Matthew Colless, Quentin Parker
& the 6dF Science Advisory Group
Version: 30 April 2000
This document describes the 6dF Galaxy Survey, a
redshift and peculiar velocity survey of the southern sky based on the
2MASS and DENIS near-infrared sky surveys. This version has been
prepared for the Schmidt Telescope Panel by the 6dF Science Advisory
Group (6dFSAG), comprising Matthew Colless (Chair, ANU), Quentin Parker
(survey manager, WFAU/IfA), John Huchra (CfA), Ofer Lahav (IoA), John
Lucey (Durham), Gary Mamon (IAP), Will Saunders (IfA), Elaine Sadler
(Sydney) and Fred Watson (instrument scientist, AAO). The Schmidt Telescope Panel saw the setting-up of a
Science Advisory Group as a key element in planning and implementing the
6dF galaxy survey, and in ensuring the fullest scientific exploitation
of the survey by the community. This decision was ratified at the AATB
meeting in September 1998. The terms of reference for the SAG are:
1 Introduction
This document reports the progress of the 6dFSAG towards these
goals.
6dF is an automated fibre positioner for the AAO Schmidt Telescope. The positioner, like the FLAIR facility that it supersedes (and unlike 2dF), operates off-telescope. The positioner uses an r-theta robot with a curved radial arm to place magnetic fibre-buttons on the curved field plates needed to match the Schmidt focal plane. The 150 fibres on each field plate are housed around the perimeter of the field plate assembly. The fibres have a 6.7 arcsec core diameter and can be positioned on target objects to a precision of better than 1 arcsec. Positioning all 150 fibres will take less than 1 hour. The field accessible to the fibres is a 6o diameter circle. There are two field plate assemblies, so that one can be configured while the other is on the telescope. The field plate assemblies are mounted at the focal plane of the telescope using an adaptation of the standard photographic plate mounting system. The change-over to a new field plate and the acquisition of a new field will take no more than half an hour. Further details of the system are given in the Functional Specification for 6dF (Watson et al. 1999).
The optic-fibre cable feeds the existing floor-mounted FISCH spectrograph, which will be upgraded to use a Marconi 1024x1024 CCD detector with 13um pixels. Figure 1 shows the excellent sensitivity of this device.
Figure 1: The quantum efficiency of the new Marconi CCD.
The current schedule has the on-telescope commissioning of 6dF beginning in November 2000 and ending in April 2001. 6dF will be available for shared-risk observing during the commissioning phase, and test observations for the 6dF galaxy survey are planned for this period. The survey proper will begin immediately the instrument commissioning phase is complete, in April 2001.
The 6dF Galaxy Survey is a near-infrared (NIR) selected survey covering the 17,000 deg2 of the southern sky with |b|>10o. The survey has two distinct phases: a redshift survey (the z-survey) and a peculiar velocity survey (the v-survey).
The selection of targets for the redshift survey is determined by three requirements:
With these requirements in mind, the primary targets for the z-survey are chosen to be a magnitude-limited, K-selected sample of galaxies from the 2MASS Extended Source Catalog (XSC) having total magnitudes K<12.8, corresponding approximately to K<13 in the standard 2MASS isophotal magnitudes. The 2MASS star/galaxy separation will be checked against the higher-resolution DENIS I-band imaging to minimise stellar contamination of the sample. The primary target sample has a mean surface density of 5.3 deg-2, based on the 2MASS XSC number counts from the Spring`99 release shown in Figure 2. This corresponds to 90,000 galaxies over the 17,000 deg2 of the survey. The survey will also include galaxies fainter than K=13 in order to make the sample complete down to H=13.5 and J=14.1 (from 2MASS), down to I=15.0 (from DENIS), and (at higher Galactic latitudes) down to about B=16.5 (from the APM galaxy catalogue). These secondary targets amount to a further 1.5 deg-2, or 25,000 galaxies over the whole survey area. In addition we will observe the optically-bright (B<18.5) QSOs and AGN in the ROSAT all-sky survey; these have a surface density of 0.6 deg-2, so add another 10000 targets to our input catalogue.
The full sample is thus 125,000 galaxies and QSOs at a surface density of 7.4 deg-2. About 11,000 of these already have literature redshifts in Huchra's ZCAT, and all 14,000 sample galaxies in the 2,000 deg2 covered by the 2dF survey will also have redshifts. Thus the number of targets for which the 6dF z-survey will need to measure redshifts is approximately 100,000 over approximately 15,000 deg2, at a mean surface density of 6.7 deg-2 or 190 per 6dF field.
