AAT Imaging and Microslit Spectroscopy in the Southern Hubble Deep Field

0.55 MB·English

by  Karl Glazebrook

AAT Imaging and Microslit Spectroscopy in the Southern Hubble
Deep Field
Karl Glazebrook1,2, Aprajita Verma3,4, Brian Boyle5,2, Sebastian Oliver3,6, Robert G.
Mann3,7, Davienne Monbleau2
6
0
0
2
ABSTRACT

n
We present a deep photometric (B- and R-band) catalog and an associated spectroscopic
a
J redshift survey conducted in the vicinity of the Hubble Deep Field South. The spectroscopy

5 yields 53 extragalactic redshifts in the range 0 < z < 1.4 substantially increasing the body of
spectroscopic work in this field to over 200 objects. The targets are selected from deep AAT

1 prime focus images complete to R < 24 and spectroscopy is 50% complete at R = 23. There is
v now strong evidence for a rich cluster at z ≃ 0.58 flanking the WFPC2 field which is consistent
2
with a known absorber of the bright QSO in this field. We find that photometric redshifts of
1
1 z < 1 galaxies in this field based on HST data are accurate to σz/(1+z) = 0.03 (albeit with
1 small number statistics). The observations were carried out as a community service for Hubble
0 Deep Field science, to demonstrate the first use of the ‘nod & shuffle’ technique with a classical
6
multi-objectspectrographandto testthe useof‘microslits’forultra-highmultiplex observations
0
along with a new VPH grism and deep-depletion CCD. The reduction of this new type of data
/
h is also described.
p
- Subject headings: Catalogs, surveys,Galaxies: evolution
o
r
t 1. Introduction containabrightQSOJ2233-606atz =2.24inor-
s
a der to facilitate studies of the connectionbetween
The Hubble Deep Field South (HDF-S) is one
: foreground galaxies and absorbing systems in the
v
ofthedeepestimagingfieldsinthesky. Itwasob-
QSO spectrum.
i
X servedin 1998(Williams et al. 2000)by the Hub-
Thedeterminationofpreciseredshiftsforextra-
r bleSpaceTelescope(HST)asacounterparttothe galacticsourceshasbeenimportantsincethetime
a northern Hubble Deep Field (HDF-N). However
of Hubble (1929). While much spectroscopy has
in contrastto the northern field it was selected to
been done in the other Hubble Deep Field, HDF-
1DepartmentofPhysicsandAstronomy,JohnsHopkins S has lagged behind. As a community service
University,Baltimore,MD21218 and in order to test new instrumentation tech-
2Anglo-Australian Observatory, P.O.Box 296, Epping niques,inparticular‘nod&shuffle’(Glazebrook&
NSW,Australia Bland-Hawthorn2001;GB01)withverysmall‘mi-
3Astrophysics Group, Imperial College London, Black- croslits’, we carried out a spectroscopic campaign
ett Laboratory, Prince Consort Road, London, SW7 2AZ,
of galaxies in, and in the vicinity of, the HDF-S
UK
in order to obtain redshifts. The plan of this pa-
4Max-Planck-Institut fu¨r extraterrestrische Physik,
per is asfollows: Section2details the prime focus
Giessenbachstraße, 85748Garching,Germany
5AustraliaTelescopeNationalFacility,POBox76,Ep- pre-imaging,the proceduresused to constructthe
pingNSW1710,Australia photometric catalogand the selectionofthe spec-
6Astronomy Centre, University of Sussex, Falmer, troscopic targets. In Section 3 we describe our
BrightonBN19QJ,UK novel spectroscopic observations and the special
7Institute for Astronomy, University of Edinburgh,
data reduction procedures used. In Section 4 we
Royal Observatory, Blackford Hill, Edinburgh EH9 3HJ,
presentourspectroscopiccataloganditsbasicpa-
UK
1
rametersandcomparewithotherworkinthisfield 213479 at 22h 33m10.97s −60◦26′00.4′′, B=8.1).
