Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://hea-www.harvard.edu/QEDT/Papers/mn_crssiii.ps Äàòà èçìåíåíèÿ: Thu Jan 11 17:55:25 1996 Äàòà èíäåêñèðîâàíèÿ: Tue Oct 2 00:34:19 2012 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: http astrokuban.info astrokuban

Mon. Not. R. Astron. Soc. 000, 000--000 (1995) Printed 10 January 1996
The Cambridge­Cambridge ROSAT Serendipity Survey ­
III. VLA observations and the evolution of Radio­quiet
and Radio­loud objects
P.Ciliegi, 1;2;3 , M.Elvis, 1 B.J.Wilkes, 1 B.J.Boyle, 4 R.G. McMahon, 5 T.Maccacaro 3
1. Harvard­Smithsonian Center for Astrophysics 60 Garden St, Cambridge MA 02138, USA
2. Dipartimento di Fisica dell'Universit`a di Milano, Italy
3. Osservatorio Astronomico di Brera, Via Brera 28, I­20121 Milano, Italy
4. Royal Greenwich Observatory, Madingley Road, Cambridge, CB3 0EZ
5. Institute of Astronomy, University of Cambridge, Madingley Road, Cambridge, CB3 0HA, UK
Accepted 1995 July 26. Received 1995 July 26; in original form 1995 April 28
ABSTRACT
We present the results of the VLA Radio observations at 1.475 GHz (20 cm) of
the Active Galactic Nuclei (AGN) in the Cambridge­Cambridge ROSAT Serendipity
Survey (CRSS), a sample of 123 faint X­ray sources with f x (0.5­2.0 keV)– 2 \Theta 10 \Gamma14
erg s \Gamma1 cm \Gamma2 . Of the 80 AGN in the sample, seven show radio emission at 5 oe level
and only two (2.5 +4:0
\Gamma1:7 %) qualify as Radio­Loud (RL) objects (ff ro –0.35). This result,
compared with 13% RL in the Einstein Observatory Extended Medium Sensitivity
Survey (EMSS) sample of AGN (flux limit f x (0.3­3.5 keV)¸ 2 \Theta 10 \Gamma13 erg s \Gamma1 cm \Gamma2 )
confirms that the fraction of X­ray selected RL AGN drops rapidly as the X­ray flux
limit is lowered.
Combining the CRSS AGN sample with that extracted from the EMSS we have
studied the X­ray Luminosity Function (XLF) and evolutionary properties for Radio­
Quiet (RQ) and Radio­Loud separately. We find that the RQ and RL AGN population
show the same cosmological evolution within the errors. In fact, when the luminosity
evolution is parameterised with a power law of the form L \Lambda
X (z) = L \Lambda
X (0)(1 + z) k , we
find k = 2:43 \Sigma 0:26 and k = 2:71 \Sigma 0:10 for RL and RQ AGN populations respectively.
In addition, the shape of the de­evolved XLF of the two classes appears to be different
both at the low luminosity (LX ! 10 44 erg s \Gamma1 ) and high luminosity ends. These results
are robust for different cosmological models (using q 0 =0.0 and q 0 =0.5) and for different
value of the threshold ff ro used to distinguish between RQ and RL objects.
Finally we find that the differences in the shape of the XLF of RQ and RL AGN can
be explained by introducing an X­ray beaming model to separate the observed X­ray
luminosity of radio quasars into relativistically beamed and isotropic contributions.
Key words: galaxies:active \Gamma galaxies:nuclei \Gamma quasars:general \Gamma radio continuum:
galaxies \Gamma X­ray: galaxies
1 INTRODUCTION
The cause of the difference between AGN that are strong ra­
dio sources (radio­loud, RL) and those which are radio­quiet
(RQ) is one of the most basic topics in the field of quasar
astronomy. Although the two classes have similar spectral
distributions (SEDs) outside the radio band (Elvis et al.
1994), their luminosity functions show differences in all the
bands in which they have been studied.
In the optical band, using the PG sample of optically
selected AGN (Schmidt and Green, 1983), Padovani (1993)
has shown that the shapes of the luminosity functions for
RL and RQ are different. As a result of these differences, the
fraction of radio­loud objects is ¸20­50% for MB Ÿ \Gamma24:5,
but falls to 7­8% at fainter absolute magnitudes.
Recently, Della Ceca et al. (1994, hereafter DC94) us­
ing the Einstein Observatory Extended Medium Sensitiv­
ity Survey (EMSS, see Gioia et al., 1990 and Stocke et al.
1991) sample of X­ray selected AGN, have determined the
X­ray luminosity functions (XLF) of RL and RQ separately.
They have obtained results very similar to those obtained

2 P.Ciliegi et al.
by Padovani (1993) in the optical domain. The shape of the
XLF of the two classes appears to be different and a flatten­
ing of the XLF of the RL sample is visible for Lx Ÿ 10 44:5
erg s \Gamma1 . As a result of this difference the expected fraction
of RL is a function of the X­ray flux limit in X­ray surveys.
They predict that this fraction is ¸13 % for fx ¸ 2 \Theta 10 \Gamma13
erg s \Gamma1 cm \Gamma2 and decreases to ¸ 2.5 % forfx ¸ 2 \Theta 10 \Gamma15
erg s \Gamma1 cm \Gamma2 .
