasars are shown as triangles. The line pairs around the periphery of the ïgure show outline spectral shapes for their locations in the (a , a ) plane. hard hard

luminosity of the sources is likely to be signiïcantly higher, depending on the amount of absorption (the lower limit on the extinction is 2õ3 mag in O). This places them at [ 25 \ M \ [22, which is within the quasar regime. Similarly, O the observed X-ray luminosity ranges from 1.3 ] 1043 to 1.2 ] 1044 ergs s~1, while the intrinsic luminosities are likely to be higher by about a factor of 2, depending on the intrinsic spectrum and the amount of absorption present. The precise allocation of these AGNs as quasars or Seyferts is not fundamentally very interesting, since this is just a conveniently chosen value on a continuous luminosity scale. For simplicity we will refer to them as red quasars for the rest of this paper. It is, however, interesting to understand why much more or much less luminous AGNs were not found in the two red quasar searches. Is this a selection eect, or is there a physical preference for red objects to cluster in a limited range of luminosity ? We will return to this question later (° 4.3). The existence of a red quasar population immediately raises important questions. What makes them red compared with usual blue quasars ? How common are they, particularly once allowance is made for their reduced ÿux due to probable obscuration ? If they are absorbed, by how much ? Where is the absorbing dusty material ? Might a larger, more obscured, AGN population exist ? 4.1. How Common are Red Quasars ? We can address the relative numbers of red quasars in a rough way. The population, although a minority, is quite common. We found seven red quasars out of 45 candidates for which our spectroscopy was adequate to produce a classiïcation. Assuming that the 45 were a random subsample of the original sample of 87, that sample would produce 14 red quasars. This is 2.4% of 575 sources with optical colors available. At our ÿux limit, Stocke et al. (1991) ïnd that 51% of all X-ray sources are AGNs. So a minimum of 13.5/288 (4.7%) AGNs in our unidentiïed sample are radioquiet red quasars (with O \ 20) at this soft X-ray ÿux level. The Boyle & di Matteo (1995) upper limit of 9% of Cambridge ROSAT Serendipity Survey (CRSS) X-ray sources being red quasars is consistent with the minority population of moderately obscured (A \ 2) quasars we have found V here. Additional red quasars could be hidden in our sample. Our sample of 45 classiïed spectra is not a random subsample of the candidate list. Figure 1b shows that we preferentially selected objects with a [ 1.3, i.e., the brighter objects (with B \ 21.5). If the ox objects for which we six attempted to get spectra turn out to be red quasars, then the fraction of red quasars among the PSPC AGNs could be almost double our ïrst estimate. There are a further 36 red candidates for which no spectroscopy was attempted. So the true occurrence rate of red quasars is uncertain by a factor of 7. To ïnd the fraction of red quasars among all the AGNs in our X-ray, ÿux-limited sample, we must allow for the 165õ300 blue AGNs in the identiïed sample, so 1%õ7% of the whole soft X-ray AGN population is red. If the red AGNs are obscured, then the intrinsic rate of occurrence of red quasars has to be calculated relative to their unreddened parent population. Obscuration by A \ 2 reduces their unobscured X-ray ÿuxes by a factor of V 2. D Since higher ÿux AGNs are rarer (the X-rayõselected AGN

show outline spectral shapes for their locations in the (a , soft a ) plane.] This is consistent with their having the moderhard ate X-ray absorption expected from their optical properties (Fiore et al. 1998).
4.

DISCUSSION

These observations show that a population of red AGNs can be extracted efficiently from the ROSAT pointed archive. Moreover, the red AGNs we ïnd are radio quiet. None of them is a radio source in the NRAO VLA Radio Sky Survey ( f \ 2.5 mJy ; Condon et al. 1998), 1.4GHz implying R \ log [ f (opt)/f (5 GHz)] \ 2.0, compared with 2 \ R \ 5L for radio-loud quasars (Wilkes & Elvis 1987). L The agreement of X-ray colors, optical continuum slope, and Ha/Hb ratios with the same value of obscuring dust and gas (A D 1õ2) argues for their being dust reddened V objects rather than instrinsically red continua. Puchnarewicz & Mason (1988) discuss a similar population of 14 candidate red, a [ 2, AGNs derived from the opt RIXOS sample, which extends to several times fainter X-ray ÿuxes. The RIXOS sample was selected from the ROSAT pointed archive based on X-ray ÿux alone. (There is one object in common between the two samples.) Two of the red RIXOS AGNs may be intrinsically red, and three have clear reddening. So there is currently a total of eight reddened AGNs available from ROSAT . The red AGNs, both from this sample and from RIXOS, are borderline quasar/Seyfert objects. The observed optical luminosities of the red AGNs are modest, lying at the high end of the traditional Seyfert luminosity range (M [ [23 B mag ; Veron-Cetty & Veron 1984 ; Schmidt & Green 1983) : from M \[20 to [22 mag.6 The dereddened, intrinsic O
6 We have used H \ 50 km s~1 Mpc~1 and q \ 0. 0 0


