Galaxies & Cosmology

Topic group at UKATC workshop on Science with ELTs 25-26 September 2000

Chaired by Bruno Leibundgut, notes taken by Bob Mann

These notes are extremely wideranging and are presented with a view to acting as an aide-memoir and guide to relevant sub-topics which need to be included in the discussions at Leiden. (Chairpeople and sub-chairs please note.)

Present were : R. Abraham, E. Barton, A. Blain, R. Carlberg, C. Cesarsky, M. Franx, R. Gilmozzi, I. Hook, O. Le Fevre, B. Leibundgut, R. Mann, J. Miller, G. Monnet, S. Morris, B. Peterson, M.Postman, R. Rebolo, A. Renzini, J. Rodriguez-Espinosa, R. Sancisi, D. Schade, W. Sutherland, A. Taylor, E. van Kampen, S. White.

Day 1.   25 September 2000

Bruno Leibundgut set out some initial issues:

Sancisi: To what z can we reach with Cepheids for each diameter ELT?
Renzini: Thats a suggestion for science goal determine H(z)!
Carlberg:  We want to know about evolution of cosmological properties with z.
Sancisi:  Where is the Dark Matter  at z=5?...
White: we know there is DM at z=5?
Miller:  What size of telescope do you need to do it?..and how do you do it?
White :  Let's start by identifying main problems
Sancisi: How much DM at each redshift (1-5) and how is it distributed?
White : Deceleration over 1z5? * assembly paths for galaxies
Franx: History of Milky Way-type galaxies
Carlberg: Enrichement history of Universe.
White:  -  Where different elements come from
               -  Reionisation history of the Universe
               -  Structure of the IGM
Hudson:  Connection bewteen black holes and AGN
Leibungut:  Gamma ray bursters (GRBs)?
Blain:  High-z supernovae?...  (Monnet:... - address both their properties and their use as a tool)
Carlberg:  Globular cluster formation at z=3-5... (Renzini:...- NGST sees them to z=6-7)
White:  Star formation modes as a function of z - variation in IMF, etc
Le Fevre :  The growth of Large-Scale Structure
Postman:  There are two types of problems: -  those issues we don't know the answer to and
                      -  those where a bigger telescope makes things practical by reducing time needed
Hudson:  Maybe only brightest objects will have been seen to high-z by 10m class
Renzini:   - Which of these will not be solved in the next ten years?
                    (e.g., FLAMES will give 10^5 stellar spectra with  resolution needed to reveal MW history).
Morris:   These are qualitative questions...error bars will shrink - we need to quantify what level of solution we'll get Renzini:  Cepheids to z=1 would give evolution of cosmological parameters with want primary distance
                   indicators to z=1 - that would be really new!
Postman:  Can get 200 km/s proper motions at 1 Mpc distance in 3 years with 10 microarcsec positions...
(many):     ...- but Gaia will do that?
Sancisi: Go through list and see what will be answered within ten years
Le Fevre: Also want to look at first stars...  (Monnet: ...that's reionisation history..)
Sancisi:  How do we know where DM is at low z?....dynamics, lensing, X-rays
Renzini:  Inefficient to use lensing at high z, due to lack of very high-z background galaxies to be lensed
Sancisi:  -  Disk dynamics impossible, as no disks(?) at high z
                 -  X-rays fail, as no clusters
                 -  So have to use velocity dispersions of ellipticals to show presence of DM
Renzini:  Steidel spectra can't give vel dispersion of structure at z=3, but will soon be feasible
Franx:  Many emission lines are from winds, so maybe don't trace stellar mass well
Carlberg:  Emission, not absorption, lines only measured with Keck for high z cluster gakaxies so far
Renzini:  Rare species missed at high z so far
White:  Don't have good IR spectra for any - too hard
Renzini:  Sample size is the problem - bigger sample, get brighter objects, and can take spectra
Barton:  Can get Fundamental Plane to z=2.5 for bright gals, 10 hr exposure on 20m (overheads):
                 -  assume de Vauc profiles and 5100A rest-frame festures to get FP...
(Franx : ...gals at z=2.5 may not have elliptical galaxy spectra..  and Balmer lines are rotationally broadened
                   and they may be all we can get at 2.5)
                 -  Observed-I TF to z=1.5 or so for typical galaxy (L*)...
(Monnet: use IFU, rather than slit to get further gain...Barton agrees)
                 -  Halpha S/N as func of z - for use in kinematic studies....for unresolved galaxies -
                      results depend on how clumpy SF is in galaxies.. use OII until it leaves optical, and then
                      switch to Halpha. At z>5, can only see extremely strong clumpy SF regions
Sancisi:  This probes TF, not DM directly * what about elliptical galaxy kinematics?
White:  -  We want line profiles of z=3 galaxies in a number of IR lines at R=10000 - that's an ELT problem!
               -  Have to be able to separate populations in galaxies, so need to have S/N to distinguish lines from
                    different stellar populations, and measure their kinematics.
