Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://hea.iki.rssi.ru/integral06/papers/D02_Beckmann01-19Jan07_20:50:38.pdf Äàòà èçìåíåíèÿ: Fri Jan 19 20:50:38 2007 Äàòà èíäåêñèðîâàíèÿ: Mon Oct 1 23:26:48 2012 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: propulsion

THE HARD X-RAY LUMINOSITY FUNCTION OF INTEGRAL DETECTED AGN
V. Beckmann1 , C. R. Shrader1 , N. Gehrels1 , S. Soldi2 , and N. Produit
1 2

NASA Goddard Space Flight Center, Astrophysics Science Division, Code 661, Greenbelt, MD 20771, USA 2 ´ INTEGRAL Science Data Centre, Chemin d' Ecogia 16, 1290 Versoix, Switzerland

ABSTRACT We have compiled a complete extragalactic sample based on 25, 000 deg2 to a limiting flux of 3 â 10-11 ergs cm-2 s-1 ( 7, 000 deg2 to a flux limit of 10-11 ergs cm-2 s-1 ) in the 20 ­ 40 keV band with INTEGRAL. We have constructed a detailed exposure map to compensate for effects of non-uniform exposure. The flux-number relation is best described by a power-law with a slope of = 1.66 ± 0.11. The integration of the cumulative flux per unit area leads to f20-40 keV = 2.6 â 10-10 ergs cm-2 s-1 sr-1 , which is about 1% of the known 20 - 40 keV X-ray background. We present the first luminosity function of AGN in the 20­40 keV energy range, based on 38 extragalactic objects detected by the imager IBIS/ISGRI on-board INTEGRAL. The luminosity function shows a smoothly connected two power-law form, with an index of 1 = 0.8 below, and 2 = 2.1 above the turn-over luminosity of L = 2.4 â 1043 ergs s-1 . The emissivity of all INTEGRAL AGN per unit volume is W20-40 keV (> 1041 ergs s-1) = 2.8 â 1038 ergs s-1 h3 Mpc-3 . These results are con70 sistent with those derived in the 2 - 20 keV energy band and do not show a significant contribution by Comptonthick objects. Because the sample used in this study is truly local (z = 0.022), only limited conclusions can ¯ be drawn for the evolution of AGN in this energy band. But the objects explaining the peak in the cosmic X-ray background are likely to be either low luminosity AGN (LX < 1041 ergs s-1 ) or of other type, such as intermediate mass black holes, clusters, and star forming regions, or they have to be at higher redshift. Key words: galaxies: active; X-rays: galaxies; surveys; galaxies: Seyfert.

exposures by ROSAT have revealed an extragalactic population of mainly broad line AGN, such as type Seyfert 1 and quasars [14, 27]. In the 2 - 10 keV range surveys have been carried out with ASCA (e.g. Ueda et al. 2001), XMM-Newton (e.g. Hasinger 2004), and Chandra (e.g. Brandt et al. 2001) and have shown that the dominant extragalactic sources are more strongly absorbed than those within the ROSAT energy band. For a summary on the deep X-ray surveys below 10 keV see Brandt & Hasinger (2005). At higher energies the data become more scarce. Between a few keV and 1 MeV, no all-sky survey using imaging instruments has been performed to date. The Rossi X-ray Timing Explorer (RXTE) sky survey in the 3 - 20 keV energy band revealed about 100 AGN, showing an even higher fraction of absorbed (NH > 1022 cm-2 ) sources of about 60% [25]. INTEGRAL offers an unprecedented > 20 keV collecting area and state-of-the-art detector electronics and background rejection capabilities. Notably, the imager IBIS enables us now to study a large portion of the sky. A first catalogue of AGN showed a similar fraction of absorbed objects as the RXTE survey [2]. The Burst Alert Telescope (BAT) of the Swift mission [11] operates in the 15 ­ 200 keV band and uses a detector similar to IBIS/ISGRI, but provides a field of view about twice the size. The BAT data of the first three months of the mission provided a high galactic latitute survey, including 44 AGN [20]. Within this sample a weak anti-correlation of luminosity versus intrinsic absorption was found as previously found in the 2 - 10 keV band [31, 17], revealing that most of the objects with luminosities LX > 3 â 1043 ergs s-1 show no intrinsic absorption. Markwardt et al. (2005) also pointed out that this luminosity corresponds to the break in the luminosity function. Related to the compilation of AGN surveys in the hard X-rays is the question of what sources form the cosmic X-ray background (CXB). While the CXB below 20 keV has been the focus of many studies, the most reliable measurement in the 10 - 500 keV has been provided by the High Energy Astronomical Observatory (HEAO 1), launched in 1977 [21]. The most precise measurement provided by the UCSD/MIT Hard X-ray and Gamma-Ray instrument (HEAO 1 A-4) shows that the CXB peaks at an energy of about 30 keV [21, 13]. The isotropic nature of the X-ray background points to an extragalactic origin, and as the brightest persistent sources are AGN, it was suggested early on that those objects are the main