Until recently the EMSS AGN sample was the only
sample of X­ray selected AGN for which complete radio in­
formation exists. Now a new, fainter sample of X­ray se­
lected AGN has been obtained using the ROSAT satel­
lite. This new sample, the Cambridge­Cambridge ROSAT
Serendipity Survey (CRSS, Boyle et al. 1995) is a well de­
fined sample of 80 X­ray selected AGN discovered serendip­
itously in 20 ROSAT PSPC fields at high Galactic latitude
(jb II j– 30 ffi ). The selection criteria were: an X­ray flux
fx (0.5­2.0 keV)– 2 \Theta 10 \Gamma14 erg s \Gamma1 cm \Gamma2 (some 10 times
fainter than the typical EMSS limit), and an off­axis an­
gle ` Ÿ 15 0 in the PSPC field. Of the 80 sources, 68 were
classified as QSOs from the presence of broad emission lines
(full width half maximum (FWHM) ? 1000 km s \Gamma1 ), while
12 were classified as narrow emission line X­ray galaxies
(NLXGs) (see Boyle et al. 1995 for more details). A full
description of the sample will appear elsewhere (McMahon
et al. 1995, in preparation).
In this paper we report the results of VLA y observations
of all 80 sources in the CRSS AGN sample. Our aim was to
obtain a complete classification of the sample members as
RL or RQ in order to determine well­constrained XLFs for
X­ray selected RQ and RL AGN separately.
The paper is organized as follows. In Section 2 we
describe our radio observations of the CRSS sample. We
present the results of these observations in Section 3, while
in Section 4 we report the X­ray luminosity function of RQ
and RL objects. Finally we present our conclusions in Sec­
tion 5. Throughout the paper a Hubble constant of 50 km
s \Gamma1 Mpc \Gamma1 is assumed.
2 VLA DATA
We observed 18 of the 20 fields of the CRSS sample with
the National Radio Astronomy Observatory (NRAO) Very
Large Array at a frequency of 1.475 GHz (20 cm) on 5 June
1994. The observing bandwidth was 25 MHz in the B con­
figuration. This combination of frequency, bandwidth and
configuration allows us to obtain a primary beam of full
width half power (FWHP)' 30 0 . The sensitivity of the VLA
decreases radially compared with the value at the centre of
the primary beam such that it decrease to ¸0.5 and ¸0.2
of its peak value at off­axis angles of ` ' 8 0 and ` ' 25 0
respectively. The synthesized beamwidth was 3.9 00 FWHP.
Because the primary beam has about the same size as the
region of the ROSAT=PSPC from which the X­ray sources
were extracted, in general we obtained one VLA observation
y The Very Large Array (VLA) is a facility of the National Radio
Astronomy Observatory (NRAO) which is operatedby Associated
Universities, Inc., under cooperative agreement with the National
Science Foundation.
for each ROSAT=PSPC field. However the VLA pointing
position was not simply made coincident with the ROSAT
central field position but was chosen instead to minimize the
off­axis angle of the X­ray sources within the radio field of
view. For one field (NGC5907, ror=WP600190) it was nec­
essary to carry out two different VLA pointings, so we have
carried out a total of 19 pointings.
Since our aim was a complete classification of the CRSS
sources into RL or RQ objects, we set the integration time
to ensure a clean discrimination between RL and RQ for
undetected objects. From Zamorani et al. (1981) a quasar is
defined as radio­loud when the radio to optical spectral in­
dex, ff ro –0.35 where ff ro=\Gammalog(L 2500 š A/L5 GHz )/5.38. Us­
ing the redshifts and optical (POSS) magnitudes of each
object from McMahon et al. (1995), we calculated the radio
flux at 1.475 GHz needed to detect a source with ff ro=0.35.
Because ff ro is computed between 2500 š A (assuming ff o=1.0)
and 5 GHz (rest frame), we have assumed a spectral slope
ff r =0.7 (fš / š \Gammaff r Zamorani et al. 1981) to calculate the
radio flux at 1.5 GHz in the observed rest frame. The flux
limits do not change significantly if we assume a flatter ra­
dio spectral index (for example ff r =0.3). Allowing for the
loss of sensitivity for off­axis sources, we derived integra­
tion times between 5 and 20 minutes. All the observations
were interspersed with nearby calibrator observations at 20
minute intervals. The primary flux density calibrator was
3C48 which was assumed to have a flux density of 15.59 Jy.
For the two remaining fields, PG1512 (ror=RP700807)
and 4U1417+42 (ror=WP700535) radio data were avail­
able in the literature or from the VLA archive. The
ROSAT PSPC field 4U1417+42 is centered on the BL
Lac object 1426+4253. This object was observed by Dr.
M. Marcha and collaborators with the VLA on 1991, at
20 cm, configuration B, bandwidth 50 MHz with an inte­
gration time of 20 minutes (VLA archive code: AM330).
The ROSAT PSPC field PG1512 is centered on the quasar
3C351. From the VLA archive we retrieved the observa­
tion obtained by Dr. J.P. Leahy in 1987, at 20 cm, B
configuration, bandwidth 25 MHz and with an integration
time of 55 minutes (VLA archive code: AL146). The source
CRSS1620.1+1724 was observed by Kellermann et al. (1989)
with the VLA at 5 GHz (6 cm), D configuration, bandwidth
50 MHz.
Altogether we have VLA data for all 80 AGN in the CRSS
sample (72 from our observations, 7 from the VLA archive
and 1 from the literature).
3 RESULTS
Each field was analysed with the NRAO AIPS reduction
package. We searched for radio sources above the 5oe local
r.m.s. noise at the optical position of each source. Because
both optical and radio positions have an error of ¸ 1 00 , we
searched for radio sources inside a circle of 5 00 radius cen­
tered on the optical position of each AGN to permit also the
detection of possible double radio sources. Whenever a radio
source at the 5oe level was found within this circle its peak
flux density was taken to be the radio flux density of the
AGN. When no radio source was detected, we determined
a 5oe upper limit at the optical position. All flux densities
were corrected for primary beam attenuation.