Carlberg:  So we need to be able to resolve 10pc regions at z=3 and get spectra for these resolution elements... White/Franx: We need more photons to do that, so want larger regions
Monnet:  Trade-off between spatial resolution and photon numbers
Sancisi:  Only need 1kpc - not 10pc
Carlberg:  r_e=0.5 kpc for LBGs, so need smaller scales for disks
Sancisi:  DLAs - need lots of photons for kinematics...
Renzini:  ...but they only absorb, so don't make many photons - VLT is trying to image absorbers in emission
White:  We haven't seen the emission from all the absorbers in the HDF, in front of QSOs
Leibundgut: Need to know how to connect DLAs to galaxies....this is structure of the IGM in the list above
Morris: Can't lens at z=5, but can it be done at z=2?
Taylor:  Depends on morphology - if fragments, not relaxed galaxies, then problems with intrinsic ellipticity...
(many):  ...always problems with intrinsic ellipticity
Monnet: We want to study lensing near the diffraction limit
Postman:  In a 2-3 arcmin field, can get background to K=28 or so
Monnet:  2010 - should be 10^9 pixels in WF detectors
Carlberg: By then also, there will be SZ detections, so can look for dark clusters - there will be some
                   genuine ones in due course...aren't yet.
Renzini:   Hypothetical objects - how do you assemble so much mass without triggering SF?
Carlberg:  We want to measure masses for z=5 clusters detected in SZ
Franx : We want to get spectra of EROs... (many) * this sufficiently different from what can be done with a 10m? Renzini: What are the limits within the next 10 years of what can be done, in the way of detecting cB58-type
                 (lensed) objects?   - cluster lenses will find many of these.
               - Lensing has different effects on point  sources, e.g. Type Ia SNe at high z....
               - We want to compare Cepheids and Type Ia SNe at z=1. This is more new than internal dynamics
                   of galaxies - this is a better application for an ELT.
Carlberg:  Track time with redshift using cosmochronology
Renzini:  Would be very model dependent - Thorium lines are tiny features in the spectrum of a red giant,
                   so can be swamped by internal motions
Carlberg:  So use globular clusters
White:  First objects issue....try to find objects at z=10 by high-res spec of blank fields to look for Lyman
               alpha emission between OH lines. NGST won't have high resolution, so better to do this from the ground.
Carlberg:  This is HDF in spectral space, but problem is that fields are small, so might not get anything
White:  IFU gives wide field, so should get something. * maybe line widths smaller at very high z, which helps
Le Fevre:  Use narrow line filter, not IFU - more efficient... (White:... OH line too close, so can't fit in filter...)
Le Fevre:  Can get filters to fit.
Leibundgut: Maybe we need to know more about what you can do with a 10m before we can assess how
                         feasible this is.
Carlberg: We want a spectrum for every pixel in HDF...Keck doesn't get anything in most pixels - so there
                  is missing light there.
White:  We have examples of objects to z=4, so can predict what we should see at z=10 * there are fossils of
                reionisation history in absorption spectra of QSOs, due to ISM of absorbers.
Miller:  Are quasars important? - in gal formation, and as tracers of interesting locations....10^4 will be found
               in surveys, so follow up with ELT to see SF properties of their envrionments.
Leibundgut:  z limit for Sloan is about 6.
Sutherland:  VISTA will push to z=7-8, by going into IR and having bigger aperture
Monnet:  This is tough - needs high spatial resolution and it is a good ELT project
White:  May have seen most of the baryons at z=2-3...  (Renzini:... may have seen only 10%...)
    get all the baryons seen in the Lyman alpha clouds we must have seen all the nucleosynthesis baryons
Renzini: But metal abundance there is too low, so the metals must be in a different phase - 20-30% of stars
                 were made before z=3, so expect metallicity to be ~1/3 z=0 where are the metals? - not in
                Lyman alpha forest or DLAs, so must be yet another phase, presumably hotter....and that's never been
                seen at low where is it?
White:  Depends on low-z Lyman forest, for which the there's only sparse data...but take general point.
Renzini:  X-ray telescopes (e.g. Xeus) are designed to find Gunn-Peterson throat in metals....maybe not best
                 done in the optical
Leibundgut: Deceleration to z=5...need IR to do Type Ia supernovae - good case for NGST.
White: Does it have right spectral resolution?....
Leibundgut: photometry from space and low-res spectroscopy
Postman: Why is this an interesting problem, given the CMB?
White:  CMB gives geometry, not equation of state, which is what deceleration parameter can give you.
Leibundgut:  Papers suggest degeneracies prevent extraction of deceleration parameter

