A&A 466, 41-61 (2007)
M. Krumpe1 - G. Lamer1 - A. D. Schwope1 - S. Wagner2 - G. Zamorani3 - M. Mignoli3 - R. Staubert4 - L. Wisotzki1 - G. Hasinger5
1 - Astrophysikalisches Institut Potsdam, An der Sternwarte 16, 14482 Potsdam, Germany
2 - Landessternwarte Heidelberg-Königstuhl, 69117 Heidelberg, Germany
3 - INAF - Osservatorio Astronomico di Bologna, Bologna, Italy
4 - Universität Tübingen, Institut für Astronomie und Astrophysik, 72076 Tübingen, Germany
5 - Max-Planck-Institut für Extraterrestrische Physik, Giessenbachstrasse, Postfach 1312, 85741 Garching, Germany
Received 16 June 2006 / Accepted 7 December 2006
We report on a medium deep XMM-Newton survey of the Marano Field and optical follow-up observations. The mosaicked XMM-Newton pointings in this optical quasar survey field cover 0.6 deg2 with a total of 120 ks good observation time. We detected 328 X-ray sources in total. The turnover flux of our sample is erg cm-2 s-1 in the 0.2-10 keV band. With VLT FORS1 and FORS2 spectroscopy we classified 96 new X-ray counterparts.
The central 0.28 deg2, where detailed optical follow-up observations were performed, contain 170 X-ray sources (detection likelihood ML>10), out of which 48 had already been detected by ROSAT. In this region we recover 23 out of 29 optically selected quasars. With a total of 110 classifications in our core sample we reach a completeness of 65%. About one-third of the XMM-Newton sources are classified as type II AGN with redshifts mostly below 1.0. Furthermore, we detect five high redshift type II AGN ( ).
We show that the true redshift distribution of type II AGN remains uncertain, since their lack of emission lines in a wide optical wavelength range hampers their identification in the redshift range 1<z<2. The optical and X-ray colours of the core sample indicate that most of the still unidentified X-ray sources are likely to be type II AGN. We calculate absorbing column densities and show that the ratio of absorbed to unabsorbed objects is significantly higher for type II AGN than for type I AGN. Nevertheless, we find a few unabsorbed type II AGN. The X-ray hardness ratios of some high redshift type I AGN also give an indication of heavy absorption. However, none of these type I objects is bright enough for spectral extraction and detailed model fitting. Types I and II AGN cover the same range in intrinsic X-ray luminosity, ( ), although type II AGN have a lower median intrinsic X-ray luminosity (log ) compared to type I AGN (log ). Furthermore, we classified three X-ray bright optically normal galaxies (XBONGs) as counterparts. They show properties similar to type II AGN, probably harbouring an active nucleus.
Key words: surveys - X-rays: galaxies - galaxies: active - galaxies: quasars: general
The Marano Field was named by an early optical quasar survey up to a limiting magnitude of by Marano et al. (1988). Based on different optical selection techniques (colour-colour diagrams, grism plates, variability analysis) they discovered 23 broad emission line quasars and defined an extensive list of quasar candidates. Zitelli et al. (1992) completed this work by presenting a spectroscopically complete sample of quasars with using this list of quasar candidates. They confirmed 54 quasars including the former 23 quasars. Between December 1992 and July 1993, ROSAT observed the central Marano Field (0.2 deg2) for 56 ks (Zamorani et al. 1999). The data revealed 50 X-ray sources with a limitingflux of erg cm-2 s-1 in the ROSAT band (0.5- 2.0 keV). Multi-colour CCD and spectroscopic data identified 42 X-ray sources (33 quasars, 2 galaxies, 3 clusters, and 4 stars). 66% of the optically selected quasars within the ROSAT field-of-view were detected as ROSAT X-ray sources. Gruppioni et al. (1999) carried out a deep radio survey at 1.4 and 2.4 GHz and detected 68 radio sources ( mJy). Follow-up observation provided redshifts for 30 objects.
Our new XMM-Newton data comprise an area three times larger, compared to the ROSAT survey, and are thus almost comparable in size to the optical quasar survey. We are reaching a survey sensitivity of erg cm-2 s-1 (turnover flux) over a contiguous area of 0.6 deg2. The XMM-Newton survey of the Marano Field is thus comparable with deep ROSAT surveys (e.g., the UDS, Hasinger et al. 1998) or with medium deep CHANDRA surveys (e.g., ChaMP, ).
The paper is organised as follows. In Sect. 2 we list the X-ray and optical data and the reduction of the data. Section 3 describes and summarises the spectroscopic classification of the X-ray sources. In Sect. 4 we make use of the spectroscopic classification and redshift determination to analyse the properties of different object classes. In this section we concentrate on a "core sample'' of objects in the central part of the field, where we reached the highest degree of completeness in the spectroscopic classification. Section 5 addresses additional objects in the Marano Field that are not X-ray detected. These objects were obtained as a control sample. Section 6 dicusses the results of the different object classes. Finally, our conclusions are outlined in Sect. 7. Unless mentioned otherwise, all errors refer to a confidence interval.
Since the area of the optical quasar survey in the Marano Field is larger than the XMM-Newton field of view, the X-ray observations have been performed as a grid of overlapping pointings with a spacing of 5 arcmin in right ascension and declination. The pointing in the north-western corner of the grid was shifted to cover the position of a deep far-infrared survey with the ISO satellite (Table 1 and Fig. 1). The ISO data are not addressed in this paper.
Table 1: XMM-Newton observations of the Marano Field. filter: T.-thin, M.-medium.
Due to the overlapping pointings some deviations from the standard XMM-Newton
data analysis procedures were necessary, and are described here.
The photon event tables from the 16 grid pointings were merged into
a single event table with a common sky coordinate frame using the XMM SAS merge task. In this coordinate frame we created 240 images
(one for each pointing, energy range, and instrument). The 5 bands
used are 0.2-0.5 keV, 0.5-2.0 keV, 2.0-4.5 keV, 4.5-7.5 keV, and 7.5-12.0 keV.
Each image has
pixels with four arcseconds binning.
For each of these images separate exposure maps and background
maps were created using the XMM SAS tasks eexpmap and esplinemap.
One of the pointings (ObsID 0129320801) had accidentally been scheduled
with "medium'' thickness filters, while all other pointings had been
observed with "thin'' filters. To correct for the somewhat
lower throughput of the medium filters at soft energies, we multiplied
all exposure maps of this pointing with a correction factor derived
from the energy conversion factors for the "thin'' and "medium'' filters.
These ECFs were taken from the SSC document SSC-LUX-TN-0059 v.3
The correction is largest in the softest band (0.2-0.5 keV), where
the effective exposure in the affected pointing is reduced by 11%.
|Figure 1: Averaged mos1 and mos2 exposure map (0.5-2.0 keV) of the pointing pattern (small white crosses). The exposure time in the central region reaches 78 ks. The black rectangle shows the area covered by the WFI R-band images. The white polygon marks the area covered by the SOFI K-band images (see Sect. 2.2).|
|Open with DEXTER|
We then added the images, exposure maps, and background maps of the
individual pointings, resulting in 15 images for 5 energy bands and
3 detectors. These images were used simultaneously as input for a
detection run with eboxdetect and emldetect.
The task emldetect applies a PSF-fit to each source found by
eboxdetect to determine the source parameters.
Since we use merged mosaic images of the field, each source image results from
the superposition of several pointings. In each pointing the source is detected
at a different position on the detector.
