A&A 487, 539-554 (2008)
G. Worseck1 - L. Wisotzki1 - F. Selman2
1 - Astrophysikalisches Institut Potsdam, An der Sternwarte 16, 14482 Potsdam, Germany
2 - European Southern Observatory, Alonso de Cordova 3107, Casilla 19001, Vitacura, Santiago, Chile
Received 7 May 2008 / Accepted 10 June 2008
We present the results of the ``Quasars near Quasars'' (QNQ) survey, a CCD-based slitless spectroscopic survey for faint quasars at on 18 fields centred on bright quasars at 2.76<z<4.69. In total 169 quasar candidates with emission lines were selected from the extracted flux-calibrated spectra on the basis of well-defined automatic selection criteria followed by visual inspection and verification. With follow-up spectroscopy of 81 candidates that were likely to reside at z>1.7 we were able to confirm 80 new quasars at on 16 of our fields. 64 of the newly discovered quasars are located at z>1.7. The overall high success rate implies that most of the remaining 88 candidates are quasars as well, although the majority of them likely resides at z<1.7 on the basis of the observed line shapes and strengths. Due to the insufficient depth of the input source catalogues needed for extraction of the slitless spectra our survey is not well defined in terms of limiting magnitude for faint quasars whose Ly emission is detectable well beyond V=22, albeit at a continuum . While not useful for characterising the evolving space density of quasars, our sample provides many new closely spaced quasar sightlines around intensely studied quasars for further investigations on the three-dimensional distribution of the intergalactic medium.
Key words: surveys - galaxies: quasars: general - galaxies: intergalactic medium - cosmology: large-scale structure of Universe
Recent large optical surveys, such as the Sloan Digital Sky Survey (SDSS) and the 2dF QSO Redshift Survey (2QZ), have revealed thousands of previously unknown quasars selected on the basis of their broadband optical colours (Croom et al. 2004; Schneider et al. 2007). Colour selection is efficient if quasar candidates are well separated from normal stars in multidimensional colour space, most notably at (UV excess) and at . However, even optical multicolour surveys are systematically incomplete at , where the colours of quasars and stars are similar (e.g. Warren et al. 1991; Richards et al. 2002a). Incompleteness in this redshift range can be significantly reduced by a better tracing of the spectral energy distributions with additional filters, e.g. by incorporating mediumband filters as in the COMBO-17 survey (Wolf et al. 2003).
Alternatively, slitless spectroscopy is a particularly efficient way to find quasars at redshifts because of the prominent Ly emission line redshifted into the optical wavelength regime. Early surveys generated hundreds of quasars by visual scanning of objective-prism photographic plates for emission-line objects (e.g. Osmer & Smith 1980; Crampton et al. 1985). Subjecting such plates to digitisation with fast measuring machines made it possible to employ the automated selection of quasar candidates (Hewett et al. 1985; Clowes et al. 1984) and to build substantial quasar samples at redshifts (Wisotzki et al. 2000; Hewett et al. 1995). The systematic CCD-based slitless survey for quasars by Schneider et al. (1994) was among the first to quantify the declining space density of high-redshift quasars (Schmidt et al. 1995).
Apparent pairs or close groups of high-redshift quasars are very attractive targets to study the three-dimensional distribution of the intergalactic medium (IGM). But high-resolution studies have so far been limited to a small number of suitable groups of bright quasars (e.g. D'Odorico et al. 2002,2006). Going fainter than immediately limits the achievable spectral resolution and S/N. A possible compromise lies in combining high-resolution spectra of bright quasars with lower resolution, lower S/N data of fainter quasars in the surroundings. Pichon et al. (2001) argued that one can significantly improve the recovery of the 3-dimensional topology of the IGM this way.
Although most known quasars were colour-selected from either SDSS or 2QZ, these surveys produced relatively few useful quasar groups because of the reduced selection efficiency at combined with a bright magnitude limit. In fact, most well-studied close groups of quasars at z>2 were found by slitless spectroscopy. Follow-up spectroscopy of candidates by Bohuski & Weedman (1979) revealed 13 z>1.5 quasars on 2.1 deg2 (Jakobsen & Perryman 1992) with two showing correlated complex intergalactic C IV absorption at 1.48<z<2.15, indicative of an elongated superstructure extending over on the sky (e.g. Jakobsen et al. 1986; Dinshaw & Impey 1996). Another group discovered by Sramek & Weedman (1978) contains now 6 QSOs at within a radius of 20 around Q 1623 + 2653, all of which are bright enough for correlation studies of the IGM (e.g. Crotts & Fang 1998). Williger et al. (1996) reported 25 quasars at within a 1 deg2 region and used these to reveal large-scale structure in the IGM (Liske et al. 2000; Williger et al. 2000). At lower redshifts, slitless surveys revealed large associations of quasars at similar redshift (Clowes et al. 1999; Crampton et al. 1990).
Groups of quasars, or more generally speaking, active galactic nuclei, have also been discovered by recent deep multi-wavelength surveys in selected fields, such as in the Chandra Deep Fields North (Barger et al. 2003; Cowie et al. 2004) and South (Szokoly et al. 2004; Wolf et al. 2004), the Marano field (Krumpe et al. 2007; Zitelli et al. 1992), or the COSMOS field (Prescott et al. 2006; Trump et al. 2007). However, the majority of the AGN thus found are too faint for follow-up studies at a spectral resolution allowing for meaningful IGM studies.
Complementary to slitless spectroscopy, quasar pairs can be found from similarities in multi-colour space. Using this approach, Hennawi et al. (2006b) confirmed 40 new associated and 73 projected quasar pairs with separations from a sample of faint quasar pair candidates selected via SDSS photometry. Follow-up spectra revealed transverse clustering of optically thick absorption systems near foreground quasars (Hennawi et al. 2006a; Hennawi & Prochaska 2007).
