Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://hea-www.harvard.edu/QEDT/Papers/bjw_qsodb.ps Äàòà èçìåíåíèÿ: Thu Aug 1 22:19:01 1996 Äàòà èíäåêñèðîâàíèÿ: Tue Oct 2 00:17:57 2012 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: quasar

The Einstein Database of IPC X­ray
Observations of Optically and Radio
Selected Quasars. 1
Belinda J. Wilkes, Harvey Tananbaum,
D. M. Worrall, Yoram Avni \Lambda , M. S. Oey and Joan Flanagan
February 22, 1995
Abstract
We present the first volume of the Einstein quasar database. The database includes
estimates of the X­ray count rates, fluxes, and luminosities for 514 quasars and Seyfert
1 galaxies observed with the Imaging Proportional Counter (IPC) aboard the Einstein
observatory. All were previously known optically­ or radio­selected objects and most
were the targets of the X­ray observations. The X­ray properties of the AGN have been
derived by re­analyzing the IPC data in a systematic manner to provide a uniform
database for general use by the astronomical community. We use the database to
extend earlier quasar luminosity studies which were made using only a subset of the
currently available data.
The database can be accessed on internet via the SAO Einstein on­line system
(``Einline'') and is available in ascii format on magnetic tape and DOS diskette.
\Lambda Deceased
1

2 The Einstein Quasar Database.
1 Introduction
Einstein observations have shown that most, if not all, quasars are luminous X­ray sources
(Tananbaum et al, 1979; Avni and Tananbaum, 1986, hereafter AT86). The Einstein data
have been used to determine the X­ray properties of individual quasars as well as to study
the characteristics of statistically well­defined or complete samples. A substantial body of
literature (references in text below) presents data on quasars selected for X­ray observation
on the basis of a wide range of characteristics (e.g., optical flux, optical luminosity, redshift,
radio characteristics, etc), providing convincing evidence for the correlation of X­ray emission
with optical emission and for the presence of excess X­ray luminosity in radio­loud quasars.
In addition, Einstein observations of statistically well­defined samples (see Section 3) have
been used to carry out analyses of quasar luminosity and evolution functions in the X­ray
band.
Most of these Einstein observations of quasars (and all of those reported in this paper)
were carried out with the Imaging Proportional Counter (Giacconi et al, 1979). The Imaging
Proportional Counter (IPC) was well suited for this observational program by virtue of its
high throughput, its better than 30'' positional accuracy for point sources, and its modest
energy resolution which enabled us to determine fluxes and luminosities over a reasonably
well­defined energy band.
A large number (¸ 1000) of quasars and active galaxies (AGN) were observed from 1978
to 1981 as part of many different observing programs with the IPC. Many of these data have
been analyzed and published elsewhere under the original scientific programs involved. Due
to differing analysis procedures and scientific aims, the results are of varying quality and
scattered throughout the literature.
In this paper we present the first volume of the Einstein quasar database. The database
includes estimates of or upper limits to the X­ray count rates and errors, fluxes, and lumi­
nosities for 514 previously known, optically­ or radio­selected quasars and Seyfert 1 galaxies
for which targeted observations were made with the IPC. The results have been derived by
re­analyzing, in a systematic manner, the current (Rev 1B) version of the IPC processed
data to provide a uniform database for general use by the astronomical community. The
data are presented in a series of tables as follows: Table 1 gives the basic information on each
quasar; Table 2 gives details of the Einstein observations; Table 3 gives the X­ray fluxes and
luminosities for five different assumed energy indices: a) 0.0; b) 0.5; c) 1.0; d) 1.5; e) 2.0;
Tables 5a,b list objects and observations missing from our database; Table 6 gives optical
magnitudes and luminosities.

B. J. Wilkes et al. 3
2 The Sample
The IPC quasar observations have been divided into two subsets since the details of the
analysis procedure are different in the two cases. The first subset is presented here and
includes primarily quasars and Seyfert 1 galaxies which were targets of IPC observations.
The objects in this ``target sample'' mostly appear on­axis in the images. The target sample
also includes several fields with multiple known quasars where the entire group is the selected
target and most appear off­axis in the image. The second subset includes radio­ and optically­
selected quasars which were observed because they lie in the field of view of another IPC
target. These latter objects, which we estimate as roughly equal in number, will be reported
in a later paper. Some non­target sources, which have already been analysed and are included
in this paper, are noted in Table 2.
The current sample contains 636 observations of 514 objects. We include all sources which
are classified as quasars or Seyfert 1 galaxies in the compilations of Hewitt and Burbidge
(1987 and 1989) and V'eron­Cetty and V'eron (1987). We also include objects classified as
quasars in the Einstein Observatory Catalog of Observations (the ``Yellow Book'', 5 th edition;
Section III, Tables 4.1 and 4.3) with the exception of objects obviously mis­classified (BL
Lacs, galaxies, etc), without a redshift in the available literature, or which were missed due
to satellite pointing problems or incorrect positions. Such objects are all noted in Table 5a
(see below for additional details).
For X­ray­bright objects observed on multiple occasions to study possible time variability,
we normally use the initial (``survey'') flux observation in order to avoid introducing biases
into our analysis. For objects observed more than once in a ``survey'' or serendipitous mode,
we normally select the longest exposure for further analysis. In a few instances the selection
of a ``primary'' observation is made on an arbitrary basis. In any case, whenever there are
multiple observations, the one selected for further analysis is indicated by a ``1'' in the Note
columns of Tables 2 and 6.
Basic information is given in Table 1 with the quasars in order of increasing source
right ascension and including various common names, optical celestial coordinates (1950),
references, redshift, notes and sample membership. The coordinate designation given in the
first column follows IAU convention, is unique to each object, and is used throughout the
paper to identify that object in the database. Letters have been added when necessary to
ensure uniqueness internal to this database, but no attempt was made to conform to the
efforts of other authors to do the same.
As a whole our sample is incomplete, although it contains a few complete subsamples: 64
PG quasars (Tananbaum et al. 1986); 33 3CR radio quasars (Tananbaum et al. 1983); and
30 Braccesi BF quasars (Marshall et al. 1984). These complete subsamples are indicated in
Table 1. The PG quasars are an unbiased subset selected from the original Bright Quasar
Survey (Schmidt and Green 1983); the current sample contains two fewer PG X­ray observa­
tions than reported by Tananbaum et al. (1986) due to the exclusion of one HRI observation