Therefore, the standard configuration of emldetect, which uses the
off-axis angle and position-angle of each source to extract the appropriate
PSF from calibration files, could not be used here. Instead emldetect
was modified to use the calibration PSF corresponding to an off-axis
throughout the entire field.
This PSF is circular symmetric and is a good representation of the point
sources in the merged images.
For this work no extent models were fitted to the sources, therefore
the source list produced by emldetect does not contain
any information whether a source is extended or point-like.
Generally, the X-ray data reduction and
analysis was performed with XMM SAS version 5.1 (from 18/06/2001) with
the abovementioned exception of an adapted version of emldetect.
The version of emldetect used here suffered from an error, which
resulted in the overestimation of detection likelihood values
(XMM-Newton-NEWS #29, 11-Mar.-2003). All likelihood values quoted here have
been corrected to adhere to the relation
where ML is the
detection likelihood and P is the probability that the detection is caused
by a random fluctuation of the background.
All count rates and derived quantities are taken from the PSF fitting of emldetect.
Using the combined expoure maps, emldetect corrects the source count rates for all
spatial variations of telescope and detector efficiency, i.e., the count rates relate
to the optical axis of each EPIC camera.
Following Cash (1979), the 68% confidence intervals for the
source positions and source fluxes were calculated
using emldetect as follows:
each parameter is varied until the statistic
For all sources 3 hardness ratios of the form HR = (CR2 - CR1)/(CR1 + CR2) were calculated between the adjacent pn-detector energy bands 0.2-0.5 keV, 0.5-2.0 keV, 2.0-4.5 keV, and 4.5-7.5 keV. For example, HR2 was calculated between the bands 0.5-2.0 keV and 2.0-4.5 keV. The different energy responses of the mos- and pn-detectors result in different hardness ratios for the same source. If not noted otherwise the pn-values are used throughout the paper.
In total we detected 328 X-ray sources with detection likelihoods . The X-ray fluxes are in the range erg cm-2 s-1 (0.2-10 keV). The X-ray flux histogram (Fig. 3) shows the X-ray detection limit of our survey. Below erg cm-2 s-1 we are unable to detect the majority of the X-ray sources. In the central region (Fig. 2) we found 252 X-ray sources. The complete X-ray source list can be found in the Online Material, Appendix A.
|Figure 2: PN exposure map (0.5-2.0 keV) of the pointing pattern. The exposure time in the central region reaches 35.5 ks. The area inside the black contour markes the "core sample'' region where the pn-detector exposure time exceeds six ks. White rectangles show the position of the spectroscopic masks for the optical follow-up observations (see Sect. 2.4).|
|Open with DEXTER|
XMM-Newton redetected all 50 ROSAT X-ray sources (Zamorani et al. 1999).
However, two ROSAT X-ray sources have low detection likelihoods
and were thus not included in the final XMM-Newton
X-ray source list (Online Material, Appendix A).
Figure 4 compares ROSAT and EPIC X-ray fluxes in the
0.5-2.0 keV band.
|Figure 3: 0.2-10 keV X-ray flux histogram of the 328 XMM-Newton detected X-ray sources in the Marano Field. Up to an X-ray flux of erg cm-2 s-1 (turnover flux) we detect more and more sources. Below that X-ray flux the detection rate decreases dramactically.|
|Open with DEXTER|
|Figure 4: Comparison between the 0.5-2 keV X-ray fluxes measured by ROSAT and XMM-Newton of all 50 ROSAT X-ray sources in the Marano Field. XMM-Newton fluxes are averages of the three individual EPIC camera fluxes. The black solid line represents equal ROSAT and XMM-Newton fluxes.|
|Open with DEXTER|
The central region of the Marano Field has been observed in the UBVRI bands with the ESO Wide-Field-Imager (WFI). Here we make use of the deep (7000 s) R-band image (Mignoli & Zamorani 1998) with a limiting magnitude (defined as turnover magnitude minus 0.5 mag, see Fig. 5). The WFI-image covers an area of 35 32 arcmin well aligned with the region of the deepest XMM-Newton coverage (Fig. 1).
|Figure 5: Magnitude histogram of the sources detected in the WFI R-band image. Completeness in detecting objects is lost for R>23.|
|Open with DEXTER|
The K-band data ( m) were obtained at the ESO New Technology Telescope (NTT) with the SOFI instrument. With an area of 754 arcmin2 the K-band mosaic covers a slightly smaller area than the WFI R-band observation and is well aligned with the deepest XMM-Newton exposure (Fig. 1). The SOFI observations consist of a mosaic of 33 jittered pointings, each covering arcmin with an exposure time of 29 min each. The limiting magnitude is (turnover magnitude minus 0.5 mag, see Fig. 6).
|Figure 6: Histogram of the K-band magnitudes in the NTT-SOFI images. Completeness in detecting objects is lost for K>20.5.|
|Open with DEXTER|
Our target selection for the optical spectroscopy was primarily based on FORS1/FORS2 R-band pre-images. WFI R-band and SOFI K-band data were used in addition to support the target selection on the pre-images in case of faint counterparts. The chosen fields represent a compromise in terms of maximum survey area and source density because of limited telescope time. First priority for spectroscopic follow up was given to candidates with a likelihood of existence ML > 10 within a 2 error radius. The total position error is . The statistical errors of the positions were calculated as described in Sect. 2.1. The systematic errors are caused by uncertainties in the spacecraft attitude, errors in the linearisation of the detector coordinates, and undersampling of the PSF (in particular for the PN camera). At this stage, we allowed a systematic position error of , see Sect. 3.2 for a more accurate estimation of the systematic position errors.
The multi-object spectroscopy masks were designed with the software fims 2 provided by ESO. For all spectroscopic targets we used straight slitlets with a nominal length of ten arcseconds and width of one arcsecond. For extended sources or closely spaced candidate counterparts, a different slit length, 6-14 arcsec, was used. Still available mask positions were filled with candidate objects not fulfilling the main selection criteria and with additional random non-X-ray emitting galaxies.
The second run was performed with FORS2 in spectroscopic mask mode (MXU) from November 27th-30th, 2002. We aimed for an exposure time of 3 1800 s per mask, but weather and time constraints required some deviations from this general scheme (Table 2). Grism GRIS_150I+27 was used for all masks.
Table 2: Details of FORS2 spectroscopic mask mode observation (see Fig. 2).
Table 3: Setups for the optical spectroscopy.To prevent second order contamination, the first mask (mask6) in the observing sequence was observed through filter GC 375, which limited the wavelength range from 3850 to 7500 Å. A comparison with the second mask without an order separating filter showed that the contamination effect is negligible. Hence, for all following masks the filter was removed from the light path resulting in a wavelength range for central targets of the final spectra from 3500 to 10 000 Å.
The reduction of the data was accomplished by a semiautomatic pipeline coded in MIDAS. It was specially designed to reduce FORS2-MXU data with as little interaction with the user as possible. After modifications this code was also used to reduce the FORS1-MOS data. The bias correction was done in the standard manner with careful attention to possible time dependence on the bias level and dark current. An ordinary flat field correction was used to rectify the pixel-to-pixel variation.
Table 4: Origin of optical counterparts to XMM sources.
|Figure 7: Optical atmospheric absorption corrected, wavelength and flux calibrated spectrum for X-ray source 480. The black spectrum marks the object's spectrum. The green line (see the online colour version of the manuscript) shows the error spectrum. Possible emission and absorption features in the spectrum are labelled. The spectral feature at 5580 Å is spurious due to incomplete substraction of a night sky line.|
|Open with DEXTER|
The extraction is based on an optimal extraction algorithm (Horne 1986) including a cosmic ray rejection. Extensive tests determined the optimal extraction parameters. Standard flux calibration was applied. Since FORS2-mask6 was observed with an additional filter, a different standard star was used. All spectra were corrected for atmospheric absorption using the standard ESO extinction correction function scaled to the given airmass.