CRSS\GammaIII 3
Of the 80 CRSS X­ray selected AGN, 6 were detected
from our observations at 1.475 GHz with fluxes ranging from
0.44 to 6.85 mJy and one (CRSS1620.1+1724) was detected
at 5 GHz with a flux of 1.09 mJy (Kellermann et al. 1989).
Of these, 6 are QSOs and one is an NLXG. The differences
between the radio and optical positions for the radio de­
tections ranges from 0 00 to 1.47 00 . All detected sources were
the only radio sources down to 3 oe within 5 00 radius of the
optical position. Most of the 5 oe upper limits are in the
range 0.25­1.00 mJy. Only in few case did the presence of a
strong (? 1 Jy at 20 cm) radio source in the field results in
higher upper limits, typically 3­5 mJy, except for one case
(CRSS1418.3+0637) where we have a 5 oe upper limit of 25
mJy. The radio data on the 80 AGN in the CRSS sample
are given in Table 1.
The X­ray flux of each object was obtained by analysing
all the X­ray sources with the IRAF/PROS z software pack­
age. A full description of the X­ray spectral analysis and
X­ray flux list will appear elsewhere (Ciliegi et al. 1995, in
preparation). The B magnitude of each object was estimated
from the O mag. and O \Gamma E color using (Evans 1988):
B = O \Gamma 0.119(O \Gamma E)
The optical luminosity at 2500 š A and the radio luminosity
at 5 GHz were computed following Zamorani et al. (1981):
log L 2500 š A = 38.011 \Gamma 0.4B + 2log(z(1+z/2))
log L5 GHz = 34.63 + log S(š) + 2log(z(1+z/2)) +
ff r log š
5000 + (ffr \Gamma1)log(1+z)
where S(š) is the observed radio flux (in Jy) at fre­
quency š in MHz (š=1.475 GHz in our observations) and
ff r is the energy slope measured at radio wavelengths. Fol­
lowing Zamorani et al. (1981) we have assumed ff r =0.7.
In Figure 1, we show ff ox vs. ff ro for all the sources
of the CRSS sample. The dashed line represents the di­
vision between radio­loud (ffro – 0:35) and radio­quiet
(ff ro ! 0:35) objects. Different symbols were used to distin­
guish between QSO (open) and NLXG (filled) and between
radio detections (circles) and radio upper limits. Only two
QSOs (CRSS1705.5+6042 and CRSS2250.0+1407, see Table
1) have ff ro – 0:35 and these only slightly so. These two ob­
jects remain the only objects with ff ro – 0:35, also assuming
a flatter radio spectral index (ff r =0.3) to derive the radio
luminosity. The number of RL objects in the CRSS sample
depends on the chosen ff ro threshold. This dependence is
discussed in x4.2.
4 THE X­RAY LUMINOSITY FUNCTION
4.1 4.1 The X­ray Luminosity Function of RL and RQ
AGN
In order to determine well­constrained XLFs for X­ray se­
lected RQ and RL AGN separately, we have combined the
CRSS data with the EMSS data. Following DC94, we have
used only the portion of the EMSS north of \Gamma40 ffi declination
and we have considered as RQ those objects with a radio
upper limit ff ro – 0:35. With these choices, the RL AGN
z IRAF is distributed by NOAO, which is operated by AURA,
Inc., under contract to the NSF.
sample consists of 45 objects (43 from EMSS and 2 from
CRSS) while the RQ sample consists of 440 objects (363
from EMSS and 77 from CRSS). These numbers show that
the fraction of RL AGN in the CRSS sample (¸2.5 +4:0
\Gamma1:7 %)
is lower than the fraction of RL AGN in the EMSS sample
(¸13%). This is in agreement with the prediction of DC94
that the expected fraction of RL should drop rapidly as the
X­ray flux limit is lowered (the X­ray flux limit for the CRSS
sample is fx(0.5­2.0 keV)¸ 2 \Theta 10 \Gamma14 erg s \Gamma1 cm \Gamma2 , some
10 times fainter than the typical EMSS limit).
The RQ object CRSS1620.1+1724 was excluded from
the CRSS sample because it is in common with the EMSS
sample (MS1617.9+1731). The 0.5­2.0 keV X­ray fluxes of
the CRSS sample were converted to the EMSS 0.3­3.5 keV
passband using S(0.3­3.5 keV) = 1.8 S(0.5­2.0 keV) which is
accurate to 2% for spectra with spectral indices in the range
0.6! ff x !1.5 (Boyle et al. 1993). Throughout this paper,
in order to maintain consistency with the analysis of DC94,
we have assumed ff x = 1.0.
Using the Ve=Va variable of the 1/Va method of Avni
and Bahcall (1981), DC94 have shown that in both the RL
and RQ sample the hypothesis of no evolution is rejected
at more than the 99.99% confidence level. We have repeated
the same analysis, combining CRSS and EMSS sample, with
the same result. Having confirmed that both RL and RQ
samples exhibit significant cosmological evolution, we then
used the maximum likelihood technique to obtain a ``best­
fit'' parametric representation for the luminosity function
and its evolution for both samples (see Boyle et al. 1993
and Marshall et al. 1984 for a complete description of this
method). As in Boyle et al. 1993, we use the two­power­law
form for the XLF
\Phi X (LX ) = \Phi \Lambda
X L \Gammafl 1
X 44
LX ! L \Lambda
X (z = 0)
\Phi X (LX ) = \Phi \Lambda
X
L (fl 1 \Gammafl 2 )
X 44
L \Gammafl 2
X 44 LX ? L \Lambda
X (z = 0)
where \Phi \Lambda
X is the normalization of the XLF and fl 1 and
fl 2 are the faint and bright end slopes respectively. LX 44 is
the 0.3­3.5 keV X­ray luminosity expressed in units of 10 44
erg s \Gamma1 .