4 The Einstein Quasar Database.
and one IPC measurement confused by nearby X­ray sources (see Table 5a). The 3CR and
BF subsets are both complete, flux­limited samples (Schmidt 1968, Tananbaum et al. 1983,
Braccesi, Lynds and Sandage 1968, Marshall et al. 1984). The remainder of our sample is
a heterogeneous mixture of quasars and Seyfert 1 galaxies which were observed by many
different investigators for various scientific reasons.
3 The Database
3.1 Source Detection and Flux Estimates
Throughout this paper, we use the standard IPC PI (pulse­height invariant), broad band
(0.16­3.5 keV) images and background maps from the data processed with the current, Rev
1B, version of the standard processing software (described in Harnden et al. 1984). Fluxes
were determined using definitive versions of the effective area table and the gain calibration
(Harnden et al. 1984). The data were analyzed using a fully automated procedure to en­
sure uniformity and consistency. A list of all the observations, including Einstein sequence
number (unique identifier for each Einstein observation), observation dates, count rate and
uncertainty, livetime, angle off­axis, and detector gain, are given in Table 2 in increasing RA
order. The coordinate designation and IPC sequence number are used in subsequent tables
to identify uniquely each object/observation pair.
Quasar optical positions are taken in order of preference from Schmidt and Green (1983)
for PG objects, from Hewitt and Burbidge (1987,1989) and from V'eron­Cetty and V'eron
(1987) (Table 1). X­ray locations are accurate to \Sigma20 00 (1oe) as determined by comparing
the X­ray and optical positions for the X­ray detected quasars in the PG sample. This is
consistent with the detailed study of position accuracy reported in the IPC Specifications
(Harnden et al. 1984). The X­ray analysis was performed using the optical position for X­
ray non­detections and the X­ray centroid for X­ray detections within 1 0 (3oe) of the optical
position. The use of the X­ray centroid allows an accurate estimate of the X­ray flux of the
source.
The presence/absence of X­ray emission was determined using a standard 2 0 :4 \Theta 2 0 :4
detect cell on the IPC broad­band (0.16­3.5 keV) image. This size of detect cell optimizes
the signal­to­noise and detection sensitivity for on­axis point sources for the Einstein broad
energy band. Background counts were estimated using a detect cell of the same size and
position on the broad­band background map, which was generated by the standard processing
software for each image by combining appropriate proportions of the instrument flat field
(DSMAP) and the background light due to diffuse X­rays (BEMAP, Harnden et al. 1984).
Observations were determined to be detections when
N box ? 3
q
B box (1)

B. J. Wilkes et al. 5
where N box = net counts (image counts minus background map counts) and B box = back­
ground map counts in the detection box and assuming the noise in the background map is
negligible.
A best estimate or 3oe upper limit on the source counts was determined using a 3 0 ­radius
circle centered on the X­ray centroid (X­ray detections) or optical position (X­ray non­
detection). This size circle ensures that the majority of counts are included and is appropriate
for use with the standard effective area calibration (Harnden et al. 1984). Background was
estimated using the same 3 0 circle at the same position on the background map. Net source
counts evaluated this way are given in Table 2. The error on the net counts was estimated
as:
oe cir =
q
T cir (2)
where T cir = total counts in a 3 0 circle and we have assumed that the uncertainty in the
background subtraction is negligible. The different expressions appropriate for detection and
flux determination along with the use of differently sized regions in the two cases lead to
the possibility that a source which gives a 3oe detection may have a flux which is known less
accurately. This situation arises for 57 of our observations (corresponding to 56 different
quasars), which are treated as positive detections but are flagged in Table 2 (``9'' in the Note
column) and Table 3 (``3'' in the Note column). Moreover, in a few cases (10) the detection
process resulted in a large (? 3oe) negative net rate for reasons not well understood (although
7 of these 10 cases are likely due to the presence of a nearby contaminating source (cf. x3.2
below); these 10 instances are indicated as unreliable via a ``6'' in the Note column of Table
2.
After this work was essentially complete a problem was found in the IPC software used
to convert counts from pulse­height (PH) to PI bins. This error enters through a roll­angle
calculation, so sources near the telescope/detector axis are affected only slightly. For the
sources more than a few arc minutes off­axis, we determined the effect of this error on
our results by making a comparison of a random subset of quasars with and without the
error corrected. We found that for ¸2/3 of the objects an additional uncertainty of Ÿ4%
is introduced by the use of inaccurate local gain factors in the conversion from broad­band
PH to broad­band PI counts. For the remaining 1/3, the additional uncertainty is ¸4­7%
(although in the case of 1 marginal detection we found a change of ¸30%). In all cases, the
changes were less than 0.8oe and this additional uncertainty has negligible effect on any of
our subsequent analyses. Similarly we would also expect a few of the objects around our
detection threshold to change their detection status.
For estimating the fluxes, we corrected the net source counts for counts lost due to mirror
scattering, the telescope/IPC point spread function (PSF), vignetting, and detector dead
time (generally ¸ 4%). The effects of mirror scattering and energy­dependent vignetting
were removed using the effective­area table appropriate for a 3 0 radius circle as a function
of its off­axis distance (Harnden et al. 1984). The fraction of counts lost due to the PSF

6 The Einstein Quasar Database.
is a strong function of their pulse height and thus differs among sources. In most cases
this correction was made to the counts in each individual pulse height (PH) channel before
converting to pulse­height invariant (PI) bins and combining them to compute the flux. The
percentage correction computed in this way is strongly peaked at 1.8%. Table 3 a,b,c,d,e
lists the broad­band and monochromatic X­ray fluxes or 3oe upper limits (in the observer's
energy frame), derived from the quoted count rates corrected as described above. Fluxes were
computed assuming Galactic absorption, and a power law energy distribution (F š / š \Gammaff x )
for ff x =0.0, 0.5, 1.0, 1.5, 2.0 respectively. This set covers the range of slopes typically found
in the IPC energy band for quasars (Wilkes and Elvis 1987).
Galactic NH values were determined by interpolation from the Bell Laboratories survey
(Stark et al. 1992) and were set for each IPC field based upon the coordinates of the center of
the IPC field. 1 For southern quasar positions, which are not covered by this survey (typically
ffi ! \Gamma42 o ), the maps of Heiles and Cleary (1979) were used, as noted by a ``1'' in the Note
column of Table 3.
Statistical errors on the fluxes and luminosities can be derived from the fractional error
in the count rate (Table 2). Errors due to the uncertainty in the spectral slope can be
estimated from the range of flux present in Table 3. For X­ray non­detections the 3oe upper
limits on fluxes and luminosities given in Table 3 are derived conservatively from the data
in Table 2, using [max(0,counts)+3\LambdaErr] as the count rate. We have used a Friedmann
cosmology with Hubble constant of H o = 50 km s \Gamma1 Mpc \Gamma1 and deceleration parameter of
q o = 0:0 to compute luminosities. In Table 3, X­ray luminosities for the 0.2--4.5 keV band
and 2 keV monochromatic are referenced to the source energy frame.
3.2 Quasar positions close to a nearby, contaminating source
Sample objects with X­ray sources detected less than 6 0 away were individually inspected
to determine the significance of contamination introduced by the nearby source. If contam­
ination was found to be insignificant (! 5% of the total counts from nearby source falling
within a 3 0 radius circle centered on the quasar position), as was usually the case for sepa­
rations greater than 4 0 , the processing was resumed as normal. Objects suffering significant
contamination from a single nearby source were processed as described below. Those with
significant contamination from more than one nearby source were not processed and are
listed in Table 5 (see below).
For objects contaminated by a single source, a 3 0 ­radius circle was centered on the tar­
get and another 3 0 ­radius circle was centered on the contaminating source. Counts were
determined for the target and contaminating source 3 0 circles excluding the region of ge­
ometric overlap (C 1 ; C 2 ), and counts were also determined for the overlap region, (C 3 ,see
Figure 1). Background counts were then determined for identical regions of the background
1 N.B. Thus all objects in a given field will use the same NH , but two different observations of the same
object may have slightly different NH values. The difference is always well within the uncertainties.