The n individual spectra of a given counterpart object were combined to form one single final spectrum. The individual spectra were firstly normalised to the same mean intensity. The normalisation factor was determined from an analysis of all spectra in a given mask. The final spectrum is the uncertainty-weighted mean spectrum.
|Figure 8: Adopted mean k correction from observed R-band to rest-frame , for broad emission line AGN.|
|Open with DEXTER|
For the large majority of our spectra the signal-to-noise ratio (SNR) is sufficient to give reliable classifications and redshifts. Like other X-ray identification surveys, we encounter the problem of difficult classification of the optical spectra below a certain SNR. For spectra with a SNR = 3-5, the identification of narrow emission lines was still possible. However, at this SNR faint broad emission lines and normal galaxy spectra are very difficult to identify. Figure 7 shows a spectrum of a narrow emission line galaxy with a SNR = 2.4 (continuum near the O II emission line), which is close to the identification limit. Only one significant narrow emission line at 7100 Å is found. The spectral shape makes it reasonable to identify this line as O II emission line. Even though this classification is likely, an unambiguous redshift determination and classification of the type cannot be given. Spectra with a SNR less then 2.5 are not identifiable. The reliability of the redshift determination and classification of the optical spectra is given by flags in Col. (9) Flags of Table 8.
Table 5: X-ray counterpart distribution.
Table 6: Properties of the core sample.The complete list of X-ray classification is given in Table 8. The columns are described as follows:
Table 7: Fractions of absorbed sources in different X-ray luminosity bins based in Fig. 19c. Only objects with are selected.
Classification of a counterpart object consists of the sequence number of the X-ray source list and a suffix (A, B) to discriminate between different candidates.
(2) RA [hh:min:s] and (3) Dec [deg:min:s]
Right ascension and declination of the optical candidate counterpart (J2000).
(4) DistOX [arcsec]
Spatial offset between the X-ray and optical positions.
SOFI K-band magnitude of the spectroscopically classified candidate, whenever available.
WFI R-band magnitude of the spectroscopically classified candidate, whenever possible.
Spectroscopic classification of the identified object. S - star, G - normal galaxy (no emission lines), N - narrow emission line galaxy with unresolved emission lines (at 6000 our spectral resolution of 21 corresponds to 1050 km s-1), B - broad emission line object (all measured line widths have km s-1), and ? - undefined object.
Spectroscopic redshift of the identified object. The redshift is taken from the literature for objects with "1 - -'' and "0 - -'' in Col. (9) Flags. Column (15) Rem. states the source of redshift determination and classification.
X-ray identification flag, redshift flag, and classification reliability flag. The first number (0,1) marks whether a spectroscopically classified object was accepted as X-ray counterpart. Objects which we consider to be the correct identification of the X-ray source are flagged by "1'', while "0'' flags objects not considered as the X-ray source.
The second (middle) flag states the redshift reliablility. A redshift flag "1'' means a reliable, well-established redshift determined by several spectral features. "0'' marks objects where the redshift determination relies on a single but reasonable spectral feature.
The third (last) flag characterises the classification reliability. A flag "1'' marks that the object type as given in "(7) Class'' is well established and reliable. Flag "0'' indicates an uncertain classification of the object type. Either high SNR spectral features of the object do not allow a proper classification or the optical spectra do not allow to give a reliable classification of the object type because of a low SNR and/or insufficient wavelength coverage of the optical spectra. The latter is illustrated in Fig. 7. The narrow O II line indicates a narrow emission line galaxy. However, the SNR of the spectrum does not allow us to judge the existence of a broad Mg II line and, hence, a classification as type I AGN. Since C IV and C III are outside the spectrum's wavelength range, Mg II is the only possible broad-line feature. The most likely classification of this object is a narrow emission line galaxy. However, the given arguments show that a classification as type I AGN cannot be excluded. Therefore, the classification flag for this object is "0''. Objects with "1 - -'' and "0 - -'' have only an X-ray identification flag since their redshift and classification relies on follow-up surveys previously done in the Marano Field (see Col. (15) Rem.).
|Figure 9: Adopted conversion term to estimate AB magnitudes at fixed rest-frame wavelength Å from observed R-band magnitudes.|
|Open with DEXTER|
|Figure 10: a) Number of objects identified as X-ray counterparts vs. difference in X-ray and optical position scaled to a 1 position error with a systematic X-ray position error of 2 arcsec. The smooth line represents the theoretical distribution. The distribution of type I AGN, which are a class of well-accepted X-ray emitters, is plotted as a dotted line. The solid line characterises the complete sample including type I AGN, type II AGN, galaxies, and stars. b) The 1 position error was adjusted to the observed distribution by reducing the systematic X-ray position error from to . The complete sample agrees well with the type I AGN distribution, which gives evidence of a marginally false match rate.|
|Open with DEXTER|
(10) log (L
XOBS [erg s-1])
Observed rest-frame X-ray luminosity (logarithmic units) in the 0.2-10 keV energy band calculated by using Eq. (1). The k-correction vanishes since we assume an energy index with (-photon index) based on Alexander et al. (2003) and Mainieri et al. (2002). The luminosity distance was computed by the analytical fit for flat cosmologies with , , km s-1 Mpc-1 following Szokoly et al. (2004).
Absolute magnitudes (in the Johnson system) were estimated only for type I AGN using the relation
The broad band spectral index roughly characterises the UV-X-ray spectral energy distribution by connecting the rest-frame points at 2500 Å and 1 keV with a simple power-law, . For each broad emission line AGN we estimated its flux at a fixed rest-frame wavelength of Å, applying the relation
(13) NH [cm-2]
X-ray absorbing hydrogen column density in units of (see Sect. 4.4).
(14) log (
LXINT [erg s-1] )
Intrinsic rest-frame X-ray luminosity (logarithmic units) in the 0.2-10 keV energy band after X-ray flux correction for the absorbing hydrogen column density. Calculation uses Eq. (1).
Remarks for individual objects: 1 - optically selected and spectroscopically classified quasar by Marano et al. (1988); 2 - optically selected and spectroscopically classified quasar by Zitelli et al. (1992); 3 - ROSAT X-ray source with spectroscopic classification and redshift determination by Zamorani et al. (1999); 4 - ROSAT X-ray source with no or wrong identification by Zamorani et al. (1999); 5 - unclassified radio objects within , Gruppioni et al. (1999); 6 - spectroscopic classification and redshift taken from Teplitz et al. (2003); 7 - radio source, spectroscopic classification and redshift taken from Gruppioni et al. (1999), and C - individual comment to an object (see Sect. 3.1).
9A: New redshift determined for this source, formerly known as Zamorani et al. (1999) X043-12.
15A: Object 15A is not regarded as the optical counterpart, since the identification of the line features is uncertain and the positional offset is rather large. A very faint object at the detection limit lies in the X-ray error circle. A lower limit of for this source was estimated by assuming a limiting magnitude of R=24.