Targeted deep surveys of sky regions around well-studied high-redshift quasars are rare. An exception are many of the fields selected for the Lyman-break galaxy survey by Steidel et al. (2003) which were centred on bright quasars to correlate the galaxies with the intergalactic absorption along the sightline (Adelberger et al. 2005). Here we describe a systematic search for apparent quasar groups at in the southern hemisphere, targeting fields centred on known bright z>2.7 quasars that had been observed at high resolution with the UV-Visual Echelle Spectrograph (UVES) at the VLT. We already reported results for two special fields (Worseck & Wisotzki 2006; Worseck et al. 2007, hereafter Papers I and II, respectively). The present paper is devoted to present the entire survey. In Sect. 2 we describe the slitless spectroscopic survey observations. Section 3 outlines the automatic reduction pipeline developed for these data. Section 4 describes the semi-automatic selection of quasar candidates. We report on the follow-up slit spectroscopy of candidates in Sect. 5, followed by a brief discussion of the properties of confirmed quasars and the remaining candidates (Sect. 6). In Sect. 7 we present the resulting quasar groups and discuss the efficiency and completeness of our survey. We conclude in Sect. 8.
The survey was carried out using the ESO Wide Field Imager (WFI, Baade et al. 1999) at the ESO/MPI 2.2 m Telescope (La Silla) in its slitless spectroscopic mode (Wisotzki et al. 2001). Since this mode has been rarely used we shortly describe its main characteristics.
The WFI is a focal-reducer type camera offering a field of view of sampled by a mosaic of 8 2k 4k CCDs with /pixel. In the slitless spectroscopic mode a low-resolution grism is placed in the converging beam of the telescope in front of the WFI in order to disperse the light of every object in the field of view. A blue-blazed grism and a red-blazed grism are available. However, the red-blazed grism (R50, dispersion 7 Å/pixel, blaze wavelength 6000 Å, Å FWHM) has a much higher throughput for 1st-order spectroscopy even in the blue (Wisotzki et al. 2001), rendering the blue-blazed B50 grism almost obsolete. We only used the R50 grism. As both grisms were originally constructed for the prime focus of the ESO 3.6 m telescope (long decommissioned) and only retrofitted into WFI, the size of the grisms does not fully match that of the WFI field of view. In a strip on the left side of the chip mosaic the light passes undispersed, effectively reducing the usable field of view to 6 of the 8 WFI CCDs, or .
During two visitor mode runs in October 2002 (5 nights) and February 2003
(3 nights) we observed in total 18 fields centred on bright high-redshift
quasars with available UVES spectra in order to find faint quasars in their
surroundings. We additionally included the Extended Chandra Deep Field South
(ECDFS) to compare our selection with the results of deeper multiwavelength
|Figure 1: Schematic view of the WFI dither pattern. The full lines denote the edges of the field of view in the dithered 600 s exposures. Dashed lines mark the 6 2k 4k chips with the inter-chip gaps. After rotating the instrument by about its origin in the focal plane (cross) three further 600 s exposures were taken.|
|Open with DEXTER|
Table 1: Observing log of the slitless survey observations. The columns list the quasar at the field centre with redshift and celestial coordinates, the night of observation, the employed filters, instrument rotations, total exposure time per band and the seeing.
The sky transparency during the October 2002 run was variable with some thick cirrus clouds passing occasionally, but mostly the sky was clear. Conditions were clear to photometric in the February 2003 run. The seeing varied over a broad range of values (see Table 1) but was mostly of the order of 1 . Since the spectral resolution in slitless spectroscopy is given by the seeing, this quantity varied by -50. We used the broadband B and V filters of the WFI to lower the effectively undispersed sky background and to reduce the degree of crowding by limiting the length of the spectra. Crowding was further accounted for by taking the spectra in two instrument rotations (0 and 10 ). For all fields (except the one centred on BR 1202-0725) three dithered 600 s exposures were taken per band and rotation, resulting in a total exposure time of 1 h per band. Figure 1 illustrates the observing pattern. The dithered exposures produced a contiguous field of view per rotation angle. The combination of the R50 grism with the broadband B and V filters resulted in a spectral coverage from the blue sensitivity cutoff at 4200 Å to 5800 Å. An unfiltered spectrum of the low-redshift emission-line galaxy HE 1250-0256 provided the input for an approximate wavelength calibration of the slitless data. The spectrophotometric standard stars HD 49798, LTT 7987 and GD 108 were observed in astronomical twilight for relative flux calibration. Blank sky fields observed during twilight provided flat fields. The V band exposures of the field centred on PKS 0528-250 were retrieved from the ESO Science Archive (PI Vanzi). Table 1 summarises the slitless survey observations.
The nominal surveyed area was 4.39 deg2. However, due to the two rotation angles of the instrument, the field edges received only 1/2 of the exposure time. Moreover, spectra located near interchip gaps are affected by dithering. The net exposure time per field is further decreased for some objects due to contamination by nearby other sources or spectral orders of bright stars (see Fig. 3 below). However, the two instrument rotations ensured that a clean spectrum of almost every object was obtained.
Direct images are necessary for object identification in the slitless data. We primarily relied on images from the Digitised Sky Survey (DSS) available for all our fields. For three fields we also employed direct WFI images taken in service mode.