Postman:  Suggest we may want to split into smaller groups for second day, to assess needs for
                     observations in each heading.

Day 2.  26 September 2000

Leibundgut: Summary of Day One.

.. Now occurred a shift of format, with different people leading discussion of main topics.

White:  First Light

                -  We need emission and absorption objects
                -  We don't know how small first light objects will be: the medium around them won't be ionised,
                     but we should soon see Lyman alpha emission - only spectral z=9, the jump will
                     be in the J band.

Peterson:  Want J and H survey to find them, then point ELT at them to study them.
White:  - Need to get quantitative feel for just how faint these objects will be, and how large a telescope we need.
               - If lines are narrow, then ELT beats NGST, which has max resolution of about 3000.
               - Evaluate two cases: (i) line emission from these objects; (ii) continuum step
Renzini:  GRB hypernovae as sites for early star formation?
               - maybe next generation of gamma ray telescopes will do that science... (White:  ...needs good
                  follow-up coverage - wide area, then deep spectroscopy)
              - SWIFT - immediate release of alerts once bursts seen
White:  Use absorption spectrum of material in front of burst
Peterson: Also want to know about host galaxies - but no idea yet what they are
Renzini:  Before ELT, should be hundreds of z>6 GRB IDs - do we need ELT to see the host object.
Postman:  Hosts have been studied with HST
Renzini:  Hypernovae (50-60 M_solar), will be seen in IR and are good tracers of early star formation.
White:  From known objects, can guess what magnitudes will be
Renzini:  Several satellites and ground-based follow-up programes will work in next decade, yielding many
                  high-z (z=6-10) IDs...and ELT will be required to image the host galaxies.....GR satellite should
                  give positions to a few arcsec Leibundgut * can the time dilation in the decay be used to estimate z?... (many):   .....isn't dispersion in intrinsic time delay too great?
White: - Absorption line spectroscopy can be done with a 10m, provided that you catch the burst quickly
                 enough - but host galaxy imaging needs ELT
             -  What is special about environments of GRBs, as compared to other SF regions?....and we should
                 answer that at lower redshift in next 10 years.
            -  Other method is high resolution absorption spectra along QSO LOS
Postman:  Sloan should get about 100 z>6 QSOs at mags of z'=20
White:  Need ELT to get high resolution spectra to detect fossils of processes out to z=20
Hudson:  Want multiple skewers through structures to get spatial structure
White:  ..... and that requires lower S/N spectra - so does that need an ELT??
Peterson:  But lines will blend if resolution low
White:  -  Statistical measures only - don't need to identify all clouds
               - Very high resolution needed for metal lines in walls of bubbles blown out by forming galaxies.
Morris:  Arjun Dey has some of the required numbers already - for case of 30m.
Peterson:  If z=10 objects have Lyman alpha lines, then profiles should show asymmetry due to absorption
                    in HII regions surrounding emission regions

Franx: Assembly of galaxies and evolution of MW-like objects

               - What happens to LBGs? - may be a question for a lot of 10m tine....these end up in clusters,
                   they're not normal spirals - are they related to them?
               - Want to measure LSS of LBGs - be done in the next 10 years
               - Progenitors of normal galaxies have R=27-28 at z=3 (Steinmetz prediction): so would then want
                   to study their clustering.