17A: Source 17 is a likely X-ray blend with large contrast between the
two individual sources. Object 17A is a unique identification of
the brighter X-ray source, whose X-ray flux is likely to be
overestimated due to blending.
|Figure 11: WFI R-band magnitude vs. difference in X-ray and optical position scaled to a position error with a systematic X-ray position error of 0.7 arcsec. Crosses mark type I AGN, asterisks type II AGN, triangle galaxies, and rectangles stars.|
|Open with DEXTER|
20A and 20B: Zamorani et al. (1999) identified source 20 with the M star 20B. Some contribution of the M star's X-ray flux cannot be excluded, but the identification with the type II AGN (object 20A) seems more likely, since the X-ray colours indicate a relatively hard spectrum.
22A and 22B: X-ray source 22 is a probable blend. Objects 22A and 22B fall on top of the two suspected point sources. We regard both optical objects as likely counterparts of the X-ray blend and distribute the measured X-ray flux evenly between the two sources.
26A: New redshift determined for this source, formerly known as Zamorani et al. (1999) X301-29.
32A: N V strongest emission line in the spectrum (brighter than Ly-). Narrow emission lines are present and the object is classified as "N'' (narrow emission line galaxy). However, all narrow emission lines have underlying broad components. Therefore, the classification flag is set to "0''.
35A and 35B: Two galaxies at equal redshift, the brighter galaxy 35A is regarded as the X-ray counterpart. The possibility that the X-ray source is a galaxy cluster cannot completely excluded, but an X-ray extent is not obvious. For 35A a new redshift is determined (update to Zamorani et al. 1999, X022-48).
42A: New redshift determined for this source, formerly known as Zamorani et al. (1999) X031-24.
46A: X-ray blend with major contribution from the southwestern component with type I AGN 46A as counterpart.
47A and 47B: Two narrow emission line galaxies at equal redshift. The possibility that the X-ray source is a galaxy cluster cannot completely excluded, but an X-ray extent is not obvious. 47A is assumed to be the optical counterpart.
63A: Unresolved narrow emission lines are present and the object is classified as "N'' (narrow emission line galaxy). However, the C IV emission line has an underlying broad component. Therefore, the classification flag is set to "0''.
66A: The identification is not unique, since the optical image reveals a possible second object inside the 1-X-ray position error circle.
133A: Narrow emission lines are present. The object is classified as "N'' (narrow emission line galaxy). However, Ly- shows a resolved, broad base. Therefore, the classification flag is set "0''.
Table 8: Optical properties of candidate counterparts of Marano XMM-Newton X-ray sources. For a detailed explanation of the individual columns see Sect. 3.
151A: Probably a spurious detection of the X-ray source. The source is kept in the source list for the formal reason of having an ML = 5.2, but is not considered further.
191A: Optical spectrum in Zitelli et al. (1992) indicates a type I AGN with typical broad emission lines. However, this object shows the lowest X-ray-to-optical flux ratio of all AGN in our sample. suggests an X-ray faint AGN. Since this object was not detected by ROSAT, it is unlikely that the extreme -ratio is due to a temporary low X-ray state of the object. Furthermore, it is one of the type I objects with intrinsic absorption ( cm-2).
217A and 217B: X-ray blend, the identification of 217A with the southeastern component seems unambiguous, the identification of 217B with the northwestern component is not unique, since a similarly bright, close-by, but still unidentified object is present at the same distance from the X-ray source.
224A and 224B: Two galaxies at equal redshift, no obvious X-ray extent.
253A and 253B: Two objects at similar redshift with difference in the optical. The brighter object 253A is regarded as the identification.
280A: Object is classified as "B'' since C III is well resolved with (FWHM). Ly- and C IV are narrow emision lines, but show strong absorption with broad underlying components. The classification flag is set "0''. Possible broad absorption line quasar.
361A: Broad absorption line quasar.
382A and 496A: Physical quasar pair, separated by , at z = 1.904, kpc. The spectra are different, i.e., the two objects are not lensed images of the same source.
437A: Spectrum, optical, and X-ray image, and relative soft hardness ratios point undoubtedly to an M star as X-ray source. However, a flux ratio is unusually high for an M star as the X-ray identification.
512A and 512B: Zamorani et al. (1999) identify the NELG 512B with the X-ray source. The fainter NELG 512A lies somewhat closer to the X-ray position. While both galaxies may contribute to the observed X-ray flux, we assume object 512A as the counterpart in the following.
582A: Spectrum, optical, and X-ray image, and relative soft hardness ratios point undoubtedly to an M star as X-ray source. However, a flux ratio is unusually high for an M star as the X-ray identification.
607A and 653A: Detected as a single X-ray source by ROSAT Zamorani et al. (1999) X404-23, X-ray source 653 is brighter and closer to X404-23 and, therefore, treated as the detected ROSAT X-ray source. The broad spectral feature at 7100 in the optical spectrum of 607A is spurious due to the zeroth order light of the neighbouring slit.
615A and 615B: Both objects, the broad emission line object 615A and the narrow emission line object 615B, are possible counterparts to the X-ray source. We regard the fainter, but positionally better matching object 615A as the counterpart.
632A: Spectrum suggests a BL-Lac object, but the object is not a radio source, classification unclear.
Since type I AGN are a well-established class of X-ray emitters with relatively low surface density, false matches should not play any role for this object class. However, when investigating the classes of optically normal and narrow emission line galaxies, the problem of false matches has to be taken into account due to the high surface densities of these objects. To check the quality of our optical X-ray counterpart identification, we applied various tests.
First, the derived false match rate will depend on the assumed
position errors, which take into account the statistical
and the systematic position error (see Sect. 2.3).
The difference in X-ray and optical position scaled to a
position error is shown in Fig. 10a.
When we assume systematic errors of 2 arcsec, the theoretical Gaussian
distribution (Eq. (7))
we calculate the false match rate
by following Sutherland & Saunders (1992) and Ciliegi et al. (2003). For every spectroscopically
classified counterpart we determine the likelihood ratio L by
The total number of false matches can then be calculated by adding up all probabilities that the counterpart is spurious. Statistically, 25% of our narrow emission line galaxy are spurious counterparts. However, after visually examining objects with low probabilities , we find that in at least two cases (422A, 485A) the X-ray position given by the source detection software is influenced by blending with another X-ray source and the counterparts are consistent with the peak of the brighter X-ray source (see Appendix C, finding chart). Taking this into account, we estimate that the false match rate is 20% for the sample of type II AGN. For the optically normal galaxies the false match rate calculated with the likelihood method is 2%.
As a further quality check we plot the position error vs. WFI R-band magnitude. False matches would be recognisable due to different distributions of the type I and type II counterparts in this diagram. However, Fig. 11 shows that all object classes occupy the same regions in the diagram. Hence, the diagram does not indicate any serious contamination by false matches even for faint counterparts.
Our survey suffers somewhat from incomplete optical coverage of the X-ray survey area and, hence, a low identification rate over the whole area. To reach conclusions for the survey of a certain statistical significance we constrain our survey area and sample size to the central 0.28 deg2 with a pn-exposure >6 ks and significance of detection of individual sources with ML >10. We refer to this as the "core sample''. See Fig. 2 for the definition of the core region. Note that the maximum exposure is 35.5 ks for the PN camera and 78 ks for each of the MOS cameras.
The core sample contains 170 X-ray sources (ML>10). No optical data are available for six out of the 170 X-ray sources which fall outside the WFI R-band image and are also not covered by VLT pre-images. A further six X-ray sources from the core sample have no optical detection in the WFI R-band image in a 3 position error circle ( ). 122 X-ray sources are new detections, while 35 are spectroscopically classified ROSAT X-ray sources, and 13 are optically unidentified ROSAT sources. Out of the 140 spectroscopically classified objects in the Marano Field, 110 are associated with X-ray sources of the core sample. A summary of the properties of the core sample is given in Table 6.