B. J. Wilkes et al. 7
map, B 1 ; B 2 ; B 3 , and net broad­band counts in the three regions N 1 ; N 2 ; N 3 . The fraction of
source counts in the overlap region is primarily a function of the point response and there­
fore depending on geometry and photon energy. Using a maximum­likelihood formulation
assuming Poisson distributions for each region and treating the fraction of the source counts
in the overlap region as a parameter to be fit, allows us to make preliminary estimates of
the source counts S 1 ; S 2 from the target object and contaminating source respectively:
S 1 = N 1 + N 1 N 3
N 1 +N 2
(3)
S 2 = N 2 + N 2 N 3
N 1 +N 2
(4)
The error oe 1 , representing 68% confidence with the formulation above, for the target
source is given by
oe 1 =
q
C 1 [(N 1 +N 2 ) 2 +N 2 N 3 ] 2 + C 3 N 2
1 (N 1 +N 2 ) 2 + C 2 N 2
1 N 2
3
(N 1 +N 2 ) 2
(5)
where we have assumed that the uncertainty in the background counts (B 1 ; B 2 ; B 3 ) is negligi­
ble. The maximum­likelihood calculation was used to determine source existence (S 1 – 3oe 1 )
as well as to estimate source intensity.
The estimate for quasar counts does not take into account the ¸1.8% of the counts
falling outside a 3 0 circle, some of which will fall in the neighboring source circle and thereby
increase the contaminating counts. The fraction of counts, x, from one source falling inside
the 3 0 circle around a nearby source, varies with source separation and with the pulse­height
(PH) distribution of the counts, which in turn depends on the detector gain and the source
spectrum. Since these sources are contaminating one another, we do not know the exact
PH distribution for either source which limits the accuracy of our corrections for this effect.
To quantify this situation, we examined the fraction of counts falling into a circle around
a putative nearby source as a function of separation distance using seven observations of
bright Seyfert galaxies and quasars covering a full range of the IPC gain: 12--18. For the
source separations of interest (2 0 \Gamma 4 0 ), the correction factor, x, was mostly dependent on the
detector gain with separation a secondary factor. Mean values appropriate for the different
gain ranges are shown in Table 4; we note that these values of x are quite comparable to
the scattering correction factor of 1.8% for isolated sources quoted in the previous section.
Given the detector gain for the observation of interest, a correction based on the first results
for S 1 and S 2 was computed and applied to N 1 and N 2 . Values for S 1 , S 2 , and oe 1 were then
recomputed.
This iterative process was repeated for five cycles which was usually more than sufficient
for convergence. On a few occasions unphysical conditions (such as negative count rates)
were encountered. We then decreased x to x \Gamma 0:005 under the assumption that the source

8 The Einstein Quasar Database.
spectrum is flatter than average, and restarted the iteration using the original net counts.
For estimates of S 1 less than 0, no iterative correction was applied since any contamination
effect is insufficient to produce a positive net count rate for the target.
In the case of N's or S's being negative and j N 1 j'j N 2 j, the estimate of oe 1 can become
unstable due to a small denominator. To guard against unreasonably high error estimates,
the error appropriate for the non­contaminated case, p
C 1 + C 3 , was substituted if its value
was lower than oe 1 .
3.3 Deep Survey Fields
A background map for each observation is generated during the standard Rev1B processing
by combining: (1) the instrument flat field (DSMAP) generated from source subtracted
deep survey fields, which accounts for cosmic ray background and non­uniformities in the
detector and (2) the diffuse background (BEMAP) generated from data taken while looking
at the bright earth, which includes contributions from the sky and from solar scattered X­
rays. The contribution of the DSMAP is determined by the exposure time and that of the
BEMAP from the source­subtracted background count rate in the image. In the standard
processing a first pass is made through the image using a local background estimate to detect
and subtract sources in order to determine the background rate and evaluate the BEMAP
contribution. A second source­detection pass is then made using the combined DSMAP
and BEMAP to find fainter sources. The final source list includes sources detected in both
passes. For standard processing no subsequent correction is made to the background map
to correct for any additional sources found in the second pass. While such a correction
would be negligible in most cases, long observations may contain a number of weak sources
which are not initially detected. This would result in an over­estimate of the background
rate and thus the level of the background map. We developed an iterative procedure to
regenerate the background map after the second source detection search, repeat the search
and generate a final background map. This procedure was applied to fields with ? 20000
seconds of exposure time with the exception of those containing obvious diffuse sources. In
these fields no second source detection pass is made due to the large uncertainties involved,
which in turn may lead to an overestimate of the background level and the failure to detect
a relatively weak point source. The change in background level was significant only for the
BF fields (sequence numbers 5390,5391,5392), which contain an unusually large number of
weak sources.
3.4 Partially obscured objects
There are a number of cases where the 3 0 circle around a quasar position partially overlaps
the IPC ribs which obscure X­rays from a source, or where the circle is partially off the
edge of the detector. For these, detection was determined from the counts available in a