No. 1, 1999

RADIO-QUIET RED QUASARS

15

log N/log S relation has a slope of [3/2 in this ÿux range ; Hasinger 1995 ; Della Cecca et al. 1992), the red population forms a larger fraction of this population, D3%õ20%. Using the fraction of the sources at the intrinsic ÿux of the red quasars allows us to compare our result with samples unaected by obscuration and with model predictions. Comparisons with the observed frequency of radioloud red quasars with broad emission lines found by Smith & Spinrad (1980 ; 178 MHz 3CR) and Stickel et al. (1996 ; 5 GHz 1 Jy sample) and with the Comastri et al. (1995) predictions are straightforward. Red quasars are found in 15% of the 3CR source sample and 6%õ20% of the 1 Jy (Carilli et al. 1998) sample. These are comparable with the 3%-20% of the "" intrinsic ÿux îî X-ray population that we ïnd, suggesting that the radioloud and radio-quiet quasar populations have similarly sized populations of moderately obscured quasars. The obscuration in the 3CR red quasars is also about A \ 2 V (Elvis et al. 1994 ; Economou et al. 1995 ; Rawlings et al. 1995). For four objects in the 1 Jy sample, Carilli et al. (1998) estimate lower limits on A from 2 to 5 based on extrapolating the radio-infrared V index to the optical. However, since such steeply rising slopes are not known among unobscured quasars, these limits are likely to be too large, and values comparable with the 3CR estimates are probably acceptable. For AGNs with N \ 1022 cm~2, the Comastri et al. H (1995) model predicts that 26% will have 1021 \ N \ 1022 H cm~2 (A D 2), which is somewhat larger but comparable Vnumbers found here. Comastri et al. predict far with the larger numbers of more obscured objects. 4.2. More Obscured Objects More obscured objects may exist. Puchnarewicz & Mason (1998) ïnd several objects with steeper optical continua, and Figure 1b shows several redder candidates and many more X-ray loud candidates with no optical spectra to date. Webster et al. (1995) suggested that A \ 5 may be V typical of their red objects, giving their putative ROSAT counterparts in our sample V \ 23 õ25, which is below the Palomar Sky Survey limit. A \ 5 corresponds to a column V of 9 ] 1021 atoms cm~2, which would reduce ROSAT PSPC count rates to 15% of their unobscured values. These objects would then be hidden as a 6% minority among the more common, lower luminosity, unabsorbed quasars if they had the same unobscured space density as normal blue quasars. Some 9% of our initial X-rayõselected sample are "" blank ïeld îî objects, i.e., have no counterpart on the Palomar Sky Surveys. A fraction of these could be more heavily obscured red quasars. Boyle & di Matteo (1995) ïnd that the CRSS sample could be missing no more than 9% (at the observed ÿux) in red quasars. Subtracting the 1% of moderately obscured quasars we have identiïed still leaves 8% that could be highly obscured. Hence the CRSS result is, perhaps surprisingly, consistent with a sizeable, heavily obscured population. The Comastri et al. (1995) model predicts a comparable population of AGNs with N D 1022 H cm~2, which is 1.1 times larger than the unobscured population. In our sample the occurrence of red quasars appears to be 4 times higher for O [ 19 mag (4/12) than for O \ 19 mag (3/32 ; Table 2), although the number of sources is small. Such a trend is expected if the quasars are heavily absorbed. An increase in N from 3 ] 1021 to 1 ] 1022 H