To study the evolutionary properties of RQ and RL
samples we have used a (1+z) power­law evolution in the
``break'' luminosity, L \Lambda
X (z) (Boyle et al. 1993):
L \Lambda
X (z) = L \Lambda
X (0)(1 + z) k
Since the maximum­likelihood analysis only give a
``best­fit'' solution without information on a ``goodness of
fit'' for the best­fit model, we must also test for the accept­
ability of the model against the data. To do this, we used
the 2­dimensional KS statistic (Peacock 1985) employing the
algorithm devised by Press et al. (1992). This statistic pro­
duces a probability P KS for the model being an acceptable
fit to the data. A 2D KS test shows that the power­law evo­
lutionary form that we have used in our analysis, is an ac­
ceptable fit to the data, with a KS probability always greater
than 10 per cent.
Because DC94 in their analysis used a different method
to study the evolutionary properties, as first step we applied
the method just to the EMSS data for RL and RQ separately
(model A and B for q0=0.0 and q0=0.5 respectively). The

4 P.Ciliegi et al.
results of the Ve=Va test, maximum likelihood analysis and
2D KS test are presented in Table 2. The quoted errors are at
1oe level, determined using the method described by Boyle,
Shanks and Peterson (1988). Table 2 shows that the evolu­
tionary parameters obtained with our analysis are in good
agreement with the results obtained by DC94 (k = 2:35 +0:22
\Gamma0:25
and k = 2:92 +0:19
\Gamma0:23 at 1oe level for RQ and RL respectively).
The small discrepancies between the ``best­fitting'' parame­
ter values derived with the maximum­likelihood method and
those obtained with the method used by DC94 must be sim­
ply due to the different analysis methods applied (Boyle et
al. 1993). The significant difference in fl 1 (the faint end slope
of the XLF) between RL and RQ, confirms the flattening of
the XLF of the RL sample for LX (z = 0) Ÿ 10 44:5 erg s \Gamma1
noted by DC94.
We then combined EMSS and CRSS samples (model C
(q0=0.0) and D (q0=0.5)). The inclusion of the CRSS does
not change the value of the evolution parameter k signifi­
cantly. However, we note an increase of k for RQ AGN and
a decrease for RL AGN. An increase of the evolution param­
eter of the same order of magnitude (\Deltak ¸ 0:2) was found
by Boyle et al. 1993, when combining the EMSS and the
ROSAT Deep Survey. For the CRSS sample, this difference
in the evolution parameter may be explained in terms of
the difference in the mean spectral index between the CRSS
and EMSS samples. The spectral index ff x enters straight­
forwardly into the determination of the evolution parameter
k in the power­law evolution model (Della Ceca et al. 1992).
If k 0 is the derived value for the evolution paramenter in the
case where ff x = 1:0, then k = k 0 + (ffx \Gamma 1). Because the
mean spectral index of the CRSS sample is ff x = 1:2 (Ciliegi
et al. 1995 in preparation) while the mean spectral index of
the EMSS sample is ff x = 1:0 (Maccacaro et al. 1988), we
have a difference between the two samples of \Deltaff ' 0:2. Al­
though the combined sample (EMSS + CRSS) of RQ AGN
is strongly dominated by the EMSS sources (¸80%), the fact
that we found \Deltak ' \Deltaff x (see models A and C in Table 2)
suggests that the difference in the evolution parameter sim­
ply reflects the difference in the mean spectral index of the
two samples. This difference may be real, reflecting a ``soft
excess'' in the ROSAT=PSPC band (¸0.1­2.4 keV) com­
pared to the Einstein IPC band (¸0.3­3.5 keV), or may
be due to calibration error in the PSPC and/or IPC instru­
ments (see Appendix B in Fiore et al. 1994 for more details).
We conclude that the differences in the evolution parameter
between RL and RQ AGN are within the 1oe error in all the
models that we have analyzed. Therefore, with the available
data we do not find evidence that the cosmological evolution
of RL and RQ AGN is different.
It is, however, clear that the shape of the XLF is differ­
ent for RL and RQ objects also using different cosmological
model (q0 =0.0 and q0=0.5). Table 2 shows that in model
C and D the values of fl 1 and fl 2 are significantly different
for RL and RQ AGN. The shape of the XLF is different not
only at low luminosity as noted by DC94, but also at the
bright end. In Figure 2 we plot a de­evolved z=0 XLF ob­
tained with the maximum likelihood analysis for the EMSS
and for the combined (EMSS + CRSS) samples (models A
and C respectively). To test if this behaviour of the XLF is
due to an error in the X­ray flux calibration of the CRSS
sources we have calculated the XLF (for q0=0.0) increas­
ing and decreasing the X­ray flux of the CRSS sources by
30% respectively. The RL and RQ evolution parameters are
still equal within their 1oe errors, while the difference in the
shape of RQ and RL XLF (fl 1 and fl 2 parameters) remains
significant.