B. J. Wilkes et al. 9
smaller box so the detection sensitivity is reduced. The source counts were measured from
an unobscured half or three­quarter circle and corrected for the fraction of area not used.
If more than half the circle was obscured more than 5% of the time (due to small motions
of the satellite), then the source was not processed and is listed as obscured or on the field
edge in Table 5. Otherwise, the source counts were used to estimate the flux or upper limit
in the usual way. The obscuration seriously decreases the area over which PH counts can be
obtained from the image and in the case of the ribs presents an additional energy dependent
effect, preventing accurate determination of a PSF correction. A mean PSF correction of
1.8% (see Section 3.1) was applied for the partially obscured objects. Quasars processed in
this way are noted in Table 2 (notes 2,3).
3.5 Quasars not processed
A number of objects could not be processed for a variety of reasons including multiple
nearby sources, excessive obscuration by ribs/edge, mispointing, no aspect solution (useful
information may still be obtainable for these objects), or unresolved computer processing
errors (which may eventually be resolved). For completeness they are listed in Table 5 along
with the reason for their omission. For the user's convenience we also list omitted objects
mis­classified as quasars in the Einstein Observatory Catalog of Observations (5 th edition),
objects with no redshift and thus questionable quasars; a set of objects listed as quasars in the
Yellow Book (Seq. 6727--6747) which in reality were targets hypothesized, but not confirmed,
as quasars with possible high X­ray luminosity, and a few objects not yet properly processed
due to coordinate errors. Table 5a lists such objects which have been omitted from this
paper, while Table 5b is comprised of objects for which results are reported here but for
which one or more observation sets (sequences) have not yet been fully analyzed.
3.6 Optical Data
Optical B and/or V magnitudes, collated primarily from the quasar catalogues (Hewitt and
Burbidge, 1987,89 (HB87,89), V'eron­Cetty and V'eron 1987 (VV87)), are listed in Table 6
along with the reference. For these 2 catalogs, magnitudes of quasars without color data are
quoted as ``V''. In many cases the original references provide additional information (such
as photographic magnitude which approximates B). To the extent possible, we rechecked
the literature for objects without color information in the catalogues. Where applicable,
we changed ``V'' magnitudes in the catalogs to B in Table 6 and indicated such objects by
a ``2'' in the Note column. Additional reference information is provided by other entries
to the Note column of Table 6 for several quasars for which the catalog magnitudes have
been superseded. In some cases, such as the PG quasars, we utilized the original reference
(e.g., Schmidt and Green 1983) and B magnitudes, which occasionally differ slightly from
the values in HB87,89 and VV87. When V magnitudes from the quasar catalogs for PG

10 The Einstein Quasar Database.
quasars are included in Table 6, the B­V color does not always agree with the color difference
quoted in the quasar catalogs. For other samples, such as Sramek and Weedman (1980) and
Anderson (1990), the available magnitudes are essentially B and are quoted as such in Table 6
without an accompanying note, even though HB87,89 and VV87 may list the same data as
V magnitudes. We calculated the 2500 š A (rest frame) optical luminosity, l o , from the B
magnitude, where available, assuming a power law energy distribution with slope (ff o ) of
0.5. This calculation was made using a magnitude to flux conversion constant of 48.36 for B
magnitudes (Hayes and Latham 1975) leading to the following equation for the emitted flux
at 2500 š A:
log[f em (2500)] = \Gamma19:34 + ff o log[ 2500
4400 ] \Gamma (1 \Gamma ff o )log[1 + z] \Gamma 0:4(B \Gamma AB + \DeltaB) (6)
When no B magnitude was available, the V magnitude was used with a conversion con­
stant of 48.60 (Oke 1974) leading to a constant of \Gamma19:44 in the equation above.
The correction for reddening along the line of sight through our Galaxy assumes the
constant gas­to­dust ratio given by Burstein and Heiles (1978) so that:
EB \GammaV = max[0; (\Gamma0:055 + 1:987 \Theta 10 \Gamma22 NH )] (7)
where NH is the Galactic value as used to determine the X­ray luminosity. Following Allen
(1973), we then take:
A V = 3EB \GammaV (8)
AB = A V + EB \GammaV = 4EB \GammaV
The correction for the presence of emission lines in the B,V filter range uses the formula:
\DeltaB = 2:5log 10 [1 +W – (1 + z) RB (–)
R
RB (–)d– ] (9)
where W – is the rest frame equivalent width in š A of the emission line, – = – rest (1+z), is the
observed wavelength of the line at redshift z and RB is the response of the B filter in š A \Gamma1 (and
similarly for V; Marshall 1983). Adopted mean equivalent widths for prominent emission
lines were taken from Wilkes (1986) and are shown in Table 7. No correction was made for
the contribution of starlight which is expected to be important at low l o (log l o Ÿ 29). This
does effect the results of our analysis (see Section 5.3).
We note that the values of log l o (the spectral luminosity at 2500 š A in units of erg s \Gamma1 Hz \Gamma1
computed using equation (6) assuming ff o = 0:5) are lower than those reported in earlier
papers (Marshall et al. 1984, AT86, Worrall et al. 1987) by 0.06 on average. This is due to
a combination of updated constants for the magnitude to flux conversion, a more accurate
reddening correction, and generally smaller corrections for emission lines due to updated

B. J. Wilkes et al. 11
equivalent­width measurements. This mean change results in a negligible change in the
resulting ff ox values of 0.002.
The effective optical to X­ray power law slope, ff ox , also listed in Table 6, was computed
following the definition:
ff ox = \Gamma
log `x
`o
log šx
šopt
(10)
where š x corresponds to 2 keV and š opt corresponds to 2500 š A in the quasar's rest frame, l x
is the spectral luminosity at 2 keV in units of ergs \Gamma1 Hz \Gamma1 assuming ff x = 0:5 (Tananbaum
et al. 1979). For multiple X­ray observations of a given object, Table 6 contains multiple
entry lines and appropriate values of ff ox . Lower limits for ff ox , corresponding to X­ray
non­detections, are so indicated in Table 6.
4 Characteristics of the Quasar Sample
As noted earlier, the quasars in the database are generally a heterogeneous sample but
contain as subsets three complete samples: PG (optically selected), 3CR (radio­selected)
and BF (optically selected). We describe the characteristics of the full set of 514 objects
with the complete samples shown individually for comparison. Figure 2 shows the redshift
distributions of the full, PG, 3CR and BF samples respectively. It is clear that the database is
dominated by low­redshift quasars with a long tail to higher redshift, the maximum redshift
being 3.53. The X­ray and optical luminosity histograms (Figures 3, 4 respectively) show
the range in spectral luminosity at 2 keV and 2500 š A respectively. The database covers ?6
orders of magnitude in both quantities.
In Figure 5 the distribution of ff ox (effective X­ray to optical slope) values for the whole
database and for each complete sample are displayed. X­ray non­detections, i.e. lower limits
on ff ox are shaded. For our set of 514 ``primary'' observations, we have 328 X­ray detections
and 186 non­detections. We further note that 5 of the non­detected objects were detected
in one of our ``non­primary'' observations. The PG sample is the most representative of the
quasar population as a whole, it has a mean ff ox of 1.5 and a range 1.0\Gamma2.0. We note however
that this sample is biased towards bright optical luminosity as a result of its relatively bright
apparent magnitude limit, which in turn affects its distribution of X­ray luminosities due
to the dependence of ff ox on optical luminosity. Figure 6 shows the same histograms but
with radio­loud objects shaded. In the optically­selected samples (PG,BF) the figures show
a tendency for the radio­loud objects to have smaller ff ox (L x relatively greater) than the
radio­quiet ones. This is a manifestation of the result that for a given optical luminosity,
ff ox is lower if an object is radio­loud (Ku, Helfand and Lucy 1980, Zamorani et al. 1981,
Worrall et al. 1987).