Figure 2: JHK galaxy counts from 2MASS.
The survey will use an adaptive tiling scheme similar to that employed by the 2dF survey. This provides complete sky coverage and allows the survey to achieve nearly uniform sampling despite the large fluctuations in number density due to clustering and the loss of close pairs due to the 5 arcmin minimum separation between fibres on the sky. For the z-survey sample, the rms variation in the number of targets per field is found to be 40% (derived from the Spring`99 XSC). This large variation means that tiling is essential to achieve relatively uniform sampling. Tiling also reduces the number of close pairs lost from the sample because of the minimum fibre separation constraint. At the limit of the 6dF survey, approximately 20% of objects would be inaccessible because of this constraint if fields were not overlapped or repeated.
With an adaptive tiling scheme like that used by the 2dF survey we can achieve about 90% sampling completeness (uniform to 5%). This level of completeness requires that we overlap the 6dF fields to the extent that the mean effective field area is reduced to 18 deg2 (this is crudely equivalent to observing half the fields once and half twice). Thus about 850 overlapping fields will be required to cover the 15,000 deg2 of the survey area not covered by the 2dF survey. This corresponds to 100,000/850 = 118 targets per field, which is well-matched to the number of available fibres (since not all fibres can be positioned on targets and some must be used for sky).
Using the 300B grating and the new detector, the 6dF spectra will cover 3500Å (from 3900Å to 7400Å) at a resolution of 3.5Å pixel-1. Based on experience with 2dF, at this resolution we can obtain galaxy redshifts with 50 km/s precision and 95% completeness at a S/N of 5-10 pixel-1. With the new CCD, we expect this S/N can be exceeded easily for objects with B=17, V=16 and R=15.5 in 1 hour integrations in median conditions (note that the minimum integration time on one field plate is set by the time taken to configure the other field plate). We will confirm this performance during the commissioning phase. Allowing for overheads (CCD readout, plate exchange, field acquisition, calibrations) of about 0.5 hour per field, we expect to average 5 fields per clear night.
Figure 3: The photometric properties of the z-survey sample.
Figure 3 gives the photometric properties of the z-survey galaxy sample. The top panel gives the J-K vs K colour-magnitude diagram, and shows the additional objects included by the J<14.3 selection. The middle panel shows the B-K vs K colour-magnitude diagram, and makes clear the very large difference between a K-selected and a B-selected sample. The bottom two panels show the sample distribution in B and R.
Although 20% of the sample are fainter than B=17, only 5% of the sample are fainter than R=15.5. Since the strongest spectral features (Mgb and NaD in absorption, H-alpha and [OIII] in emission) are in the V and R bands, and because the fainter, bluer objects have stronger H-alpha while the redder absorption line objects have higher surface brightnesses, we expect that the overall redshift completeness will be greater than 90%. The expected redshift distribution for the 6dF survey, based on the luminosity function of Loveday (2000, MNRAS, 312, 557), is shown in Figure 4. The median redshift is 0.05, the mean is 0.055, and 90% of the galaxies have z<0.1.
Figure 4: The predicted redshift distribution for the 6dF survey.
The primary observational goal of the second phase of the 6dF survey is to measure peculiar velocities for an all-sky sample of galaxies. Peculiar velocities for early-type galaxies will be obtained via the Dn-sigma relation using the photometry from 2MASS and DENIS to give diameters (Dn) and the v-survey spectroscopy to give velocity dispersions (sigma). The target sample will consist of a subset of the early-type galaxies with cz<15,000 km/s identified in the z-survey. Since 50% the objects in the z-survey sample will have redshifts less than 15,000 km/s (see Figure 4), and since preliminary analysis of the 2MASS and DENIS images finds that at least 60% are bulge-dominated, we estimate that there are approximately 30,000 potential targets for the v-survey (i.e. about 1.8 deg-2).
The optimal sampling of these objects for the v-survey is still being studied, however the approach envisaged is to observe that half of the survey fields having surface densities above the median. Reducing the sky coverage by a factor of two in this way allows the v-survey to be completed in half the time, while still adequately sampling the peculiar velocity field over the volume out to 15,000 km/s. Concentrating on the richer groups and clusters also allows the survey to take better advantage of the reduced distance errors obtained by grouping objects. Given the clustering of the sample, this strategy yields approximately 70 galaxies per 6dF field in the mean. Although this number is reduced because of the fibre-separation constraint, we estimate that at least 50 galaxies per 6dF field could still be observed, giving 15,000 galaxies in the 300 richest fields over the southern hemisphere.
