ASCA deep surveys

ASCA deep surveys

N ELSEVIER D|B|N||| PROCEEDINGS SUPPLEMENTS Nuclear Physics B (Proc. Suppl.) 69/1-3 (1998) 600-609 ROSAT/ASCA Deep Surveys G. Hasinger a aAstrophys...

641KB Sizes 1 Downloads 10 Views

N

ELSEVIER

D|B|N||| PROCEEDINGS SUPPLEMENTS Nuclear Physics B (Proc. Suppl.) 69/1-3 (1998) 600-609

ROSAT/ASCA Deep Surveys G. Hasinger a aAstrophysikalisches Institut P o t s d a m An der Sternwarte 16 14482 Potsdam, G e r m a n y ROSAT deep and shallow surveys have provided an almost complete inventory of the constituents of the soft X-ray background which led to a population synthesis model for the whole X-ray background with interesting cosmological consequences. According to this model the X-ray background is the "echo" of mass accretion onto supermassive black holes with absorbing molecular tori, integrated over cosmic time. Deep surveys with ASCA are in agreement with this model. A new determination of the soft X-ray luminosity function of active galactic nuclei (AGN) is consistent with pure density evolution, and the comoving volume density of AGN at redshift 2-3 approaches that of local normal galaxies. This indicates that many larger galaxies contain black holes, which were active in the past. The cosmic history of black hole density and star forming rate indicates, that the bulk of black holes was produced before most of the stars in the universe. However, only more sensitive and higher angular resolution X-ray surveys in the harder energy bands, where the maximum of the energy density of the X-ray background resides, will provide the acid test of this picture.

1. I n t r o d u c t i o n The soft X-ray background (XRB) has practically been resolved by deep ROSAT pencil b e a m survey observations into discrete sources and at the faintest fluxes - source fluctuations [1,2]. The ultradeep ROSAT H R I survey now reaches a surface density of ~ 970 d e 9 - 2 at a flux of 10 -t5 e r g c m - 2 s - 1 [3]. Counterparts of the weakest X-ray sources are optically very faint (R < 24) and require very good, unconfused X-ray positions and high-quality optical spectra. The large majority of optical counterparts in the ROSAT deep surveys turned out to be AGN [4]. X-ray selection is therefore the most efficient means to construct large, almost unbiased samples of distant AGN. At the faintest X-ray fluxes there is still a debate a b o u t the existence of a possible new population of nearby X-ray active, but optically innocent narrow emission line galaxies (NELG), whose Xray luminosity may be powered by star forming processes [5,6]. These findings are, however, challenged by the higher resolution d a t a in the Lockm a n Hole [4]. For a more detailed review see [7,8]. The AGN resolved in the soft X-ray band have a quite steep spectrum[I] and can only contribute 0920-5632/98/$19.00 © 1998 Elsevier Science B.V. All rights reserved. PlI S0920-5632(98)00303-X

a minor fraction in the harder X-ray band, where the bulk of the XRB energy density resides. This led to the postulation of a substantial population of X-ray sources with intrinsically hard spectral9], which should be readily detectable with harder Xray telescopes. First imaging surveys with ASCA are just becoming available but complete optical identifications are difficult to complete. In section 2-4 the status of ROSAT and ASCA deep surveys and their optical identification are reviewed. So far all results are consistent with the assumption t h a t AGN provide the m a j o r contribution to the whole X-ray background. In section 5 a new determination of the AGN X-ray luminosity function is discussed, which leads to the exiting speculation t h a t black holes should have been the first building blocks of almost all galaxies, most of which must have been active at high redshifts. 2. T h e R O S A T U l t r a - D e e p

Survey

An area of ~ 0 . 3 d e g 2 in the Lockman Hole was chosen as the field for the deepest X-ray survey ever, because it provides a minimum of interstellar absorption. Over the last 10 years it has become one of the best studied sky regions

G. Hasinger/Nuclear Physics B (Proc. Suppl.) 69/1-3 (1998) 600-609

601

ROSAT IogN-IogS function 1000(

,.U""...'

........

'

........

'

........

'

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

'

........

•. . . . . . ' . . . .

......::.~ .,~..

v~

~"'k.. =,.:.

z 0.01

00001 1 0E-16

........