12 The Einstein Quasar Database.
5 X­ray and Optical Luminosity Relations
The study by AT86 of the dependence of the X­ray to optical luminosity ratio on optical
luminosity and redshift is extended here using our significantly larger sample of optically­
selected quasars and Seyfert 1 galaxies.
5.1 Object Selection
The optically­selected sample was assembled from the database by excluding certain sources:
objects with flags ``X'' or ``R'' in the Hewitt and Burbidge (1987, 1989) catalogue of quasars,
indicating X­ray or radio­selection, were excluded unless there was also an ``O'' or ``C'' flag
(indicating optical or UV­excess selection) or unless they were part of the PG or BF samples.
Eight radio­loud objects flagged ``C'' by Hewitt and Burbidge are excluded from our ``optical''
sample and are listed in Table 8. Six of these are included in the PHL Catalog of very blue
objects (Haro and Luyten 1962). For all 8 objects follow­up redshift studies and classification
as quasars only resulted from independent discovery of strong radio emission. As noted in
Table 1, 0112­017 (also UM 310) is included in the optically­selected sample even though it
is not flagged ``C'' or ``O'' by Hewitt and Burbidge, because it was independently selected in
a prism survey (Lewis, McAlpine, and Weedman 1979).
This selection process ensures that radio­loud objects are included as optically selected
only when they are independently selected in an optical survey. Our optically­selected sample
is comprised of 343 objects (179 detections and 164 upper limits) from the original 514.
5.2 Subsample Characteristics
Figure 7 compares the distribution in redshift and log l o for the current sample and that of
AT86. One obvious difference between the two samples is our inclusion of lower luminosity
objects. A KS­test for two­dimensional data (Press and Teukolsky 1988) finds that the
distribution of z, log l o values differ between the current sample and that of AT86 (0.2% of
a chance occurrence).
5.3 Dependence of X­ray Luminosity on Optical Luminosity
We use the DB regression analysis method described in AT86 with the same functional form
for the mean ff ox ,
! ff ox (log l o ; z) ?= A z (Ü (z) \Gamma 0:5) +A o (log l o \Gamma 30:5) +A (11)
where Ü (z) is the lookback time in units of the Hubble time (or z=[1 + z] for the q o = 0
cosmology assumed here).

B. J. Wilkes et al. 13
For consistency with Avni and Tananbaum (1982; hereafter AT82) and AT86, we assume
that the ff ox residuals follow a Gaussian distribution about ! ff ox ? and we compute the
value of ff ox for each source using an X­ray spectral index of ff x = 0:5. Figures 8a,b show 90%
confidence contours for two interesting parameters (\DeltaS = 4:6) in the A o vs A z and A o vs A
planes, respectively, for the new optically selected sample of 343 sources (solid line) and the
earlier AT86 sample of 154 sources (dotted line). The revised contours are consistent with
the earlier results but provide tighter constraints on the parameter values due to the larger
sample. The decrease in size of the contours is not as great as would be expected (square­
root of ratio of number of sample objects in each linear dimension) if the larger sample were
drawn from exactly the same population as the smaller. We attribute this to the differences
between the ranges of z and log l o in the samples (see x5.2). Note that measurements of l o
and l x revised with respect to AT86 have an insignificant effect on the fitted results. 143
sources in the current database are included in the set of 154 sources used by AT86. We find
consistency with the AT86 result using the new measurements for these 143 sources.
Our results, which are consistent with A z = 0, confirm with smaller errors the result of
AT82 and AT86 that ! ff ox ? depends primarily on log l o rather than z, although we cannot
rule out a small z dependence. Our formal 1oe errors for two interesting parameters are:
A o = 0:10 \Sigma 0:04, A z = 0:006 \Sigma 0:17, A = 1:54 \Sigma 0:04, with a best­fit Gaussian standard
deviation of oe = 0:25 \Sigma 0:02 for the dispersion in the spectral index ff ox .
Assuming no dependence on z, again with 1oe errors for 2 interesting parameters, we find,
! ff ox ?= (0:11 \Sigma 0:02)(log l o \Gamma 30:5) + (1:53 \Sigma 0:02) (12)
*
log
/
l x
10 26:5
!+
= (0:71 \Sigma 0:05) log
/
l o
10 30:5
!
+ (0:018 \Sigma 0:065) (13)
Figure 9 shows the values of log l x and log l o for our sample objects with the fit of
eq. 13 superimposed. The scatter of the data about the best­fit line suggests visually that
something more complicated than a linear dependence of l x on l o may give a better fit; this is
investigated in a forthcoming paper (Avni, Worrall and Morgan 1993). Figure 10 illustrates
that the X­ray and optical flux densities are correlated, confirming that a correlation between
log l x and log l o is not an artifact of common distance spreading along each axis.
AT86 found that although a Gaussian form to the residuals was acceptable, a skew
distribution with a longer tail at high ff ox (low l x ) and a shorter tail at low ff ox (high l x ) gave
a better fit. This change was found to reduce the normalization but not the shape of the
dependence on l o . DB regression analysis can also be applied using a non­parametric form
for the residuals (Avni et al. 1980). This method gives a similar dependence of ff ox on log l o
and, by fitting the mean ff ox for different bands of log l o , the shape of the dependence of ff ox
on log l o can be investigated. The non­parametric approach has been investigated further
and applied to the present sample by Avni, Worrall and Morgan (1991; 1993) who show how
the resulting uncertainties (although not the best fits) depend on the allowed range for the