anddiscussthepossiblegalaxyclusteratz ≃0.58. In order to maximise the reliability of the final
The finding charts, images, spectra, photomet- catalogue without compromising the faint source
ric and spectroscopic catalogs presented in this detection, we flagged sources which lie in areas
paper are all available from the Anglo-Australian moststronglyaffectedbystraylightorinvignetted
Observatory web site at: regions as ‘masked’ to minimise the number of
spurious extractions. Only sources lying within
http://www.aao.gov.au/hdfs/Redshifts
unmasked regions are considered as targets and
for further analysis. The magnitudes presented in
2. Imaging Observations & Spectroscopic
the final catalogue are SExtractor MAG AUTO mag-
Selection
nitudesdeterminedusinganautomatedadapative
We obtained pre-imaging data1 (prior to the aperturetechnique. Thefinalphotometriccatalog
HST campaign) from which our sample of tar- is given in Table 1 (electronic edition only).
getsforspectroscopicfollow-upwereselected. Im- We extracted an extragalactic catalogue us-
ages in B and R were taken with the Anglo Aus- ing star-galaxy separation and analysed the re-
tralianTelescopePrimeFocusCCDcamera(0.391 sulting extragalactic number counts. We chose
′′
/pixel) in May 1999 at two pointings contain- to use the object’s magnitude and a measure of
ingtheWFPC(‘AAT-WF’)andSTIS(‘AAT-ST’) surface brightness (magnitude/FWHM2) as our
fields respectively. When stacked and mosaicked, star-galaxy separator as it provides a cleaner cut
they covera contiguousareaof 12.5′×7′ centered at faint magnitudes than SExtractor’s classifier
on 22h 33m18.68s −60◦31′45.8′′ (J2000)) which or the commonly used core:total indicator. Our
includes the HST deep fields and their flanking number-magnitude analysis showed the counts of
fields. Allimagesweretakeninphotometriccondi- galaxies to be still increasing as a power-law to
′′
tionsand0.8 seeing,exceptfortheB-bandAAT- R=24, consistent with close to 100% complete-
′′
ST data (2.2 through thick clouds). A finding ness at this depth. The formal magnitude limit
chart for our pointing is given in Figure 1. of the images, measured from the noise in ran-
′′
Theimageswerede-biased,flat-fieldedandmo- dom 0.8 diameter apertures placed on blank re-
saiced following standard CCD procedures. All gions is R=25.30, 25.16 (WF, ST) for a 3σ de-
images were registered to the USNO-A2.0 refer- tection. This is also consistent with the images
enceframeusingx-andy-shiftsandrotation. The being highly complete at R = 24. The B-data
′′
internalastrometricoffsetbetweensourcesinover- reaches 25.18 (WF) and 23.97 (ST, 2 aperture
lapping frames is measured to be ≈ 0.12′′. An — due to the poorer seeing). At R = 24 galaxies
all-sky photometric solution was determined from overwhelminglydominatestarsandsowechoseto
observations of 7 Landolt (1992) fields using 29 select spectroscopic targets from a purely R mag-
stars. B-band images taken in non-photometric nitude limited catalog to this depth. R = 24 was
conditions and were calibrated using overlap ar- somewhat deeper than we expected to reach with
eas with photometric images (accurate to < 0.04 spectroscopyevenunderthebestconditions,how-
magsresidualscatter). Asourcecataloguewasex- ever we expected to go deeper for objects with
tractedusing Sextractorversion2.2.1(Bertin and strong emission lines and we also desired a high
Arnouts, 1996). Sextractor was run in the stan- surfacedensitytotestthe‘microslit’spectroscopy
dardmannerusingalowdetectionthresholdof1σ described below so our final strategy was to ob-
andaminimumof3connectedpixelstodefinethe serve a lot of objects and tolerate a lower com-
extraction. We used a local determination from pleteness. Thefinallistofobjectsthatmadeiton
theskytoaccuratelyhandledatawithvariablesky to the slit-mask was based solely on geometrical
levels. Thisisparticularlyimportantsincetheim- considerations.
ages suffer from strong stray light pollution from Finally, we note that the spectroscopic selec-
an off-image star to the north of the field (HD tionwasdoneonapreliminaryversionofthefinal
astrometricandphotometrictablepresentedhere.
1Webrieflydescribetheimaginghereandfurtherdetailsare We find that the positions and magnitudes repro-
givenontheprojectwebsite.
ducewellbetweenthepreliminarycatalogandthe
2
finalversion(<0.4arcsecand<0.03magsystem- 3.1. Nod & Shuffle
atic differences).
Thenod&shuffletechniqueforhighlyaccurate
sky-subtraction is extensively described in GB01.