cm~2 cuts the ROSAT ÿux by a factor of 2.4 but reddens the V band by 3.8 mag (a factor 33). So the optically fainter sources might well be redder. However, the RIXOS red quasars (Puchnarewicz & Mason 1998) show no correlation of optical slope with m : the three steepest slopes are all in V the brighter half of the sample of 14. Gunn & Shanks (1998) have pointed out that while redshifting the ultraviolet into the optical increases the eects of reddening, the corresponding shift of hard X-rays into the soft ROSAT band decreases the eectiveness of reddening. To estimate an accurate fraction of this potential hidden population needs a larger sample, including more absorbed, fainter objects. At even larger column densities (1023õ1024 cm~2) the Comastri et al. (1995) model predicts nearly 4 times the unobscured population. More heavily obscured quasars could be found in hard X-ray surveys from ASCA (Ueda et al. 1998) and the Beppo-SAX HELLAS survey (Fiore et al. 1999). 4.3. L imited L uminosity Range It is striking that while ROSAT surveys that are deïned simply by an X-ray ÿux limit ïnd AGNs spanning over 3 decades in X-ray luminosity (e.g., CRSS, Boyle et al. 1997 ; RIXOS, Puchnarewicz et al. 1996), both RIXOS and this survey ïnd red quasars in only 1 decade of luminosity, and this decade is the lowest one in which RIXOS and CRSS AGNs are found. A two-tail Kolmogorov-Smirnov test shows that the chance that red and nonred AGNs from RIXOS come from the same luminosity distribution is only D2%. This suggests that predominantly lower luminosity AGNs are obscured. (Note that the observed amount of obscuration only decreases the observed X-ray luminosity by a factor of D2, and so does not itself cause the low observed luminosities.) Similar suggestions have been made before : Lawrence & Elvis (1982) found that only AGNs below L D 1044 ergs X s~1 (2õ10 keV) showed obscuration. Occasional examples of highly obscured type 2 (i.e., narrow-line) quasars have been reported (Stocke et al. 1982 ; Almaini et al. 1995 ; Shanks et al. 1995), and careful searches have found broad Ha in most cases (Halpern, Eracleous, & Forster 1998), making them similar to the red quasars found here. Searches among the fainter objects in our sample and searches at higher energies (e.g., the Beppo-SAX HELLAS survey ; Fiore et al. 1999) will be more eective at ïnding a highluminosity red quasar population. The Comastri et al. X-ray background models assume luminosity functions for the obscured objects that are identical to those of the unobscured objects except for normalization and so predicts high-luminosity red quasars. If instead obscured AGNs occur preferentially at low luminosity, this will substantially aect the model predictions. We would, for example, expect the obscured population to be more numerous and at lower redshift. If there is a real deïcit of high-luminosity red quasars, then one possibility to explain this lack might have been that as an AGN became more luminous, the continuum ionized the obscuring medium, rendering it transparent to X-rays. However, ionized absorbers are also more common at lower luminosities (Laor et al. 1994). So most likely, highluminosity AGNs have fewer lines of sight with intervening material regardless of ionization state. (Interestingly, this is in the same sense as the Baldwin eect : that higher luminosity quasars have weaker C III] j1909 emission lines.) Any


