arX

iv:a

stro

-ph/

0702

511v

1 2

0 Fe

b 20

07

The SkyMapper Telescope and The Southern Sky Survey

S. C. KellerAB, B. P. Schmidt, M. S. Bessell, P. G. Conroy, P. Francis, A.Granlund, E. Kowald, A. P. Oates, T. Martin-Jones, T. Preston, P. Tisserand,A. Vaccarella, M. F. Waterson

A Research School of Astronomy and Astrophysics, Australian National University, Cotter Rd,

Weston, ACT 2611, AustraliaB Email: stefan@mso.anu.edu.au

Abstract: This paper presents the design and science goals for the SkyMapper telescope. SkyMapperis a 1.3m telescope featuring a 5.7 square degree field-of-view Cassegrain imager commissioned forthe Australian National University’s Research School of Astronomy and Astrophysics. It is located atSiding Spring Observatory, Coonabarabran, NSW, Australia and will see first light in late 2007.The imager possesses 16k×16k 0.5 arcsec pixels. The primary scientific goal of the facility is to performthe Southern Sky Survey, a six colour and multi-epoch (4 hour, 1 day, 1 week, 1 month, 1 year sampling)photometric survey of the southerly 2π steradians to g ∼23 mag. The survey will provide photometryto better than 3% global accuracy and astrometry to better than 50 mas. Data will be supplied to thecommunity as part of the Virtual Observatory effort. The survey will take five years to complete.

Keywords: telescopes — surveys — techniques: photometry

1 Project Overview

SkyMapper is amongst the first of a new generationof dedicated, wide-field survey telescopes to start op-eration in the coming five years. It is now feasibleto tessellate the field of view of a low f-number tele-scope with CCDs and achieve areal coverage compara-ble to that obtained with photographic Schmidt cam-eras. The all-sky photographic surveys of the Palomarand UK Schmidt telescopes (Reid et al. 1991; Holmberg et al.1974) are currently unsurpassed in the optical. Suchphotographic surveys have several shortcomings how-ever, they do not provide photometry to better than0.2 mag. or astrometry to better than 0.5 arcsecs.

Several groups are now actively pursuing digitalmulti-colour surveys of the sky, notably Pan-Starrs(operational 2007; Kaiser et al. 2002) and the LargeSynoptic Survey Telescope(first light in 2012; Claver et al.2004). The only survey already in progress is the SloanDigital Sky Survey (SDSS; York et al. 2000). SDSShave mapped close to π steradians of the northern skyin five colours. The gamut of scientific applications forSDSS data spans properties of the asteroid population(Ivezic et al. 2001) to the discovery of the most dis-tant quasar (Fan et al. 2001). In the infrared, VISTA(Lewis et al. 2006) plans to survey the southern skystarting in early 2007. Other large aperture survey in-struments are planned such as ESOs VST (first light2007; Cappellaro et al. 2004) and the Lowell Obser-vatorys Discovery Channel Telescope (first light 2009Sebring et al. 2004) with narrower fields of view de-signed for targeted deeper surveys.

SkyMapper will perform the multi-colour, multi-epoch Southern Sky Survey (S3). Reflective of our sci-ence goals we have devised a filter set for the S3 thatis optimised for stellar astrophysics. We have soughta filter set that best discriminates the important stel-

lar parameters of effective temperature, surface gravityand metallicity. This is not to the detriment of non-stellar science as these areas are well served by a seriesof broadband filters.

The SkyMapper focal plane will feature a mosaicof 32 2k×4k CCDs coupled to high speed device con-trollers for rapid, low noise readout. The facility willoperate in an automated manner and require minimaloperational support. On site there is a scheduling anddata quality assurance system. Data is transferred viaa high speed link to the Australian National Univer-sity’s Supercomputing Facility where the data reduc-tion pipeline resides. SkyMapper will see first light in2007.

2 The SkyMapper Telescope

2.1 Site

The SkyMapper telescope is located 20m north-eastof the summit of Siding Spring Mountain, near Coon-abarabran, NSW (altitude 1169m). Figure 1 shows aplan of the observatory site. To minimise the facil-ity’s impact on the seeing obtained with the telescope,large heat sources such as power supplies, computersand instrumentation within the enclosure are cooledby chilled water supplied from an equipment pad 40mdownhill from the telescope.

2.2 Enclosure

The SkyMapper enclosure is 6.5m in diameter and 11mhigh with three internal levels as shown in Figure 2.The first level houses the electronics and control com-puters and is thermally isolated from the rest of thebuilding. Four sets of vent shutters are set into the

1

2 Keller et al.

Figure 1: Plan of the SkyMapper site on Siding Spring Mountain.

walls of the second level. When observing these shut-ters enable passive ventilation of the observing spaceabove via the mesh floor of the observing space. Thevent shutters will have controlled opening angles toflush the observing space while not introducing exces-sive wind buffeting.

During the day the volume of levels 2 and 3 willbe sealed and actively cooled to the expected ambi-ent temperature at the start of observing for the nightahead. The interior of the dome will be finished mattblack to minimise scattered light reaching the detector.

The SkyMapper enclosure features internal and ex-ternal arrays of weather sensors. These will preventoperation in inclement weather. Should mains powerfail, there will be sufficient power in the enclosure’sUPS to safely shutdown the facility.

Lightning is a significant risk for the facility asSiding Spring Mountain is exposed to severe summerthunderstorms (the vicinity receives approximately 3strikes per km2/yr). The mountain top is largely out-cropping bedrock of high resistivity. Our lightningmitigation consists of a series of six air terminals thatconnect to a radial ground plane of buried copper braidthat radiates 20m from the enclosure. In addition,the telescope control computer will receive informa-tion from a commercial lightning warning service andwill close the enclosure in the case of an impendingthunderstorm.

2.3 Site Conditions

The prevailing astronomical conditions at the site arean important input to the design of the project. Wehave examined the weather logs of the Anglo-Australian

Figure 2: The SkyMapper enclosure.

SkyMapper & the Southern Sky Survey 3

Figure 3: Seeing at Siding Spring derived from logsof the Anglo-Australian Telescope.

Telescope (AAT) from 2000-20051 . Siding Spring Ob-servatory’s (SSO) weather patterns show little seasonalvariation, with an average of 2110 hours (64%) of thetime useable throughout the year, of which 1135 hours(35%) is photometric.