3. Spectroscopic Observations
In a nutshell the technique uses an unilluminated
Thespectroscopicobservationsweremadewith part of the CCD as a storage area for an image
the Anglo-Australian Telescopes Low Dispersion of the sky. Observing consists of taking an image
Survey Spectrograph (Wynne & Worswick 1988). ofthe objects throughthe slits,clockingthe CCD
This instrument (LDSS) was designed for multi- voltagepatternsothisimageis‘shuffled’intothe
slitimagingspectroscopyandoffersawidecircular storage, nodding a few arcsecs, taking an image
(12.3 arcmin diameter) field of view. Data was of the sky through the slits, and repeating. The
takeninOctober1998duringacommissioningrun result is quasi-simultaneous images of the object
to test the following new features: andskyadjacentontheCCD.Thisisdonerapidly
totrackskyvariations,e.g. forHDF-Swetook30
1. A new deep-depletion CCD detector from secimagesatobjectandskypositions(plusa2sec
MIT Lincoln Labs (see Burke et al. 2004 deadtime to allow for telescope settling). The se-
fordetailsonthesedevices)hadjustbeenin- quenceisrepeated30×givingan1800sec(900sec
stalledofferingimprovedredsensitivity. The on object) exposure. Readout noise only becomes
deep depletion of high-resitivity p-type Sil- important during the final clock-out, charge shuf-
icon causes pixels to be up to 40µm deep, fling is essentially noiseless as long as the CCD
thisallowsmoredepthforaphotontobede- has good charge-transfer efficiency (CTE). Sky-
tected and improves the quantum efficiency subtractionthenconsistsofwindowingtheskyre-
>8000˚A where Silicon is becoming increas- gionoftheimageandthensubtractingitfromthe
ingly transparent. It also reduces the effect object region which is offset by an exactly known
offringing. We obtaineda2048×4096pixel number of pixels. With the setup employed here
device (15 µm pixels giving 0.395 arcsec / the systematic sky residuals do not exceed 0.04%
pixel). The red quantum efficiency peaked (measured in GB01).
at 87% at 7000˚A and maintained 16% at Many schemes are possible for laying out po-
10,000˚A.(Tinney & Barton 1998). sitions of slits and storage areas on the detector.
Whetheronecanshufflealongdistanceorashort
2. A new red-optimized grism based on a Vol-
distance depends on the CTE and the number of
umePhaseHolographic(VPH)gratingfrom
charge traps on the detector (each trap causes a
Kaiser Optical Systems was tested. VPH
trail of charge when shuffled). For example Abra-
gratingsoffer improvedthroughput (Barden
ham et. al. (2004) in their later Gemini Multi-
etal. 1998). This wasthe firsttestofalow-
ObjectSpectrograph(GMOS)observationsused2
resolutionVPHgratingdesignedforredshift
arcseclongrectangularslitlets,and‘microshuffled’
measurements in a multi-object system and
the charge only one slit length. The storage was
proved successful. The peak efficiency was
immediately below the detector soeffectively 50%
measured to be 82% at 6700˚A (Glazebrook
oftheGMOSFOVwasavailabletoslits. Forthese
1998). The grating delivered a dispersion of
small shuffles CTE is not critical, also the GMOS
2.6˚AperpixelwiththeMIT/LLCCDanda
CCDshadhundredsofchargetrapssosmallshuf-
onearcsecslitprojectedto ∼3 pixelsgiving
fles was desirable.
8˚A spectral resolution.
For our observations the MITLL3 CCD only
3. The use of ‘nod & shuffle’ (GB01) for accu- had 12 traps and the CTE was 99.9999% which
rate sky-subtraction for the first time in a allowed us to make large ‘macroshuffles’ across
multi-object configuration. the whole device. We chose to follow the scheme
shown in Figure 2 which is motivated by the fact
4. The use of very small slits (‘microslits’), en- that the MITLL3 device is physically much larger
abledby nod& shuffle, to allowlargemulti- than the LDSS focal plane. One third of the de-
plex, as described below. vice (1365 pixels = 9.0 arcmin) is used for imag-
3
ing,the restforstorage. 75%ofthe LDSS FOVis axis) in order to maximize the number of tar-
available for slits, which is advantageous, but we gets allocated within the Hubble camera fields.
have to shuffle 1365 pixels. This is not a problem We found the typical overall number of slits al-
with the high CTE, tests in the lab showed sig- located by our algorithm was not sensitive to this
nificant image degradation did not happen until choice. The position angle is not optimal for at-
severalhundredmacroshuffleoperationshad been mosphericdispersion,howeversinceweweredoing
performed. Finally, we note that the FOV ad- red spectroscopy this would not have been a sig-
vantagein macroshufflingis only gainedwhen the nificant source of light loss. Using the red VPH
FOVgreatlyunderfillstheCCDandwouldnotbe grating the 2048 pixels in the dispersion (CCD
applicable, for example, to GMOS. row) direction gives 5300˚A wavelength coverage.