1. INTRODUCTION The Galactic X-ray sky is dominated by accreting binary systems, while the extragalactic sky shows mainly active galactic nuclei (AGN) and clusters of galaxies. Studying the population of sources in X-ray bands has been a challenge ever since the first observations by rocket borne X-ray detectors [12]. At soft X-rays (0.1 ­ 2.4 keV) deep


source of the CXB (e.g. Setti & Woltjer 1989). In the soft X-rays this concept has been proven to be correct through the observations of the ROSAT deep X-ray surveys, which showed that 90% of the 0.5 - 2.0 keV CXB can be resolved into AGN [27]. At higher energies (2 - 10 keV), ASCA and Chandra surveys measured the hard X-ray luminosity function (XLF) of AGN and its cosmological evolution. These studies show that in this energy range the CXB can be explained by AGN, but with a higher fraction of absorbed (NH > 1022 cm-2 ) objects than in the soft X-rays (e.g. Ueda et al. 2003). A study based on the RXTE survey by Sazonov & Revnivtsev (2004) derived the local hard X-ray luminosity function of AGN in the 3­20 keV band. They showed that the summed emissivity of AGN in this energy range is smaller than the total X-ray volume emissivity in the local Universe, and suggested that a comparable X-ray flux may be produced together by lower luminosity AGN, non-active galaxies and clusters of galaxies. Using the HEAO 1-A2 AGN, Shinozaki et al. (2006), however, obtained a local AGN emissivity which is about twice larger than the value of Sazonov & Revnivtsev (2004) but consistent with the estimates by Miyaji et al. (1994) which was based on the cross-correlation of the HEAO 1-A2 map with IRAS galaxies. With the on-going observations of the sky by INTEGRAL, a sufficient amount of data is now available to derive the AGN hard X-ray luminosity function. In this paper we present analysis of recent observations performed by the INTEGRAL satellite, and compare the results with previous studies. In Section 2 we describe the AGN sample and in Section 3 the methods to derive the number-flux distribution of INTEGRAL AGN are presented together with the analysis of their distribution. Section 4 shows the local luminosity function of AGN as derived from our data, followed by a discussion of the results in Section 5. Throughout this work we applied a cosmology with H0 = 70 km s-1 Mpc-1 (h70 = 1), k = 0 (flat Universe), matter = 0.3, and 0 = 0.7, although a 0 = 0 and q0 = 0.5 cosmology does not change the results significantly because of the low redshifts in our sample.

used here, but it ensures completeness of the sample at a significance limit of 5 (see Section 3). The list of 73 sources has been reported in Beckmann et al. (2006c). 22 of the sources have Galactic latitudes -10 < b < +10 (14, if we only consider the sources with significance 5 ). In addition to the sample presented here, 8 new INTEGRAL sources with no identification have been detected in our survey with a significance of 5 . These un-identified sources, most of them in the Galactic Plane, are not included in this work. The significances used have been derived from the intensity maps produced by the OSA software. Fluxes are determined by integrating the best-fit spectral model over the 20­40 keV bandpass. The uncertainty in the absolute flux calibration is about 5%. The extracted images and source results are available in electronic form1.

3. NUMBER-FLUX DISTRIBUTION OF INTEGRAL AGN 3.1. Completeness of the Sample In order to compute the AGN number-flux relation it is necessary to have a complete and unbiased sample. Towards this end, one must understand the characteristics of the survey, such as the sky coverage and completeness for each subset of the total sample. Because of the in-homogeneous nature of the survey exposure map, we applied a significance limit rather than a flux limit to define a complete sample. The task is to find a significance limit which ensures that all objects above a given flux limit have been included. To test for completeness, the Ve /Va -statistic has been applied, where Ve stands for the volume that is enclosed by the object, and Va is the accessible volume, in which the object could have been found (Avni & Bahcall 1980). In the case of no evolution Ve /Va = 0.5 is expected. This evolutionary test is applicable only to samples complete to a well-defined significance limit. It can therefore also be used to test the completeness of a sample. We performed a series of Ve /Va -tests to the INTEGRAL AGN sample, assuming completeness limits in the range of 0.5 up to 16 ISGRI 20 ­ 40 keV significance. For a significance limit below the true completeness limit of the sample one expects the Ve /Va -tests to derive a value Ve /Va < Ve /Va true , where Ve /Va true is the true test result for a complete sample. Above the completeness limit the Ve /Va values should be distributed around Ve /Va true within the statistical uncertainties. It appears that the sample becomes complete at a significance cutoff of approximately 5 , which includes 38 AGN. The average value is Ve /Va = 0.43 ± 0.05. This is consistent with the expected value of 0.5 at the 1.5 level, suggesting no evolution and a uniform distribution in the local universe. It is unlikely that cosmological effects have an influence on the result, as the average red1