I

........

I

........

1.0E-14

I

. . . . . . . . . . . . . . . . . . . . . . . . 1 0E-12

,

,,

,,,

1.0E-10

S (0.5-2 keV) [cgs] Figure 1. Total log(N)-log(S) function determined from various ROSAT surveys. Filled circles give the source counts from the 207 ksec PSPC observation in the Lockman Hole[3]. Open circles are from the HRI ultradeep survey observation (1.112 Msec). The data are plotted on top of the source counts (solid line) and fluctuation limits (dotted area)[1]. The dotted line at intermediate fluxes refers to the total source counts from the RIXOS survey[10]. The solid line at bright fluxes was determined from the ROSAT All-Sky Survey bright source catalogue[11].

over a wide range of frequencies: U B V R I and Kband imaging from Kitt Peak, Mt. Palomar and M a u n a Kea (Keck and UH), a 16hr VLA 20cm mosaique[12], a 120ksec ASCA hard X-ray observation[13] and deep IR mosaiques with ISOCAM and I S O P H O T aboard ISO (PIs: Cezarsky, Taniguehi) have been taken. The Lockman Hole is also targeted in the Heidelberg CADIS NIR survey, in Luppino's wide field weak lensing survey and in a Beppo-SAX AO-2 deep survey observation. The ROSAT Deep Survey (RDS) is a 207 ksec P S P C exposure, and the ROSAT Ultradeep Survey an 1.11 Msec ROSAT H R I exposure in the Lockman Hole[3]. The final limiting sensitivity for the detection of discrete sources is about 2 x 10 -15 erg crn -2 s -1 for the P S P C and about a factor of two fainter with the HRI. Figure 1 shows the P S P C and H R I log(N)-log(S) function[3] which is in very good agreement with d a t a

published previously. The HRI source counts reach a surface density of 970 + 150 deg -2, about a factor of two higher than any previous X - r a y determination. About 70-80% of the X - r a y background measured in the 1-2 keV band has been resolved into discrete sources now.

3. Optical identifications Early optical identification programs concentrated on medium-deep ROSAT P S P C observations in AAT deep optical QSO fields and could quickly identify an impressive fraction of faint X ray sources as classical broad-line AGNs (mainly QSOs)[14,15]. There was, however, mounting evidence that a new class of sources might start to contribute to the XRB at faint X - r a y fluxes. The faintest X ray sources in ROSAT deep surveys on average show a harder s p e c t r u m than the identified QSOs[1]. In medium-deep point-

602

G. Hasinger /Nuclear Physics B (Proc. Suppl.) 69/1-3 (1998) 600-609

Lockman Hole Deep Survey "~.. "..... "-

• '"...... •

"

\

~

qk

' ".~.



m •

'~.. •

.

1.OE-14

+ ....... + o

o

18" ; + *++ +~m+~+++...,÷.,.

d-.. '"..

15

O~b'.,.q

~ +

o-I+ "... + "'.. -I-

20

R Magnitude

Figure 2. Correlation between X-ray flux and optical R magnitude for all objects in the Lockman Hole survey. Filled circles give spectroscopically identified AGN, open circles unspecified galaxies, open squares candidate clusters and group of galaxies. Asterisks show coronal stars and plus signs unidentified objects, in this case always the brightest optical counterpart in the 90~o error circle is shown. The catalogue of sources down to X-ray fluxes of 5.5 x 10-15 has been published[4]. Fainter objects are subject of the ongoing multislit spectroscopy and therefore provides a fair sample of the data.

ings a number of optically "innocent" naxrowemission line galaxies (NELGs) at moderate redshifts (z<0.4) were identified as X - r a y sources, which was in excess of those expected from spurious identifications with field galaxies[16,15,5]. Roche et al.[17] have found a significant correlation of X - r a y fluctuations with optically faint galaxies. Finally, in an a t t e m p t to push optical identifications to the so far faintest X - r a y fluxes, McHardy et al.[6] claim t h a t the surface density of broad-line AGN flattens dramatically at fluxes below 5 × 10 -15 e r g c m - 2 s - 1 , while the N E L G number counts still keep increasing, so t h a t those would eventually dominate the XRB below a flux of 10 -15 e r g crn - 2 s - 1 . While this is obviously an interesting possibility, it is useful to remind t h a t all these findings are based on identifications near the limit of deep P S P C surveys, at fluxes where our simulations