Using DIMMmeasurements (Wood et al. 1995) themedian free air seeing at SSO is approximately 1.1arcsecs. This assumes the turbulence is not near theground and therefore the actual seeing may be some-what worse. This is borne out by the logs of the AATwhich are summarised in Figure 3. This shows 68% ofthe time the seeing is less than 1.75 arcsecs.

2.4 Optical design

The SkyMapper telescope optical system was designedby Electro Optics Systems, Australia. It features a1.33m primary mirror and a relatively large 0.69m sec-ondary. This results in a collecting area equivalent toan unobstructed aperture of 1.13m. The telescope isa modified Cassegrain design, optimised for wide-fieldoperation between 340nm and 1000nm. The design inshown in Figure 4.

The primary mirror is of Astrosital glass ceramicfabricated by LZOS, Russia. The secondary mirror(sourced from SAGEM, France) is carried on an ac-tuated hexapod system that allows for tip-tilt, x&yand telescope focus movement. The telescope controlsystem uses the hexapod to automatically adjust forgravity induced flexure and focus change due to tem-perature.

The primary and secondary mirrors possess pro-tected aluminium coatings. This increases the longevityof the reflective surface, particularly in the UV. Thethree transmissive corrector elements are single-layeranti-reflection coated and composed of fused silica tomaximise the UV throughput of the instrument.

The EOS SkyMapper telescope is a compact, stiffand rugged structure with high mechanical natural fre-quencies. We expect the telescope to point to < ±3arcsec and track to 0.5 arcsec RMS for 5 minutes in

1C. McCowage private communication

Figure 4: SkyMapper optics.

Table 1: SkyMapper Imager ParametersTelescope working f/ratio f/4.7878

Telescope Focal Plane Scale 0.0302mm/arcsec

Detector Pixel Projection 0.5 arcsec/pixel

Number of Pixels in Mosaic 268 435 456

Mosaic Dimensions 256.34×258.75mm

Mosaic Field of View 2.373o×2.395o

Mosaic Fill Factor 91.05%

Sky Coverage w. Fill Factor 5.68 square degrees

winds of <5m/s. Both of the telescope principal axesare driven by direct on-axis DC ring motors withoutintervening gearboxes and direct on-axis incrementalencoders. The system is therefore free of backlash.

3 The SkyMapper Imager

The SkyMapper Imager has been designed in-house atRSAA (see Granlund et al. (2006) for more details).Due to the large focal plane, the instrument is com-paratively large and heavy for the size of telescope.To achieve an optimal outcome, the designs of the tele-scope and imager have proceeded in parallel, with closecooperation between the RSAA and Electro-Optic Sys-tems design teams. Important imager parameters aresummarised in Table 1.

Figure 5 shows the imager instrument from abovelooking down on the rotator interface plate. The auto-guider and science filters can be seen through the beamaperture. Also seen is the Shack-Hartmann system

4 Keller et al.

Figure 5: The SkyMapper Cassegrain Imager asseen from above.

Figure 6: The SkyMapper Cassegrain Imager asseen from below.

that is used for automatic collimation and focusing.Figure 6 shows the underside of the imager. The Im-ager vacuum jacket is at the centre, flanked by closed-cycle helium cryocoolers and twin detector controllers.The instrument weighs ∼300 kg.

A large mechanical shutter has been fabricated bythe University of Bonn, Germany. The shutter is lo-cated below the filter stack and above the vacuumjacket window. The shutter is composed of two bladesthat form a moving slot of variable width to provideuniform exposure of the focal plane. The shortest ex-posure is about 1ms and exposure homogeneity is 0.3%in a 100ms exposure (Reif et al. 2004).

SkyMapper has six interchangeable filters, each 309-mm square and up to 15mm thick. Table 4 lists thespecifications of the filter set for the S3. Our motiva-tion for the choice of filters is detailed in Section 5.2.The filter set is constructed of coloured glass wherepossible. Coloured glass filters show better through-put homogeneity across the focal plane than can becurrently achieved with multi-layer interference filters.

Figure 7: The SkyMapper Cassegrain Imager vac-uum jacket showing the CCD mosaic.

The filter glass was sourced from Macro Optica, Rus-sia and Schott, Germany. Only the r and i band filtershave additional short wavepass coatings to define thebandpass. An optimised anti-reflection coating will beapplied to each filter by Optical Surface Technologies,USA.

The Imager vacuum jacket carries the CCD mo-saic as shown in Figure 7. The vacuum jacket win-dow is 25mm thick fused silica with a broadband anti-reflection coating. A steady flow of dry air is passedover this window to prevent condensation.

3.1 The SkyMapper Imager CCD Mo-saic

The SkyMapper Imager CCD mosaic is a 4 x 8 ar-ray of E2V CCD44-82 detectors. Each CCD detectorhas 2048 x 4096, 15 micron square pixels. The deepdepletion devices are back-illuminated and 3-side but-table. They possess excellent quantum efficiency from350nm-950nm (see Figure 8), near perfect cosmetics,and low-read noise.

All 32 CCDs are carried on a 10mm thick Invarcarrier plate. The surface of the carrier plate has beenmachined to 10µm flatness and on to this surface theprecision pads of the E2V devices are mounted. Ouraim for focal plane flatness is to match the single pixelgeometric depth of field in the f/4.78 beam of ±32µm.

A portion of the top of the CCD detector mosaicassembly can be seen in Figure 7. The inner two rowsof CCDs are butted together with a 1.5mm gap be-tween the rows and a 0.5mm gap between columns. Alarger gap exists between the centre and outer rowsof about 4mm. This assembly gives a filling factor of91%.

SkyMapper & the Southern Sky Survey 5

3.2 CCD Controllers

The SkyMapper Cassegrain Imager will utilise two cus-tomised versions of the newly developed 16-channelSTARGRASP controllers2 developed for Pan-STARRSby Onaka and Tonry et al. from the University ofHawaii. A hybrid 300MHz PowerPC CPU FPGA isused in the controller for digital signal processing. Eachcontroller also possesses 256MB of embedded memoryfor buffering the high data flows. The controllers willhave integrated Gigabit Ethernet optical-fibre portsfor fast readout, which will enable the SkyMapper Im-ager to easily meet its requirement of reading out theentire focal plane in <20s (450kpix/s). It is our goalto read out the focal plane in 10s, i.e. 1Mpix/sec.This new detector controller technology will also re-duce the size and weight of the controller hardwaresubstantially and increase reliability over legacy sys-tems. The controllers are to be attached to either sideof the Detector Vacuum Enclosure (see Figure 7).