2. THE INTEGRAL AGN SAMPLE Observations in the X-ray to soft gamma-ray domain have been performed by the soft gamma-ray imager (20­ 1000 keV) ISGRI [18] on-board INTEGRAL. The data used here are taken from orbit revolutions 19 137 and revolutions 142 - 149. The list of sources was derived from the analysis as described in Beckmann et al. (2006a,2006c). The analysis was performed using the Offline Science Analysis (OSA) software version 5.0 distributed by the ISDC (Courvoisier et al. 2003). Additional observations performed later led to further source detections within the survey area. We extracted spectra at those positions from the data following the same procedure. It is understood that most of those objects did not result in a significant detection 3 in the data set

http://heasarc.gsfc.nasa.gov/docs/integral/spi/pages/agn.html


shift in the sample is z = 0.022, with a maximum redshift ¯ of z = 0.13. A positive cosmological evolution would result in a slightly higher value than 0.5. We would like to remind that we use the Ve /Va test not to determine any cosmological effects, but use it to see at what significance level it returns a stable value.

3.2. Deriving the Area Corrected Number-Flux Distribution

A correct representation of the number flux distribution (i.e. log N>S versus log S , see Beckmann et al. 2006b) for the sample presented here must account for different exposure times comprising our survey, and the resulting sensitivity variations. We determine here the number density and thus the number of AGN above a given flux has to be counted and divided by the sky area in which they are detectable throughout the survey. We therefore first determined the exposure time in 64, 620 sky elements of 0.63 deg2 size within our survey. In each sky bin, the exposure is the sum of each individual exposure multiplied by the fraction of the coded field of view in this particular direction. The dead time and the good time intervals (GTI) are not taken into account but the dead time is fairly constant (around 20%) and GTI gaps are very rare in IBIS/ISGRI data. We excluded those fields with an exposure time less than 2 ks, resulting in 47, 868 sky elements with a total coverage of 9.89 sr. The flux limit for a given significance limit should be a function of the square root of the exposure time, if no systematic effects apply, but this assumption cannot be made here. The nature of coded-mask imaging leads to accumulated systematic effects at longer exposure times. In order to achieve a correlation between the exposure time and the flux limit, we therefore used an empirical approach. For each object we computed what we will call its 5 equivalent flux f5 , based on its actual flux fX and its significance s: f5 = fX 5/s. We found a correlation between these f5 values and exposure times, which has a scatter of < 0.2 dex. The correlation was then fitted by a smooth polynomial of third degree. This function was then used to estimate the limiting flux of each individual survey field. It must be noted that the individual limits are not important, but only the distribution of those flux limits. Based on the flux limits for all survey fields, we are now able to construct the number flux distribution for the INTEGRAL AGN, determining for each source flux the total area in which the source is detectable with a 5 detection significance in the 20 - 40 keV energy band. The resulting correlation is shown in Figure 1. The errors on the data points indicate 1 uncertainties in the flux of the objects and the Poissonian errors for the number of objects per steradian.

Figure 1. Number flux distribution per steradian of INTEGRAL AGN with a detection significance > 5 . Blazars have been excluded. The maximum likelihood slope as described in Section 3.3 is 1.66 ± 0.11. 3.3. The Slope of the Number-Flux Distribution We applied a maximum-likelihood (ML) algorithm to our empirical number-flux distribution to obtain a power-law approximation of the form N (> S ) = K · S - . We note that we are fitting the "integrated" N (> S ) function, as distinct from the "differential" number-flux function. The latter entails binning the data, and thus some loss of information is incurred. The advantage of fitting the differential distribution is that a simple least squares procedure may be employed. However, given the modest size of our sample, the expected loss of accuracy was considered unacceptable. For this analysis, we used the complete sub-sample of 38 sources for which the statistical significance of our flux determinations was at a level of 5 or greater. The dimmest source among this sub-sample had fX = 5.6 â 10-12 ergs cm-2 s-1 , and the brightest had fX = 3.2 â 10-10 ergs cm-2 s-1 . We derived a ML probability distribution, which can be approximated by a Gaussian, with our best fit parameters of = 1.66 ± 0.11. A normalization of K = 0.44 sr-1 (10-10 ergs cm-2 s-1 ) was then obtained by performing a least-squares fit, with the slope fixed to the ML value. The uncertainties in the final log N - log S primarily manifest in the normalization and it should not affect the slope significantly. Furthermore, the uncertainty in the detection limit will affect mainly the low flux end of the Log(N) ­ Log(S) distribution. The high flux end is less sensitive to scatter, since it is based on a larger sky area. To make a more quantitative assessment, we have recomputed the ML Log(N) ­ Log(S) calculation for scenarios in which the exposure time ­ flux limit curve shifted in amplitude and pivoted about the 700 ks point where we have the highest density of measurements. For those scenarios, we found that the inferred Log(N) ­ Log(S) slope varied by less than about 5%, which is contained within the range of our quoted 1-sigma uncertainty. The amplitude varied by as much as 7% in the extreme case, but for the pivoted cases, by only a few percent. We thus conclude that the maximum uncertainty resulting from possible systematics in our ef-


fective area correction is bounded by about 5% in slope and 7% in amplitude.