16

KIM & ELVIS

Vol. 516

physical model of a quasar would need to explain this dierence. 4.4. Physical Properties We can say only a little about the physical properties of the red quasars from this data. The consistency of the optical reddening indicators with the X-ray colors suggests that the same obscuring material covers both emitting regions and that it lies outside the broad emission line region. Smith & Spinrad (1980) suggested that the redness of the red 3CR quasars is intrinsic to the continuum emission process based on the lack of an absorption feature at j \ 2200 A, which is a typical characteristic of Milky Way dust (e.g., Bless & Savage 1972). The detection of 21 cm H I absorption toward a large fraction of the 1 Jy (Stickel et al. 1996) red quasars (Carilli et al. 1998) argues for dust reddening in those objects. Since our sample of radio-quiet quasars is relatively nearby (with redshifts up to 0.3), we cannot check for this feature directly. However, this explanation is in contrast to the observed Balmer decrement and the X-ray colors. It is possible that the reddened 3CR sources contain dusty, ionized absorbers, as is seen in 3C 212 (Mathur 1994 ; Elvis et al. 1994) and IRAS 17020]4544 (Komossa & Bade 1998), where the dust composition may dier depending on, for example, the quasar continuum shape. Ultraviolet observations are needed to investigate the j \ 2200 A feature but are probably infeasible at present. 4.5. Other Red AGNs Some previous studies have considered red AGN-like objects in X-ray surveys. The RIXOS survey (Puchnarewicz & Mason 1998) found that 9% (14/160) of their AGNs were red. However, only three of the RIXOS sources have Balmer decrements that require reddening, so the true occurrence rate of reddened objects may be similar to what we have found. The fainter ÿux limit of the RIXOS survey may render more absorbed objects visible. Certainly, the steeper optical slopes ([2.5 [a [ [4.0) of half the opt RIXOS sample suggest greater reddening. Kruper & Canizares (1989) studied red AGNs in Einstein X-rayõselected samples and indirectly concluded that these are red because of the presence of host galaxies. Benn et al. (1998) arrive at a similar conclusion for low-frequency, selected radio-loud quasars. However, no galaxian starlight features are seen in our spectra, and the host galaxy cannot explain the observed Balmer decrements in our sample. The Kruper & Canizares objects are not as red as our samples, having B[I \ 1.5 õ2.5 mag. If we take R[I to be 0.5õ1.0 mag (this is the R[I range of the samples in their Table 2), B[R would be less than 2.0, our deïning threshold. In fact, none of their objects with measured R magnitudes exceed B[R \ 2.0. The narrow-line X-ray galaxies (NLXGs) found plentifully in deep ROSAT surveys (e.g., Boyle et al. 1995a) are also normally assumed to be obscured AGNs (e.g., Hasinger et al. 1998 ; Schachter et al. 1998). X-rayõabsorbed NLXGs could contribute signiïcantly to the cosmic X-ray background if they are more common at fainter X-ray ÿux levels, as was suggested by McHardy et al. (1998 ; see also Hasinger et al. 1998 for a cautionary note). Although they have similar X-ray luminosities to the NLXGs, the X-rayõ selected red quasars are not simply the same population,

however. The red quasars have the normal, broad optical emission lines of quasars, while NLXGs have either none or only extremely weak ones (Boyle et al. 1995b ; Figs. 2a and 2b). Moreover, NLXGs usually exhibit blue optical continua (for example, O[E \ 2 for ïve out of six NLXGs in CRSS ; Boyle et al. 1997), while red quasars have red optical continua (O[E [ 2). Further X-ray and optical study of these objects may let us understand whether they are two separate populations or are related by, e.g., special viewing geometry or scattering of a blue continuum. The Palomar survey of the nuclei of bright galaxies (Ho, Filippenko, & Sargent 1997) is based on a sample of galaxies selected for their nonnuclear properties and so is less biased against ïnding red AGNs than most other optical search methods. The Palomar AGNs are low-luminosity AGNs, allowing us to see whether the high incidence of red AGNs at lower luminosities continues to increase at even lower values. The Palomar survey ïnds that 18% (8/44) of Seyfert nuclei have A [ 2 based on their narrow-line Balmer decrements (HoVet al. 1997). However, almost all of these are LINERs or type 2 Seyferts, which may have large reddening toward the broad-line region. Only one Seyfert (NGC 7479) shows any evidence for a broad-line component. Broad emission lines are extremely hard to detect at these ÿux levels, but the suggestion is that the middling luminosities toward the quasar/Seyfert borderline are particularly prone to moderate obscuration.
5

. CONCLUSIONS

Radio-quiet red quasars can be found in substantial numbers. They comprise at least 1%, and potentially 7%, of the soft X-ray population in a ÿux limited survey. Correcting the X-ray ÿuxes to their intrinsic values puts them among brighter AGNs, where they form 3% of the population. Allowing for blank ïeld sources, as much as 20% of ROSAT selected quasars may be red at a given unobscured ÿux. The size of this population is consistent with previous upper limits, with the Comastri et al. (1995) model for the X-ray background, and with the size of the radio-loud 3CR and 1 Jy red quasar populations. Red quasars seem to be preferentially lower luminosity objects on the quasar/ Seyfert borderline but not at higher or lower luminosities. Such a bias against obscured high-luminosity objects would aect X-ray background estimates for this population and would need explaining in a physical model of quasars. We stress that the quasars we ïnd have broad optical lines. They are not NLXGs, which by contrast have predominantly narrow optical permitted lines and blue continua. The optical slopes, Ha/Hb ratios, and X-ray colors are all consistent, with reddening by A D 2 assuming standard V Milky Way dust properties. So the same obscuring material probably covers each of the emitting regions. A signiïcant population of more highly obscured (A \ V 5) quasars could well exist and be consistent with the results here, with earlier ROSAT limits, and would be as predicted by the Comastri et al. (1995) model. Hard X-ray surveys will soon settle the question of the size of any such population. Using a minor reïnement of the technique presented here, red quasars can be found with high (20%) efficiency in the ROSAT data. We thank L. Angelini for helping us to use WGACAT and M. Irwin for use of the APM online catalog. We also