Both detector controllers will be connected to asingle pixel server computer that transfers the data tolocal storage. The pixel server is commanded by theComputerised Instrument Control And Data Acqui-sition software system (CICADA: Young et al. 1999)developed by RSAA for all observatory instruments.The RSAA’s Telescope Automation and Remote Ob-serving System (TAROS: Wilson et al. 2005) in turncontrols CICADA to configure the instrument and ac-quire the required exposure.

4 Science Goals

The science that can be done with a comprehensivesurvey such as the Southern Sky Survey cannot befully described or predicted. Data from previous all-sky surveys has been used in thousands of papers andimportant discoveries continue to be based upon themdecades after their completion. In the following sec-tion we outline a series of scientific questions that wehave identified SkyMapper will have high impact in ad-dressing. These science goals define the requirementsfor SkyMapper’s Southern Sky Survey.

4.1 What is the distribution of so-lar system objects beyond Nep-tune?

The existence of quiescent comet-like objects orbitingthe Sun in a region beyond Neptune was first pro-posed by Leonard (1930) following the discovery ofPluto. The objects in this region remained undiscov-ered until 1992 when the first Trans-Neptunian Ob-ject (TNO) was discovered (Jewitt & Luu 1993). Sub-sequent searches have, to date, uncovered more than1000 TNOs. TNOs are believed to be debris left overfrom the formation of the solar system. The distri-butions of their orbits and masses provide importantclues to the process of outer planet formation and tothe origin and nature of comets. An all-sky survey will

2see http://www.stargrasp.org/

Figure 8: The spectral response of SkyMapper sci-ence CCDs. The shaded area encloses the mini-mum and maximum measured response for a setof six devices.

overcome the observational biases that exist in cur-rent work based around the ecliptic. Little is knownabout any larger TNOs, especially those on scatteredorbits that take them out of the ecliptic. Models ofKenyon & Luu (1999) predict many large TNOs, somewith masses larger than Pluto. The recent discovery of2003UB313, a scattered object with a diameter slightlyin excess of Pluto (Brown et al. 2006), demonstratesthe possibilities for future discovery.

The requirements for this science program are asfollows. By going to r ∼21.5 over the entire sky westand to discover ∼2000 TNOs (Trujillo et al. 2001).The motions are typically a few arcsec per hour atopposition. To distinguish nearby objects from distantones, at least 3 observations are required. We envisagetwo epochs separated by ∼4hrs on the first night witha third epoch 1-3 days later. A fourth epoch afterone month will be sufficient for extrapolation of theorbit to the next year. Astrometric accuracy is notparticularly stringent: < ±0.2 arcsec will suffice.

4.2 What is the history of the young-est stars in the solar neighbour-hood?

In the last two decades it has been possible to defini-tively discern coeval, comoving associations of youngstars within the solar neighbourhood. By looking atthe space motions and spatial distribution of youngobjects Kastner et al. (1997) revealed the TW Hyaassociation. With the advent of all-sky X-ray anddeep proper motion catalogues further close young as-sociations were found (η Cha; Mamajek et al. (1999),Tucana-Horologium; Torres et al. (2000) and Zuckerman & Webb(2000)). These associations of young stars have beenrecently ejected from the molecular clouds of their birth,most likely within the Sco-Cen association, a massiveregion of star formation in the southern Milky Way(Zuckerman & Song 2004).

Because these young stars are unobscured and closeto the Sun, they are ideal targets for high resolution

6 Keller et al.

imaging of protoplanetary disks with Spitzer, SOFIAand IR adaptive optics imagers. These objects offer usa view of how stars and planetary systems form, onenot obscured by the material from which they formed.While several comoving associations have been foundfrom the Hipparcos and Tycho catalogues, these sur-veys are limited to the brightest members of the clusterluminosity function.

A key requirement for this science program is tohave good proper motions for objects across the sky aswell as accurate colour selection down to r=18. Thiswill enable us to select nearby candidates for planetarydisk searches. Relative astrometry to 50mas over a fiveyear baseline will provide proper motions to 4masyr−1

accuracy. This is sufficient to find relevant objects to120pc.

4.3 What is the shape and extentof the dark matter halo of ourgalaxy?

Dark matter is the dominant gravitational componentof our Universe. We have few clues to its nature. Oneclue is its distribution. It is now clear that dark matterin the cores of galaxies does not produce central cuspsof material as predicted by theory (Navarro & Steinmetz2000). Even less is known about how dark matter isdistributed in the outer regions of galaxies. For theMilky Way we would like to know the extent and shapeof the dark matter: is it spherical, flattened or triax-ial? The answers to these questions place constraintson the nature of dark matter and the formation of ourgalaxy.

Visible dynamical tracers must be used to probethe distribution of dark matter in the galactic halo.Perhaps the best such tracers are RR Lyrae stars butsuch stars are rare: approximately one per 10 squaredegrees beyond 50kpc and one per 100 square degreesbeyond 100kpc (Ivezic et al. 2001). Blue horizontalbranch stars are 8-10 times more common, but requirea good surface gravity discriminant to separate fromthe bulk of higher surface gravity main sequence starsand blue stragglers in the field. Intrinsically luminousF and G giants can be seen to greater distances butare as rare as RR Lyraes and similarly need a goodsurface gravity indicator.

Requirements for this program are a series of atleast two exposures over short (∼ 1 week) time frameto discriminate RR Lyraes. 90% of RR Lyrae stars canbe recovered with a g−r colour to 0.1mag precision and6 epochs in g with a precision of 0.1mag. Blue hori-zontal stars can be discriminated on the basis of colourinformation with a gravity indicator to distinguish con-taminating blue stragglers and main-sequence A-typestars (see Section 5.2). Selection of blue horizontalbranch stars and giants requires photometry accurateto ∼0.03mag.