4. THE LOCAL LUMINOSITY FUNCTION OF AGN AT 20 ­ 40 KEV The complete sample of INTEGRAL AGN with a detection significance 5 also allows us to derive the density of these objects in the local Universe as a function of their luminosity. In order to derive the density of objects above a given luminosity, one has to determine for each source in a complete sample the space volume in which this source could have been found considering both the flux limit of each survey field and the flux of the object. We have again used the correlation between exposure time and flux limit as discussed in the previous section in order to assign a 5 flux limit to each survey field. Then the maximum redshift zmax at which an object with luminosity LX would have been detectable in each sky element was used to compute the total accessible volume
N

Va =
i=1

i Vi [z 4

Figure 2. Differential luminosity function of AGN with log LX = 0.5 binning. The line shows a fit to a smoothly connected two power-law function with a turnover luminosity at L = 2.4 â 1043 ergs s-1 . Blazars have been excluded. The density describes the number of objects per Mpc3 in a given luminosity interval. of h-2 ergs s-1. The 1 errors have been determined 70 by applying a Monte Carlo simulation which simultaneously takes into account the flux errors on the individual sources, the error induced by deriving an average luminosity per bin, and the statistical error of the density based on the number of objects contributing to the density value. Each simulated data set included 9 luminosity values with a density value for each of them. These values where then fit by the smoothly connected two power-law function as described above. The scatter in the resulting parameters gave the error estimates as shown above. The parameter values describing the differential luminosity function are consistent with values derived from the 2 - 10 keV XLF of AGN as shown by e.g. Ueda et al. (2003), La Franca et al. (2005), and Shinozaki et al. (2006). For example the work by Ueda et al. (2003) reveals for a pure density evolution model the same values (within the error bars) for 1 and 2 , but a higher log L = 44.11 ± 0.23. The higher value can be easily explained by the different energy bands applied. A single power law with photon index of = 2 in the range 2 - 40 keV would lead to L(2-10 keV) /L(20-40 keV) = 2.3, assuming no intrinsic absorption. This has, of course, no implications for the XLF at higher redshifts. The values are also consistent with the luminosity function for AGN in the 3 - 20 keV band as derived by Sazonov & Revnivtsev (2004) from the RXTE all-sky survey. Information about intrinsic absorption is available for 32 of the 38 objects (89%) from soft X-ray observations. This enables us to derive the luminosity function for absorbed (NH 1022 cm-2 ) and unabsorbed sources, as shown in Figure 3. The absorbed sources have a higher density than the unabsorbed sources at low luminosities, while this trend is inverted at high luminosities. The lu-

max,i

(LX )]

(1)

with N being the number of sky elements in which the object would have been detectable and i the solid angle covered by sky element i, and Vi the enclosed volume based on the maximum redshift at which the object could have been detected in this sky element. Figure 2 shows the luminosity function in differential form. Blazars have been excluded because their emission is not isotropic. The redshifts in the sample range from z = 0.001 to z = 0.129 with an average redshift of z = 0.022. Thus the luminosity function is truly a lo¯ cal one. The data points are independent of each other. In case one of the luminosity bins would suffer from incompleteness compared to the other bins, this would result in a break or dip in the differential luminosity function. The errors are based on the number of objects contributing to each value. The differential XLF shows a turnover around LX = (5 - 10) â 1043 ergs s-1 . Because our study is based solely on low redshift objects, we are not able to constrain models involving evolution with redshift. Nevertheless we can compare the XLF presented here with model predictions from previous investigations. XLFs are often fit by a smoothly connected two power-law function of the form [19]

d(LX , z = 0) =A d log LX

LX L



1

+

LX L



2

-1

(2)

We fit this function using a least-squares method. The +1.5 best fit values we obtained are A = 0.7 -0.5 â 10-5 h3 Mpc-3 , 1 = 0.80 ± 0.15, 2 = 2.11 ± 70 0.22, and log L = 43.38 ± 0.35 with L in units