suggest that the P S P C d a t a start to be severely confused, with some likelihood of misidentification[3]. Also, spectroscopic optical identifications in these samples are limited to R < 22. In the Lockman Hole Deep survey, optical counterparts have magnitudes in the range R=19-24. With the excellent H R I positions (of order 2-4 arcsec) typically only one counterpart is within the X-ray error box. In Fig. 2 the X-ray fluxes of our sources are plotted against the magnitude of their optical counterparts or, in case of no identification, the brightest optical candidate in the error box. Long-slit and multislit spectra of the 1-3 candidates closest to the X-ray source have been taken using the Palomar 200" 4shooter and the Keck LRIS instruments. Reliable spectroscopic identifications are now practically complete[4] in a 0.30 deg 2 field to a flux limit of 1.1 • 10 -14 cgs and in the central 0.14 deg 2 of the

G. Hasinger/Nuclear Physics B (Proc. Suppl.) 69/1-3 (1998) 600-609

field down to a flux limit of 5 . 5 . 1 0 -1~ cgs and in a smaller area, which we continue to work on, down to 10 -15 cgs. The majority of the optical counterparts are active galactic nuclei with broad emission lines (QSOs and Seyfert galaxies) in the redshift range 0.08-4.5, whereas only a very small fraction of non-AGN N E L G s is found. D a t a on the highest redshift X-ray selected QSO will be presented elsewhere[18]. A large fraction of the faint X-ray sources are optically resolved low luminosity AGN (Seyfert galaxies), some of which show clear evidence of gas and dust obscuration in the X-ray and optical band (i.e. Seyfert 1.5-2 galaxies), which are very hard to select by any other means than X-rays. The d a t a of McHardy et al.[6] is roughly consistent with our findings, if we consider three selection/incompleteness effects, which are partly discussed in their work: a.) their optical identifications are restricted to R < 22, while we have optical counterparts as faint as R=24. b.) they call some fraction of objects NELGs, which due to our higher quality spectra reveal some evidence for AGN activity, and c.) some fraction of their N E L G s may be misidentified due to source confusion. 4. A S C A

Surveys

and

the

hard

X-ray

background The X-ray background has a significantly harder spectrum t h a n that of the sources resolved in the soft band. This led to the assumption that a large fraction of the background flux is due to obscured AGN, as originally proposed by Setti and Woltjer[19]. A model following the unified AGN schemes, assuming a mixture of absorbed and unabsorbed AGN spectra folded in with cosmological AGN evolution models, is quite successfully explaining the shape of the background spect r u m over the whole X-ray band as well as a number of other observational constraints[9]. This model predicts that in the hard X - r a y band most of the contribution to the XRB conies from significantly absorbed objects, which are almost absent in the soft band, even at the faintest ROSAT limit. As a consequence, a significant test for this model would be the comparison of its pre-

603

dictions with the results of optical identifications of a complete sample of X ray sources selected at low fluxes in the hard X - r a y band. Hard X-ray surveys are just becoming available now. A large sky survey survey (LSS) near the North Galactic Pole was performed with the ASCA satellite[20,21]. The limiting X-ray flux of the LSS is 10 -13 erg s -1 cm -2 (2 10 keV) (300 times deeper than the HEAO-1 A2 survey which was the deepest survey in the 2 10 keV baud). The LSS covers an area of 6 square degrees which is the widest among the surveys with this flux level in the 2-10 keV band. 50 sources are detected in the 2-10 keV band above a significance of 4 sigma. The average photon index of the sources is close to that of the CXB in the 2-10 keV band[20]. Therefore tile LSS seems to be disclosing the presence of the postulated hard X-ray population. Recently, ASCA d a t a have been used [22] to derive the 2-10 keV log(N)-log(S) down to fluxes slightly below l O - 1 3 e r g c m - 2 s -1. This data, based on 60 X - r a y sources detected in ASCA images, is shown in Figure 3 together with the predictions for various classes of objects from the Comastri et al. model. As discussed in Cagnoni et al. [22], if one uses the ROSAT log(N)-dog(S) and the average spectral properties of the ROSAT sources to predict the ASCA log(N)-log(S), the prediction falls a factor ~ 2 below the d a t a at a flux of the order of ~ l O - a 3 e r g crn - 2 s -Z . Oil the other hand, as shown in the figure, the d a t a are in good agreement with the predictions of the Comastri et al. model. The dashed line in Figure 3 shows that a significant fraction of the AGNs in this ASCA survey, and an even larger fi'action at fainter fluxes, should be substantially absorbed and therefore their optical counterparts are expected to have optical spectra typical of Seyfert 2 galaxies. A program to optically identi~" these sources has already started, but tile large positional uncertainty ( ~ 2 arcrnin) together with tile relatively faint expected magnitudes of the optical counterparts will make it difficult to obtain unambiguous identifications. The deepest ASCA surveys[13,25,26] resolve source counts down to limiting 2-10 keV fluxes of 5 x 1() - H erg c m - 2 s -1 . At source densities