Using the 1200V grating and the new detector, the 6dF spectra will cover the strong H-beta (4860Å), Mgb (5174Å) and Fe (5207Å, 5270Å) absorption features over the redshift range of interest at a resolution of approximately 0.9Å pixel-1. With the detector upgrade, we estimate that spectra with S/N>20 pixel-1 can be obtained down the 6.7 arcsec fibres in total exposure times of 3-4 hours. This resolution and S/N will give central velocity dispersions with a precision of at least 10% for galaxies with velocity dispersions greater than 50km/s. We expect to be able to observe 2 v-survey fields per clear night.
The photometry with which the velocity dispersions will be combined to obtain Dn-sigma distance estimates will come directly from the 2MASS and DENIS surveys from which the target samples are drawn. At the v-survey limit, the 2MASS and DENIS surveys give total magnitudes to better than 0.1 mag and effective radii with a precision of approximately 10%. Velocity dispersion uncertainties of 10% translate (for a Dn-sigma relation slope of 1.2) to an error in the predicted Dn of 12%. Since the measurement error in Dn is always significantly less than the measurement error in the effective radius, the errors in the velocity dispersions will dominate the measurement error contributions to the uncertainties in the Dn-sigma distances. The 2MASS and DENIS photometry is therefore sufficiently accurate for measuring Dn-sigma distances.
There are opportunities to observe additional targets in both the z- and v-surveys, since not all fibres will be used in all fields.
The z-survey will use almost all 150 fibres in most fields. However there will usually be a few fibres which cannot be allocated to z-survey targets, and in low-density fields there may be a few tens of unused fibres. The broad wavelength range of the z-survey spectra makes them useful for observing a range of other interesting sources, provided the sources are bright enough for useful spectra to be obtained in the 1 hour integration time. Examples of potential targets include IRAS Faint Source Survey galaxies and NVSS/SUMSS radio sources.
The v-survey will not use all 150 fibres even in the most clustered fields. However the limited spectral range of the observations means there are fewer types of additional sources worth observing, despite the 3-4 hour integration times. Probably the most effective use of the remaining fibres is to extend the v-survey to fainter sources, which will yield larger and deeper redshift samples in these highly-clustered fields. About 20,000 additional targets could be surveyed in the course of the v-survey.
It is highly desirable to extend the 6dF survey to the whole sky in order to allow mapping of the density and velocity fields for the whole local volume of the universe. A 2MASS-based redshift survey in the northern hemisphere to K<12.2 has already begun under John Huchra. This survey consists of a compilation of redshifts from previous surveys and the literature that is being supplemented by observations from the Mt Hopkins 1.5m telescope and the Sloan redshift survey. However there is no plan or capability for extending this to a peculiar velocity survey.
We are currently investigating the possibility of placing a clone of 6dF at one of the northern-hemisphere Schmidts at Palomar and Kiso. Given that a northern z-survey is already being done (albeit to a brighter limit), the northern 6dF could perhaps concentrate solely on the v-survey, and so could be carried out on the same time-scale as the southern 6dF v-survey.
The main scientific goals of the z-survey are:
The major advantage of this survey over all preceding redshift surveys is the near-infrared (NIR) selection of the target sample. NIR selection is based on the luminosity of the old stellar population, which means that the luminosity is integrated over the galaxy's star-formation history and is therefore the most direct measure of the stellar mass. Thus, compared to surveys selected in the optical or far infrared, NIR selection avoids over-weighting those galaxies with high current star-formation rates (SFRs).
Furthermore, by using the integrated light of the old stellar population, the galaxy sample is not dominated by details of the recent star-formation history, and so permits much more reliable comparisons with models of galaxy formation and the evolution of structure. Because the survey sample is large and covers a representative volume of the universe, we can fully characterise the galaxy population by determining the various conditional distribution functions: the luminosity function for each galaxy type, N(L|T); the luminosity function variation with environment, N(L|E); the morphology--density relation, N(T|E); and so on. We can also establish the large- and small-scale structure from the mass-weighted galaxy power spectrum, as opposed to the SFR-weighted power spectrum, and so obtain a direct comparison to N-body models for the formation of structure.