|Figure 2: Flat-fielded sky-subtracted and combined slitless spectroscopic WFI field in the vicinity of HE 0940-1050. The spectra were taken with the B filter and without rotation (i.e. north is up and east is left on the image). The dispersion runs horizontally from left to right. The three dithered 600 s exposures were combined to fill the inter-chip gaps. The (linear) intensity of each pixel is weighted by its effective exposure time. The total dimension of the combined image is .|
|Open with DEXTER|
The slitless data were reduced with a semi-automatic pipeline running under ESO-MIDAS, but largely consisting of our own dedicated software modules. For several parts of the data reduction we could use the extensive toolbox developed for the slitless spectroscopy reduction pipeline of the Hamburg/ESO Survey (HES, Wisotzki et al. 2000). This was supplemented by new software where necessary. Each exposure was reduced separately before combining the extracted spectra. Briefly, the data reduction comprised the following steps:
|Figure 3: Masking of unwanted spectral orders. The upper panel shows a region of a sky-subtracted 600 s B band exposure of the field around HE 0940-1050. A bright star in this region generates the indicated different spectral orders in the slitless exposure that may contaminate 1st-order spectra of nearby sources. The white stripe marks a vertical inter-chip gap. The lower panel displays the corresponding region of the variance image. Black rectangles show masked spectral orders of the bright star and other objects. Dark vertical stripes correspond to masked bad columns on the chips.|
|Open with DEXTER|
|Figure 4: Sensitivity curves of the WFI B and V filters derived from the spectrophotometric standard HD 49798. At 4700 Å one triggered photoelectron per second per 6.7 Å pixel corresponds to a flux density of . Dotted lines mark the considered wavelength range.|
|Open with DEXTER|
The magnitude limit of our survey obviously depends on both the depth of the input source catalogues and on the obtainable S/N in the slitless spectroscopy data. We estimated the completeness of the source catalogues as a function of magnitude by considering the differential source counts. Figure 5 shows the differential surface densities per mag obtained from the DSS source catalogues in the 18 fields. From the breaks in the differential number counts we conclude that the input catalogues are reasonably complete to , with 4 fields being shallower by 0.5 mag. Similar limiting magnitudes were obtained in the three fields with direct WFI data.
The limiting magnitude of the slitless survey material was determined by
evaluating the continuum S/N of the extracted spectra. The V magnitude of
each object was calculated by integrating the flux over the V band which is
completely covered by the spectra. We estimate that the V magnitudes of
objects are accurate to 0.1 mag. Fainter sources have larger
uncertainties due to their lower S/N. The S/N is a function of the
wavelength-dependent system sensitivity (Fig. 4) and of
the source spectral energy distribution. We defined the median S/N in the
V band as a characteristic value. Recall that the extraction yielded a noise
array along with each spectrum, so that the S/N can be estimated rather
accurately. In Fig. 6 we plot the median S/N of all extracted
1800 spectra in one particular field,
as a function of V magnitude. One can see that a S/N of 3 is obtained
a S/N of unity is reached.
|Figure 5: Differential source density of the survey fields in mag bins derived from the DSS plates. The colours indicate different estimated limiting magnitudes ( : CTQ 0460, Q 0302-003, Q 1209+093, PKS 2126-158, : BR 1117-1329, BR 1202-0725, HE 0940-1050, HE 2243-6031, PKS 0528-250, Q 0420-388, Q 1451+123, Q 2139-4434, : CTQ 0247, Q 0055-269, HE 2347-4342, Q 0000-263, Q 0002-422, Q 0347-383).|
|Open with DEXTER|
Some objects appear to have too low S/N for their V magnitude; these
generally have shorter net exposure times induced by dithering and field
rotation. After excluding these outliers we fitted a simple log-linear
relation for the S/N as a function of magnitude,
We conclude that at a magnitude of , our survey should be reasonably complete in terms of the input catalogues, and that the slitless spectra at these magnitudes are generally still good enough to search for emission-line objects. We demonstrate below that the selection of even fainter z>2.4 quasars with prominent Ly lines was also possible, in some fields down to or beyond (corresponding to a continuum S/N of 1 in the slitless spectra). However, it is clear that only the most promising candidates will be selected at these faint levels. Moreover, we have seen that at , already the input catalogues are heavily affected by approaching to the limits of the DSS itself. We return to a more quantitative assessment of (in)completeness of the quasar samples in Sect. 7.2 below.
|Figure 6: Integrated WFI V magnitude vs. median S/N in the WFI spectra in the field of HE 0940-1050 (dots). The line shows the fitted log-linear relation of the magnitude and the S/N for this field.|
|Open with DEXTER|
|Figure 7: Visibility of strong emission lines over the covered wavelength range. Dashed lines mark gaps without major emission lines where quasar candidates are missed.|
|Open with DEXTER|
Quasar candidates were semi-automatically selected from the calibrated slitless spectra on the basis of prominent emission lines. Figure 7 shows the redshift ranges in which major quasar emission lines can be observed in the slitless spectra. The detectability of the various emission lines depends on their equivalent widths and the S/N in the data (Sect. 7.2). Moreover, quasar candidates could not be selected in two narrow low-redshift ranges due to the lack of major emission lines (see Fig. 7).
We first estimated a (pseudo-)continuum for each spectrum by applying a simple median filter with a window width of 400 Å. The continuum-subtracted spectra were searched automatically for emission line objects by cross-correlating them with Gaussian emission line templates. The fields around Q 0000-263, Q 0302-003 and Q 2139-4434 were chosen for testing the continuum subtraction and the line detection algorithm, ensuring that all emission line objects selected by eye were also found automatically. Finally, we chose two Gaussians with dispersions Å and of Å in order to match emission lines with small and large widths, respectively. Only features detected with a total line and an observed equivalent width of Å were kept.
At our spectral resolution, the continuum subtracted spectra of normal stars
often appeared to show ``emission lines'' that in reality were artefacts of poor
continuum estimation and oversubtraction. The initial samples of emission line
objects contained therefore 1/4 of all extracted spectra. However, it
was quite straightforward to beat down this number significantly. First of
all, the equivalent widths of these features were generally quite small.
Secondly, most of the spuriously selected stars had rather red colours.
We quantified a colour measure by defining a ``hardness ratio''
We finally decided to treat emission line candidates with large equivalent widths differently from those with weak line detections. A strong-lined object was defined as one with a line and Å. If the line was detected in the B band, the object needed to be moderately blue, ; this criterion removed most of the red stars. If the line was detected in the V band, no further restriction was applied. A line detection with small equivalent width, Å, needed always to be accompanied by a blue but meaningful spectrum, given by the conditions and .