Hudson:  Would also want pairwise velocities from spectra to get some mass info.
Franx:  Need simulations of surface brightness of these objects to assess requirements
White:  Want LBG spectra in IR, to get rest frame optical, as well as UV, as now
Le Fevre:  Also want to look in near/mid-IR to get census of galaxies including the dusty ones
Franx:  Maybe submm-selection with ALMA would be best
Blain:  ALMA gets CO lines (and redshifts) for most galaxies, but not all
Cesarsky:  FIR-selection would be best, but no instrument expected
Renzini:  -  Want to know what LBGs look like, at level of 10^4 pixels per galaxy
                 -   Latest "Madau plot" claims to have seen whole SF history, so no great hidden population
                       of Star-forming galaxies at high-z
White:  Also want spatially-resolved spectrum to get kinematics
Postman:  Not easy for even 100m telescope, based on HDF population
Sancisi:  What is known of spectra of LBGs now?
Franx:  Major question is whether there can be a good mass estimator
Monnet: Difficult if mergers go on....need bound systems
Franx:  ALMA may be major competition - get kinematics (and masses) from CO lines
Renzini:  - Really want stellar absorption lines, not gas absorption, which could be from gas flows...can also
                     get rotation curves
                 - Don't know how ALMA can get 5 mas - should be 50mas?
Le Fevre:  Also want census of sources, to get LF, etc
Renzini:  Need to prioritise what is the most important thing - look at LBGs in detail, or extend their LF
                  by a couple of magnitudes?
Carlberg:  ALMA gives SF-selection, so need to complement with selection by old stellar pops, and probe
                   mass and Fundamental Plane, etc....and that will need an ELT, rather than a 10m.
Franx:  Neglected physics of LBGs - know this for local galaxies, but have little idea at high-z.
Peterson:  Want to follow stellar abundances and populations as a function of z - e.g. tracking possible
                    variation of the IMF.
White:  Keck does better on cB58 than ELT will do on unlensed LBGs, which must make us cautious
               about what can be done.
Franx:  Missed properties of halos in which galaxies sit - but that's DM, as discussed yesterday
Sancisi: Out to ow far can 10m telescopes get Fundamental Plane? (Franx: z=1 or so ..)..
Renzini:  No Fundamental Plane for star-forming gals, so when most galaxies are making stars
                  the FP will vanish
White:  Presumably the older population is lurking underneath, just hidden by younger stars
Renzini:  Could be 1-1.5 GYr worth of older stars - and SIRTF will be very important in looking
                  for these stars, by rest frame NIR
Franx:  Few non-SF galaxies found at z=3 (White: some radio galaxies only), which must limit the star
               formation happening at z=10, suggesting there's not one burst at z=10, followed by passive evolution
Leibundgut: Our basic task is to see how popn we know at z=3 ties in to z=0 popn.
White:  .... that will need stellar masses.

Le Fevre: Growth of Structure

                        -  How do we form the clusters we see today? - e.g. Coma - we have little knowledge
                             above z=1...some info at z=3, but little more.
                        -  Problem is we see only brightest galaxies
                        -  Need significant volumes to be probed to get real census of galaxy population and structure

White: What is the scientific goal here? - what have we learnt in the 20 years since Peebles started
               measuring clustering in Zwicky catalogue?
Postman:  Learning about bias, and nature of DM from evolution of redshift
Carlberg:  Small separation correlation function gives merging - that's physics, not just statistics
Sancisi:  - What is the prediction at high-z?
                - are there telling obsevations to be made?
                - what if clustering is seen at level not expected?
Le Fevre:  First step is to measure clustering, then interpret
Hudson:  Want to study morphology-density relation, as probe of galaxy evolution
Carlberg:  Not clear that we have a good theory yet - semi-analytical models will be ruled out by current data
White: Relationship between galaxy and mass distributions will be more complicated at z=3 than now,
              so comparing clustering of different classes of galaxy at z=3 will yield info about galaxy
              formation and evolution
Renzini: But this can be done with 10m? - large numbers of galaxies, if tip of iceberg, as far as LF
                 is concerned
White:  Need census of different types of galaxy
Postman:  We got it wrong at low-z, when used only bright galaxies in the 80s - needed to go deeper
                   to see full richness of structure
Le Fevre:  Weak lensing is only way to link galaxy and mass distribution fully
White: Cross-correlation of gals and absorption line spectra - assuming abs line popn is closer to
              mass than galaxies
Le Fevre: Evolution of bias?
Postman:  Higher order clustering possible with enough galaxies, and predictions strong at high z,
                    so gives powerful handle on bias
Renzini:  What will be done by 2010?
Postman:  Will be done out to z=1 or so
White:  Better statistics on populations we know now, but little on other types of galaxy that we
               can't get spectra for now
Renzini:  -  99% of mass locally is in 1% of galaxies, so massive galaxies must be the
                      most interesting to study
                 -  By z=1, we can assess whether the same holds, using 10m telescopes
                 -  But at what point do we move to a regime where mass is dominated by the dwarfs