Figure 12 shows R-band magnitude histograms of the core sample X-ray sources that were spectroscopically classified and only optically identified, respectively. The ( , ML)-distribution of X-ray sources in the core region is given in Fig. 13. The identification ratio of the core sample is 65%. In the next subsections we always refer to this sample.
|Figure 12: WFI R-band magnitude histogram for the core sample in the XMM-Newton Marano Field survey. The solid line represents the R-magnitude distribution of the spectroscopically classified X-ray counterparts. The dashed line shows the distribution of the optically identified but not spectroscopically classified X-ray counterparts. Six X-ray sources, which have no counterpart within a 3 error circle, are assigned to the bin at R=25.|
|Open with DEXTER|
|Figure 13: Likelihood vs. 0.2-10 keV X-ray-flux of the 252 central X-ray sources (pn-detector exposure time 6 ks). Crosses mark type I AGN, asterisks type II AGN, triangle galaxies, and rectangles stars. Dots indicate X-ray sources with no spectroscopic classification. The vertical line indicates ML = 10.|
|Open with DEXTER|
|Figure 14: X-ray diagnostic diagram based on hardness ratios. HR2 is calculated by using the 0.5-2.0 keV and 2.0-4.5 keV band, HR3 2.0-4.5 keV and 4.5-7.5 keV. Only X-ray sources that have are plotted. Labels: crosses - 63 type I AGN, asterisks - 24 type II AGN, triangles - 4 galaxies, and open circles - 24 unidentified X-ray sources with ML > 10. Large symbols represent objects with , small symbols . The solid line with box symbols (at ) represents an unabsorbed power-law X-ray spectrum (corrected for galactic absorption) with different photon index in steps of 0.2 (squares). Different hydrogen column densities ( /cm ; see Sect. 4.4) are plotted as a solid line with diamonds for and z=1.|
|Open with DEXTER|
The small number of optically normal galaxies span a large range in HR2. Two have HR2 values that belong to the softest in the whole core sample. The other two have X-ray spectra similar to those of soft type II AGN or very hard type I AGN.
In addition to the spectroscopically classified objects, we also plot unidentified objects from the core region. Out of the 60 total unidentified X-ray sources, only 24 meet the selection criterion of . Based on the rather clear separation between type I and II AGN, one may assign a likely classification to the yet unidentified sources. Among the 24 unidentified sources with reliable X-ray colours, the numbers of type I and type II AGN candidates appear to be similar.
|Figure 15: Redshift distribution of AGN in the Marano Field. a) Complete sample of 64 optically selected quasars ( ) by Zitelli et al. (1992); b) ROSAT X-ray sources (29 type I AGN by Zamorani et al. 1999); c) 35 type I AGN newly detected by XMM-Newton (core region); d) redshift distribution of the core sample: dotted line: all 70 XMM-Newton detected type I AGN, solid line: 31 XMM-Newton detected type II AGN, and dashed line: 5 XMM-Newton detected optically normal galaxies.|
|Open with DEXTER|
Previous extensive studies of AGN in the Marano Field enable us to compare these samples with our XMM-Newton detections. The optical survey by Zitelli et al. (1992) covers 0.7 . The selected quasars, which are all of type I, show an almost flat distribution in redshift (Fig. 15a) up to z=2.8. The ROSAT survey in the field (Zamorani et al. 1999) recovered most of the optically selected quasars at redshifts up to (Fig. 15b). The newly detected ROSAT AGN, with few exceptions, are type I AGN, which is expected due to the limited capability of ROSAT to detect absorbed sources.
In the core region of the XMM-survey we have detected 23 of the 29 broad emission line quasars of the optically selected sample of Zitelli et al. (1992). The detection rate of optically selected quasars remains constantly high over all redshifts. However, looking at the type I AGN newly discovered with XMM-Newton (Fig. 15c), it is apparent that the X-ray selection tends to detect quasars at lower redshifts than the optical surveys. This is particularly obvious from the redshift distribution of the ROSAT detected quasars, representing the brightest X-ray sources in the field. But the mean redshift of the XMM-Newton detected type I AGN is also lower than that of the optically selected sample , despite the fact that the XMM-Newton-observations are deeper in terms of the surface density of quasars than the optical survey. Possible reasons for these differences are discussed in Sect. 6.
|Figure 16: Observed X-ray luminosity (0.2-10 keV) vs. redshift. Labels: crosses - 70 type I AGN, asterisks - 30 type II AGN, and triangles - 5 galaxies.|
|Open with DEXTER|
|Figure 17: Observed X-ray luminosity (0.2-10 keV) vs. hardness ratio. HR2 is calculated by using 0.5-2.0 keV & 2.0-4.5 keV bands. Only X-ray sources that have are plotted. Labels: crosses - 63 type I AGN, asterisks - 24 type II AGN (grey marked have ), and triangles - 4 galaxies. The diagram separation into different regions is based on Szokoly (2004) with an adjusted threshold in the HR2-value.|
|Open with DEXTER|
The redshift distribution according to object class of the XMM-Newton detected core sample is given in Fig. 15d. We find that almost half of the new XMM-Newton sources are classified as type II AGN with redshifts mostly below 1.0. Type I AGN extend over a wide range of redshifts with a maximum at . Type II AGN are comparable to type I AGN in number density at low redshifts, but are mostly found below z=1. Five type II quasars at z > 2.2 have been identified. Optically normal galaxies without emission lines are found at z < 0.9.
For the X-ray sources with measured redshifts X-ray luminosities can be computed. The coverage of the survey in redshift - X-ray luminosity space is plotted in Fig. 16. Szokoly (2004) showed that the different object classes identified in X-ray surveys occupy different regions in a diagram of hardness ratios versus observed X-ray luminosities. In Fig. 17 we follow a scheme similar to that adopted by Szokoly (2004) and consider objects with an X-ray luminosity log as quasars (QSOs). The majority of type I AGN (70%) are actually type I QSOs. Most of type II AGN (71%) are low X-ray luminosity objects. Only type II objects with redshifts (marked in grey in Fig. 17) have X-ray luminosities of type II QSOs.
High-redshift type II QSOs with intrinsic absorption are found to be indistiguishable from non-absorbed type I AGN on the basis of their X-ray spectral hardness ratios, since the absorbed part of the spectrum is shifted out of the observable spectral window towards lower energies. This explains the emptiness of the upper right corner of Fig. 17 labelled "type II QSO'', a classification that applies to low redshift objects only.
Optically normal galaxies vary clearly in X-ray luminosity. The HR2-soft objects have very low X-ray luminosities. The two HR2-hard normal galaxies are found in the same region as the softest type II AGN, but are harder than type I AGN.
Szokoly (2004) use HR2=-0.2 as a threshold for the separation of type I and type II objects. Assuming a , this value corresponds to a hydrogen column density /cm -2=1022, 1023 for z= 0.25 and 2.1, respectively. For their CHANDRA observation they computed in the 0.5-10 keV band and their hardness ratio was based on the 0.5-2 keV and 2-10 keV bands. Because their definition of the X-ray bands differs from our study, the majority of their objects have higher hardness ratios (compared to our HR2). Furthermore, the XMM-Newton pn-detector, which was used for calculating the hardness ratios, has a higher efficiency in the 0.5-2 keV band compared to the CHANDRA detector. Therefore, we lowered the HR2 threshold from HR2 = -0.2 in Szokoly (2004) to HR2 = -0.45. Our threshold corresponds to /cm , and for and z= 0.25, 1, and 2.1. This is about two times lower than the cutoff Szokoly (2004) are using.