Figure 3. Cumulative AGN luminosity function for 19 absorbed (NH 1022 cm-2 ; triangles) and 12 unabsorbed sources (octagons). As an example some error bars indicating the Poissonian error are shown. minosity where both AGN types have similar densities is about L(20-40 keV) = 3 â 1043 erg s-1 . Based on the luminosity function, the contribution of the AGN to the total X-ray emissivity W can be estimated [25]. This can be done by simply multiplying the XLF by the luminosity in each bin and integrating over the range of luminosities (1041 ergs s-1 < L20-40 keV < 1045.5 ergs s-1). This results in W20-40 keV (> 1041 ergs s-1 ) = (2.8 ± 0.8) â 1038 ergs s-1 h3 Mpc-3 . Please note that absorption 70 does not affect the luminosities in this energy range and therefore the values given here are intrinsic emissivities. Figure 4 shows the emissivity per unit volume in discrete luminosity bins. Apparently, the objects with LX > 1044 ergs s-1 do not contribute significantly to the total emissivity in the 20 - 40 keV energy band.

Figure 4. Emissivity of INTEGRAL AGN per unit volume and logarithmic luminosity bin. The objects with LX > 1044 ergs s-1 do not contribute significantly to the total X-ray emissivity W20-40 keV (> 1041 ergs s-1 ) = (2.8 ± 0.8) â 1038 ergs s-1 h3 Mpc-3 . 70

10-2 deg-2 above a threshold of 10-11 ergs cm-2 s-1 in the 20 - 50 keV energy band, where we get a consistent value of (1.2 ± 0.2) â 10-2 deg-2 . Comparing the total flux of all the objects in the AGN sample (f20-40 keV = 2.6 â 10-10 ergs cm-2 s-1 sr-1 ) with the flux of the X-ray background as presented by Gruber et al. (1999) shows that the INTEGRAL AGN account only for about 1% of the expected value. This is expected when taking into account the high flux limit of our sample: La Franca et al. (2005) have shown that objects with f2-10 keV > 10-11 ergs cm-2 s-1 contribute less than 1% to the CXB in the 2 - 10 keV energy range. This flux limit extrapolates to the faintest flux in our sample of f20-40 keV = 5.6 â 10-12 ergs cm-2 s-1 for a = 1.9 power law spectrum. We compared the unabsorbed emissivity per unit volume of our objects W20-40 keV (> 1041 ergs s-1) = 2.8 â 1038 ergs s-1 h3 Mpc-3 with that observed by RXTE 70 in the 3­20 keV band. Assuming an average power law of = 2, the extrapolated value is W3-20 keV (> 3 1041 ergs s-1) = (7.7 ± 2.2) â 1038 ergs s-1 h70 Mpc-3 , which is a factor of 2 larger than the value measured by RXTE [25] but consistent within the 1 error. If we apply the conversion to the 2 - 10 keV energy band, we derive the intrinsic emissivity W2-10 keV = (6.4 ± 1.8) â 1038 ergs s-1 Mpc-3 , consistent with the value derived from the HEAO-1 measurements (W = (5.9 ± 1.2) â 1038 ergs s-1 Mpc-3 ; Shinozaki et al. 2006). This showed that the local X-ray volume emissivity in the 2­ 10 keV band is consistent with the emissivity from AGN alone. It has to be pointed out that the value derived from our sample and the one based on HEAO-1 data are higher than the one based on the RXTE All-Sky Survey (W = (2.7 ± 0.7) â 1038 ergs s-1 Mpc-3 ; Sazonov & Revnivtsev 2004).

5. DISCUSSION A simple power-law model fitted to the number flux distribution (Fig. 1) has a slope of = 1.66 ± 0.11. Even though the difference from the Eucledian value is not statistically significant, at a 1.5 level, a deviation from this value could have two reasons. The difference might indicate that the area density at the low flux end of the distribution has been slightly overcorrected. One has to keep in mind that only a few sources derived from a small area of the sky are constraining the low flux end. Another reason for the difference could be that the distribution of AGN in the very local universe is not isotropic, caused e.g. by the local group and other clustering of galaxies. Krivonos et al. (2005) studied the extragalactic source counts as observed by INTEGRAL in the 20-50 keV energy band in the Coma region. Based on 12 source detections they determine a surface density of (1.4 ± 0.5) â