Essentially all Ly detected quasars, but also some of the C IV and [O III] detections have large equivalent widths and therefore fell into the first set of criteria. Quasars at lower redshifts generally have weaker lines (C IV, C III], Mg II and [O III]), but are distinctly blue. Note that our highly conservative colour cuts were only applied after the emission line object selection; these cuts would by no means have sufficed to select quasar candidates on the basis of their colours.
After these cuts we visually inspected the selected objects to cull the quasar candidates from the remaining sample of slitless spectra, dominated by three types of spurious detections: (i) M stars with strong TiO absorption bands that our selection criteria did not catch; these features are always at the same wavelengths and could be recognised with good confidence after some training; (ii) low-redshift emission line galaxies with [O III] visible at ; at the low resolution of the WFI spectra, the [O III] doublet is blended, resulting in an asymmetric single line that could be unambiguously identified in almost all cases (see Fig. 8, for examples). Since emission line galaxies are interesting in their own right, we kept them in a separate bin; (iii) defective spectra were overlapping 1st order spectra of two close objects, or spectra with unrecognised zero order contamination or spectra with unphysical breaks at 5000 Å resulting from an imperfect splicing of the two bandpasses. Any remaining spectral overlaps were easily identified by comparing the spectra obtained at the two instrument rotation angles. All emission line candidates that passed the visual check of the one-dimensional spectra were finally verified on the slitless WFI images to check for any remaining flaws that could result in a spurious emission line. For objects at the field edges detected in only one instrument rotation (Fig. 1) this visual examination was essential. In total, 387 emission line objects were selected from the 29 000 objects detected on the DSS images of the 18 survey fields.
|Figure 8: Slitless spectra (black) and corresponding noise arrays (green/grey) of four emission line galaxies showing blended [O III] emission. H can be recognised as well.|
|Open with DEXTER|
|Figure 9: Slitless WFI spectra of two newly discovered quasars near HE 0940-1050. The upper left panel shows a close-up of the combined slitless B-filtered exposures in the field of HE 0940-1050 taken without rotation (Fig. 2). The dispersion runs horizontally from left to right. The lower left panel shows a direct WFI B band image of the corresponding sky region north-east of HE 0940-1050 ( ). The indicated objects display prominent emission lines in the slitless data. The right panels show the automatically reduced and calibrated one-dimensional B band spectra of the two quasars (black) and their noise arrays (green/grey). Both of them display Ly emission blended with N V at the indicated redshifts.|
|Open with DEXTER|
Visual inspection of the sky-subtracted slitless images revealed three additional candidates that were missed in the DSS catalogues, mainly due to source blending. These fields were reprocessed with the source catalogues of available direct WFI data. At the higher resolution of the direct WFI data, these candidates were resolved and their spectra were successfully extracted. Figure 9 presents the slitless B band data and the corresponding direct WFI image of a region near HE 0940-1050 where two prime quasar candidates were discovered. On the DSS image of this field, QNQ J09434-1053 could not be resolved from the nearby foreground object. The WFI catalogues did not reveal further undetected candidates, confirming our expectation that the fraction of candidates lost due to blended sources is very small.
In total, our semi-automated search for emission line objects yielded 390 sources on the surveyed 4.39 deg2. Of those, 38 we rejected as dubious due to very low S/N. The remaining sample of 351 objects consists of 37 known quasars (including 15 of the 18 targeted central quasars), 169 quasar candidates, and 145 candidate low-redshift star-forming galaxies. The classification was purely based on the slitless spectra. No morphological selection was made.
Most of the emission line galaxies were identified by their blended [O III] doublet. However, due to low S/N at this low resolution , we could not determine whether the H lines seen in the spectra are broad or narrow, so that this sample of low-redshift objects might still contain low-redshift AGN. At the bright end of our survey (), the surface density of emission line galaxies is 1.1 deg-2, which compares well with other slitless surveys (Bongiovanni et al. 2005; Salzer et al. 2002). As most of the deeper surveys for emission line galaxies select H emission (e.g. Pascual et al. 2001), a comparison at fainter limits is difficult.
In total, we rediscovered 37 previously known quasars in the survey fields with detectable emission lines in the observed wavelength range. Three of the central quasars (Q 0000-263, BR 1117-1329 and BR 1202-0725) reside at too high redshift to be detected in emission. Excluding the central bright quasars, 10 out of 12 known quasars are rediscovered in the range of high selection probability (, 1.7<z<3.6, see Sect. 7.2 below). One of the two missed quasars had been recorded in one WFI rotation only, whereas the other one was not contained in the shallow DSS source catalogue of the Q 0302-003 field (Fig. 5). The redshifts of the rediscovered quasars measured in the slitless data are consistent (within ) with the published values (see Sect. 6.2 and Fig. 11 below).
We estimated redshifts of the 169 quasar candidates from the slitless data, and the candidates were grouped by redshift confidence. The redshifts were considered to be almost secure if either two emission lines with a consistent redshift were present in the spectral range, or if a single line could be confidently identified due to its shape and/or that of the adjacent continuum. In particular, Ly emission blended with N V could often be unambiguously identified (for examples see Fig. 9). Bright low-redshift candidates displayed lines with small equivalent widths (likely C III] or Mg II) superimposed on a blue continuum. Redshift assignments of fainter quasar candidates were less certain due to the similar equivalent widths of the C IV, C III] and Mg II emission lines. We decided in all doubtful cases to assign the highest plausible redshifts to the candidates.
For eight of our fields we retrieved WFI+grism R band data from the ESO Science Archive (taken in the course of another programme, PI Vanzi). We reduced the data with our pipeline, yielding calibrated R band spectra of our candidates in the range 5800 Å 7400 Å. Due to the much higher sky background, however, the R band spectra are considerably noisier than the B and V band spectra. Therefore, only emission lines of brighter objects of our sample were unambiguously detected. Nevertheless, the R band spectra confirmed the previously adopted low redshift assignments of 8 candidates. No systematic search for further emission line objects was performed among the R band spectra.