4.2 Dependence of the XLF of RQ and RL on the
definition of Radio­loudness
Until now, we have used the value ff ro=0.35 to discriminate
between RL and RQ objects. This value represents a natural
division of X­ray selected AGN in RQ and RL. In fact, as
shown by DC94, the ff ro distribution of EMSS AGN is a
clear bimodal distribution with a minimum at ff ro ' 0.34 ­
0.37 (see Figure 3a). It is clear in Figure 3b that the addition
of the CRSS sample does not change this. The combined ff ro
distribution has still a bimodal distribution with a minimum
at ff ro '0.35. However, other authors have used different
values of ff ro to discriminate between RL and RQ, testifying
to the fact that the underlying physical differences between
the two classes are not well understood. For example, Stocke
et al. (1992), studying optically selected samples of quasar,
concluded that the most likely dividing value between the
RQ and RL populations is ff ro ¸0.20.
To test whether our results depend on the value of ff ro
chosen to separate RQ from RL, we have re­calculated the
XLF for RQ and RL using different values of ff ro . In table
2 we report (models E to G) the ``best­fit'' parameters for
RQ and RL XLF using ff ro = 0.2, 0.4 and 0.5. When we use
ff ro =0.20 (model E), some sources have inadequate radio
upper limits (ff ro –0.20). We assumed all these (108 sources
in the EMSS sample and 36 in the CRSS sample) to be RQ
AGNs in models E and to be RL AGNs in model E 0 . In
Figure 4 we show the de­evolved z=0 XLF for models E to
G. The shape of the RL and RQ XLFs remains significantly
different for all values of ff ro . In the ff ro=0.2 models, the
difference is only marginal if we assume that all the sources
with inadequate upper limits are RQ AGNs (model E) and
disappears if we assume all these sources to be RL AGNs (in
model E 0 the parameters fl 1 fl 2 and logL \Lambda
X are all consistent
within the 1oe errors). However these models are unrealistic
due to the simplifying assumption that all the sources with a
radio upper limit ff ro –0.20 are RQ or RL AGNs (we expect
that only about 13% (¸ 20 objects) of these sources are RL
AGN).
The major effect of changing the threshold ff ro on the
XLF is on the faint end slope of the XLF for RL AGN (see
parameter fl 1 in Table 2 for models E ­ G). An increase in
the ff ro threshold results in the loss of the faintest bin in the
RL XLF (Figure 4). This is due to the strong correlation
between ff ro and the X­ray luminosity (Figure 5) first noted
by Zamorani et al. (1981) for radio and optically selected
QSOs. Increasing the threshold ff ro (dotted lines in the top
panel of Figure 5) excludes mainly objects with low X­ray
luminosity.
Changing the threshold ff ro does not affect the evolu­
tion parameter k: the two populations of objects show the
same cosmological evolution in all the models that we have
analysed. Therefore the conclusion that RQ and RL AGN
show the same cosmological evolution and have different
shapes of their XLFs, is robust to changes in the cosmo­
logical model used (models A ­ D), and to the value of ff ro

CRSS\GammaIII 5
used to discriminate between RL and RQ objects (models E
­ G).
4.3 The different shape of RQ and RL XLF : a
possible explanation
We now investigate how the different shapes of the RQ and
RL XLFs could arise from the RL objects having an addi­
tional mechanism producing X­rays.
The difference between RL and RQ AGN in the X­ray
band have been studied by many authors. Zamorani et al.
(1981) showed that, for a given optical luminosity, the X­ray
luminosity of RL is ¸ 3 times stronger than for RQ. This
property is clearly present in Figure 5. Worrall et al. (1987)
confirmed this property and showed that, for a given optical
luminosity, the X­ray emission is expected to be higher for
RL AGN with flat radio spectra (that is, with a dominant
compact radio emission) than for RL AGN with steep radio
spectra. Also the shape of the X­ray spectra appear to be
different in RL and RQ. Wilkes and Elvis (1987) showed that
RL objects have flatter X­ray spectra (ff x ¸0.5 in the 0.3­
3.5 keV band) compared to RQ objects (ff x ¸1.0). Lawson
et al. (1992), using EXOSAT data, showed that RL objects
have X­ray spectral indices consistent with a unique index
(i.e. consistent with a dispersion oe=0.0), whereas RQ objects
show a large spread in indices (oe ?0.10)
All these differences between RQ and RL AGN can be
explained if we consider a simple two component scenario
for quasar X­ray emission. In this scenario (first proposed
by Zamorani et al. 1981) all quasars have a central ``energy
machine'' which provides at least part of the optical and X­
ray emission. Any mechanism proposed to explain this emis­
sion should contain one or more variable parameters which
can produce the large observed dispersion in the X­ray lu­
minosities (Figure 5), in the level of X­ray loudness (ff ox ,
Zamorani et al. 1981), and in the X­ray spectral indices of
RQ AGN (Lawson et al. 1992). In addition, RL quasars must
have a second X­ray producing mechanism to explain their
higher observed average ratio between X­ray and optical lu­
minosity, the lack of RL objects with low X­ray luminosity
(see Figure 5), the flatter X­ray spectra of RL AGN (Wilkes
and Elvis 1987) and the low dispersion in the X­ray spectral
indices of RL AGN (Lawson et al. 1992).
In the framework of the ``unified interpretation'' for RL
AGN (Blandford and K¨onigl 1979), the differences between
radio core­dominated (CDQs) and lobe­dominated quasars
(LDQs) are due to orientation alone. Based on the corre­
lation of the X­ray luminosity with nuclear and lobe radio
luminosity (Worrall et al. 1987, Browne and Murphy 1987)
and on the fact that soft X­ray energy indicies have been
found to be systematically flatter for CDQs than for LDQs
(Canizares and White 1989, Boroson 1989, Shastri 1991),
Shastri (1991) suggested that the ``radio­linked'' component
of the X­ray emission in RL AGN is orientation­dependent
and relativistically beamed.