The luminosity function derived from the INTEGRAL 20 - 40 keV AGN sample appears to be consistent with the XLF in the 2 - 20 keV range. A turnover in the XLF at 2.4 â 1043 ergs s-1 is observed (Fig. 2). Below this luminosity also the fraction of absorbed AGN starts to exceed that of the unabsorbed ones, although the effect is significant at only on a 1 level (Fig. 3). Both effects have been seen also in the 2 - 10 keV [31, 17, 29] and in the 3 - 20 keV band [25]. This implies that we do detect a similar source population as at lower energies. Even though it has to be taken into account that the low luminosity end of the XLF is based only on a small number of objects, below this luminosity the distribution between active and normal galaxies becomes blurred. Additionally objects like Ultra-luminous Xray sources (ULX) and star-forming galaxies could provide the necessary emission. One interesting case in this category is the detection of NGC 4395 with a luminosity of L(20-40 keV) = 1.4 â 1040 ergs s-1, consistent with measurements by XMM-Newton which showed L(2-10 keV) = 1.5 â 1040 ergs s-1 [33]. The central engine of this galaxy has been classified as an "ultraluminous" source, possibly associatied with an intermediate mass black hole with MB H = (4 +6 â 104 M -3 [22]. In addition, NGC 4395 harbors a ULX with L(2-10 keV) = 1039 ergs s-1 at a distance of 2.9 kpc from the center of the galaxy [8]. If a larger fraction of absorbed AGN is necessary to explain the cosmic X-ray background at 30 keV as indicated by HEAO 1 A-4 measurements [13], the fraction of absorbed sources could be correlated with redshift. It has for example been proposed that there is an evolution of the population leading to a higher fraction of absorbed sources at higher redshifts. It should be noted however that this effect is not clearly detectable in the 2 - 10 keV range. The fraction of absorbed sources seems to depend on luminosity [31, 30], as is also seen in the 20 - 40 keV band (Fig. 3). But some studies come to the conclusion that there is no evolution of intrinsic NH [31, 30], while others find the fraction of absorbed sources increasing with redshift [17]. La Franca et al. also find that a combination of effects (the fraction of absorbed AGN decreases with the intrinsic X-ray luminosity, and increases with the redshift) can be explained by a luminosity-dependent density evolution model. They further show that the luminosity function of AGN with low luminosities as those presented here peaks at z 0.7 while high luminosity AGN peak at z 2. Unified models also predict, depending on the applied model, a fraction of absorbed AGN of 0.6 - 0.7 compared to the total population for high-flux low-redshift objects [30]. Worsley et al. (2005) examined Chandra and XMMNewton deep fields and come to the conclusion that the missing CXB component is formed by highly obscured AGN at redshifts 0.5 - 1.5 with column densities of the order of fX = 1023 - 1024 cm-2 . Evidence for this scenario is also found in a study of Chandra and Spitzer data [24]. Combining multiwavelength data, this work estimates a surface density of 25 AGN deg -2 in the infrared in the 0.6 deg2 Chandra/SWIRE field, and

only 33% of them are detected in the X-rays down to f0.3-8 keV = 10-15 ergs cm-2 s-1 . The work also indicates a higher abundance of luminous and Compton-thick AGN at higher redshifts (z 0.5). This source population would be missed by the study presented here, because of the low redshifts (z = 0.022) of the INTEGRAL ¯ AGN. Several studies [31, 30] propose that the absorbed AGN needed to explain the CXB should be Compton thick, and therefore would have been missed at 2 - 10 keV. This argument does not hold for the INTEGRAL observations, where the impact of absorption is much less severe than at lower energies. The effect on the measured flux of a source with photon index = 2 for Compton thick absorption (NH = 1024 cm-2 ) is only a 5% decrease in flux (40% for NH = 1025 cm-2 ). It is therefore unlikely that many Compton-thick objects have been missed by the INTEGRAL studies performed to date. One possibility would be, that they are among the newly discovered sources found by INTEGRAL. The fraction of unidentified objects among the INTEGRAL discovered sources is approximately 50%. Eight such sources without crossidentification have a significance above 5 in the data set discussed here. Thus, if they are ultimately identified as AGN, they would have to be considered in this study. It should be pointed out though, that most of the sources discovered by INTEGRAL are located close to the Galactic plane and are more likely to belong to the Galaxy: the Second IBIS/ISGRI Soft Gamma-Ray Survey Catalog [5] lists 55 new sources detected by INTEGRAL, of which 93% are located within -10 < b < +10 . Among these 55 sources, 3 are listed as extragalactic sources, 18 are of Galactic origin, and 29 have not been identified yet. In addition, those objects that have been classified as AGN based on soft X-ray and/or optical follow-up studies, are no more likely to be Compton-thick objects than the overall AGN population studied here. Only four AGN (NGC 1068, NGC 4945, MRK 3, Circinus galaxy) detected by INTEGRAL have been proven to be Compton thick objects so far, and none of them showed absorbtion of NH > 5 â 1024 cm-2 . In order to clarify this point, observations at soft X-rays of those objects without information about intrinsic absorption are required for all INTEGRAL detected AGN. At present 23 % of the INTEGRAL AGN are missing absorption information. A first indication of what the absorption in these sources might be, can be derived from comparison of the INTEGRAL fluxes with ROSAT All-Sky Survey (RASS) Faint Source Catalogue data [34]. In order to do so we assumed a simple power law with photon index = 2.0 between the ROSAT 0.1 - 2.4 keV band and the INTEGRAL 20 - 40 keV range and fit the absorption. In the six cases where no detection was achieved in the RASS, an upper limit of f(0.1-2.4 keV) 10-13 ergs cm-2 s-1 has been assumed, resulting in a lower limit for the absorption NH > (5 - 11) â 1022 cm-2 . In Fig. 5 we show the distribution of intrinsic absorption. It has to be pointed out that the estimated values can only give an idea about the distribution of intrinsic absorption and should not be taken literally, as the spectral slope between the measure-