14 The Einstein Quasar Database.
underlying distribution of residuals, and who point out that errors may be underestimated
if care is not taken in applying such methods (see also Anderson 1985).
Margon et al. (1992) have found a very similar dependence of ff ox on log l o to our fit of
eq. 12. Their slope is 0:11 \Sigma 0:01 (error presumed to be 1oe for 1 interesting parameter) from
fitting measurements of 146 objects which were observed with the IPC and taken from a
complete quasar sample. While some of their objects are in common with our analysis, it is
significant that the analysis method they employ, `image stacking', is quite different from ours
and yet the results agree very well. Marshall (1992) also obtains a similar result to that given
here from a further analysis of the quasar sample of AT86. Marshall's treatment includes
the uncertainties in the detected fluxes as well as the probability distribution associated with
the non­detections.
Low­luminosity objects show increased dispersion in their dependence of l x on l o , probably
due to a combination of varying degrees of starlight contamination which would effect l o and
intrinsic absorption (believed to be largest in low­luminosity sources) having been ignored
in the calculations of l x (Worrall 1987). If we restrict our sample to exclude the lowest
luminosity AGN (log l o ! 29:0), we find that the results change marginally in the direction
of a larger A o (i.e., a flatter dependence on l o ; see Figure 11a,b). The new best­fit parameters
are A o = 0:15 \Sigma 0:04, A z = \Gamma0:08 \Sigma 0:17, A = 1:49 \Sigma 0:04, with oe = 0:23 \Sigma 0:02. Still no
dependence on redshift is required by the fit. This change is consistent with contamination
by starlight at low l o steepening the dependence of l o . This is clear in Figure 9 where the
low luminosity objects have systematically high l o .
5.4 Dependence on X­ray Spectral Index
Although we assumed a value of X­ray spectral energy index ff x = 0:5 in our analysis, for
comparison with earlier work, it is now known that, for the brightest X­ray AGN measured
with the IPC, a slope of 0.5­0.7 is more typical for radio­loud objects and ¸ 1:0 for their
radio­quiet counterparts (Wilkes and Elvis 1987; Canizares and White 1989; Worrall 1989;
Brunner et al. 1989). Figure 12 shows that our conclusion that ff ox depends primarily on log l o
rather than redshift is relatively insensitive to our choice of ff x . A change in ff x affects A z
and A but not A o . This is because ff ox jointly depends on ff x and z through the K­correction
term of the X­ray luminosity calculation (see the form of eq. 6) whereas there is no joint
dependence on ff x and l o . [However, note that in fits to eq. 11 with A z explicitly set to zero,
changes in ff x affect both A o and A; this is because if there is no redshift term in the fitting
function to accommodate the joint dependence of ff ox on ff x and z, the optical­luminosity
term becomes indirectly affected due to a dependence of z on l o (fig. 7).]

B. J. Wilkes et al. 15
5.5 Difference between Radio Loud and Radio Quiet Sources
Figure 13 shows separately the redshift vs log l o figures for the 291 radio­quiet and 52 radio­
loud sources of the optically­selected sample. A KS­test finds that these subsamples differ in
their distributions of z, log l o (0.5% of a chance occurrence), although the incompleteness of
our samples precludes any cosmological significance from being attributed to this difference.
We exclude low­luminosity objects and compare samples for log l o ? 29:0 in order to be
somewhat comparable with Worrall et al. (1987) who actually used an even higher threshold
of log l o ? 29:95. The radio­loud and radio­quiet subsets are consistent with A z = 0 and
the same dependence on log l o (i.e., same A o ), but fit very different normalization constants,
A (fig. 14a, b). This confirms earlier results that radio­loud sources produce more X­rays
relative to their optical luminosity (i.e., radio­loud sources have a lower value of A, and thus a
smaller [flatter] ff ox ; Ku, Helfand and Lucy, 1980; Zamorani et al. 1981; Worrall et al. 1987).
We see from fig. 14b that allowing for a different spectral slope for the two populations
(0.5 for radio loud and 1.0 for radio quiet; see x5.4) does not account for the difference in
normalization.
Acknowledgements
We would like to thank Richard Kondo for his work on coding the automated analysis pro­
cedures and Paul Martenis and Susannah Hopkins for assistance in its execution. We also
thank Rick Harnden, Frank Primini, and John McSweeney for assistance with computation
of the appropriate background levels for deep survey exposures; Mark Birkinshaw and Mar­
tin Elvis for a careful reading of the manuscript; and Karen Modestino for her professional
expertise in preparation of the text and tables. This work was supported by NASA contracts
NAS8­30751 and NAS8­39073.

16 The Einstein Quasar Database.
6 References
Allen, C.W. 1973, Astrophysical Quantities, p264
Anderson, S.F. 1985, Ph.D. Thesis, University of Washington
Anderson, S. 1990, private communication
Avni, Y., Soltan, A., Tananbaum, H., and Zamorani, G. 1980, ApJ, 238, 800
Avni, Y. and Tananbaum, H. 1982, ApJ, 262, L17
Avni, Y. and Tananbaum, H. 1986, ApJ, 305, 83
Avni, Y., Worrall, D.M., and Morgan, W.A.,Jr. 1991, Institute of Mathematical Statistics
Bulletin, 20, 260.
Avni, Y., Worrall, D.M., and Morgan, W.A.,Jr. 1993, in preparation.
Braccesi, A., Formiggini, L., and Gandolfi, E. 1970, A&A, 5, 264
Braccesi, A., Lynds, R. and Sandage, A. 1968, ApJ, 152, L105
Bregman, J.N., Glassgold, A.E., Huggins, P.J., and Kinney, A.L. 1985, ApJ, 291, 505.
Brunner, H., Worrall, D.M., Wilkes, B.J., and Elvis, M. 1989, in Proc. 23rd ESLAB Sympo­
sium, (Noordwijk: ESA SP­296), p. 905
Burbidge, E.M. and Strittmatter, P.A. 1972, ApJ, 174, L57
Burstein, D. and Heiles, C. 1978, ApJ, 225, 40
Canizares, C. R. and White, J. L. 1989, ApJ, 339, 27
Cohen, A.M., Porcas, R.W., Browne, I.W.A., Daintree, E.J., and Walsh, D. 1977, MNRAS,
84, 1
Colla, G., et al. 1970, A&AS, 1, 281
Formiggini, L., Zitelli, V., Bonoli, F., and Braccesi, A. 1980, A&AS, 39, 129
Giacconi, R., et al. 1979, ApJ, 230, 540
Green, R.F., Schmidt, M., and Liebert, J. 1986, ApJS, 61, 305
Harnden, F.R., Fabricant, D.G., Harris, D.E. and Schwarz,J. 1984, SAO Report No. 393
Haro, G. and Luyten, W.J. 1962, Bol. Obs. Ton. Y. Toc., 13, N22, 37
Harris, D.E., et al. 1993, NASA Research Publication, in press
Hayes, D.S. and Latham, D.W. 1975, ApJ, 197, 593
Hazard, C. 1992, private communication.
Heiles, C. and Cleary, M. N. 1979, Aust. J. Phys. Astrophys. Suppl., 47,1
Hewitt, A. and Burbidge, G. 1987, ApJS, 63,1
Hewitt, A. and Burbidge, G. 1989, ApJS, 69,1
Hoag, A.A. and Smith, M.G. 1977, ApJ, 217, 362
Hunstead, R.W., Murdoch, H.S., and Shabbrook, B.B. 1978, MNRAS, 185, 149
Kristian, J. and Sandage, A. 1970, ApJ, 162, 391
Ku, W.H­M., Helfand, D.J., and Lucy, L.B. 1980, Nature, 288, 323.
Lewis, D.W., McAlpine, G.M., and Weedman, D.W. 1979, ApJ, 233, 787.