In the past years, several authors developed models to
explain this scenario (Kembhavi, Feigelson and Singh 1986,
Browne and Murphy 1987, Kembhavi 1993). The general
feature of these models is that the two components of the X­
ray emission mentioned above have the following properties:
(a) the first component unrelated to the radio emission (the
``radio­quiet mechanism'') is intrinsically very dominant and
has a steep X­ray spectrum with ff x ¸1.0 (b) the additional
component in RL quasar associated with the radio emission
(the ``radio­linked mechanism'') has a flat X­ray spectrum
(ff x ¸0.5) and, although intrinsically weak relative to the
other component, it makes a significant contribution to the
X­ray emission due to the effect of relativistic beaming when
the direction of the jet with which it is associated is oriented
close to the line of sight.
In this ``X­ray beaming'' model, the total X­ray luminosity
Lx of RL AGN can be written as:
Lx = Lxb + Lxu
where Lxu is the unbeamed X­ray luminosity associated
with the radio­quiet mechanism which occurs in both RQ
and RL and Lxb is the beamed X­ray luminosity which is
dominant in core­dominated RL due the beaming effect. In
this scenario we do not expect to have different shapes of
the XLF for RQ AGN and RL AGN, if for the latter we use
only the unbeamed luminosity.
If we call F the ratio of the beamed X­ray component to the
isotropic component
F = Lxb
Lxu
the beamed and unbeamed X­ray luminosity are:
Lxb = Lx F
1+F and Lxu = Lx
1+F
In this model the radio luminosity from extended radio
regions in quasars is unbeamed (Lru) and has a steep ra­
dio spectrum, while that from the compact region has a flat
spectrum and is boosted due to the relativistic motion of
the emitting region. The observed compact radio luminosity
can be written as Lrb(`) = Lrb(90 ffi )gr (fi; `) where Lrb(90 ffi )
is the luminosity when the angle ` with the line of sight is
90 ffi and gr (fi; `) is the beaming factor defined as
gr (fi; `) = 1
2
[(1 \Gamma ficos`) \Gamma(2+ff r ) + (1 + ficos`) \Gamma(2+ff r ) ]
where v=cfi is the bulk velocity and ff r is the radio spec­
tral index. For a given radio quasar, the radio beamed and
unbeamed luminosity are observables. Assuming that the
beamed X­rays arise in the compact radio source, which is
also the source of the beamed radio radiation, then it is
possible to obtain the X­ray beaming factor gx (which is de­
fined in the same way as the radio beaming factor gr with
ff r replaced by the X­ray spectral index ff x) and hence the
factor F for the X­ray emission (see Kembhavi 1993 for more
details).
Unfortunately we do not have good enough radio data
to estimate F . However we can use the relation between F
and Lx (found by Kembhavi 1993) to estimate the factor F
and then the unbeamed X­ray luminosity for each RL AGN
in our sample. We have used the relation
LogLx = 0.124\ThetaLog(Lxb =Lxu ) + 27.71 (1)
to obtain the unbeamed X­ray luminosity of each RL AGN.
In this relation Lx is the monochromatic luminosity at 2keV
in units of erg s \Gamma1 Hz \Gamma1 . x
x In Kembhavi (1993) it is reported the relation LogLx =
0.124\ThetaLog(L xb =Lxu ) + 20.71 with Lx monochromatic luminosity
at 2 keV in units of erg s \Gamma1 Hz \Gamma1 . He used a sample of 126 radio
quasars obtained by Browne and Murphy (1987) for which pub­

6 P.Ciliegi et al.
Using Lxu we have re­calculated the XLF for RL AGN.
The ``best­fit'' parameters for this ``unbeamed'' XLF are re­
ported in Table 2 under the model C ? . A comparison of
model C ? with model C­RQ shows that the two XLFs now
have the same slope at the bright end (fl 2 parameter) while
remain some differences at the faint end, although now the
fl 1 parameters are consistent at less than 2oe level.
Because the LogLx ­ LogF relation shows a large scat­
ter (see Kembhavi 1993) and the author did not report the
errors on the correlation coefficients, we have also investi­
gated the effect on the unbeamed XLF for RL AGN using
relations between LogLx and LogF slightly different from
that shown above. We find, for example, that for
LogLx = 0.13\ThetaLog(Lxb =Lxu ) + 27.52 (2)
the unbeamed XLF for RL AGN and the XLF for RQ AGN
have the same shape. The ``best­fit'' parameter for this sec­
ond unbeamed XLF for RL AGN are reported in Table 2 un­
der the model C ?? . A comparison of model C ?? with model
C­RQ shows that the parameters fl 1 and fl 2 are consistent
within the 1oe errors and that also the evolution parameters
k are consistent within the errors. The de­evolved z=0 XLF
for RQ AGN (model C) and for unbeamed RL AGN (model
C ?? ) is plotted in Figure 6.
Using the relation (2) we find that the RL AGN in our
sample show a large range in the factor F , and that only
for four objects does the beamed X­ray luminosity give a
significant contribution to the total X­ray luminosity. There
are only four objects with LogF ?1. In table 3 we report the
properties of these four objects. On the basis of the X­ray
beaming model and on the relation between radio and X­ray
emission, we expect that these sources are AGNs with flat
radio spectra and with a dominant compact radio emission.