Follow-up spectroscopy of 81 quasar candidates was obtained with the Focal Reducer/Low Dispersion Spectrograph 2 (FORS2, Appenzeller et al. 1998) on ESO VLT UT1/Antu in Visitor Mode on November 17-19, 2004 and in Service Mode between April and July 2005. We restricted the follow-up campaign to candidates with a high-confidence slitless WFI redshift . Quasar candidates likely residing at lower redshifts (C III] or Mg II seen in the slitless spectra) were not followed up. The field of PKS 0528-250 was not selected for follow-up and also the field of Q 1451+123 was not included due to the lack of high-quality candidates. Longslit spectra with a 1 slit kept at the parallactic angle were taken either with the 300V grism or the 600B grism, resulting in a spectral resolution of 10 Å FWHM and 4.5 Å FWHM, respectively. Each resolution element is oversampled by a factor of 3 (300V: 3.36 Å/pixel, 600B: 1.5 Å/pixel). No order separation filter was employed to maximise the UV coverage in the spectra, leading to possible order overlap at Å in the spectra taken with the 300V grism. Exposure times were adjusted to yield in the quasar continuum and ranged between one and 40 min. The sky was clear to photometric during the Visitor Mode run, but the seeing varied strongly during the nights, resulting in slit losses for some of the 53 targeted objects. The remaining 28 prime quasar candidates were observed in Service Mode under variable conditions. The spectra were calibrated in wavelength against the FORS2 He/Ne/Ar/HgCd arc lamps and flux-calibrated via spectrophotometric standard stars taken each night. However, absolute spectrophotometry was not achieved due to the narrow slit and the sometimes variable or poor sky transparency in the Service Mode observations. Table 2, available in the online edition of the Journal, lists the quasars confirmed in the spectroscopic follow-up observations.
The VLT spectra were reduced with standard IRAF tasks. The bias value was taken from the overscan regions. Dome flatfields were used to correct pixel sensitivity variations. After sky subtraction by fitting both spatial regions close to the target by 2nd-order Chebychev polynomials, the spectra were extracted with the optimal extraction algorithm by Horne (1986). The root-mean-square residuals of the wavelength calibration with low-order Chebychev polynomials were 0.05 Å and 0.2 Å for the 600B and the 300V grism, respectively. The spectra were corrected for atmospheric extinction assuming average conditions.
In order to approximately correct the spectra for slit losses we carried out aperture photometry on the FORS2 aquisition images taken in the B filter, which were calibrated against photometric standard stars (Landolt 1992) taken each night. Table 2 provides the airmass-corrected Johnson B magnitudes and their calculated standard deviations. Apparent magnitudes were also derived by integrating the flux-calibrated slit spectra. Slit losses were estimated by calculating the expected loss over the used 1 slit for a Gaussian PSF with a FWHM equal to the seeing during the observations that we measured on the aquisition images. Figure 10 presents the correlation of the integrated and photometric magnitudes. The integrated magnitudes are consistent with the photometric ones after correcting for slit losses. The two outliers are the two quasars QNQ J09425-1048 and QNQ J21294-1521 that were observed in strongly variable clearly non-photometric conditions.
|Figure 10: Blue magnitudes of the quasar candidates observed with the VLT. The abscissa shows the photometric B magnitudes obtained from the aquisition images, the ordinate the difference to the magnitudes integrated from the slit spectra. Full (open) circles correspond to measurements with (without) correction for slit losses. The line marks a perfect 1:1 relation between and .|
|Open with DEXTER|
Redshifts of all candidates observed with FORS2 were determined by taking into account all detectable emission lines. Peak positions of the lines were measured by eye and errors were estimated by considering the S/N of the lines, blending with other emission lines and the presence of absorption lines, sky residuals or line asymmetries. We confirmed that in many objects high-ionisation lines (Ly, N V, Si IV+O IV], C IV) are blueshifted with respect to low-ionisation lines (Richards et al. 2002b; Gaskell 1982; Tytler & Fan 1992; McIntosh et al. 1999). However, the low-ionisation lines O I+Si II and C II were often weak and noisy, resulting in larger individual redshift errors. Mg II could be measured only for quasars at () taken with the 300V (600B) grism. Therefore, we estimated the systemic redshift of each quasar by weighting the measurements of individual lines, giving a lower weight to high-ionisation lines or discarding them completely in case of large blueshifts. The redshift uncertainty of each quasar was estimated from the redshift scatter between the remaining lines and their centroiding errors. The adopted quasar redshifts and their estimated errors are listed in Table 2.
|Figure 11: Comparison of the redshifts obtained with FORS2 and WFI. Circles (+ signs) denote quasar candidates with almost secure (uncertain) WFI redshifts. Previously known quasars in the fields have been added with their published redshift in comparison to the redshift measured in the slitless spectra ( signs). The line marks a 1:1 relation.|
|Open with DEXTER|
|Figure 12: VLT/FORS2 spectra of the 11 discovered quasars in the vicinity of HE 0940-1050. The spectra are shown in black together with their noise arrays (green/grey). The spectra have been scaled to yield the measured photometric B magnitudes. Major emission lines are labelled. The small inserts show the corresponding slitless WFI spectra in the same units.|
|Open with DEXTER|
With the follow-up spectroscopy we confirmed that 80 of our 81 candidates are broad-line AGN in the redshift range . Only one of the candidates turned out to be a low-redshift galaxy with narrow emission lines (object J03490-3820, (J2000) = 03 49 00 56, (J2000) = -38 2031 9, z=0.2848). 12 of the confirmed quasars, located in the fields around Q 0302-003 and HE 2347-4342, were already presented and discussed in Papers I and II. In Fig. 12 we show the spectra for the quasars in one additional, particularly rich field. 11 quasars were found in the vicinity of HE 0940-1050. Appendix A, available in the online edition of the Journal, lists the spectra of the remaining 57 newly discovered quasars.