The typical wavelength range observed for a slit
3.2. Microslits & Mask design on the mask center was 5000–10000˚A. Slits were
allowedtobeasfaras1.5arcminoff-axisresulting
The unconventional approach taken with the
in wavelengthcoverageshifts of upto 590˚A.Thus
spectroscopicmaskswasthatinsteadofusingrect-
it was only possible to accommodate one object
angularslitlets we used small circularholes in the
at each CCD row (i.e. one tier). Slit allocation
mask at the location of target galaxies. The logic
allowedaminimumonepixelgapbetweenobjects
was that the usual reason to employ an extended
to cleanly delineate spectra, thus the maximum
rectangular slit several arcsecs long is to sample
possiblenumberofholeswas348. Runningtheal-
neighboringskyforsky-subtraction. Withthenod
gorithm on the R<24 HDF-S catalog for a mask
& shuffle technique this need is obviated. Thus
centerof22h 33m 11s −60◦33′ 16′′gaveanalloca-
one only needs apertures large enough to receive
tionof225objects(aftertakingoutsomespaceto
the light from typical targets in the expected see-
allow for larger holes for alignment stars) in the
ing. NormalLDSSslitletsareoforder5–10arcsec
3 arcmin × 9 arcmin field. This compares with
in length, thus with 1 arcsec ‘microslits’ we could
a normal LDSS mask setup without nod & shuf-
obtain a 5–10× larger multiplex.
fle where one normally only gets spectra of 20–30
Ideallywewouldhaveusedsquareholessothat
objects.
thespectralPSFwouldbe uniformacrossthetar-
The drawback of this strategy is that the ob-
get but due to hardware limitations of our par-
jectsareoutoftheslitsforhalfthetime whenthe
ticular mask cutter it turned out to be much eas-
telescope is in the sky position. An alternate nod
ier to simply drill circular holes. The PSF varia-
& shuffle approachwould be to use slightly larger
tion is not important for one dimensional spectra
slitlets so the object is still in the slit in both nod
summed across the slit and does not affect nod
positions (e.g. Abraham et al. 2004). However
& shuffle sky subtraction as objects and sky are
this means ingeneralone canonly observehalf as
observed through exactly the same slits (this is
many objects because the slits are twice the spa-
confirmed by the essentially Poisson-limited sky
tial size, for a high sky density survey there is no
subtractioninthefinalspectra). Noteweplanned
net difference in the number of objects observed
todotheobservationsinthebestAATseeing(.1
to a given depth in a given amount of telescope
arcsec)andsowecutonearcsecholes(150µmdi-
time.
ameter). The masks were available immediately
before the run but there was no opportunity to
3.3. Observations
measure them until after the run. Using a micro-
scope we later measured the diameters to be on The observations were carried out over the
average0.7arcsec,thiswouldhaveresultedunfor- nightsof13–16th October,1998attheAAT.Con-
tunately inlightlossgiventhe actualseeing. This ditionswerephotometricandtheseeingwas1–1.5
non-optimal size lessened our final spectroscopic arcsec. The red VPH grating was used with a
depth. GG475 blocking filter, so the spectra are poten-
Slit allocation was done using a custom algo- tially contaminated by second order for > 9000˚A,
rithmwhichsimplymaximizesthenumberofnon- though this effect was not in practice seen in any
overlappingtargetspectraforagivenpositionan- spectra. The filter was also not anti-reflection
gle. We chose 90◦ (i.e. along the STIS–WFPC2 coatedcausingsomeghostsinthespectralimages.