By observing objects selected from the ROSAT all-sky survey, we will be able to generate a very large sample of optically-bright QSOs and AGN. This sample will complement existing UVX-selected samples and would be about 100 times larger than the catalogue of bright QSOs by Schmidt & Green (1983). It will be used to construct the joint optical-X-ray luminosity function for AGN, and to study its evolution with redshift. It will also be used for QSO clustering studies (complementing the 2dF QSO redshift survey), and as a source list for absorption-line studies of bright QSOs.
The galaxies in the 6dF z-survey can cross-identified with the radio sources from the NVSS/SUMSS all-sky survey to give redshift and spectroscopic data for an expected 8000 radio galaxies. Together with the 4000 radio galaxies identified in the 2dF galaxy redshift survey, this will give an increase of more than an order of magnitude in the number of low-redshift radio source identifications. Since the 6dF galaxies are bright enough for morphological classifications, the z-survey will yield, for both AGN and star-forming galaxies, by far the best determination of the local radio luminosity function as a function of Hubble type, and a proper benchmark for measuring the evolution of these radio-source populations.
The 6dF survey complements the 2dF survey in the primary sample selection criteria (NIR vs optical) and covers a comparable volume (10x the sky coverage over the volume interior to the median redshift of the 2dF survey). Compared to the IRAS PSCz survey, the 6dF survey has very different sample selection (NIR vs FIR), greater depth and much denser sampling (more than 5x larger sample). The outstanding advantage, however, is that NIR sample selection means 6dF will yield the first survey of large-scale structure that is not skewed by current star-formation. Additionally, the redshift survey provides the crucial volume-limited sample of early-type galaxies for the peculiar velocity survey.
The extra parameter provided by the v-survey is the galaxy's internal velocity dispersion, which is a dynamical measure of the galaxy's mass that we can compare to the measure obtained from the NIR luminosity. Thus we will be able to determine the mass distribution as a function of type and environment, N(M|T,E), and likewise the mass-to-light ratio distribution, N(M/L|T,E). Despite their fundamental nature, these distribution functions have not yet been determined for representative samples of galaxies.
As well as masses, measurements of the internal velocities also yield distances (D) to a precision of at least 20% from the Fundamental Plane or Dn-sigma relations for early-type galaxies. For galaxies at the outer limit of the survey (cz=15000 km/s), this results in peculiar velocities, v=cz-D, with a precision of only 3000 km/s for individual galaxies. However our very large sample allows us to group galaxies together, and so determine the local velocity field to higher precision. For example, by grouping galaxies into 100 independent redshift shells of thickness 1.5 h-1 Mpc, we can measure the velocity of each shell to a precision of better than 150 km/s. This will allow us to track the convergence of the mean bulk motion of the local universe to the CMB frame out to a scale of 150 h-1 Mpc with a velocity resolution of 150 km/s and a scale resolution of 1.5 h-1 Mpc. This will reveal the precise nature of the mass concentrations that are the proximate cause of the motion of the Local Group with respect to the CMB. Splitting each shell into 10 subsets, we can measure higher-order moments of the velocity field from 1000 independent cells, each having a peculiar velocity determined to better than 500 km/s.
The volume probed by the 6dF survey is large enough to provide, for the first time, a fair sample of the velocity field in the universe on all scales of interest. Note that the velocity field is a very direct constraint on the mass distribution, since (i) it constrains the mass power spectrum as opposed to the galaxy power spectrum, and (ii) it is less affected by non-linearities than the density field. The velocity field also is more responsive to long-wavelength modes, and so constrains the power spectrum on larger scales than the density field.
Figure 5 shows the precision with which the 6dF survey will be able to determine the bulk motion on different scales, and hence constrain cosmological models. The black curve shows the mean bulk motion in spheres of a given radius that is predicted for the current `consensus' Lambda-CDM model, while the coloured curves correspond to a range of plausible cosmological models consistent with existing observations. The grey region shows the 90% confidence level constraints obtained from the 6dF v-survey assuming the consensus cosmology. World-models outside this region are strongly rejected. By comparison, almost all the current generation of peculiar velocity observations are consistent with all the alternative models shown in the figure.
Comparison of the velocity and density fields from this NIR-selected survey provides a measure of Omega0.6/b which is less affected by biasing than estimates from optical/FIR samples. Moreover with our large sample we can determine relative biasing between different morphological types and biasing as a function of local environment. The dense sampling of the 6dF survey also opens up the possibility of studying non-linear and stochastic biasing models.
Figure 5: Constraints on cosmological models from the v-survey.