4.4 Extremely metal-poor stars: howdid our Galaxy evolve?

What were the characteristics of the first stars, whendid they light up the Universe, and how did they forminto galaxies like the Milky Way? The formation ofthe first stars is a key moment in the history of theUniverse yet it lies approximately 13 billion years inthe past (Naoz, Noter & Barkana 2006), beyond thereach of modern-day telescopes. However, our galaxyholds a fossil record of the first generations of stars.Since the Big Bang gave rise to a universe of hydro-gen, helium and an admixture of light elements, suc-cessive generations of stars have synthesized and re-turned to the interstellar medium the ∼4% by massof heavier elements that comprise, for instance, theSun. By using metallicity as a proxy for age, the mostmetal-poor stars are candidates for the first generationof stars. These candidates will then require follow-upspectroscopy to determine such quantities as radial ve-locities and fine abundance analysis.

Requisite for this program is accurate (σ < 0.03mag) multi-colour photometry from the near-UV tothe red for stars between 12−18thmagnitude. Largesky coverage is essential as such stars are extremelyrare. The choice of filters should provide optimal de-termination of effective temperature, surface gravityand metallicity.

4.5 High redshift QSOs: when didthe first stars in the Universeform?

We know that within 1.5 billion years after the BigBang, some galaxies had formed; this is evidenced bygalaxies observed to z∼6. Even at this epoch mostof the intergalactic medium was ionized as evidencedfrom the lack of continuum absorption redward of Lyman-α (the Gunn-Peterson effect; Gunn & Peterson (1965)).It must be concluded that some objects must have ex-isted earlier that produced sufficient UV flux to ionisenearly all the baryonic matter in the Universe.

The SDSS located a series of z>5.8 quasars in-cluding the most distant object at z=6.42 (Fan et al.2001). The most distant QSO does seem to showcontinuous Lyman-α absorption (Becker et al. 2001).This is our first glimpse of the interface between theionised intergalactic medium we see locally and theneutral material before. The description of this sur-face tells us about the physics of the era of reionisation(Wyithe & Loeb 2004).

So far three high redshift objects from the SDSSshow the Gunn-Peterson effect. By covering the en-tire southern sky we stand to take this to a dozen ormore such objects. Requirements for this program aresurvey sensitivities matching those of SDSS. Multipleepochs are essential for good cosmic ray rejection.

The union of the above requirements provides thedefining (most difficult to achieve) constraints of thesurvey. They are summarised in Table 2.

SkyMapper & the Southern Sky Survey 7

Table 2: Southern Sky Survey defining requirementsu v g r i z

Sensistivity σ = 0.03 mag σ = 0.03 mag σ = 0.1 mag σ = 0.1 mag 3σ 7σu <20 vs < 20 g = 21.2 r = 21.0 i = 22.5 z = 20.5

detection detectionCadence 6 epochs 6 epochs 3 epochs for 3 epochs for

hours to hours to cosmic-ray cosmic-ray

years years rejection rejection

Systematic 0.03 magAstrometry 50 mas over 3 years

4.6 Non-Survey Science

Planetary transits - SkyMapper allows monitoringof a large field of view to high stability (relative flux tobetter than 0.01mag) making it a highly competitivefacility for this purpose.

Supernovae - We are proposing to use time thatdoes not meet the survey’s seeing and sky backgroundconstriants to undertake a supernova survey. Such asurvey will provide continuous coverage of 1250 squaredegrees of sky, and discover approximately 100 SN Iato z<0.085 per year. In addition to allowing a system-atic exploration of the statistical properties of TypeIa supernovae, it will also serve to populate the SN IaHubble Diagram in this low redshift regime. Similarly,for core collapse supernovae we will increase statisticswith which to explore the energetics and nucleosyn-thetic production of these explosions. The supernovasurvey will lead to the detection of many hypernovaewith which to explore the relationship between theseobjects and Gamma-Ray Bursts.

5 Southern Sky Survey Design

The requirements of Table 2 are met by the designof the Southern Sky Survey that we now detail. Ourscience goals emphasize the need for wide sky coverage.We have defined 4069 fields covering the 2π steradiansof the southern sky.

5.1 Survey Coverage and Cadence

To meet our science goals we must capture variabil-ity on short and long time scales. Since most variableastronomical objects have time scales for variabilityranging from hours (RR Lyraes, asteroids), months(supernovae, long period variables) and years (QSOs,stellar parallax and proper motions) our survey designwill ensure that each field is observed in each filter onsix epochs with approximately the following cadence, t= 0, +4 hours, +1 day, +1 week, +1 month, +1 year.

The six epochs will be spatially offset slightly fromeach other according to a dither pattern. This willensure that chip gaps and defects do not occult thesame area of sky. At completion, 90% of the sky willbe covered at least five times and 100% imaged threeor more times.

The survey is to be completed in five years to re-alize its full potential. Together with the weather con-straints detailed in Section 2.3, we have defined ourexposures to be 110 seconds (plus 20 second overheadbetween expsures) in each filter. Two data releasesare planned: a first data release when 3 epochs in eachfilter have been obtained and calibrated and a secondwhen the complete set of 6 epochs in each filter havebeen obtained. The expected survey depths in 1.5 arc-sec seeing for a 5σ detection in AB magnitudes aregiven in Table 3.

Table 3: SkyMapper Main Survey depth in ABmagnitudes (signal-to-noise of 5 - 110 second ex-posures).

u v g r i z

1 epoch 21.5 21.3 21.9 21.6 21.0 20.6

6 epochs 22.9 22.7 22.9 22.6 22.0 21.5

5.2 Optimised for Stellar Astrophysics- filter set design

The filter set we have designed for S3 provides bothinter-operability with existing systems such as SDSSand key information for the science goals discussed inSection 4. The specification of the filter set is given inTable 4. In Figure 9 we show the expected throughputof the SkyMapper Telescope and Cassegrain Imager.We have taken as our basis the Gunn & Thuan griz fil-ter set (Thuan & Gunn 1976) utilized by SDSS3. Whencompared to the SDSS system the SkyMapper systemfeatures twice the throughput in u and three times thethroughput in z, albeit with our smaller aperture. Thecleanly separated i and z filters are optimal for discov-ery of high redshift QSOs (Lyman alpha in z equatesto a redshift of greater than 5.8). Photometric red-shifts for z<0.5 galaxies are facilitated by broadbandphotometry over the optical spectrum.