4
The totalexposure time obtainedon the HDF- trum. Reversing the procedure, i.e. flat fielding
S field was 12 hours (half on sky). The main cal- before sky subtraction will actually result in an
ibrations taken were arc spectra for wavelength erroneous flat field correction.
calibration and white-light dispersed flat fields
(throughthemask). Theflatfieldsservedthedual 4. Spectra and Redshift catalog
purpose of removing pixel sensitivity variations
Redshifts were determined from the final set of
and allow the spectra location on the detector to
225targetspectrabycarefulvisualinspectionand
be mapped empirically.
are given in Table 2 along with key spectral fea-
3.4. Data Reduction turesusedforidentificationandasubjectivequal-
ity system — Q = 4 denotes a dead certain red-
The format of the data shares similarities both
shift (> 99% confidence), Q = 3 is ‘quite certain’
withclassicalmultislitspectraandwithfiberspec- (∼ 80% confidence), Q = 2 is a ‘possible iden-
tra. Eachsmallcircularapertureproducesaspec- tication’ (∼ 50% confidence), Q = 1 is a single
tral trace (‘tramline’) on the CCD image. A por-
emission line redshift assumed to be [OII] (which
tionofthisdataisshowninFigure3. Wereduced
is the best a priori guess given the likely redshift
the data using a mixture of IRAF (Tody 1986)
range at R < 24 and wavelength coverage) and
tasks and custom Perl Data Language (Glaze-
Q=0 represents no identification.
brook & Economou 1997) scripts. First the indi-
Outof the 225objects there were(24,19,22,8,
vidualimagesweresky-subtracted,usingtheshuf-
152) objects with Q = (4,3,2,1,0) respectively.
fleoffsettomaptheskypixelstotheobjectpixels.
Thus 73identifications (33%)were made ofwhich
Theimages(+arcsandflats)werethencorrected
20weregalacticstars. Thecompletenesswas79%,
for small alignment shifts throughout the night
63%, 46% at R < 21, 22, 23 respectively. Many
andregisteredandstackedwithacosmicrayfilter.
of the faint R > 23 identifications were for ob-
The flat-field was used to map the spectra tram-
jectswithstrongemissionlines. Randomexample
line locations and also the PSF width as a func-
spectra of Q = 4 and Q = 3 objects are shown
tion of wavelength for optimal extraction of 1D
in Figure 4 to illustrate the quality. FITS format
object, sky (for reference), arc and flat-field spec-
spectra(andvariancespectra)areavailableonthe
tra. The flat-field spectra were then normalized
web site. In the final list we identify five galaxies
withapolynomialfitinthe spectraldirectionand
and two stars within the deep WFPC-2 field and
dividedintotheobjectspectratocorrectthemfor
one galaxy in the STIS field.
high frequency pixel/wavelengthsensitivity varia-
tions. An overallwavelengthsolution was derived
4.1. Comparison with other HDF-S red-
from the arc and sky spectra (1˚A RMS).
shifts
Care was taken to propagate variance arrays
(withinitialerrorscalculatedfromthe CCDread- Spectroscopy in the HDF-S vicinity has also
noiseandgainusingphotonstatistics)throughev- been done by Tresse et al. (1999; T99), Vanzella
ery step of the data reduction so errors could be et al (2002; V02) and Sawicki & Mall´en-Ornelas
assigned to the final spectra. (2004; SM03). T99 published redshifts obtained
with the NTT for nine objects within 1 arcmin of
One important point about the data reduction
the QSO with I < 22.2, three of these are on our
is that the shuffled sky must be subtracted before
mask of which we identify one (#20) as a star,
flat-fielding. This is a key advantage of nod &
however this is only a Q = 2 confidence. The
shuffle: accurate sky-subtraction does not require
other two we do not identify. V02 obtained 50
flat fielding. In the shuffled image the relevant
redshifts using the VLT in the range IAB = 20–
pixel responseis that ofthe originalpixel andnot
the storage pixel. So the correct procedure is to 25 (75% complete at IAB < 22.5) with an em-
phasis on color selected z > 2 galaxies. We find
subtract the sky first. The pixel response is the
three identifications in common, all agree at the
same and the sky subtracts correctly. Then ap-
∆z ≤ 0.001 level. SM03 obtained 97 ‘secure’ red-
plying the flat field will correct the flat field error
shifts for sources with IAB < 24 using the VLT.
in the object correctly. If the flat field is imper-
Five of their objects are also identified by us, all
fectly measured this only affects the object spec-
5
are high confidence redshifts (except for one sin- to allthe galaxiesfromthis work,SM03,T99 and
gle emission line redshift) and all agree to within V02 (after removing all duplicate objects). It is
∆z ≤ 0.001. There is one object (#204) which clear that there is a compact cluster of galaxies
appearsbothinV02,SM03andinourcatalog,all at this redshift flanking and overlapping with the
at the same redshift. If we exclude the objects in SW corner of the WFPC2 field. The cluster is
common with the other spectroscopic catalogs we of order one arcmin in size which is fairly typi-
find we have identified 45 new extragalactic ob- cal for large clusters at these redshifts. Taking an
jects in the HDF-S and flanking fields. Excluding approximatecluster center of (−0.5,−0.5)arcmin
all objects in common there are now 206 unique wrt the WFPC2 field we count 16 galaxies within
galaxies with spectroscopic redshifts from all the a one arcmin radius and with 0.56 < z < 0.60.