No. 1, 1999

RADIO-QUIET RED QUASARS

17

thank F. Fiore and F. Nicastro for providing their program to calculate X-ray spectral indices and F. Fiore once again for supplying the X-ray slopes of the radio-quiet sample in Figure 3. The support by the FLWO, MMT, and CTIO stas in setting up and operating various instruments were

invaluable to this study. The HEASARC/GSFC PIMMS program was used to calculate PSPC count rates. This work was supported by NASA grants NAG5-3066 (ADP), NAG5-6078 (LTSA), and NASA contract NAS8-39073 (ASC).

REFERENCES Almaini, O., Boyle, B. J., Griffiths, R. E., Shanks, T., Stewart, G. C., & Kruper, J. S., & Canizares, C. R. 1989, ApJ, 343, 66 Georgantopoulos, I. 1995, MNRAS, 277, L31 Laor, A., Fiore, F., Elvis, M., Wilkes, B. J., & McDowell, J. C. 1994, ApJ, Avni, Y., & Tananbaum, H. 1986, ApJ, 305, 83 435, 611 Bade, N., Fink, H. H., Engels, D., Voges, W., Hagen, H.-J., Wisotzki, L., & Lawrence, A., & Elvis, M. 1982, ApJ, 256, 410 Reimers, D. 1995, A&AS, 110, 469 Lipovetzky, V. A., Markarian, B. E., & Stepanian, J. A. 1987, in Obser Benn, C. R., Vigotti, M., Carballo, R., Gonzalez-Serrano, J. I., & Sanchez, vational Evidence of Activity in Galaxies, ed. E. Ye. Khachikian, K. J. S. F. 1998, MNRAS, 295, 451 Fricke, & J. Melnick (Dordrecht : Reidel), 17 Bless, R. C., & Savage, B. D. 1972, ApJ, 171, 293 Maccacaro, T., Gioia, I. M., Wolter, A., Zamorani, G., & Stocke, J. T. 1988, Bohlin, R. C., Savage, B. D., & Drake, J. F. 1978, ApJ, 224, 132 ApJ, 326, 680 Boyle, B. J., & di Matteo, T. 1995, MNRAS, 277, L63 Mathur, S. 1994, ApJ, 431, L75 Boyle, B. J., McMahon, R. G., Wilkes, B. J., & Elvis, M. 1995a, MNRAS, McHardy, I. M., et al. 1998, MNRAS, 295, 641 285, 511 McMahon, R. G., & Irwin, M. J. 1992, in Digitized Optical Sky Surveys, õõõ. 1995b, MNRAS, 276, 315 ed. H. T. MacGillivray & E. B. Thomson (Dordrecht : Kluwer), 417 Boyle, B. J., Wilkes, B. J., & Elvis, M. 1997, MNRAS, 285, 511 Neugebauer, G., Green, R. F., Matthews, K., Schmidt, M., Soifer, B. T., & Bregman, J. N., Glassgold, A. E., Huggins, P. J., & Kinney, A. L. 1985, ApJ, Bennet, J. 1987, ApJS, 63, 615 291, 505 Pfeermann, E., et al. 1987, Proc. SPIE, 733, 519 Carilli, C. L., Menten, K. M., Reid, M. J., Rupen, M. P., & Yun, M. S. 1998, Puchnarewicz, E. M., et al. 1996, MNRAS, 281, 1243 ApJ, 494, 175 õõõ. 1997, MNRAS, 291, 177 Comastri, A., Setti, G., Zamorani, G., & Hasinger, G. 1995, A&A, 296, 1 Puchnarewicz, E. M., & Mason, K. O. 1988, MNRAS, 293, 243 Condon, J. J., Cotton, W. D., Greisen, E. W., Yin, Q. F., Perley, R. A., Rawlings, S., Lacy, M., Sivia, D. S., & Eales, S. A. 1995, MNRAS, 274, 428 Taylor, G. B., & Broderick, J. J. 1998, AJ, 115, 1693 Sandage, A. R. 1965, ApJ, 141, 1560 Della Ceca, R., Maccacaro, T., Gioia, I. M., Wolter, A., & Stocke, J. T. Sanders, D. B., et