A large part of the science potential of SkyMapperis bound to the study of stellar populations. It wasimportant, therefore, for us to consider how we can

3http://www.astro.princeton.edu/PBOOK/camera/camera.htm

8 Keller et al.

Figure 9: The expected throughput ex-atmosphereof the filter set for the Southern Sky Survey.

best characterize stellar populations using broad to in-termediate band photometry. To do this we computedthe synthetic colours of stars spanning metallicity, sur-face gravity, effective temperature and extinction fromPietrinferni et al. (2006) model atmospheres. For anygiven point in the colour-colour space, the numberof synthetic models within an error ellipsoid providesthe measure of uncertainty in the derived quantities.These simulations were performed as a function of pho-tometric uncertainty and filter choice.

For G stars and hotter the main discriminatingpower is derived from the near-UV. We incorporate aStromgren u filter and an intermediate band filter sim-ilar to DDO38, we term v, that covers the spectrumfrom 3670-3980A. These two filters are able to breakthe degeneracy between surface gravity and metallicityas illustrated by the following important applications.

Table 4: SkyMapper filter pass bandsFilter 50% Cut-on 50% Cut-off FWHM

edge (A) edge (A) (A)u 3250 3680 430v 3670 3980 310g 4170 5630 1460r 5550 7030 1480i 7030 8430 1400z 8520 9690a 1170a

a z filter is unblocked at red end, the red edge islimited by the CCD sensitivity cut off in the

near-IR.

5.2.1 Application 1: Blue Horizontal Branch

Stars

In Figure 10 we consider the uncertainty in the derivedstellar surface gravity as a function of temperature for

stars of log g = 2.5 and 4.5 with photometric uncer-tainties of 0.03 magnitudes per filter. For hot O andB stars the intensity of the Balmer jump feature isnot significantly affected by changes in surface gravity.Consequently we have limited ability to discern sur-face gravity for such stars. Similarly, for stars coolerthat 5000K (G6-7V) the Balmer jump disappears as aspectral feature. However, for A type stars we expectto determine the surface gravity to ∼ 10%. This rangein temperature is inhabited by blue horizontal branchstars (BHBs) - standard candles for the Halo.

A sightline through the halo inevitably contains lo-cal disk main-sequence stars and blue stragglers in thetemperature range of interest. However, these contam-inants are of significantly higher surface gravities thanthose of BHBs. The u and v filters measure the BalmerJump and the effect of H− opacity which increaseswith surface gravity. As can been seen from Figure11, the u−v colour index discriminates between BHBsand their contaminants. The separation is 0.5mag incolour. On the basis of our photometric selection wewill be able to discern a sample of BHBs to a distanceof 130kpc with less than 5% contamination.

4.4 4.2 4 3.8 0

0.2

0.4

0.6

0.8

log Teff

A0 A5 F0

1σ

un

cert

ain

ty in

log

g

Figure 10: The one sigma uncertainty in surfacegravity derived under photometric uncertainties of0.03 mag from the SkyMapper filter set. Solidsquares are log g=4.5, open squares log g=3.5 andcircles log g=2.0. Dashed vertical lines show thetemperatures corresponding to spectral types onthe main sequence.

5.2.2 Application 2: Extremely metal-poor

Stars

Amongst cooler stars, the u and v filters capture themagnitude of metal line blanketing, a general suppres-sion of the continuum blueward of ∼4000A. In Figure12 we show the uncertainty in derived metallicity as afunction of temperature for stars of metallicity [Fe/H]= -4 from the S3 filter set.

We are able to determine [Fe/H] best at high metal-licity and the uncertainty rises to ±0.7dex at [Fe/H]=-4. At extremely low metallicities the density of starsdrops by a factor of ten for every factor of ten drop inmetallicity. Consequently, the challenge in finding ex-tremely metal-poor stars (EMPs) is to keep the verysmall subset of interesting objects clean from photo-metric outliers. In this regard, the S3 filter set pro-vides improved separation over existing filter sets be-

SkyMapper & the Southern Sky Survey 9

Figure 11: u−v vs. g−i for stars of solar metal-licity and a range of surface gravity. Blue hori-zontal branch stars are well separated from main-sequence and blue straggler stars.

tween the bulk of halo stars of [Fe/H]> −2 and those of[Fe/H]< −4. This provides us a distinct advantage inisolating extremely metal-poor stars from photometriccolours.

4 3.9 3.8 3.7 3.6 0

0.5

1

log Teff

1σ

un

cert

ain

ty in

[Fe

/H]

F5 G0 G5

Figure 12: The one sigma uncertainty in metal-licity derived under photometric uncertainties of0.03 mag from the SkyMapper filter set. Solidsquares represent [Fe/H]=-4 models, open squares[Fe/H]=-2 and circles [Fe/H]=0.

Our ability to discern metallicity is not clearly ex-pressed in a colour-colour plot as the vector of gradmetallicity is not closely aligned with any colour. Thebest correspondence between grad metallicity and ourcolours is found with the v−g colour shown in Fig-ure 13. We see that v−g has little dependence onsurface gravity from early F to mid G; however thisbreaks down as we enter the cooler K-type range. Theevolved extremely metal-poor stars do not possess tem-peratures as cool as K0 and those cooler than K0 andstill on the main-sequence are sufficiently lacking in lu-minosity to make their numbers small over the surveyvolume.

We have calculated the number of halo stars inthe survey brighter than g=18.1 (where we reach asignal-to-noise ratio of 33 in the first data release)by integrating the Bahcall and Soniera galactic model(Bahcall & Soneira 1980) over the survey area with

(v-g

)-0.9

(g-i)

g-i

log[Fe/H]=-1-2-3-4

Solar, logg=2

3

45

-5.4 Frebel et al. 2006

Figure 13: v−g vs. g−i for stars of solar metal-licity and a range of surface gravity (solid lines)and for log g=4 and metallicities to [Fe/H]=-4.Overlaid are the computed colours of HE1327-2326(Frebel et al. 2005) and the sample of extremelymetal-poor stars from Cayrel et al. (2004).

galactic latitude greater than 15 degrees. We thenscale by the results of the Hamburg ESO Survey (Christlieb2003) to find the number of stars with [Fe/H]<-4. Thereare on order of a thousand such stars, including ahundred of metallicities less than -5. Follow-up willbe tractable; to recover 90% of the stars of [Fe/H]>−4 will require intermediate resolution spectroscopy of∼2600 stars. High signal-to-noise 8m spectra will berequired on the distilled sample of extremely metal-poor stars for detailed elemental abundance studies.