the peak at 30 keV, as Compton thick AGN are apparently less abundant than expected [30]. But this picture might change if we assume all INTEGRAL AGN lacking soft X-ray data and without counter parts in the RASS to be Compton thick. In addition the sample presented here might be still too small to constrain the fraction of obscured sources, and the missing Compton thick AGN could be detectable when studying sources with f(20-40 keV) < 10-11 ergs cm-2 s-1 .

6. CONCLUSIONS The extragalactic sample derived from the INTEGRAL public data archive comprises 63 low redshift Seyfert galaxies ( z = 0.022 ± 0.003) and 8 blazars in the hard X-ray domain. Two galaxy clusters are also detected, but no star-burst galaxy has been as yet. This INTEGRAL AGN sample is the basis for the first luminosity function above 20 keV. 38 of the Seyfert galaxies form a complete sample with significance limit of 5 . The number flux distribution is approximated by a powerlaw with a slope of = 1.66 ± 0.11. Because of the high flux limit of our sample the objects account in total for less than 1% of the 20 - 40 keV cosmic X-ray background. The emissivity of all AGN per unit volume W20-40 keV (> 1041 ergs s-1) = 2.8 â 3 1038 ergs s-1 h70 Mpc-3 appears to be consistent with the background estimates in the 2­10 keV energy band based on the cross-correlation of the HEAO 1-A2 map with IRAS galaxies [23]. The luminosity function in the 20 - 40 keV energy range is consistent with that measured in the 2 - 20 keV band. Below the turnover luminosity of L = 2.4 â 1043 ergs s-1 the absorbed AGN become dominant over the unabsorbed ones. Similar results have been derived by Sazonov et al. (2007) using a larger INTEGRAL data set and a slightly different energy band (17 - 60 keV), including 91 AGN above the 5 significance limit. The fraction of Compton thick AGN with known intrinsic absorption is found to be small (8%) in our AGN sample. For the sources without reliable absorption information we derived an estimate from the comparison with ROSAT All-Sky Survey data and find that the data do not require additional Compton thick objects within the sample presented here. It has to be pointed out though, that the sources without RASS counterpart could be Compton thick which would increase the ratio of this source type to 13% in the complete sample. Evolution of the source population can play a major role in the sense that the fraction of absorbed sources among AGN might be correlated with redshift, as proposed for example by Worsley et al. (2005). Over the life time of the INTEGRAL mission we expect to detect of the order of 250 AGN. Combining these data with the studies based on Swift/BAT, operating in a similar energy band as IBIS/ISGRI, will further constrain the

Figure 5. Distribution of intrinsic absorption for all INTEGRAL AGN (blazars excluded), as measured in the soft X-rays. The shaded area shows the reliable measurements, the other values are based on comparison of ROSAT All-Sky Survey and INTEGRAL data. Lower limits on absorption have been excluded. ments is unknown and the observations are not simultaneous. Nevertheless apparently none of the RASS detections and non-detections requires an intrinsic absorption of NH > 2 â 1023 cm-2 . Therefore it appears unlikely that a significant fraction of INTEGRAL AGN will show an intrinsic absorption NH > 1024 cm-2 . However, if we assume that the RASS non-detections are all Compton thick AGN, the fraction of this class of sources rises from 6% to 16% when considering all 63 non-blazar AGN seen by INTEGRAL, and from 8% to 13% for the complete sample with 38 objects. This is in good agreement with the fraction of 11% of Compton thick AGN as seen in the Swift/BAT survey [20]. The picture is less clear when referring to the optical classification. Here the INTEGRAL survey finds 12 Seyfert 1 (33%), 14 Seyfert 2, and 10 intermediate Seyfert 1.5 in the complete sample, while the Swift/BAT survey contains only 20% of type 1 Seyfert galaxies. It should be pointed out though that the classification based on intrinsic absorption gives a more objective criterion in order to define AGN subclasses than the optical classification with its many subtypes. The finding of the BAT survey that virtually all sources with log LX < 43.5 are absorbed, cannot be confirmed by our study, in which we detect a fraction of 33% of sources with NH < 1022 cm-2 among the sources below this luminosity. This also reflected in the observation that although the absorbed sources become more dominant below this luminosity, the trend is not overwhelmingly strong (Fig. 3). Most investigations to date have been focused on the Xrays below 20 keV, and INTEGRAL can add substantial information to the nature of bright AGN in the local Universe. Considering the expected composition of the hard X-ray background, it does not currently appear that the population detected by INTEGRAL can explain


hard X-ray luminosity function of AGN. But we will still be limited to the relatively high flux end of the distribution. Because of this INTEGRAL and Swift/BAT will most likely not be able to test evolutionary scenarios of AGN and thus will be inadequate to explain the cosmic X-ray background at E > 20 keV. Future missions with larger collecting areas and/or focusing optics will be required to answer the question of what dominates the Universe in the hard X-rays.