al. 1988, ApJ, 325, 74 1992, ApJ, 389, 491 Schachter, J. S., Fiore, F., Elvis, M., Mathur, S., Wilson, A. S., Morse, J. A., Dressler, A., & Shectman, S. A. 1987, AJ, 94, 899 Awaki, H., & Iwasawa, K. 1998, ApJL, 503, L123 Economou, F., Lawrence, A., Ward, M. J., & Blanco, P. R. 1995, MNRAS, Schmidt, M., et al. 1997, A&A, 329, 495 272, L5 Schmidt, M., & Green, R. F. 1983, ApJ, 269, 352 Elvis, M., Fiore, F., Mathur, S., & Wilkes, B. J. 1994, ApJ, 425, 103 Seaton, M. J. 1979, MNRAS, 187, 785 Elvis, M., Kim, D.-W., & Nicastro, F. 1999, in preparation Shanks, T., Almaini, O., Boyle, B. J., Done, C., Georgantopoulos, I., GrifEngels, D., Hagen, H.-J., Cordin, L., Koehler, S., Wisotzki, L., & Reimers, ïths, R. E., Rawling, S. J., & Stewart, G. C. 1995, Spectrum : Newsletter D. 1998, A&AS, 128, 507 of the Royal Obs., 7, 7 Fiore, F., Elvis, M., Giommi, P., & Padovani, P. 1998, ApJ, 492, 79 Smith, H. E., & Spinrad, H. 1980, ApJ, 236, 419 Fiore, F., LaFranca, F., Giommi, P., Elvis, M., Comastri, A., Matt, G., & Stickel, M., Rieke, G. H., Kuhr, H., & Rieke, M. J. 1996, ApJ, 468, 556 Molendi, S. 1999, in preparation Stocke, J. T., Liebert, J., Maccacaro, T., Griffiths, R. E., & Steiner, J. E. Gregg, M. D., Becker, R. H., White, R. L., Helfand, D. J., McMahon, R. G., 1982, ApJ, 252, 69 & Hook, I. M. 1996, AJ, 112, 407 Stocke, J. T., Morris, S. L., Gioia, I. M., Maccacaro, T., Schild, R., Wolter, Gunn, K. F., & Shanks, T. 1998, Astron. Nachr., 319, 66 A., Fleming, T. A., & Henry, J. P. 1991, ApJS, 76, 813 Halpern, J. P., Eracleous, M., & Forster, K. 1998, ApJ, 501, 103 Tananbaum, H., et al. 1979, ApJ, 234, L9 Hasinger, G. 1995, in MPE Rep. 263, Rontgenstrahlung from the Universe, Trumper, J. 1983, Adv. Space Res., 2 : 4, 241 ed. H. U. Zimmermann, J. Trumper, & H. Yorke (Garching : MPE), 291 Ueda, Y., et al. 1998, Nature, 391, 866 Hasinger, G., Burg, R., Giacconi, R., Schmidt, M., Trumper, J., & ZamoVeron-Cetty, M.-P., & Veron, P. 1984, ESO Special Rep. 1 rani, G. 1998, A&A, 329, 482 Voges, W., et al. 1996, IAU Circ., 6420, 2 Hewett, P. C., Foltz, C. B., & Chaee, F. H. 1995, AJ, 109, 1498 Webster, R. L., Francis, P. J., Peterson, B. A., Drinkwater, M. J., & Mascl, Ho, L. C., Filippenko, A., & Sargent, W. L. W. 1997, ApJS, 112, 315 F. J. 1995, Nature, 375, 469 Kellerman, K. I., Sramek, R., Schmidt, M., Shaer, D. B., & Green, R. 1989, White, N., Giommi, P., & Angelini, L. 1995, WGACAT, http :// AJ, 98, 1195 heasarc.gsfc.nasa.gov/W3Browse/all/wgacat.html Kollgaard, R. I., Feigelson, E. D., Laurent-Muehlson, S. A., Spinrad, H., Wilkes, B. J., & Elvis, M. 1987, ApJ, 323, 243 Dey, A., & Brinkmann, W. 1995, ApJ, 449, 61 Zitelli, V., Mignoli, M., Zamorani, G., Marano, B., & Boyle, B. J. 1992, Komossa, S., & Bade, N. 1998, A&A, 331, L49 MNRAS, 256, 349