6 Photometric Calibration

To calibrate the S3 to the required global accuracyof 0.03 magnitudes we will perform a shallow surveyof the entire southern sky under photometric condi-tions (the Five-Second Survey). Exposure times of ap-proximately five seconds in each filter will enable usto obtain photometry for stars from 8.5 to 15.5 mag-nitudes. The Five-Second Survey will consist of a setof at least three images of each field in all filters. Thenetwork of standards will provide the photometric andastrometric catalogue to which the deeper Main Sur-vey images will be anchored. With all-sky coverage,the Five-Second Survey will enable the Main Surveyimages (and any other images taken with the tele-scope) obtained in non-photometric conditions to becalibrated.

The major impediment to the derivation of accu-rate photometry from wide field of view instruments isthe influence of scattered light. Magnier & Cuillandre(2004) studied the systematic errors in the flat-fieldstructure of the CFHT12K imager. By dithering astandard star sequence over the CCDs of the mosaicthey were able to form a map of the photometric resid-uals (observed - expected) over the focal plane (a pho-tometric superflat). In the case of the CFHT12K im-ager systematic errors in the flatfield amount to 5%peak-to-peak: around half due to pixel scale change

10 Keller et al.

across the camera, another half due to the effect ofscattered light. The scattered light component arisesfrom ambient light scattered off surfaces within the en-closure and telescope reaching the detector. We havesought to minimise the scattered light by finishing theinterior surfaces matt black, however some scatteredlight will inevitably remain.

Forming a photometric superflat is a direct way toremove the signature of scattered light. During com-missioning, we will establish a series of six calibrationfields at declination ∼-25deg (to avoid the zenith key-hole of the telescope) around the sky. By observingthe calibration fields with a series of dithers and rota-tor angles we will be able to characterise and correctfor departures from photometric flatness as a functionof position on the detector array. These fields will alsoserve as our secondary standard star fields. DuringFive-Second Survey operations we will acquire stan-dard star fields every 90 minutes. At such time thetwo highest fields will be observed, thus providing arange of airmasses from ∼ 1 to 1.8.

With full coverage of right ascension a photomet-ric “ring closure solution” can be formed (see for ex-ample Pel & Lub (2006)). The ring closure solutionensures that the differential flux ratios between cal-ibration fields in all bandpasses are very accuratelydetermined. This, finally, leaves the zeropoints of theflux in each filter to be determined. We will define theabsolute zero point of the SkyMapper system from theWalraven photometric system (Pel & Lub ibid) andthe associated stars in the Next Generation SpectralLibrary4. To do this we will first use theWalraven pho-tometry to accurately zeropoint the spectrophotomet-ric standards. We will then use the spectrophotomet-ric standards and with our photometric bandpasses,to derive absolute zeropoints. Initially our photomet-ric bandpasses will be those measured by a laboratorymonochromometer convolved with the CCD spectralresponse and adjusted to match the observations ofstandard stars. Ultimately, we plan to use a monochro-matic flat + NIS standard (Stubbs et al. 2006) to char-acterise the response of our system.

7 Astrometric Calibration

Our astrometric calibration uses as its basis the UCAC-2 astrometric catalogue (Zacharias et al. 2004). Thecalibration has two stages. Firstly, UCAC2 stars incommon to Five-Second images are used to define animage world coordinate system (WCS). The UCAC2provides between 40 and several hundred stars perSkyMapper CCD. Our derived positional uncertain-ties are better than 50 milliarcsecs for bright stars.The Main Survey images are then astrometrically cal-ibrated against the stars of the Five-Second Survey.

We implement the ZPN5 coordinate representa-tion. This representation utilizes a radial term to char-acterise distortions across the image and a simple lin-ear transformation to remove the effects of rotation,

4Gregg et al. http://lifshitz.ucdavis.edu/ mgregg/gregg/ngsl/ngsl.html5Zenithal/azimuthal polynomial projection see

Calabretta et al. (2004)

scale, and any shear caused by refraction. Since mostdistortion is radial from the optical axis, the ZPN sys-tem is able to linearise a distorted optical plane witha single term. Linearising the telescope system in thisway provides robust coordinate transformations acrossthe images with no difficulties in CCD corners or inimages with few astrometric standards.

Any systematic astrometric offsets will be appliedvia a lookup table after processing a large portion ofthe survey. These are not expected, but would be dif-ficult to accurately quantify if they depend on the ro-tator angle of the telescope. Differential atmosphericrefraction will be corrected for once the colours ofeach object are determined. Over five years and 36epochs, it will be possible to derive proper motions to±4mas/yr.

8 SkyMapper Data Acquisition

& Reduction

The acquisition and reduction of SkyMapper data ismanaged by the Scheduler and Science Data PipelineSystems respectively. A schematic system level dia-gram is given in Figure 14. We now discuss the twosystems in detail.

8.1 The Scheduler System

The on-site Scheduler System is responsible for schedul-ing the required calibration and science data. In thecourse of a typical night, first bias frames will be ac-quired then sky flatfields for as many filters as timepermits. As each exposure is completed, the data ischecked for quality and system changes. In the caseof the bias frames we check that the bias levels arenominal. This is a simple check of the health of thedetector system. Before astronomical twilight a focus-ing sequence will be performed. Alternatively, whenavailable, the Shack-Hartmann system will be used todetermine the telescope focus.

At the end of evening astronomical twilight, Sky-Mapper will begin the task of acquiring S3 images.The required instrument configuration is supplied bythe Scheduler System to the instrumentation via theTAROS interface. At detector readout the next instru-ment configuration is issued by the Scheduler Systemto the telescope in order to minimise latency. TheCCD devices are read out to a 5TB RAID array andthe local disk. The copy of the image on local disk isused for data quality checks so as not to impede thewriting of data to the RAID array.