HDF-S catalogs considered here. The redshift distribution of galaxies in this circle
Anotherinterestingcomparisoniswithcatalogs is shown in the lower half of Figure 6 and it can
based on photometric redshifts. Here the WFPC- now be seen that there is again a single promi-
2 field allows accurate photometric redshifts be- nent spike at z ≃ 0.58. Since the overdensity is
cause of the depth and 3 colors. In order to com- compactin two dimensions onthe sky and also in
pare we define the fractional error on the redshift redshift this is likely to be a cluster rather than a
as x = z/(1+z), comparing with the catalog of filamentoflargescalestructure. Forthe16galax-
Gwyn (1999) we find that σx = 0.03 between the ies we find z = 0.574±0.008 corresponding to a
catalogs (5 galaxies, 2 stars). We also compare velocity dispersion σv = cσz/(1+z) = 1500 km
with the ‘Stony-Brook’ catalog2 which is based s−1. This is typical of a rich cluster of galaxies
on the HDF-N methods of Ferna´ndez-Soto et al. and we find it is insensitive to the exact choice of
(1999). Here we find a big discrepancy in object radius. These parameters are consistent with the
#871 which Stony-Brook identifies as a star and cluster containing the QSO absorber reported by
we identify as a z = 0.7 galaxy, albeit with very T99 on it’s outskirts.
low confidence (Q = 2). Excluding this object
we again find σx = 0.03 between the catalogs (4 5. Summary
galaxies, 2 stars).
To summarize the paper:
4.2. Redshift distribution and rich cluster
1. We present a significant new set of photo-
near WFPC2 pointing
metric and spectroscopic data on the HDF-
The resulting redshift distributions are com- S.
pared in Figure 5. V02 reported an over-density
near the WFPC2 pointing at z ≃ 0.58 after com- 2. We have demonstrated a new mode of spec-
bining our data (based on our initial WWW re- troscopic data taking with high-multiplex
port) with theirs. It can be seen that the redshift per unit area on the sky using the tech-
spike is at it’s most prominent in our dataset. If niqueof‘microslits’inconjunctionwithnod
we look at the combined redshift distribution in & shuffle.
Figure5 itcanbe seenthatthere areanumber of
3. The spectroscopic catalog presented here is
spikes, as is typical for narrow field redshift sur-
the third large catalog of redshifts in the
veys, and that the z = 0.58 spike is not the most
HDF-S and flanking fields. 53 redshifts are
prominent of them. However the STIS QSO dis-
obtained of which 45 are new.
playsastrongabsorptionline systematz =0.570
(T99) coincident in redshift with a bright spiral 4. Comparison with existing photometric red-
galaxy so it is interesting to explore this further. shifts based on WFPC-2 data show these to
Thiscanbedonebylookingatthedistributionon be in reasonable agreement (σz/(1 + z) =
the sky. Figure 6 shows a comparison of the loca- 0.03) for z < 1 galaxies, albeit with small
tions of the galaxies at 0.56<z <0.60 compared number statistics.
2Thisdataispubliclyavailablefrom 5. Wepresentstrongevidenceforarichcluster,
http://www.astro.sunysb.edu/astro/hdfs/wfpc2 compact in all three spatial dimensions, at
6
z ≃ 0.58 flanking the WFPC2 pointing and Glazebrook, K., & Bland-Hawthorn, J. 2001,
whichverylikelycontainstheknownabsorb- PASP, 113, 197 (GB01)
ing galaxy of the QSO at this redshift.
Gwyn, S. D. J. 1999, ASP Conf. Ser. 191: Pho-
Preliminary versions of this catalog have al- tometric Redshifts and the Detection of High
ready been used for scientific studies using the Redshift Galaxies, 191, 61
HDF-S data. Mann et al. (2002) used it for a
Hubble, E. 1929, Proceedings of the National
study ofthe relationshipbetweenfar-infraredand
Academy of Science, 15, 168
other estimates of galaxy star-formation rates.