In Fig. 11 we compare the redshifts derived from the FORS2 spectra and the slitless WFI spectra, respectively. The slit spectra confirmed the redshifts determined from the slitless WFI spectra for the majority of candidates (54/81). The WFI redshifts of the other 27 objects in the follow-up had been overestimated (note that we had systematically assigned the highest plausible redshifts). 64 of the 80 confirmed quasars are above our low-redshift cutoff z= 1.7 set for inclusion in the follow-up.
The scatter between slit and slitless spectroscopic redshifts for the 54 quasars correctly placed in z is . The main cause for this scatter is optical distortions of the grism causing the astrometric transformation to produce residual variations of the wavelength zeropoint across the WFI field of view. Of the 27 candidates whose redshifts were overpredicted, only 5 had been assigned a high confidence redshift. This suggests that the majority of remaining candidates with high redshift confidence likely reside at those redshifts. After completion of the follow-up observations we compared the WFI spectra of quasars whose redshifts had been overestimated to the WFI spectra of the remaining candidates and adjusted the redshift estimates where applicable.
|Figure 13: Slitless spectra (black) and corresponding noise arrays (green/grey) of 10 quasar candidates in the field of PKS 0528-250 with indicated coordinates, approximate V magnitudes and estimated redshifts with confidence level (! = secure, + = plausible, : = uncertain). Major emission lines are labelled.|
|Open with DEXTER|
With 169 candidates in our sample and 81 of them followed up by VLT slit spectroscopy, 88 further possible quasars remain unconfirmed. For the benefit of potential future users we list these in Appendix B, available in the online edition of the Journal. We provide the slitless spectra of these 88 quasar candidates and their celestial coordinates, magnitudes, WFI redshifts, and our assessment of redshift confidence. For 23 of these we judge the WFI redshifts to be most likely correct. For 28 candidates we call our redshift estimate plausible; for 37 they are uncertain and could be identified with any of the major emission lines, although we hold it unlikely that many Ly quasars will be hidden among them. As an example we show in Fig. 13 the slitless spectra of 10 quasar candidates in the field of PKS 0528-250 that was completely left out of the spectroscopic follow-up. Three of these candidates have a secure redshift z>2.5. The remaining candidates on the other fields are either faint with noisy WFI spectra or they likely reside at redshifts z<1.7.
We summarise the overall redshift distribution of the quasar sample in Fig. 14. The open histogram shows the redshift distribution of all detected known quasars, newly discovered quasars and remaining candidates assuming that the inferred WFI redshifts of the candidates are correct. In total, 205 quasars and candidates are shown. The gap at expectedly arises from the lack of visible quasar emission lines in the observed spectral range (cf. Fig. 7). The very small number of low-redshift () AGN is also to this effect. For , no clearly preferred redshifts are discernible from this histogram. In Sect. 7.2 we present an attempt to quantify our redshift- and magnitude selection function. First we discuss our results in terms of the main rationale driving this survey, the construction of new apparent groups of high-redshift quasars.
The main goal of this survey was to reveal new groups of quasars for further studies of the three-dimensional distribution of the intergalactic medium. In the follow-up observations we concentrated on likely high-redshift (z>1.7) quasars, with some success: 80% of the newly established quasars match that condition. In addition to the central quasars, we recovered 85% of all detectable previously known quasars in these fields (22/26).
Including the remaining candidates, the number of quasars with per field varies substantially, between 2 (near CTQ 0247, Q 1209+093 and Q 1451+123) and 11 (near Q 0347-383). Some of our fields are thus almost devoid of high-redshift quasars except the central object, but other fields we found to be quite richly populated. These fields may in the future become interesting targets for multiple line of sight absorption spectroscopy. A list of currently known quasars in our survey fields is provided in Table C. 1, available in the online edition of the Journal. We briefly discuss some of the most prominent cases:
|Figure 14: Redshift histograms of confirmed previously unknown quasars (cross-hashed), rediscovered quasars (hashed) and remaining quasar candidates (open).|
|Open with DEXTER|
|Figure 15: Distributions of quasars in three of the survey fields (upper panels: HE 0940-1050, middle panels: Q 2139-4434, lower panels: HE 2243-6031). Left panels: angular separation distribution of quasars and quasar candidates with respect to the central quasar as a function of redshift. Asterisks and star symbols mark known quasars and newly discovered confirmed quasars, respectively. Crosses show remaining WFI quasar candidates. Symbol size indicates apparent optical magnitude. Right panels: quasar distribution on the sky centred on the targeted known quasar in the field with indicated redshifts. Objects at (z<1.7) are shown in black (red/grey). The rectangle marks the nominal contiguous slitless WFI field of view ( ) without rotating the instrument, for simplicity centred on the central quasar. New quasars located slightly outside the nominal field of view are due to field rotation and small offsets of the actual field centre.|
|Open with DEXTER|
We have shown above that quasars displaying strong emission lines can be selected from the slitless spectra with very high efficiency; only 1 out of 81 candidates followed up with the VLT turned out not to be a quasar. Although in all parts of the survey we optimised the procedure towards efficiency, also at the expense of completeness, nevertheless it was interesting to see how our emission-line based candidate selection performed with respect to the quasar population displaying broad emission lines. We address this issue in two ways: first we present an estimate of the survey selection function from Monte Carlo simulations. Then we use this selection function to determine quasar surface densities from our sample and compare these numbers with the results of other surveys.
We define the survey selection function as the recovered fraction of the broad-line quasar population matching a given broad band flux limit. The main variables governing the detectability of quasars in a survey such as ours are (i) the continuum magnitudes; and (ii) the equivalent widths of the detectable emission lines (Gratton & Osmer 1987; Schmidt et al. 1986). Quasar surveys based on emission lines tend to preferentially select strong-lined quasars, even more so near their survey limits. In order to quantify this selection effect, we first generated several template quasar spectra via a Monte Carlo routine. The quasar emission lines of Ly, Si IV+O IV], C IV, C III] and Mg II were modelled as Gaussian profiles with mean equivalent widths and line dispersions from the quasar template spectrum by Vanden Berk et al. (2001) superposed on a power law continuum with a mean spectral index . We incorporated variations in the model parameters by assuming Gaussian distributions with a standard deviation of 20% around the mean. We then generated 200 random spectral templates in the quasar rest frame. The neglect of non-Gaussian line shapes and intrinsic or intervening absorption does not have a big effect on our low-resolution spectra.