The Five-Second Survey holds top priority in sched-uling as this data must be obtained under photometricconditions. A manually set flag tells the Scheduler Sys-tem if the night is likely to be photometric based onpredicted weather conditions from the Australian Bu-reau of Meteorology MLAPS model. If this flag is set,the scheduler will commence taking any required (andpossible) Five-Second Survey data. After writing theacquired data to disk the Scheduler System performsa check of the photometric zeropoint and the surface

SkyMapper & the Southern Sky Survey 11

Figure 14: A system level view of the SkyMapper Scheduler and Science Data Pipeline.

brightness in one CCD image. Rapidly fluctuating ze-ropoints or zeropoints below a nominal value indicatethe presence of cloud.

Should the conditions become non-photometric dur-ing the night the Scheduler System will switch to theMain Survey. The Main Survey can continue underuniform diminished transparency such as might occurwith light cirrus. If the transparency drops below athreshold then the data is flagged of poor quality andthe field is made available for rescheduling. Due totime constraints for quality checks we do not attemptto monitor the uniformity of transparency across all32 CCDs of a field within the Scheduler System. Thischeck is performed in the Science Data Pipeline de-scribed below.

Another critical aspect of image quality, the seeing,is also monitored by the Scheduler System. The MainSurvey will have a seeing limit of ∼1.75 arcsecs (the68th percentile of seeing). The remainder of the timewill be given to a poor seeing program. In the case ofthe Five-Second Survey, an upper limit of 2.5 arcsecswill suffice before the poor seeing program is pursued.

Data is trickled throughout the night and proceed-ing morning to the ANU Supercomputing Facility (ANU-SF) via a gigabit link. A winter night of Five-SecondSurvey data (a maximal constraint) can be transferredover several hours. At completion of observing thedome will close and the telescope will park. The en-closure air conditioning will aim for the next night’s ex-

pected temperature as predicted by the MLAPS model.

8.2 The Science Data Pipeline Sys-tem

Data reduction is managed by the Science Data PipelineSystem (SDPS) running at the ANUSF, Canberra. TheSDPS has been designed specifically for S3 data. Fig-ure 14 outlines the stages of the SDPS for data fromthe Five-Second and Main Surveys. The system con-sists of a series of modules that perform the requiredcalibration, photometric standardisation, catalogue gen-eration, photometry, and data delivery via the Inter-net.

Pipeline modules are implemented in C/C++ fornumerically intensive processes, with Perl scripting tofacilitate process flow. Pipeline progress is tracked,and results recorded, in a PostgreSQL database. Indi-vidual modules are queued for execution on the Aus-tralian Partnership for Advanced Computing SGI Al-tix 3700Bx2 cluster housed at ANUSF.

The SDPS stages are as follows. Firstly, for eachcompleted night we check the nightly median bias again-st the median bias generated from the bias frames ob-tained on many (up to 30) nights preceding the nightin question. If the nightly median bias is outside nom-inal bounds the system attempts to generate a newmedian bias to apply. A similar sanity check is ap-plied to the flatfields obtained. If a comparison of the

12 Keller et al.

nightly flatfields with the current authorized flatfieldis outside bounds, such as might occur due to a changein the optical system of the telescope caused by, for ex-ample, the deposition of dust, the system will attemptto generate a new median flatfield to apply to this andsubsequent nights’ data. When flatfield change occurs,there may not be sufficient numbers of flatfields to becombined to form an authorized flatfield and severalnights’ worth of data may not be able to be processeduntil such time as sufficient flatfields are obtained andan authorized flatfield produced.

In addition to the bias and flatfield calibration files,fringe frames are generated for i and z. Skyflats aregenerated from the median of science images in eachband. Skyflats are produced to remove most of theeffects of scattered light in the system. We then derivea photometric superflat for each filter to characterizethe residual effects of scattered light. The photomet-ric superflat is formed by imaging of our calibrationfields (Section 6). We form a median image from theresiduals between our reference photometry and thatobtained from images after bias, flat, fringe and skyflatapplication. It is necessary to regenerate the fringe,skyflat and superflat whenever the flatfield changesas all are intimately related to the optical path. Forquality control, the processes that generate calibrationdata require the astronomer to authorize the new cal-ibration files before they are applied to any data.

A world coordinate system is defined for each CCDimage as described in Section 7. Once the Five-SecondSurvey data are in place, we test uniformity of trans-parency across each Main Survey image by derivingthe zeropoints of all CCDs and examining their devi-ations from nominal. Significant spatial variation ofzeropoints indicates the presence of non-uniform cloudattenuation. Such an image is rejected from furtherprocessing.

At this stage the image is ready for extractionof photometry and astrometry. The SDPS has twostreams for the two data types, the Five-Second andMain Surveys.

For each night of five-second data the zeropointsof all images are considered. Excessive variation indi-cates the night was not photometric and the data isdiscarded. Aperture photometry is performed on theimages. We then derive the nightly transformationequations from the photometry derived from the cali-bration fields that are interleaved with our five-secondobservations. We then apply the transformation equa-tions to the data residing in the five-second database.

Main Survey images are reduced when three (firstdata release) or six (second data release) images ofa field have been obtained in all six filters and cor-responding five-second data exists. Firstly, the zero-point offset between each image and the correspondingFive-Second survey photometry is found on a chip bychip basis. The photometry for each image is thencorrected for this offset to draw the potentially non-photometric Main Survey data onto the standard in-strumental magnitude system.

We combine the set of images in each filter. Indoing so, we mask saturated stars and grow their as-sociated bleeding, remove satellite trails and remove

cosmic ray tracks. Objects within the combined im-age are detected to 7 sigma. Transient objects willnot be retained in the median frame. In order to cap-ture transients we mask the objects found from themedian image in each component image and then lookfor detections greater than 10 sigma. A final catalog isconstructed as the union of all detections in the g,r,i,zbands, as well as any transient objects.

Photometry is performed upon the complete cata-log of objects. For each object, flux will be measuredvia PSF-fitting, aperture (over a range of apertures),Petrosian, and Kron photometry6. In addition, pho-tometry will also be forced to occur in fixed positionmode (useful at low SNR), and where the centroid ofthe object is allowed to drift. Classification of objecttype will occur utilising Source Extractor’s neural net-work. Galaxy photometry is obtained in two modes:firstly, using the galaxy shape as determined from theimage with the most signal and, secondly, using theshape parameters from each median filter image. Theresulting object positions, photometry and shape pa-rameters are stored in the S3 database.