The hardness ratio diagram (Fig. 14) supports the view that the majority of type II AGN and a small fraction of type I AGN are obscured sources. Hence, the observed X-ray luminosity does not represent the intrinsic object X-ray luminosity. The significant deficit of soft photons as compared to a power-law spectrum reflects the existence of an absorbing component that is expressed by the hydrogen column density . For most of the sources the number of detected counts is not sufficient to extract a spectrum and fit a power-law model with and the photon index as free parameters. We, therefore, applied a technique that uses the measured hardness ratios to calculate the value for each source with a set of fixed power-law indices. Mainieri et al. (2002) found a mean value of for 61 type I and type II AGN in the Lockman Hole. The majority of type I and II AGN are found in the range of . The finding is confirmed by Mateos (2005). They find with a . Therefore, we use the observed pn-, mos1-, and mos2-hardness-ratios (0.2-0.5 keV, 0.5-2.0 keV, 2.0-4.5 keV, and 4.5-7.5 keV) and performed three runs to determine with the values , and 2.3 for all objects.
First of all, we computed a grid of model hardness ratios for
all EPIC instruments with Xspec
(using the models wabs, zwabs, and powerlaw).
As input the galactic absorption in the line of sight in the field
cm-2, the redshift
of the object and a grid of hydrogen column densities
We then computed the
values for the deviations of the
measured hardness ratios and their model values, summed over
the three instruments and the three hardness ratios HR1, HR2, and HR3.
The procedure was applied only to
those sources that had at least five out of nine hardness ratios
The of a source is determined by finding the minimum of
We tested our procedure by performing an individual Xspec -fit for the brightest type II AGN, which has sufficient X-ray data quality. The hardness ratios of 32A (z=2.727) indicate the highest absorption among the brightest type II AGN. A spectral fit with wabs, zwabs, powerlaw, and as a free parameter determined and ). The best fit , for a fixed value of , with the hardness-ratio -minimum fit is . A spectral fit with a fixed results in . The given in Table 8 was performed with the hardness-ratio -minimum fit and finds . Hence, both fit methods give comparable results, at least for high SNR sources.
Figure 18 shows the computed hydrogen column densities
for type II AGN and optically normal galaxies.
|Figure 18: Calculated intrinsic hydrogen column density histogram for the central Marano Field. The solid line shows the distribution of type II AGN that have reasonable fits in hydrogen column density. Objects with are set to . The dotted line represents only type II AGN that show intrinsic absorption with a significance higher than in hydrogen column density. Optically normal galaxies are plotted as a dashed line. Dotted line - 14 type I AGN; solid line - 30 type II AGN; dashed line - 5 galaxies.|
|Open with DEXTER|
|Figure 19: a) Histogram of the absorbed (observed) X-ray luminosity (0.2-10 keV) for the core sample. Dotted line - 70 type I AGN, solid line - 30 type II AGN, and dashed line - 5 galaxies. b) Histogram of the X-ray luminosity (0.2-10 keV) corrected for intrinsic absorption (core sample). Same notation as in a). c) Corrected X-ray luminosity (0.2-10 keV) vs. hydrogen column density (core sample). Crosses - type I AGN, asterisks - type II AGN, and triangles - galaxies. The upper right corner is suggested as being defined as the "type II QSO region'' by Mainieri et al. (2002). Objects with are set to .|
|Open with DEXTER|
The computed hydrogen column density of type II AGN object 145A is the highest in our sample ( ). Since this corresponds to the highest value in the model grid, no reliable error estimate can be given for this object. However, the extremely high column density for this object is confirmed by the fact that it is one of the few sources detected in the 4.5-7.5 keV EPIC images, but it is not visible in the softer bands.
Based on the hydrogen column densities, we calculated the unabsorbed (intrinsic) X-ray luminosities by computing a correction factor for the observed X-ray flux. Figure 19 shows the comparison of the absorbed (a) and unabsorbed (b) X-ray luminosity distributions. Even after correcting the X-ray luminosity, type II AGN have lower median X-ray luminosities than type I AGN, although type I and II AGN cover the same X-ray luminosity range. In contrary to the almost flat distribution of type II AGN with a median of , type I AGN show a significant peak at in observed and intrinsic X-ray luminosity. Type I AGN 5A exhibits the highest X-ray luminosity in our sample.
The hydrogen column density as a function of the intrinsic X-ray luminosity is shown in Fig. 19c. Mainieri et al. (2002) suggest labelling the region defined by and as the "type II QSO region''. They proposed this classification based on 61 AGN identified from XMM-Newton sources in the Lockman Hole. Following this classification our sample includes 10 type II QSOs.
In our case this region of the plot is also populated by nine type I QSOs, which formally have intrinsic hydrogen column densities . However, most of these detections of intrinsic have low significance; only four of all type I QSOs show absorption at the level (see Table 7). For about half of the type II AGN, significant intrinsic absorption ( ) was measured, and no dependence of the absorbed fraction on luminosity is evident (Table 7).
The distribution of our core sample in the -plane is illustrated in Fig. 20. In general, type I AGN show higher X-ray fluxes and are brighter in the R-band. Type II AGN are found at lower X-ray fluxes and have fainter R-band counterparts. Two of the optically normal galaxies have X-ray-to-optical flux ratios similar to type I and type II AGN. Another two are among the objects with lowest X-ray-to-optical flux ratios in the sample.
|Figure 20: Observed X-ray flux (0.2-10 keV) vs. R-band magnitude. Dashed lines indicate the values 0.1, 1.0, and 10. Labels: crosses - 63 type I AGN, asterisks - 25 type II AGN, triangles - 5 galaxies, rectangles - 3 stars, and circles - 21 spectroscopically unidentified sources (R-band lower limits are plotted as circles at R=24). Note that for a few objects of the core sample no WFI R-magnitudes were available and therefore these are not plotted here.|
|Open with DEXTER|
Our types I and II AGN show a large variety in X-ray flux and R-band magnitude. They are detected at X-ray fluxes from to erg cm-2 s-1 and in R-band magnitudes from down to the detection limit of . There is one type I AGN (191A) at an unusually low value , a factor of 100 below the mean value of type I AGN. The highest ratio in X-ray to optical flux is found for the type II AGN object 39A with .
Most of the stars are found at star-typical with R < 17. Nevertheless, we also detected two M stars with and . Unidentified sources with WFI R-band data are added to Fig. 20. We plotted unidentified objects that have no detection in the R-band catalogue at R = 24.
Figure 21 shows that type II AGN tend to be fainter and redder in the optical window than type I AGN. This figure shows a general trend for both type I and type II AGN to become redder for fainter R-magnitudes. The lack of faint blue objects can be explained by the K-band detection limit of K=20, where we lose completeness (see Fig. 6). However, the lack of bright red objects cannot be caused by any detection bias. The type II AGN have redder R-K colours, although with some overlap with the reddest type I AGN in the sample. These trends can be explained by an increasing contribution of the host galaxies for fainter type I AGN and type II AGN. The faint R-magnitudes and high R-K values indicate higher optical obscurations in type II AGN. This is in agreement with previous X-ray studies of these objects that revealed a significantly higher ratio of absorbed to unabsorbed objects compared to type I AGN.