Mock slitless WFI quasar spectra were created by shifting the templates to the desired redshift, degrading them to the WFI resolution followed by adding Gaussian noise that varied with wavelength according to the overall throughput of the spectrograph (Fig. 4). The S/N in the simulated spectra, related to the V magnitude by Eq. (1), was normalised in the quasar continuum near the maximum throughput at 5400 Å. We then performed Monte Carlo simulations of the detection rates of quasars as a function of redshift ( ) and continuum S/N values ( ). At each considered redshift and S/N, WFI spectra of the 200 quasar templates were generated and subjected to our automated selection routines (Sect. 4.1). Each quasar template was simulated 100 times in order to quantify the impact of noise on the detection rate per object. We also performed the subsequent visual screening done for the real candidates on several hundred simulated spectra to investigate visual selection effects, finding that 90% of the automatically selected candidates survived the selection by eye.
|Figure 16: Survey selection function as a function of redshift and apparent V magnitude. Contours correspond to median simulated detection probabilities . Full and open circles mark newly discovered confirmed quasars and known quasars, respectively. Crosses denote remaining candidates.|
|Open with DEXTER|
Figure 16 presents the resulting selection function, displayed as a contour map of the median detection probability as a function of redshift and V magnitude. Overplotted are the quasars and quasar candidates from our survey. The selection function obviously varies with the visibility of different emission lines in the spectrum. Due to its high equivalent width Ly emission is detectable for essentially 100% of quasars brighter than , and still 80% of the simulated spectra are recovered at (continuum S/N of 1). At redshifts below 2.5 the selection function obtains lower values because of the lower equivalent widths of C IV, C III], and Mg II, and we have already commented upon the gap at where no emission line is visible in the range of our spectra. The selected quasars and quasar candidates fill the full range of the selection function spanned by the simulated redshifts and magnitudes (S/N). The known quasars at the field centres are clearly separated from the other quasars and candidates in magnitude and redshift. Note, however, that the survey selection function in Fig. 16 assumes that all quasar spectra were extracted from the slitless data. Due to photometric incompleteness (Sect. 3.2), the detection probability at is lower, as we will describe in the following.
With the survey selection function from above we were able to compare our number counts of quasars with those of other surveys. We consider two surveys reaching similar or greater depths. The only comparably deep slitless survey is the Palomar Scan Grism Survey (PSGS, Schneider et al. 1999) which yielded 39 quasars on 1.10 deg2 at . Disregarding the known bright quasars at the field centres we detected 102 quasars (known or newly confirmed) and 88 highly promising candidates on 4.39 deg2. Assuming that the candidates are quasars, the total surface density for our survey is 20% higher. The wavelength ranges and therefore the redshift ranges are not exactly equal between the two surveys, so that we simply state that the results are in good agreement.
A more quantitative comparison is possible by comparing our survey to the photometric COMBO-17 survey (Wolf et al. 2003). Since COMBO-17 is flux-limited in the R band whereas our survey is best defined in V, we assumed a mean quasar colour of V-R=0.2 to convert the COMBO-17 surface densities. After correction for the selection function, the cumulative surface number density of quasars found for our survey is 48 deg-2 at V<21, and 89 deg-2 at V<22. The corresponding values for COMBO-17 are 51.3 deg-2 and 93.5 deg-2, respectively. Thus again, the numbers are in excellent agreement, even at magnitudes as faint as . Looking closer at the differential number counts we find that they agree very well at 20.5 < V < 21.5( in our survey vs. in COMBO-17), but that our survey drops by a factor of 1.5 in the subsequent magnitude bin at 21.5 < V < 22.5( vs. ). It is not surprising to find that our survey is incomplete at these magnitudes, probably as a combination of incompleteness already in the input catalogues and missing true quasars among the noisy spectra.
As we had included the ECDFS from COMBO-17 in our survey observations, we could also perform an object by object comparison in this well-studied field. The slitless spectra confirmed the general trend of the above survey selection function. We rediscovered 7 out of 10 R<21 COMBO-17 quasars at 1.3<z<3.6located in the common survey area. The three missed quasars are quite faint (R>20) and reside at , where the selection function of our survey drops (Fig. 16). At z>2.5 only one out of three 21<R<22 quasars was rediscovered, indicating photometric incompleteness at the faint end of the DSS catalogue.
Slitless spectroscopic surveys are successful in finding closely projected quasar groups at high redshift, especially in redshift ranges where colour surveys are less effective. However, previous targeted slitless surveys for quasar groups were few, always pointed at one individual field only (e.g. Williger et al. 1996; Jakobsen & Perryman 1992), and also generally shallower than our survey by 1 mag. Thus, the resulting surface maps of quasars were sparse, but probed the large-scale distribution of quasars. Our survey significantly increases the number of high-redshift quasar groups with separations in the southern hemisphere. The centroiding of our fields on known bright quasars enables follow-up studies on the three-dimensional distribution of the IGM. Our faint survey limit ( -22) results in a high surface density of high-redshift quasars despite the comparatively small survey area per field.
More recently, Hennawi et al. (2006b) selected faint quasar candidates in the vicinity of confirmed SDSS quasars using the SDSS photometry. Their survey was dedicated to find close quasar pairs with separations , especially binary quasars at the same redshift. They confirmed 40 new binary quasars and 73 projected quasar pairs. However, 89 of their candidates turned out to be stars, most likely due to broadband colour similarities between stars and quasars in certain redshift ranges. In comparison, our high success rate of almost 100% shows that slitless spectroscopy is highly efficient in finding high-redshift quasars. A smooth selection function (Fig. 16) is essential to discover high-redshift quasar groups in selected fields. However, given the small surface density of high-redshift quasars, we were unlikely to find quasars at very small separations to the central quasars in our fields.