The available radio data show that two of these sources are
indeed core­dominated radio sources with flat radio spec­
tra. MS0038.8\Gamma0159 (4C 02.04) is a core­dominated radio
source with a radio spectral index ff r =0.04 between 6 and
20 cm (Perley 1982), while MS2134.0+0028 (PKS2134+004)
is a well known optically violently variable (OVV) core­
dominated radio source (Murphy, Browne and Perley 1993,
Browne and Perley 1986). Moreover, in the framework of the
Synchrotron Self­Compton model, Ghisellini et al. (1993)
derived for MS2134.0+0028 a very small angle (OE = 0:1 ffi )
between the axis of the jet and the line of sight. For the
other two sources MS0012.5\Gamma0024 and MS2247.8\Gamma0703 the
only radio information are from the VLA snapshot observa­
tions at 6 cm (Stocke et al. 1991) with no information on
the nature of these radio sources.
The change in the XLF that we have obtained by intro­
ducing the X­ray beaming model is due to these four objects.
Because of their large value of F , these objects are shifted
from the XLF bright bins to the fainter bins, causing a large
change in the XLF due to the poor statistics (there are only
one and three objects in the two fainter bins of the RL AGN
XLF shown in lower panel of Figure 2).
Therefore, we can conclude that the differences in the
lished Einstein X­ray observations exists. In Browne and Murphy
(1987) the monochromatic X­ray luminosity at 2 keV are in W
Hz \Gamma1 with 19.14Ÿ LogLx Ÿ22.13. Kembhavi used the same sam­
ple, with the same interval of LogLx but he assumed that the
luminosities were expressed in erg s \Gamma1 Hz \Gamma1 . There is a clear
mistake in the units used by Kembhavi (1993).
shape of XLF between RQ and RL AGN can be explained
introducing the X­ray beaming model where the ``radio­
linked'' component in RL objects is orientation­dependent,
but larger samples of X­ray selected AGN are needed to
strengthen this conclusion.
5 CONCLUSION
Using the Very Large Array (VLA) we have obtained sen­
sitive radio observations for all 80 X­ray selected AGN in
the Cambridge­Cambridge ROSAT Survey. Seven of these
sources show radio emission at 5oe level. Only two AGN qual­
ify marginally as radio­loud on a standard radio to optical
spectral index criterion. These two objects represent only
2.5 +4:0
\Gamma1:7 per cent of the sample compared with 13% RL in
the EMSS (flux limit fx(0.3­3.5 keV)¸ 2 \Theta 10 \Gamma13 erg s \Gamma1
cm \Gamma2 ), confirming the prediction of Della Ceca et al. 1994
that the expected fraction of X­ray selected RL AGN should
drop rapidly as the X­ray flux limit is lowered (the CRSS
flux limit is fx (0.5­2.0 keV)¸ 2 \Theta 10 \Gamma14 erg s \Gamma1 cm \Gamma2 )
We have combined the CRSS sample with the EMSS
sample and studied the X­ray luminosity functions for RQ
and RL separately using the maximum likelihood method.
To study the evolutionary properties of RQ and RL sam­
ples, we have used a power law evolutionary form L \Lambda
X (z) =
L \Lambda
X (0)(1 +z) k . From our analysis we found that the RQ and
RL populations show the same cosmological evolution. Us­
ing the best­fit evolution parameters we have computed the
de­evolved X­ray luminosity function of RQ and RL AGN.
The shapes of the de­evolved XLF of the two classes ap­
pear to be different both in their low luminosity and high
luminosity slopes.
These results are robust against: (i) possible errors due
to the calculation of the ROSAT X­ray flux (increasing and
decreasing the flux of the CRSS sources of 30%); (ii) for dif­
ferent cosmological models (using q0=0.0 and q0=0.5); (iii)
for different value of the threshold ff ro used to distinguish
between RQ and RL objects.
Finally we have investigated the possibility of explain­
ing the difference between the XLFs of the two classes of ob­
jects in terms of an additional beamed radio­linked compo­
nent producing X­rays. This component, intrinsically weak,
becomes dominant when the direction of the jet with which
it is associated is oriented close to the line of sight. Using
the relation LogLx = 0.13\ThetaLog(Lxb =Lxu ) + 27.52 we have
calculated the unbeamed X­ray luminosity for each RL AGN
in our sample. We find that the shape of the XLFs for RQ
and RL AGN is the same if for the latter we use only the
unbeamed component.
This results supports the two component scenario for
quasar X­ray emission with the radio­linked component in
RL AGN being orientation­dependent.
ACKNOWLEDGMENTS
P.C. acknowledges R. Della Ceca for useful discussion and
suggestions. A part of the analysis of the radio data pre­
sented in this paper was done at the ``Istituto di Radioas­
tronomia del CNR'', Bologna, Italy. P.C. wishes to thank L.
Padrielli, director of the Institute for the hospitality and C.

CRSS\GammaIII 7
Fanti, P. Parma and C. Gruppioni for helpful suggestions
in the use of the AIPS software package. P.C. acknowledges
the Smithsonian Fellowship program for partial support dur­
ing this work. RGM acknowledges the support of the Royal
Society. This work has received partial financial support
from the Italian Space Agency (ASI contract 191/3 AXG),
from NASA contract NASA5­30934(RSDC) and from NASA
grant NAGW­2201(LTSA). The optical identifications of the
CRSS were made at the William Herschel Telescope at the
Observatorio del Roque de los Muchachos operated by the
Royal Greenwich Observatory.