In our survey type I and type II objects do not separate as clearly as seen in a similar plot in Mainieri et al. (2002). Therefore, we checked whether the reddest and faintest type I objects are reliably classified. The most extreme type I objects are 69A, 84A, and 585A. Their optical spectra clearly show broad emission lines, but their continua are redder than typical type I spectra. In addition, object 585A shows only weak emission lines.
|Figure 21: Colour magnitude diagram of spectroscopically classified X-ray sources. Labels: crosses - 56 type I AGN, asterisks - 20 type II AGN, and triangles - 4 galaxies. Only objects with detections in both R and K are included. The absence of blue, faint objects (below the solid line) is due only to the K-magnitude limit.|
|Open with DEXTER|
The optically normal galaxies are brighter in R-magnitudes than the typical type II AGN, but show similar R-K. All classified galaxies have R<21.5. The majority of type II AGN and galaxies have , consistent with the spectrum of the host galaxy being the dominating component. A considerable fraction of type I AGN also have red R-K colours. These are mostly low luminosity objects, where the host galaxy may also dominate the optical continuum.
|Figure 22: Redshift distribution of the additional objects. Labels: solid line - 37 NELGs, and dotted line - 8 optically normal galaxies.|
|Open with DEXTER|
Figure 22 illustrates the redshift distribution of the additional non-X-ray emitting objects in the Marano Field. Narrow emission line galaxies (NELG) outnumber normal galaxies substantially. This is due to the fact that a large fraction of the spectra without emission lines did not have sufficient SNR to determine a redshift. There is a non-X-ray emitting NELG peak at z =0.5-0.9. At z > 1.0 the detection of NELG and normal galaxies drops dramatically.
Despite the low number of objects, from Figs. 22 and 15d we can compare the redshift distributions between X-ray emitting type II AGN and non-X-ray emitting NELG. Both populations show more or less the same distribution up to z = 1.2. The two groups peak at z =0.5-0.9.
The similar redshift distributions could mean that the X-ray emitting NELGs are drawn from the same population of galaxies as the control sample. On the other hand, the similar redshift distribution could also be due to a large fraction of false matches of NELGs to our X-ray sources. A detailed discussion of the number of false matches is given in Sect. 3.2 and shows that false matches make only a minor contribution to our sample of narrow line AGN.
In Sect. 4.2 we showed that the X-ray selected type I AGN peak at lower redshifts than the optically selected sample. It is interesting to see whether this difference in the redshift distributions is due to a redshift dependence of the QSO spectral energy distributions (SED). As a parameter which characterises the SED, we calculated the optical to X-ray broad band spectral index between the UV luminosity density at 2500 and the X-ray luminosity density at 1 keV (see Sect. 3).
No significant correlation of with redshift (Fig. 23) or X-ray luminosity can be found. However, there is a very significant ( ) correlation of the optical luminosity and (Fig. 24). This correlation has been found in various samples observed with EINSTEIN (Avni & Tananbaum 1986), ROSAT (e.g., Green et al 1995; Lamer et al. 1997), and CHANDRA (e.g., Steffen et al. 2006). For a large sample of optically selected AGN, Steffen et al. (2006) computed the bivariate linear regression coefficients of as a function of and z. They find that is correlated with , but find no significant correlation with z (see also Avni & Tananbaum 1986 for similar results on earlier EINSTEIN data).
If and optical luminosity are correlated, a non-linear relation between X-ray luminosity and optical luminosity is expected. Therefore, we plotted the X-ray luminosities versus the optical luminosity densities (Fig. 25) and computed the linear regression coefficients between and . Both variables and are measured quantities and neither of them can be regarded as the independent or dependent variable. We used the ordinary least-squares (OLS) bisector algorithm as described by Isobe et al. (1990), which is symmetric regarding the choice of independent and dependent variable.
We find a best-fit regression . This slope is marginally (2 significance) flatter than the expected for a linear relation. With the same method Steffen et al. (2006) find a slightly flatter slope .
A correlation flatter than implies that the optical luminosities in the sample are spread over a wider range of values than the X-ray luminosities. This might explain the above-mentioned discrepancies of the redshift distributions of X-ray and optically selected QSOs. The low luminosity objects are more likely to be detectable in X-rays, while the highest luminosity objects are relatively more luminous in the optical, and therefore detectable at higher redshifts in optical surveys.
|Figure 23: UV to X-ray spectral index vs. redshift for type I AGN in the core sample.|
|Open with DEXTER|
|Figure 24: UV to X-ray spectral index vs. optical luminosity density at for type I AGN in the core sample.|
|Open with DEXTER|
|Figure 25: Observed 0.2-10 keV X-ray luminosity vs. optical luminosity density at for type I AGN in the core sample (crosses). The solid line indicates the best-fit linear regression with a slope . The dotted line with slope marks a linear relation.|
|Open with DEXTER|
Type II AGN show only narrow emission lines and the optical continuum radiation is dominated by the host galaxy. The spectroscopical classification of X-ray sources relies on emission lines like Ly-, C IV, C III, Mg II, O II, and others. Analysis of the optical spectra of type II AGN with z > 0.6 in our sample, in Szokoly et al. (2004), and in Caccianiga et al. (2004) indicates that almost all objects show no Mg II emission lines. Only in a few cases marginal Mg II emission is recognised. Effectively, no strong spectral features are present in the wavelength range between (C III]) and ([O II]).
Since the UV host galaxy continuum is usually very faint, the detection of type II AGN in optical imaging is already hampered by the lack of a strong UV continuum and emission lines, if the imaging bandpass falls into this rest frame range. For R-band imaging (5800-7300 ) this is the case for the redshift range . Our multi-object spectroscopy covers a useful range of Å. This range is reduced, if an object is not centred on the mask, but offset from the centre in the dispersion direction. Hence, in many cases only one emission line would be detectable for type II AGN in the redshift range 1.0-2.0. With a low SNR spectrum this would usually not be sufficient for a clear spectroscopical classification. For a few sources, which were undetected or very faint in the R-band images, we were able to position a MXU slit using the K-band counterparts (e.g., 20A, 63A, and 463A). In all of these cases a high redshift type II AGN could be identified.
Figure 14 supports the assumption that a large fraction of unidentified sources consists of type II AGN. Above HR2=-0.1 only type II AGN are found as counterparts for identified X-ray sources. In this region 50% of the X-ray sources are not classified and are expected to be type II AGN. Furthermore, Fig. 20 gives more evidence that the unidentified sources overlap with the type II AGN region. A ratio is clearly a sign of AGN activity, since normal galaxies and stars usually have . Both type II AGN and unidentified objects are found at AGN-typical , with low X-ray fluxes and faint optical counterparts.
We conclude that the majority of the unidentified X-ray sources are likely to consist of type II AGN, most of them presumably with redshifts z>1. Therefore, the intrinsic redshift distribution of the type II AGN is uncertain. The number of unidentified sources is sufficient to fill the observed gap between .
Five objects in the core sample have been spectroscopically classified as galaxies without any emission lines. The soft X-ray radiation of the low X-ray luminosity galaxies (objects 120A, 241A) can be explained by a halo of X-ray emitting hot gas around elliptical galaxies (; White & Davis 1998). Three optically normal galaxies have X-ray-to-optical flux ratios and X-ray luminosities typically found for AGN. Objects of this type have been named X-ray bright optically normal galaxies (XBONGs; Comastri et al. 2002). The objects 8A, 49A, and 204A have intermediate redshifts (0.3<z<0.9) and fairly high X-ray luminosities (log ), rather hard X-ray spectra, and faint optical counterparts (R>19). The high X-ray luminosities and indicate active nuclei. The computed hydrogen column densities show moderate absorption.