G. Hasinger /Nuclear Physics B (Proc. Suppl.) 69/1-3 (1998) 600-609

604

1000 ~ " ~ ' ....... \ ~ . . '-,. ~

lOO = .

........ UnabsorbedAGNs ___ Absorbed AGNs . . . . . . . Clusters

".,.. "x,x.

.=

10

", • ~x.\ \ x ~

%

,

I

-o

1 c. A

0.1

0,01

0.001

' \~-~ " ' , - ~ . . . . . . . .

lO-IS

lO-t4

......

. . . . . . . .

h

I0-I~

I0-~

. . . . . . . .

, 10-11

,'..,N,

,':,-.

lO-=O

S(2 - 10 keY) [cgs]

Figure 3. The 2-10 keV log(N)-log(S) (from [23]. The circles represent the counts derived from the 60 ASCA sources discussed in [22]; the square shows the preliminary results of a different ASCA survey in a more limited sky area [25]; the triangle shows the extragalactic surface density from the HEAO-1 survey [24]. The curves show the predictions for various classes of objects from the model discussed in [9]. of typically 100 deg - 2 these surveys are heavily confused due to ASCA's limited angular resolution. Therefore optical identifications will remain ambiguous, until much higher angular resolution data is available. Results of an ASCA deep survey in the Lockman Hole, where due to the existence of much deeper ROSAT HRI observations practically all ASCA sources have well-defined soft Xray counterparts, will be presented elsewhere[27]. An analysis of the spatial fluctuations in deep ASCA images[28] probes the 2-10 keV X-ray source counts down to a flux limit of 2 x 10 -14 er9 c m - 2 s - 1 , using special care to correct for the extended wings of the ASCA point spread function through simulations. The source counts are found to be consistent with [13,25,22], however, they disagree with the deep survey counts by Georgantopoulos et al. [26]. This analysis resolved about 35% of the extragalactic 2-10 keV X-ray background.

5. A n e w A G N X-ray l u m i n o s i t y function (XLF) Using data from medium deep ROSAT fields combined with the Einstein Medium Sensitivity Survey[29], Boyle et al.[30] could derive the AGN XLF and its cosmological evolution. Their data is consistent with pure luminosity evolution proportional to ( l + z ) 2'7 up to a redshift z,~a= ~ 1.5, similar to what was found previously in the optical range. This result has been confirmed and improved later on by more extensive or deeper studies of the AGN XLF, e.g. the RIXOS project[31] or the UK deep survey project[32]. All these studies agree that at most half of the faint X - r a y source counts can be explained by classical broadline AGN based on the pure luminosity evolution models. However, the limited information at bright X-ray fluxes and in particular the uncertain crosscalibration between the ROSAT and Einstein surveys severely limits the accuracy of the AGN XLF[30,31].

G. Hasinger/NuclearPhysics B (Proc. Suppl.) 69/1-3 (1998) 600-609

605

ROSAT AGN . . . .

i

. . . .

i

. . . .

a"

i

.

.

.

.

.

.

.

.

i

. . . .

i

. . . .

,; .....

~e~

"%""

"

© e¢~

",.

'\

\

-8

!

'\

.\ \.

\ \

\ \.

-lO

'\

~I1 \.\

. . . .