The Five-Second Survey represents 75 terabytesworth of data, the Main Survey another 150 terabytes.This data is retained, in raw and reduced forms, atthe ANUSF’s Mass Data Store, a robotic tape archiveof PB size. It is estimated that the database of pho-tometry for the Main Survey will amount to of order5TB.

9 Data Products

The first SkyMapper data product will be the Five-Second Survey database. This will be released when aring closure solution has been established for the cali-bration fields (see Section 6). These photometric datawill be accessible from the survey website. The web-site will offer the ability to perform simple relationalqueries on the database of photometry.

We envisage two Main Survey data releases foreach field; one when three epochs in all filters havebeen obtained and a second when the final set of sixobservations have been obtained. The data release willstaged after adequate quality control has been under-taken. In addition to database queries of the MainSurvey photometry, external users will have access tothe reduced, combined images.

Acknowledgments

SkyMapper is funded by the Australian National Uni-versity and supported by grants from the AustralianResearch Council. Weather information from the AATlogs was kindly provided by Chris McCowage and StevenLee of the AAO.

6using modified Source Extractor codeBertin & Arnouts (1996)

SkyMapper & the Southern Sky Survey 13

References

Bahcall, J. N., & Soneira, R. M. 1980, ApJs, 44, 73

Becker, R. H., et al. 2001, AJ, 122, 2850

Bertin, E., & Arnouts, S. 1996, A&AS, 117, 393

Brown, M. E., Schaller, E. L., Roe, H. G., Rabinowitz,D. L., & Trujillo, C. A. 2006, ApJL, 643, L61

Calabretta, M. R., Valdes, F., Greisen, E. W., & Allen,S. L. 2004, ASP Conf. Ser. 314: Astronomical DataAnalysis Software and Systems (ADASS) XIII, 314,551

Cappellaro, E., Capaccioli, M., Mancini, D., Sedmak,G., Baruffolo, A., Cascone, E., & Greggio, L. 2004,Baltic Astronomy, 13, 683

Cayrel, R., et al. 2004, A&A, 416, 1117

Christlieb, N. 2003, Reviews in Modern Astronomy,16, 191

Claver, C. F., et al. 2004, Proc. SPIE, 5489, 705

Fan, X., et al. 2001, AJ, 122, 2833

Frebel, A., et al. 2005, Nature, 434, 871

Granlund, A., Conroy, P. G., Keller, S. C., Oates,A. P., Schmidt, B., Waterson, M. F., Kowald, E.,& Dawson, M. I. 2006, Proc. SPIE, 6269, 69

Gunn, J. E., & Peterson, B. A. 1965, ApJ, 142, 1633

Holmberg, E. B., Lauberts, A., Schuster, H. E., &West, R. M. 1974, A&AS, 18, 463

Ivezic, Z., et al. 2001, AJ, 122, 2749

Jewitt, D., & Luu, J. 1993, Nature, 362, 730

Kaiser, N., et al. 2002, Proc. SPIE, 4836, 154

Kastner, J. H., Zuckerman, B., Weintraub, D. A., &Forveille, T. 1997, Science, 277, 67

Kenyon, S. J., & Luu, J. X. 1999, AJ, 118, 1101

Kurucz, R. L. 1979, ApJs, 40, 1

Leonard, F. C. 1930, Leaflet of the Astronomical So-ciety of the Pacific, 1, 121

Lewis, J. R., Irwin, M. J., Gonzalez-Solares, E. A.,Bunclark, P. S., Hodgkin, S. T., Evans, D. W., &McMahon, R. G. 2006, ASP Conf. Ser. 351: Astro-nomical Data Analysis Software and Systems XV,351, 255

Magnier, E. A., & Cuillandre, J.-C. 2004, PASP, 116,449

Mamajek, E. E., Lawson, W. A., & Feigelson, E. D.1999, Publications of the Astronomical Society ofAustralia, 16, 257

Naoz, S., Noter, S., & Barkana, R. 2006,astro-ph 0604050

Navarro, J. F., & Steinmetz, M. 2000, ApJ, 528, 607

Pel, J. W., & Lub, J. 2006, in The Future of Photomet-ric, Spectrophotometric and Polarmetric Standardi-sation, ed. C. Sterken, Blankenberge, Belgium, ASPConf. Series.

Pietrinferni, A., Cassisi, S., Salaris, M., & Castelli, F.2006, ApJ, 642, 797

Reid, I. N., et al. 1991, PASP, 103, 661

Reif, K., Klink, G., Muller, P., & Poschmann, H. 2004,in Scientific Detectors for Astronomy, The Begin-ning of a New Era, 367

Sebring, T. A., Dunham, E. W., & Millis, R. L. 2004,Proc. SPIE, 5489, 658

Stubbs, C. W., et al. 2006, ArXiv Astrophysics e-prints, arXiv:astro-ph/0609260).

Thuan, T. X., & Gunn, J. E. 1976, PASP, 88, 543

Torres, C. A. O., da Silva, L., Quast, G. R., de la Reza,R., & Jilinski, E. 2000, AJ, 120, 1410

Trujillo, C. A., Jewitt, D. C., & Luu, J. X. 2001, AJ,122, 457

Wilson, G., et al. 2005, ASP Conf. Ser. 347: Astro-nomical Data Analysis Software and Systems XIV,347, 563

Wood, P. R., Rodgers, A. W., & Russell, K. S. 1995,Publications of the Astronomical Society of Aus-tralia, 12, 97

Wyithe, J. S. B., & Loeb, A. 2004, Nature, 427, 815

York, D. G., et al. 2000, AJ, 120, 1579

Young, P. J., Roberts, W. H., & Sebo, K. M. 1999,ASP Conf. Ser. 172: Astronomical Data AnalysisSoftware and Systems VIII, 172, 115

Zacharias, N., Urban, S. E., Zacharias, M. I., Wycoff,G. L., Hall, D. M., Monet, D. G., & Rafferty, T. J.2004, AJ, 127, 3043

Zuckerman, B., & Webb, R. A. 2000, ApJ, 535, 959

Zuckerman, B., & Song, I. 2004, ARA&A, 42, 685