The properties of our XBONGs are in agreement with other studies (Silverman et al. 2005; ). Even though different scenarios are discussed, the nature of XBONGs still remains unclear (Brandt & Hasinger 2005). Comastri et al. (2002) assume that the non-detection of optical emission lines and the hard X-ray colours are due to a heavily absorbed AGN embedded in a galaxy whose X-ray emission is due to a scattered/reprocessed nuclear component. The three XBONGs do not show harder HR2 values than type II AGN and also do not have the highest column densities of the AGN sample. Hence, a heavily absorbed AGN without scattered/reprocessed X-ray radiation is ruled out. Scattered or reprocessed radiation is necessary to explain the X-ray and optical observations. Following Komossa et al. (1998), the observed X-ray luminosity of a scattered emission by a warm reflector is only a hundredth or thousandth of the intrinsic X-ray luminosity. For our objects with observed log , this would require intrinsic X-ray luminosities of log , which would make them by far the most luminous X-ray emitters in the sample.
Another possible explanation for these objects is given by Severgnini et al. (2003). They studied three low redshift XBONGs with log and find values similar to our sample. In their interpretation the faint emission lines of an obscured or unobscured AGN up to an intrinsic log can be overwhelmed by a host galaxy with an absolute magnitude . However, XBONGs 8A and 49A are not consistent with the scenario mentioned by Severgnini et al. (2003). The AGN of 49A with an intrinsic log should be optically too bright to be hidden by a galaxy of . Moreover, source 8A, which is also detected as a radio source (Gruppioni et al. 1999), has an intrinsic log . Therefore, it is more X-ray luminous by almost one order of magnitude than the examples in Severgnini et al. (2003). The absolute magnitude of the host galaxy is , but an optically much brighter galaxy is needed to hide the emission lines of such a powerful AGN. By adding a template type I spectrum to the measured spectra of objects 8A and 49A, we estimated that any hidden type I AGN in these objects would have values of 0.8 or less.
Regarding the X-ray and optical colours, our XBONGs are very similar to type II AGN. Therefore, it is likely that these sources have a narrow line type II spectrum, which is intrinsically weak or dust absorbed and not detected above the continuum of the host galaxy. This result is consistent with Caccianiga et al. (2004), who state that more accurate re-observation (high resolution data and/or better spectral coverage) of hard X-ray emitting galaxies will reveal narrow emission lines and, therefore, their real AGN nature. Severgnini et al. (2003) also claim possible misclassification of type II AGN as XBONGs. However, some of these objects are still classified as XBONGs after a high resolution observation with better spectral coverage.
As for type II AGN, the observed redshift distribution of XBONGs could be due to an observational bias. XBONGs are found with brighter R-magnitudes than typical type II AGN up to z=1. But missing emission line features make them optically even more difficult to identify than type II AGN. For a reliable classification they have to be optically brighter, thus limiting their maximum redshift.
With a total of 120 ks good observation time we detect 328 X-ray sources. Among 140 spectroscopic classifications of 187 optical counterparts (in a 3 position error with ) to 328 X-ray sources (not completely covered by optical data), we find 89 broad emission line objects, 36 narrow emission line objects, 6 galaxies, and 9 stars. In the central region of the Marano Field we reach an identification completeness of 65%.
While the redshift distribution of the optically selected QSOs in the field is basically flat up to , the distribution of the XMM-Newton sources peaks at . Using our sample of XMM-Newton sources classified as type I AGN, we investigate possible causes for this tendency of deep X-ray surveys to discover faint populations at comparably low redshifts. We find no significant correlation of the optical to X-ray SED slope with redshift. As is widely reported in the literature, is tightly correlated with optical luminosity. A different representation of this correlation is the non-linear dependency . The best-fit regression implies that the optical luminosities in a typical sample spread over a wider range than the X-ray luminosities. Therefore, the less luminous objects of the population are more easily detected in X-rays than in the optical. On the other hand, an increase of optical luminosity is, on average, not accompanied by a proportional increase in X-ray luminosity. Hence, the luminous (and more distant) objects are detected more efficiently in optical surveys.
In the core region of the field we classified 31 new type II AGN. Most of them are found at redshifts z<1.5; additionally we find five high redshift type II AGN at z>2.2. Fifteen objects can be classified as type II QSOs with intrinsic X-ray luminosities erg s-1.
We show that the optical identification of type II AGN is very difficult in the redshift range z=1-2 due to the absence of suitable emission lines in the optical window. Therefore, their intrinsic redshift distribution remains unclear. We demonstrate that the use of K-band data for MXU slit positioning reveals type II AGN or XBONGs that would have likely been missed in R-band images. The X-ray selected type II AGN have a very similar redshift distribution to non-X-ray emitting narrow emission line galaxies, which have been spectroscopically classified in the same field as a control sample.
The intrinsic hydrogen column densities, as derived from X-ray hardness ratios, show that the fraction of absorbed X-ray sources is much higher for type II AGN than for type I. Nevertheless, we find a few unabsorbed type II AGN and some evidence for absorption in high redshift type I AGN. However, due to the faintness of the sources, the significance of absorption in the individual type I AGN is low. Furthermore, at high redshifts statistical fluctuations in the X-ray spectrum can lead to high values of spuriously measured values (e.g., Akylas et al. 2006). If we only include detections of intrinsic in our analysis, only 4 absorbed type I AGN remain and no dependency of absorbed fraction on redshift is obvious. In the CHANDRA data from Chandra Deep Field South (CDFS), Tozzi et al. (2003) find hints of an increase of absorbed fraction with redshift. However, these data probably also suffer from the uncertainties mentioned above. Using XMM data in the same field, Dwelly & Page (2006) find little evidence that the absorption distribution is dependent on either intrinsic X-ray luminosity or redshift. Our type I and type II AGN cover the same range in absorption corrected X-ray luminosity. However, the mean corrected X-ray luminosity is smaller for type II AGN than for type I AGN.
Three of the XMM-Newton classifications are X-ray bright optically normal galaxies (XBONGs), which show X-ray luminosities typical for AGN, but no optical emission lines. Their X-ray luminosities of log are comparable to the mean type II AGN X-ray luminosity. They do not show harder X-ray spectra and do not reveal higher hydrogen column densities than the average type II AGN. We conclude that the objects are very similar to type II AGN. However, their narrow emission lines are not detected, since they are either intrinsically weak or obscured by dust.
Mirko Krumpe is supported by the Deutsches Zentrum für Luft- und Raumfahrt (DLR) GmbH under contract No. FKZ 50 OR 0404. Georg Lamer acknowledges support by the Deutsches Zentrum für Luft- und Raumfahrt (DLR) GmbH under contract No. FKZ 50 OX 0201.
Table A.1: X-ray source list.
Table B.1: Optical properties of candidate counterparts of marginal Marano Field XMM-Newton X-ray sources. For a detailed explanation of the individual columns see Sect. 3.
137A: Broad absorption line quasar.
315A and 315B: As counterpart candidates for X-ray source 315, an M star (315A) and a NELG (315B) are found. The hardness ratios of 315 indicated an obscured X-ray object. Therefore, we regarded 315B as the identification.
Table B.2: X-ray source list for marginal X-ray sources. The explanation of the individual columns is given in Appendix A.
(2) RA [hh:min:s] and (3) Dec [degmin:s]
Right ascension and declination of the spectroscopically classified object.
Spectroscopic redshift of the classified object.
SOFI K-band magnitude of the spectroscopically classified object, whenever possible.
WFI R-band magnitude of the spectroscopically classified object, whenever possible.
Classification of the optical identified object. S - star, G - normal galaxy (no emission lines), and N - narrow emission line galaxy (unresolved emission lines).
Table D.1: Optical data of additional objects.