B. J. Wilkes et al. 17
Margon, B., Anderson, S.F., Wu, X., Green, P.J., and Foltz, C.B. 1992, in ``X­ray Emission
from AGN and the Cosmic X­ray Background'', eds. W. Brinkman and J. Tr¨umper,
(Garching: Max­Planck Institut), 81
Marshall, H. L. 1992 in ``Statistical Challenges in Modern Astronomy'' eds. E. D. Feigelson
and G. J. Babu, [Springer­Verlag]
Marshall, H. L. 1983, Ph.D. Thesis, Harvard University
Marshall, H. L., Avni, Y., Braccesi, A., Huchra, J. P., Tananbaum, H., Zamorani, G., and
Zitelli, V., 1984, ApJ, 283, 50
Oke, J.B. 1974, ApJS, 27, 21
Peterson, B.A., Bolton, J.G., and Shimmins, A.T. 1973, Astrophys. Lett., 15, 109
Press, W.H. and Teukolsky, S.R. 1988, Computers in Physics, July/August Issue, 74
Rafanelli, P. and Schulz, H. 1983, A&A, 117, 109
Schmidt, M. 1968, ApJ, 151, 393
Schmidt, M. and Green, R. F. 1983, ApJ, 269,352
Sramek, R.A. and Weedman, D.W. 1980, ApJ, 238, 435
Stark, A.A., Gammie, C.F., Wilson, R.W., Bally, J., Linke, R.A., Heiles, C. and Hurwitz,
M. 1992, ApJS, 79, 77
Tananbaum, H., et al. 1979, ApJ, 234, L9
Tananbaum, H., Wardle, J. F. C., Zamorani, G. and Avni, Y. 1983, ApJ, 268, 60
Tananbaum, H., Avni, Y., Green, R. F., Schmidt, M. and Zamorani, G. 1986, ApJ, 305, 57
Usher, P.D. 1978, ApJ, 222, 40
Usher, P.D. 1981, ApJS, 46, 117
Vaucher, B.G. and Weedman, D.W. 1980, ApJ, 240, 10
V'eron, M.P., V'eron, P., Algie, R.L., and Gent, H. 1976, A&A, 47, 401
V'eron­Cetty, M.­P. and V'eron, P. 1987, ESP Scientific Repo No. 5
Wilkes, B. J. and Elvis, M. 1987, ApJ, 323, 243
Wilkes, B. J. 1986, MNRAS, 218, 331
Willis, A.G. and deRuiter, H.R. 1977, A&AS, 29, 103
Worrall, D.M. 1987, ApJ, 318, 188
Worrall, D.M. 1989, in Proc. 23rd ESLAB Symposium, (Noordwijk: ESA SP­296), p. 719
Worrall, D.M., Giommi, P., Tananbaum, H., and Zamorani, G. 1987, ApJ, 313, 596
Zamorani, G., et al. 1981, ApJ, 245, 357

18 The Einstein Quasar Database.
Table 1: Quasars and Seyfert 1 galaxies included in the database.

32 The Einstein Quasar Database.
Table 2: Observational details.

B. J. Wilkes et al. 45
Table 3: X­ray Fluxes and Flux Densities.
a. ff E = 0.5

B. J. Wilkes et al. 81
Table 4: Correction Factor, x as a function of detector gain
Gain Correction
Factor, x
Ÿ 14.0 0.025
14.0 ! Gain Ÿ 15.0 0.020
? 15.0 0.015

82 The Einstein Quasar Database.
Table 5: Quasar and Seyfert 1 galaxy observations not included in our sample due to pro­
cessing problems.

B. J. Wilkes et al. 85
Table 6: Optical magnitudes and ff ox .

98 The Einstein Quasar Database.
Table 7: Mean equivalent widths for the prominent emission lines
Line Wavelength Equivalent Width
( š A) (W – , š A)
Hfi 4861 47
MgII 2798 27
CIII] 1909 17
CIV 1549 32
Lyff 1215 84

B. J. Wilkes et al. 99
Table 8: Objects flagged as color­selected by Hewitt and Burbidge, but excluded from our
optically selected sample
Name Other Names
0017+154 PHL2871, 3CR9
0056­001 PHL923, PKS
0226­038 PHL1305, PKS
0730+659 W1 0730+659
1223+252 TON616, 4C25.40
2128­123 PHL1598, PKS
2134+004 PHL61, PKS
2135­147 PHL1657, PKS

100 The Einstein Quasar Database.
7 Figure Captions
SOURCE 1 SOURCE 2
Region C1 Region C2
C3
Figure 1: Schematic diagram showing the respective regions of overlapping source circles
from which counts C 1 ; C 2 ; C 3 are extracted.

B. J. Wilkes et al. 101
0
1
2
3
4
0
20
40
60
80
100
z
N
0
1
2
3
4
0
20
40
60
80
100
N
a
0
1
2
3
4
0
10
20
30
40
z
N
0
1
2
3
4
0
10
20
30
40
N
b:
PG
0
1
2
3
4
0
1
2
3
4
5
6
z
N
0
1
2
3
4
0
1
2
3
4
5
6
N
c:
3CR
0
1
2
3
4
0
1
2
3
4
5
z
N
0
1
2
3
4
0
1
2
3
4
5
N
d:
BF
Figure 2: Histograms of the number of objects at each redshift for a) the full database and
b) the PG, c) the 3CR and d) the BF samples. X­ray upper limits are shown shaded.

102 The Einstein Quasar Database.
24
26
28
30
0
50
100
150
200
N
24
26
28
30
0
50
100
150
200
a
24
26
28
30
0
5
10
15
20
25
30
N
24
26
28
30
0
5
10
15
20
25
30
b:
PG
24
26
28
30
0
5
10
15
20
25
N
24
26
28
30
0
5
10
15
20
25
c:
3CR
24
26
28
30
0
5
10
15
N
24
26
28
30
0
5
10
15
d:
BF
Figure 3: The range in 2 keV X­ray luminosity present in a) the full sample of 514 sources
and b) the PG, c) the 3CR and d) the BF sub­samples. X­ray upper limits are shown shaded.
Log l x is spectral luminosity at 2 keV with l x in units of erg s \Gamma1 Hz \Gamma1 assuming ff x =0.5.