We performed the ``Quasars near Quasars'' (QNQ) survey, a CCD-based slitless spectroscopic survey for faint quasars at around 18 well-studied bright quasars at 2.76<z<4.69, covering a total area of 4.39 deg2. In order to analyse the slitless data we developed a data reduction pipeline that performs an optimal extraction of the spectra and that is able to cope with contaminating spectral orders of slitless grism data. From the 29 000 extracted flux-calibrated spectra we selected 169 previously unknown quasar candidates and >100 likely low-redshift emission line galaxy candidates on the basis of emission features that fall in the covered wavelength range 4200 Å 5800 Å. A semi-automatic selection routine limited potential biases of purely visual selection and allowed to quantify selection effects.
Follow-up spectroscopy confirmed 80 out of 81 selected quasar candidates on 16 fields. 64 of these newly established quasars reside at z>1.7. The highest redshift quasar is QNQ J22484-6002 at z=3.586. The brightest newly discovered high-redshift quasar is QNQ J11197-1340 (z=2.220, B=18.3). Given the high success rate of the follow-up, the vast majority of the remaining 88 candidates will be quasars as well, although most of them likely reside at lower redshifts.
The primary aim of this survey was to provide new groups of quasars for medium-resolution spectroscopy in the vicinity of well-studied known quasars. Originally, we had not expected to reach as faint as in the survey observations. At these faint magnitudes our survey is not well defined because of photometric incompleteness in the source catalogues needed for automatic extraction of the slitless spectra. Together with the fact that not all selected quasar candidates could be included in the spectroscopic follow-up, our survey probably is of limited use for constraining quasar evolution. However, focusing on survey efficiency rather than on completeness at the faint end was justified in order to accomplish the main goal of this survey. In fact, the faintest quasars discovered will for now remain beyond the limits for obtaining high-quality spectra. But at least the quasars are well suited for follow-up studies with current (e.g. ESI at Keck) or upcoming (e.g. X-shooter at VLT) high-throughput spectrographs at 8-10 m-class telescopes.
Together with the central quasars in the fields already observed at high resolution, these quasar groups can be used to perform a tomography of the intergalactic medium. Large-scale clustering of the Ly forest or correlations of metal line systems can be investigated as well (Williger et al. 2000; D'Odorico et al. 2002,2006). Some of the discovered quasars reside at similar redshifts or approximately at the same redshift of the central quasar in the field, giving potentially insights to quasar clustering along overdense filaments in the plane of the sky. We have identified two quasars coinciding with a damped Ly absorber on one central line of sight, as well as a large-scale group of quasars at z=2.70-2.78. Moreover, our study provides new foreground quasars to investigate the transverse proximity effect of quasars (Papers I and II). In combination with the already available high-resolution spectra of the central quasars, medium-resolution spectra of these quasar groups will offer great opportunities to study the large-scale cosmic web in three dimensions.
We thank the staff of the ESO observatories La Silla and Paranal for their professional assistance in obtaining the data presented in this paper. This research has made use of the NASA/IPAC Extragalactic Database (NED) which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration. G.W. was partly supported by a HWP grant from the state of Brandenburg, Germany. G.W. and L.W. acknowledge support by the Deutsche Forschungsgemeinschaft under Wi 1369/21-1.
Table 2: Observing log of the spectroscopic follow-up. The quasars reported in Papers I and II have been included for completeness. The columns list the quasar field, the name of the confirmed quasar, its right ascension and declination, its redshift with estimated error, its B magnitude with error, night of observation, the grism used, the exposure time, the seeing and the sky conditions during observation.
|Figure A.1: VLT/FORS2 spectra (black) and noise arrays (green) of 57 quasars discovered in the QNQ survey. Spectra presented in Worseck & Wisotzki (2006), Worseck et al. (2007) and in Fig. 12 of this paper are not shown. The spectra have been scaled to yield the measured photometric B magnitudes. The small inserts show the corresponding slitless WFI spectra in the same units. The spectra are ordered columnwise by right ascension of the field and redshift of the quasar.|
|Open with DEXTER|
Table B.1: Remaining quasar candidates not included in the spectroscopic follow-up. The columns list the central quasar in the field, the object (format: HHMMSS.S DDMMSS, J2000), the approximate V magnitude and the redshift with redshift confidence.
|Figure B.1: RA-ordered list of slitless WFI spectra of 88 unknown quasar candidates without follow-up spectra. The spectra are shown in black together with their noise arrays (green/grey) if the S/N is low. The (approximate) flux densities are given in . Labelled are the celestial coordinates (format: HHMMSS.SDDMMSS, J2000), the integrated V magnitude and the estimated redshift. Ambiguous redshifts guessed from a single emission line are marked by a colon. Plus signs mark plausible redshift estimates. Exclamation marks stand for almost secure redshifts based on the shape of a single line or the detection of two lines either in the principal spectral range 4200 Å 5800 Å or the R band spectra if available. J005840.0-263314 is a prominent BAL QSO.|
|Open with DEXTER|
|Figure B.1: continued. Note that the field of PKS 0528-250 with 7 bona fide quasar candidates was not included in the follow-up campaign with FORS2.|
|Open with DEXTER|
|Figure B.1: continued.|
|Open with DEXTER|
|Figure B.1: continued. The spectrum of J214326.2-442036 might be contaminated by a nearby galaxy. Francis et al. (2004) list J214124.0-443301 as a quasar candidate with uncertain redshift, whereas they identified J214119.4-441256 as a galaxy at z=0.062 which is not supported by the slitless WFI spectrum.|
|Open with DEXTER|
|Figure B.1: continued.|
|Open with DEXTER|
Table C.1: List of confirmed quasars and AGN within a radius < around the central quasars in the survey fields. The columns list the name of the field, the name of the confirmed quasar, its right ascension and declination, its redshift, its optical magnitude with filter, the angular distance to the central quasar and the discovery paper.