The 6dF survey is unique in several respects when compared to other existing or planned surveys:
The 6dFSAG believes the 6dF galaxy survey will have a significantly stronger scientific impact if both the redshift and peculiar velocity phases are completed as rapidly as possible. The management model for the AAO Schmidt approved at the most recent AAT Board meeting envisages continuing operations at the level of 18 dark/grey nights per lunation for all 13 lunations each year (i.e. 234 nights per year), with 6dF available in every lunation. In order to complete both phases of the galaxy survey in a timely fashion, the 6dFSAG advocates that 75% of the available time on the AAO Schmidt (i.e. 175 dark/grey nights per year) be dedicated to the survey. Although this commits a large fraction of the available time, it is important to remember that all current photographic surveys are expected to be complete by the end of 2001. Moreover, non-survey observations (both common-user 6dF observations and photographic or CCD imaging) could still be carried out on the Schmidt at the level of 59 nights per year. Over-rides (e.g. observations requiring the best seeing conditions) would also be possible if carried out at a low level.
With survey observations on 175 nights per year, and assuming that 50% of the time is spectroscopically usable (the fraction found for the 2dF survey), the durations of the two phases of the survey are as follows:
Thus if survey observations are supported for 175 nights per year, we can complete both phases of the survey in under 4 years. With survey observations anticipated to commence from April 2001, an allocation of 175 nights per year means that the durations of the two phases of the survey would be:
The 2MASS near-infrared sky surveys, on which target selection for the 6dF galaxy survey will primarily be based, is currently scheduled to have completed the southern sky by early 2002. The DENIS will finish, with approximately 93% of the southern sky covered, at the end of 2000. Both surveys are approaching completion, with 2MASS having covered 91% of the sky, and DENIS 77% (see Figure 6). We are already using the publicly available 2MASS data to refine our selection criteria and experiment with using total rather than isophotal magnitudes. There is no problem with selecting targets in time for the start of the redshift survey in April 2001.
The 6dF survey time-line proposed above is highly competitive with other large redshift and velocity surveys, especially given the unique NIR selection and sky coverage of the 6dF survey. It may also be a good scientific strategy to begin the v-survey before completing the whole of the z-survey. For example, it would be interesting to first carry out both redshift and velocity observations for all the rich clusters within 15,000 km/s.
Figure 6: 2MASS and DENIS sky coverage as of April 2000.
Three categories of staff are required to carry out this survey:
Although it is difficult to predict the level of staffing needed to effectively perform the survey observations, the 6dFSAG expects that a minimum of three on-site observers will be required. The 6dFSAG strongly recommends that the survey manager should also be on-site during the commissioning phase and the initial survey observations. This recommendation is based on the experience of the 2dF survey, where the presence of key members of the survey team was essential in the early phases of the project. Once survey observations become routine, it would then be appropriate to review the on-site operational requirements.
While there will be opportunities for students to be involved in the survey, it would not be appropriate for them to carry out observing and data reduction duties at the level required of a survey observer or postdoctoral assistant.
The survey will be carried out under the auspices of the AAO and the WFAU at Edinburgh, with the 6dFSAG providing scientific advice and oversight. The WFAU will support the survey manager, the AAO will support the survey observers (and the instrument), and the 6dFSAG will work with the AAO and the WFAU to find funding for postdoctoral assistants.
The data from the 6dF galaxy survey will be reduced using software based on the successful 2dF data reduction pipeline. The required modifications to the 2dF software will be made at the AAO. The 6dFSAG will be responsible for providing software for the determination of redshifts and spectral types (using modified versions of the software used for the 2dF redshift survey), and for automating existing codes for measuring velocity dispersions and line indices so that they can deal with the large data-flow from the peculiar velocity survey.
The survey products will be:
The survey will be non-proprietary, and the 6dFSAG envisages that the products will be made public as soon as is consistent with quality assurance. Quality assurance will be the responsibility of the 6dFSAG and the survey manager. We propose that preliminary subsets of the data publicly available be made available during the course of the observations for both surveys. The final catalogue and spectral database for the z-survey would be released within 6 months of completing the observations; for the v-survey (where considerable additional analysis is required to derive the galaxy distances and peculiar velocities) the final release would occur within 12 months of completing the observations. The data will be made available on the WWW by the AAO and WFAU, with mirror sites at other locations as appropriate. If there is the demand and it is economical to do so, the data could also be made available in some hardware form such as CDROM.