B. J. Wilkes et al. 103
28
30
32
34
0
50
100
150
200
N
28
30
32
34
0
50
100
150
200
a
28
30
32
34
0
5
10
15
20
25
N
28
30
32
34
0
5
10
15
20
25
b:
PG
28
30
32
34
0
5
10
15
20
N
28
30
32
34
0
5
10
15
20
c:
3CR
28
30
32
34
0
2
4
6
8
10
12
N
28
30
32
34
0
2
4
6
8
10
12
d:
BF
Figure 4: The range in 2500 š A optical luminosity present in a) the full sample and b) the
PG, c) the 3CR and d) the BF sub­samples. X­ray upper limits are shown shaded. Log l o is
the spectral luminosity at 2500 š A with l o in units of erg s \Gamma1 Hz \Gamma1 derived from the optical
magnitude under the assumption ff o =0.5.

104 The Einstein Quasar Database.
1.0
1.5
2.0
0
50
100
150
N
1.0
1.5
2.0
0
50
100
150
a
1.0
1.5
2.0
0
5
10
15
20
N
1.0
1.5
2.0
0
5
10
15
20
b:
PG
1.0
1.5
2.0
0
5
10
15
N
1.0
1.5
2.0
0
5
10
15
c:
3CR
1.0
1.5
2.0
0
5
10
15
N
1.0
1.5
2.0
0
5
10
15
d:
BF
Figure 5: The range in effective optical to X­ray slope (ff ox ) present in a) the full sample
and b) the PG, c) the 3CR and d) the BF sub­samples. Lower limits are shown shaded.

B. J. Wilkes et al. 105
1.0
1.5
2.0
0
50
100
150
N
1.0
1.5
2.0
0
50
100
150
1.0
1.5
2.0
0
50
100
150
1.0
1.5
2.0
0
50
100
150
a
1.0
1.5
2.0
0
5
10
15
20
N
1.0
1.5
2.0
0
5
10
15
20
1.0
1.5
2.0
0
5
10
15
20
1.0
1.5
2.0
0
5
10
15
20
b:
PG
1.0
1.5
2.0
0
5
10
15
N
1.0
1.5
2.0
0
5
10
15
1.0
1.5
2.0
0
5
10
15
1.0
1.5
2.0
0
5
10
15
c:
3CR
1.0
1.5
2.0
0
5
10
15
N
1.0
1.5
2.0
0
5
10
15
1.0
1.5
2.0
0
5
10
15
d:
BF
RQ
D
RQ
ND
RL
D
RL
ND
Figure 6: The range in effective optical to X­ray slope (ff ox ) present in a) the full sample
and b) the PG, c) the 3CR and d) the BF sub­samples. RQ = radio­quiet, RL = radio­loud,
D = X­ray detections, ND = X­ray non­detections.

106 The Einstein Quasar Database.
(a)
(b)
Figure 7: The Distribution in redshift and log l o of the 343 sources in a) the new optically­
selected sample, and b) the earlier, smaller, sample of 154 sources from AT86. Log l o is the
logarithm of the spectral luminosity at 2500 š A in units of ergs s \Gamma1 Hz \Gamma1 . X­ray detections
are shown as filled symbols; open symbols are X­ray upper limits.

B. J. Wilkes et al. 107
(a) (b)
Figure 8: The 90% confidence (\DeltaS=4.6) contours for (a) A o , A z , (with A free), and (b) A o ,
A, (with A z free), from fitting the dependence of ff ox on z and log l o (eq. 11). Contours using
the previously­analyzed smaller dataset of AT86 are shown (dotted lines) for comparison.

108 The Einstein Quasar Database.
Figure 9: The relation between X­ray and optical luminosity for the new optically­selected
sample of 343 sources, with best­fit model assuming no redshift dependence. Luminosities
are in units of ergs s \Gamma1 Hz \Gamma1 at 2 keV and 2500 š A for l x and l o , respectively. Detections are
indicated by filled squares and X­ray upper limits by open squares with arrows attached.

B. J. Wilkes et al. 109
Figure 10: A dependence of X­ray flux density (¯Jy) on optical flux density (mJy) shows that
the log l x , log l o correlation of figure 9 is not induced merely by common distance spreading
along each axis. Detections are indicated by filled squares and X­ray upper limits by open
squares with arrows attached.

110 The Einstein Quasar Database.
(a) (b)
Figure 11: The 90% confidence (\DeltaS=4.6) contours for (a) A o , A z , and (b) A o , A, from fitting
the dependence of ff ox on z and log l o , where low luminosity AGN (log l o ! 29:0) have been
excluded. As in fig. 8, contours using the previously­analyzed smaller dataset of AT86 are
shown (dotted lines) for comparison.

B. J. Wilkes et al. 111
(a) (b)
Figure 12: The 90% confidence (\DeltaS=4.6) contours for (a) A o , A z , and (b) A o , A, from
fitting the dependence of ff ox on z and log l o , for three different assumptions about the X­ray
spectral index: ff x = 0:5 (solid line; same as figure 10), ff x = 1:0 (dotted line), and ff x = 1:5
(dashed line)

112 The Einstein Quasar Database.
Figure 13: The Distribution in redshift and log l o of objects in the new optically selected sam­
ple, divided into radio­quiet (upper) and radio­loud (lower) subsamples. Following Zamorani
et al. 1981, a source is radio loud if its spectral index between 5 GHz and 2500 š A in the source
frame is larger than 0.35. log l o is the logarithm of the spectral luminosity at 2500 š A, in units
of ergs s \Gamma1 Hz \Gamma1 . X­ray detections are shown as filled symbols; open symbols are X­ray upper
limits.

B. J. Wilkes et al. 113
(a) (b)
Figure 14: The 90% confidence contours (\DeltaS=4.6) for the 272 radio­quiet (ff x = 0:5,
solid line; ff x = 1:0, dashed line) and 49 radio­loud (ff x = 0:5, dotted line) quasars with
log l o ? 29:0. The low­luminosity sources are excluded because they bias the fit and are
predominantly only radio quiet (fig. 13). (a). A o vs A z . The radio­quiet and radio­loud
subsamples are consistent with the same A o and A z = 0. (b). A o vs A. The radio­quiet
and radio­loud subsamples are consistent with the same A o (dependence on log l o ), but fit
different normalization constants, confirming earlier results that a source which is radio loud
will produce more X­rays relative to its optical luminosity.