[15] Hasinger G. 2004, Nucl. Phys. B (Proc. Suppl.), 132, 86 [16] Krivonos R., Vikhlinin A., Churazov E., et al. 2005, ApJ, 625, 89 [17] La Franca F., et al. 2005, ApJ, 635, 864 [18] Lebrun F., et al. 2003, A&A, 411, L141 [19] Maccacaro T., Della Ceca R., Gioia I. M., et al. 1991, 374, 117 [20] Markwardt C. B., Tueller J., Skinner G. K., et al. 2005, ApJ, 633, L77 [21] Marshall F. E., et al. 1980, ApJ, 235, 4 [22] McHardy I. M. M., Gunn K. F., Uttley P., & Goad M. R. 2005, MNRAS, 359, 1469 [23] Miyaji T., Lahav O., Jahoda K., & Boldt E. 1994, ApJ, 434, 424 [24] Polletta M., et al. 2006, ApJ, 642, 673 [25] Sazonov S. Y., Revnivtsev M. G. 2004, A&A, 423, 469 [26] Sazonov S. Y., Revnivtsev M., Krivonos R., Churazov E., & Sunyaev R., 2007, A&A, 462, 57 [27] Schmidt M., et al. 1998, A&A, 329, 495 [28] Setti G., Woltjer L. 1989, A&A, 224, L21 [29] Shinozaki K., Miyaji T., Ishisaki Y., et al. 2006, AJ, 131, 2843 [30] Treister E., Urry C. M. 2005, ApJ, 630, 115 [31] Ueda Y., Akiyama M., Ohta K., & Miyaji T. 2003, ApJ, 598, 886 [32] Ueda Y., Ishisaki Y., Takahashi T., et al. 2001, ApJS, 133, 1 [33] Vaughan S., Iwasawa K., Fabian A. C., & Hayashida K. 2004, MNRAS, 356, 524 [34] Voges W., et al. 2000, IAUC, 7432, 3 [35] Worsley M. A., et al. 2005, MNRAS, 357, 1281

ACKNOWLEDGMENTS VB would like to thank Olaf Wucknitz for providing software to handle the 0 > 0 cosmology. This research has made use of the NASA/IPAC Extragalactic Database (NED) which is operated by the Jet Propulsion Laboratory, of data obtained from the High Energy Astrophysics Science Archive Research Center (HEASARC), provided by NASA's Goddard Space Flight Center, and of the SIMBAD Astronomical Database which is operated by the Centre de Donnees astronomiques de Stras´ bourg. This research has also made use of the Tartarus (Version 3.1) database, created by Paul O'Neill and Kirpal Nandra at Imperial College London, and Jane Turner at NASA/GSFC. Tartarus is supported by funding from PPARC, and NASA grants NAG5-7385 and NAG5-7067.

REFERENCES [1] Avni Y., Bahcall J. N. 1980, ApJ, 235, 694 [2] Beckmann V., Gehrels N., Shrader C. R., & Soldi S. 2006a, ApJ, 638, 642 [3] Beckmann V., Soldi S., Shrader C. R., & Gehrels N. 2006b, proc. of "The X-ray Universe 2005", San Lorenzo de El Escorial (Madrid, Spain), 26-30 September 2005, ESA-SP 604, astro-ph/0510833 [4] Beckmann V., Soldi S., Shrader C. R., Gehrels N., & Produit N., 2006c, ApJ, 652, 126 [5] Bird A., et al. 2006, ApJ, 636, 765 [6] Brandt W. N., et al. 2001, AJ, 122, 2810 [7] Brandt W. N., Hasinger G. 2005, ARA&A, 43 [8] Colbert E. J. M., Ptak A. F. 2002, ApJSS, 143, 25 [9] Comastri A., Fiore F., Vignali C., et al. 2001, MNRAS, 327, 781 [10] Courvoisier T.J.-L., Walter R., Beckmann V., et al. 2003, A&A, 411, L53 [11] Gehrels N., et al. 2005, ApJ, 611, 1005 [12] Giacconi R., Gursky H., Paolini R., & Rossi B. 1962, Phys. Rev. Lett., 9, 439 [13] Gruber D. E., Matteson J. L., Peterson L. E., & Jung G. V. 1999, ApJ, 520, 124 [14] Hasinger G., Burg R., Giacconi R., et al. 1998, A&A, 329, 482