2 . . . .

4

I . . . .

I . . . .

44

I . . . .

I . . . .

46

t

. . . .

48

log LX (0.5-2 keV) [erg/s] Figure 4. The AGN X-ray luminosity function derived from a joint analysis of the ROSAT Deep and Ultradeep Survey, RIXOS and the ROSAT Bright Survey. The comoving AGN volume density is displayed as a bivariate function of luminosity and redshift. Redshift shells are 0-0.2 (filled circles), 0.2-0.4 (open circles), 0.4-0.8 (filled squares), 0.8-1.6 (open squares), 1.6-2 (filled triangles), and 2-4.5 (asterisks).

We can now determine a new AGN soft Xray luminosity function based on ROSAT surveys alone and covering a very wide range of limiting fluxes. For this purpose the deep survey sample described above has been combined with optical identifications from shallower wide angle surveys, i.e. the ROSAT Bright Survey (RBS[33]), derived flom the R O S A T All-Sky Survey bright source catalogue[11], and the RIXOS AGN sample[M]. The accuracy of the relative flux scale for these surveys is d e m o n s t r a t e d by the logN-logS function in Fig. 1, where the different surveys agree within --~ 10% in the overlap regions over a flux range of a b o u t six orders of magnitude. Fig. 4 shows the binned AGN XLF in different redshift shells. Consistent with Boyle et al., we find strong cosmological evolution in the sense that high-redshift AGN are much more abundant or more luminous than their local counterparts. C o n t r a r y to the Boyle et al. findings,

however, the new XLF is not consistent with pure luminosity evolution. For the first time we see evidence for strong cosmological evolution of the space density of low-luminosity AGN (e.g. Seyfert galaxy) XLF out to a redshift 1 -2, incompatible with pure luminosity evolution. Surprisingly, however, a pure density evolution model of the form ¢ ( z ) = ¢ ( 0 ) . (1 + can fit the d a t a well. Fig. 5 shows the luminosity function, de-evolved under the assumption of pure density evolution. A smoothed broken power law model of the fl)rm 2- 10 -~

O(L, z)/dlogL = (1 + z) s°3 /2.03 +/0.r,L [Mpc -:~] has been fit to the de-evolved XLF out, to redshifts of 1.92. Here l = Lx/5.1043. The model for z=0 is shown in figure 5 by a solid line. The model for the z=2 XLF over the observed luminosity range is shown by the thick dashed line, its extrapolation to lower hnninosities by the (lash-

606

G. Hasinger /Nuclear Physics B (Proc. Suppl.) 69/1-3 (1998) 600-609

ROSAT AGN de-evolved XLF

".%

-\

qb~...

\

\ ,..1 "

-to

-10

38

40

42

44

46

48

log LX (0.5-2 keY) [erg/s]

Figure 5. X-ray Luminosity functions for AGN and galaxies from ROSAT surveys. The local galaxy luminosity function (open circles) has been derived from the ROSAT Bright Survey[33] and from a volume-limited sample of local galaxies[34]. A log-normal distribution has been fit to the galaxy XLF (dotted curve). The AGN luminosity function (filled circles) has been de-evolved using a pure density evolution model (see text). A smoothed broken power law model (solid line) has been fit to the data. The dashed curve indicates the model AGN luminosity function observed in the redshift range 2-3, its dash-dotted extension to lower luminosities is the model based on pure density evolution (see text). dotted line. Integrating the luminosity function to z ..~ 2 we we actually overpredict the soft X-ray background flux and the soft X-ray logN-logS function and fluctuations. One has to stress however, that this is only a first-order treatment of the data. The effect of intrinsic X-ray absorption and the detailed shape of the AGN spectra has to be included in the derivation of the XLF in order to obtain a self-consistent population synthesis model for the XRB spectrum[9]. Also, the so far unobserved low luminosity range of the high-redshift AGN XLF provides a significant contribution to the XRB and the faint number counts and is therefore crucial in understanding the composition of the XRB. Even deeper observations, preferrably at higher X-ray energies will thus be necessary.