REFERENCES
Avni Y., Bachall J.N., 1980, ApJ, 235, 694.
Blandford R.D, and K¨onigl A., 1979, ApJ, 232, 34.
Boroson T. A., 1989, ApJ, 343, L9.
Boyle B.J., Shanks T., Peterson B.A., 1988, MNRAS, 238, 957.
Boyle B.J., Griffiths R.E., Shanks T., Stewart G.C., and Geor­
gantopoulos I., 1993, MNRAS, 260, 49.
Boyle B.J., McMahon R.G., Wilkes B.J. and Elvis M., 1995, MN­
RAS, 272, 462
Browne I.W.A. and Murphy D.W., 1987, MNRAS, 226, 601.
Browne I.W.A. and Perley R.A., 1986, MNRAS, 222, 149.
Canizares C.R., and White J.L., 1989, ApJ, 339, 27.
Ciliegi et al., 1995, in preparation.
Della Ceca R. Maccacaro T., Gioia I.M., Wolter A. and Stocke
J.T, 1992, ApJ, 389, 491.
Della Ceca R., Zamorani G., Maccacaro T., Wolter A., Griffiths
R., Stocke J.T., and Setti G., 1994, ApJ, 430, 533.
Elvis M. et al., 1994, ApJSS, 95, 1.
Evans, 1988, PhD Thesis, Univ. of Cambridge.
Fiore F., Elvis M., Mc Dowell J., Siemiginowska A. and Wilkes
B.J., 1994, ApJ, 431, 515.
Ghisellini G., Padovani P., Celotti A., and Maraschi L., 1993,
ApJ, 407, 65.
Gioia I.M., Maccacaro T., Schild R.E., Wolter A., Stocke J.T.,
Morris S.L., and Henry J.P., 1990, ApJS, 72, 567.
Kellermann K.I., Sramek R., SchmidtM., Shaffer D.B., and Green
R.F., 1989, AJ, 98, 1195.
Kembhavi A.K., Feigelson E.D. and Singh K.P., 1986, MNRAS,
220, 51.
Kembhavi A.K. 1993, MNRAS, 264, 683.
Lawson A.J., Turner M.J.L., Williams O.R., Stewart G.C., and
Saxton R.D., 1992, MNRAS, 259, 743.
Maccacaro T., Gioia I.M., Wolter A., Zamorani G., and Stocke
J.T., 1988, ApJ, 326, 680.
McMahon et al., 1995, in preparation.
Marshall H.L., Avni Y., Braccesi A., Hucra J., Tananbaum H.,
Zamorani G., Zitelli V., 1984, ApJ, 283, 50
Murphy D.W., Browne I.W.A. and Perley R.A., 1993, MNRAS,
264, 298.
Padovani P., 1993, MNRAS, 263, 61.
Peacock J.A., 1985, MNRAS, 217, 601.
Perley R.A., 1982, AJ, 87, 859.
Press W.H., Teukolsky S.A., Vetterling W.T.and Flannery B.P.,
1992, Numerical Recipes, Second Edition, Cambridge: Cam­
bridge Univ. Press.
Schmidt M. and Green R.F., 1983, ApJ, 269, 352.
Shastri P., 1991, MNRAS, 249, 640.
Stocke J.T., Morris S.L., Gioia I.M., Maccacaro T., Schild R.E.,
Wolter A., Fleming T.A., and Henry J.P., 1991, ApJS, 76,
813.
Stocke J.T., Morris S.L., Weymann R.J., and Foltz C.B., 1992,
ApJ, 396, 487.
Wilkes B.J. and Elvis M., 1987, ApJ, 323, 243.
Worrall D.M., Giommi P., Tananbaum H., Zamorani G., 1987,
ApJ, 313, 596.
Zamorani G. et al., 1981, ApJ, 245, 357.
Figure Captions
Figure 1. The X­ray to optical spectral index (ff ox )
plotted against the radio to optical spectral index (ff ro ).
The dashed line represents the divison between radio­loud
(ffro – 0:35) and radio­quiet (ffro ! 0:35) objects.
Figure 2. The de­evolved X­ray luminosity function
for RQ AGN (open symbols) and RL AGN (filled symbols).
Upper panel: data from the EMSS sample (model A). Lower
panel: data from EMSS + CRSS sample (model C)
Figure 3. Distribution of ff ro for the CRSS AGN (up­
per panel) and for CRSS + EMSS sample (lower panel). The
radio detections are show as shaded histogram while the 5oe
radio upper limits are show as solid histogram.
Figure 4. The de­evolved X­ray luminosity function for
RQ AGN (open symbols) and RL AGN (filled symbols) for
different values of the dividing point between RQ and RL
populations. Upper panel: dividing point at ff ro =0.20 assum­
ing that all the sources with radio upper limit ff ro –0.20 are
RQ AGNs Central panel: dividing point at ff ro =0.40 Lower
panel: dividing point at ff ro=0.50
Figure 5. Distribution of the CRSS and EMSS AGN
in the ff ro ­ log LX plane. Upper panel: in this panel are
show only the objects with a radio detection. The dotted
lines represent ff ro = 0.20, 0.35, 0.40 and 0.50, the values
used to descriminate between RQ and RL (see x4.2 for more
details). Lower panel: Radio detections plus 5oe upper limits.
Figure 6. The de­evolved X­ray luminosity function
for RQ AGN (open symbols) and RL AGN (filled symbols).
For RL objects we used only the unbeamed X­ray luminosit
obtained from the relation LogLx = 0.13\ThetaLog(Lxb =Lxu ) +
27.52 (see x4.3 for more details).