Vanzella et al. (2002) used this catalog to- Landolt, A. U. 1992, AJ, 104, 372
gether with theirs to estimate cosmological star-
formations rate history. The reader is encouraged Mann R. G., Oliver S., Carballo R., Franceschini
tousethiscatalogforfurtherstudiesoftheHDF-S A., Rowan-Robinson M., Heavens A. F., Kon-
and its associated QSO. tizas M., Elbaz D., Dapergolas A., Kontizas
E., Granato G. L., Silva L., Rigopoulou D.,
Gonzalez-Serrano J. I., Verma A., Serjeant S.,
Efstathiou A., van der Werf P. P., 2002, MN-
Based on data from the Anglo-Australian Ob- RAS, 332, 549
servatorywithoutwhosededicatedstafftheLDSS
Sawicki, M., & Mall´en-Ornelas,G. 2003, AJ, 126,
upgradeprojectcouldnothaveproceeded. Nod&
1208 (SM03)
shuffle with LDSS was inspired by the pioneering
ideas of Joss Bland-Hawthorn to whom KG is in-
Tinney, C. G., Barton, J., 1998,AAO Newsletter,
debtedformanyusefuldiscussions. Wewouldlike
87, 9
to especially thank Lew Waller, Tony Farrell and
John Barton for their technical work. We would Tody, D. 1986, ”The IRAF Data Reduction and
also like to thank Roberto Abraham for his help Analysis System” in Proc. SPIE Instrumenta-
riding in the AAT Prime Focus cage during these tion in Astronomy VI, ed. D.L. Crawford, 627,
observations,Chris Tinney for observing support, 733
Pippa Goldschmidt for her help with the imaging
Tresse, L., Dennefeld, M., Petitjean, P., Cristiani,
reductions and Sam Barden (NOAO), Jims Arns
S., & White, S. 1999, A&A, 346, L21 (T99)
andBillCoulburn(Kaiser)for their helpwith the
VPH grating design and manufacture.
Vanzella, E., et al. 2002,A&A, 396, 847 (V02)
REFERENCES Williams, R. E., et al. 2000, AJ, 120, 2735
Abraham, R. G., et al. 2004, AJ, 127, 2455
Wynne, C. G., & Worswick, S. P. 1988, The Ob-
servatory,108, 161
Barden,S.C.,Arns,J.A.,&Colburn,W.S.1998,
Proc. SPIE, 3355, 866
Bertin, E., & Arnouts, S. 1996, A&AS, 117, 393
Burke, B. E., et al. 2004, Scientific Detectors for
Astronomy, The Beginning of a New Era, eds.
Amico, P. , Beletic, J. W. , Beletic, (Kluwer),
41
Ferna´ndez-Soto, A., Lanzetta, K. M., & Yahil, A.
1999,ApJ, 513, 34
Glazebrook, K. & Economou, F., 1997, Dr Dobbs
Journal, 1997, 9719 (Fall issue)
This2-columnpreprintwaspreparedwiththeAASLATEX
Glazebrook, K. 1998, AAO Newsletter, 87, 11 macrosv5.2.
7
Fig. 1.— Figure depicting follow-up imaging surveys of the HDFS. The extent of the AAT images is shown
in black. The bright star HD 213479which causes stray light in the AAT fields can also be seen.
8
Fig. 2.— Illustration of the particular LDSS nod & shuffle layout we used. The 2048×4096 CCD detector
(grey rectangle) is physically larger than the circular image (12.3 arcmin diameter) delivered by the LDSS
camera. For nod & shuffle only the middle third of the chip (1365 pixels high) is illuminated by the mask,
the lower third is used to store the sky image and the upper third is blank (it hold the object image when
the shuffling chargeup anddownby 1365pixels). Thus only 9.0arcminverticallyis availablefor slits which
is almost the full field.
9
Fig. 3.—Examplemicroslitdata. Thisshowsazoomonasmallportionofa2Ddataframe(∼30microslits)
where the only processing applied at this step is the subtraction of the shuffled sky. Each circular microslit
produces a spectral trace. Several bright objects with absorption lines and faint objects with emission lines
can be seen. The data shares features both with fiber data and classical multislit data. The apertures are
unresolvedspatially,howeverunlikefiberdatathespacingbetweenaperturesisuneven(itreflectsthespacing
between objects) and the wavelength zeropoints vary too (reflecting the spatial position of the objects on
the other axis).
10