6. D o e s e v e r y g a l a x y c o n t a i n a b l a c k h o l e ? This question has originally been posed by Rees[35]. With recent data from HST and from the deep X-ray surveys we can now shed new light on this question. In fig. 6 we plot the volume density evolution of X-ray selected AGN (this work) and optically selected AGN as a function of redshift[36]. The AGN space density shows a large variation with redshift (a factor of several hundred) with a marked peak at z=2-3. Note that only the AGN space density derived from a pure density evolution model shows such a large variation against cosmic time, while the volume luminosity derived from a pure luminosity evolution model shows a much smaller variation[37]. The extrapolated comoving volume density of X-ray selected QSO and Seyfert galaxies at z=2-3, al-

G. Hasinger/Nuclear Physics B (Proc. Suppl.) 69/1-3 (1998) 600-609

607

AGN density vs. star forming rate I ....

' ....

I .........

I . . . . . . .

' ' 1 ' ' ' '

....

'1

. . . . . . . . .

I . . . . . . . . .

I

/+ ?+

-.

. ..7''"...:..iii(..

o

.....

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

e~

T II

0

....

Illll

IIIIltllI,JII

....

[I

.......

2

I , , , , , , , J , l , , , , , , , , , I

4

redshift Figure 6. The filled circles give the volume density of X-ray selected AGN referred to an X-ray luminosity of logLx = 44.25. The dashed curve sketches the volume density of optically selected QSO[36] with MB < --26, increased by a factor of 250 to account for the different luminosities in the different wavebands. The open circles show the star formation rate as a function of redshift from Madau[40] and Connolly et al.[41], corrected for dust obscuration[42]. The dotted curves give models for the star formation rate derived from the chemical evolution of Ha clouds[43]. The latter two curves are in arbitrary units, scaled to the AGN density at low redshifts.

beit the systematic uncertainties discussed above, approaches t h a t of the low-redshift normal galaxies (see Fig. 5) as well as t h a t of the recently discovered population of Lyman-limit galaxies at high redshifts[38]. This indicates that a substantial fraction of normal galaxies may host a central supermassive black hole, which turns into an AGN as soon as it is fuelled, e.g. by interactions. The amount of interactions is much larger at higher redshifts t h a n today, which can explain the sharp drop of activity towards low z. Dorm a n t remnants of AGN are indeed found in almost every nearby galaxy with a spheroidal component. A tight relation between black hole mass and bulge mass has been found, indicating that there is a causal connection between the size of the central black hole and the number of stars in a galaxy [39]. Recently the history of global star formation in

the universe been determined e.g. through observations in the Hubble deep field, but also from the chemical evolution of Ly~ clouds. Fig. 6 gives the star forming rate in arbitrary units as a function of redshift[40,41], corrected for dust obscuration following Pettini et a1.[42]. This flmction shows only a moderate variation with cosmic time, with a possible peak around a redshift of unity and a steep decline towards lower redshifts, the shape of which is strikingly similar to the low-redshift behaviour of the AGN volume density. The pronounced peak of the AGN density indicates, that the bulk of the supermassive black holes in the universe must have been present by a redshift of ~3, while the bulk of star formation was still going on until a redshift of 1. A central black hole must therefore have been among the first building blocks of a typical galaxy. In this picture the X-ray background is ,just the to-

608

G. Hasinger/Nuclear Physics B (Proc. Suppl.) 69/1-3 (1998) 600-609

tal emission of all black hole accretion processes, integrated over cosmic space and time. Following the argument by Soltan[44], the total mass accreted onto black holes integrated over cosmic time can be estimated from the AGN number counts and thus from the X-ray background radiation. This corresponds to an average mass in dormant black holes of 107-10SMo per galaxy[45] and is surprisingly close to the average black hole mass measured for local galaxies[39]. The scenario described here suggests an attractive solution to the X-ray background puzzle, where the XRB is just the "echo" of mass accretion processes onto supermassive black holes in the centers of nearly all galaxies. However, the acid test of this picture has to come from deep survey observations at higher energies, where the bulk of the XRB energy density resides and where large numbers of obscured AGN are to be detected. New X-ray missions to be launched in the next few years, as for example AXAF and XMM will provide faint samples of hard X-ray selected sources, with unique positional uncertainty (of the order of 1-2 arcsec for AXAF and 5-10 arcsec for XMM). These faint hard samples will be complemented by the bright all-sky sample from the ABRIXAS survey[?]. The optical identifications of these sources with the 8-10m class telescopes will be able to conclusively test the predictions of our current models and, hopefully, finally settle the question of the X-ray background. I acknowledge the grant 50 OR 9403 5 by the Deutsche Agentur fiir Raumfahrtangelegenheiten (DARA). I also warmly thank my collaborators in the ROSAT Deep Survey project, Riccardo Giacconi, Maarten Schmidt, Joachim Triimper and Gianni Zamorani for fruitful discussions and the permission to discuss data in advance of publication. In particular I am grateful to M. Schmidt and T. Miyaji for help with the luminosity function. REFERENCES

1. Hasinger G. et al., 1993, A&A 275, 1 2. Branduardi-Raymont G. et al., 1994, MNRAS 270, 947 3. Hasinger et al., 1998, A&A, 329, 482 4. Schmidt M. et al. 1998, A&A, 329,495

5. Griffiths R.E. et al., 1996, MNRAS 281, 71 6. McHardy I.M. et al., 1997, MNRAS, submitted, (astro-ph/9703163) 7. Hasinger G. 1996, A&A Suppl. 120, C607 8. Hasinger G., 1998, AN 319,37 9. Comastri A., Setti G., Zamorani G. and Hasinger G. 1995, A&A, 296, 1 10. Mason, K. et al., 1996, priv. comm 11. Voges, W., et al., 1995, IAU Circ # 6102 12. de Ruiter H. et al., 1997, A&A 319, 7 13. Inoue H. 1996, MPE report 263,323 14. Shanks T. et al. 1991, Nature 353, 315 15. Georgantopoulos I. et al., 1996, MNRAS 280, 276 16. Boyle B.J. et al., 1995, MNRAS 272,462 17. Roche N. et al., 1995, MNRAS 273, 15 18. Schneider D., et al., 1998, AJ in press 19. Setti G. and Woltjer L. 1989, A&A 224, L21 20. Ueda et al., 1998a, Nature in press 21. Ueda et al., 1998b, AN 319, 47 22. Cagnoni I., Della Ceca R. and Maccacaro T. 1998, ApJ, 493, 54 23. Ogasaka Y. 1997, in X ~ray Imaging and Spectroscopy of Cosmic Hot Plasmas, F. Makino and K. Mitsuda eds., 25. 24. Georgantopoulos I. et al., 1997, MNRAS 291, 203 25. Hasinger G. and Zamorani G., 1997, astroph/9712342 26. Piccinotti G. et al. 1982, ApJ 253,485 27. Ishisaki et al., 1998, in prep. 28. Gendreau K.C., Barcons X. and Fabian A.C., MNRAS in press (astro-ph/9711083) 29. Gioia I.M. et al., 1990, ApJ Suppl. 72, 567 30. Boyle B.J. et al., 1994, MNRAS 260, 49 31. Page M.J. et al., 1996, MNRAS 281,579 32. Jones L.R. et al., 1997, MNRAS 285,547 33. Hasinger et al., 1997, AN 318,329 34. Schmidt K.-H., Boller T., Voges W., MPE report 263,395 35. Rees M., 1989, Rev. Mod. Astronomy 2, 1 36. Schmidt M., Schneider D.P., Gunn J.E., 1995, AJ 110, 68 37. Boyle B.J., Terlevich R.J., 1998, MNRAS 293, 49p 38. Steidel, C. et al., 1996, AJ 112, 352 39. Faber S.M. et al., 1997, AJ 114, 1771 40. Madau P., 1996, astro-ph 9612157

G. Hasinger/Nuclear Physics B (Proc. Suppl.) 69/1-3 (1998) 600-609

41. Connolly A.J., Szalay A.S., Dickinson M., Subbarao M.U., Brunner R.J., 1997, ApJ 486, Lll 42. Pettini M., et al., 1997, ApJ 478, 536 43. Pei Y.C., Fall S.M., 1995, ApJ 454, 69 44. Soltan A., 1982, MNRAS 200, 115 45. Chokshi A., Turner E.L., 1992, MNRAS 259, 421 46. Triimper J., Hasinger G., Staubert R., 1998, AN 319, 113

609