Schade:  If we know the age distribution of stars in galaxies, then we can assess where these stars are
                 (when they are formed?: Ed) as a function of redshift
Morris:  Renzini's argument would lead one to study only cDs locally
White:  No, space density of LBGs nearer to that of L* galaxies than Abell clusters
White: Only way to test theories is to look at predicted properties of different classes of galaxies
Le Fevre: Assembly of clusters
White:  Want to find clusters out to z=2 - seeing ROSAT-like clusters there would rule out Gaussian picture
Hudson:  Also want to look at groups, if there aren't clusters at z>2
White:  Start with X-rays, then SZ and NIR follow-up
Renzini:  Wide NIR surveys should find clusters to z=1.5 or so
Le Fevre:  Work looking around radio galaxies - suggestion to date is that we're nearly finding z=2-3 clusters
Renzini:  Want internal velocity dispersions of z=3 structures, to see if these are bound structures
White:  LBG distribution doesn't really show clusters, as such, just clustering
Le Fevre:  -  Rich protoclusters will be rare - 1/ sq deg - so needs a lot of work
                    -  Also get pointers from weak lensing maps
Franx:  Weak lensing needs understanding of population between observer and lens
Leibundgut:  Is it interesting to look at merger rates?
Carlberg:  It's the link between LSS and galaxy evolution
White:  That needs good imaging for all redshift range of interest

Postman: Galaxies and black holes:

        -  BH mass correlates with central velocity dispersion of bulge, and size of bulge (with slightly more
             scatter) galaxy and BH know about each other.
        -  This is on the basis of about 30 galaxies
        -  Does BH have some role in formation of galaxy?....does gal/BH relation hold to high redshift?
        -  10mas=70pc at z=1 or greater - need the best resolution, to get as close to the blackhole as possible.

White:  Really needed HST - ground-based mass estimates weren't accurate enough, as resolution
               not good enough.
Postman:  Limit is 10^8 M_solar at, an ELT can go to lower masses or get
                   10^8 M_solar BH at z=0.1.
Renzini:  Goal is to determine  m_BH-sigma relation out to z=1?...  (Postman: yes)
               -  but we suppose that little BH evolution occurred at at z<1, as AGN activity is low
White:  -  But not clear that all ways of accreting mass will give light in the similar way.
              -  Want to test model predictions of build-up of bulge mass since z=1.
Renzini:  Of all known AGN, only one has a bona fide binary BH...wouldn't you expect more with
                 lots of merging?
Franx:  Presence of gas helps merging of black holes..
Renzini:  ... not clear by how much
Postman:  Should be able to go out to z=2 relatively easily...
Monnet:   ... but would have to go to the NIR, which would make things harder
Franx :  Could use emission lines, which would push to higher redshifts.
White:  This is concentrating on hard aspects of AGN to observe - isn't there an easier
               property of AGN we can use? Sancisi:  Is this a case for 100m, versus 30m?
(many):  Yes
Leibundgut: What's the catch-phrase? - growth of BH?
Postman: More than that - could be key to important phase of galaxy evolution
Sancisi:  10m telescopes haven't addressed this yet - will they cover it once they have adaptive optics? hard to make case for 100m here...
Monnet:  AO on 8-10m work will only start in earnest in 2-3 years
White:  Could well be AO+10m project initially?
Leibundgut:  How do BH form in the first place? - in detail....when should this happen?
Renzini:  We see quasars out to z=6...
White:   - Bulge-BH relation suggests bulge and BH formation are linked intimately.
               - Need to do calculation to assess quantitative feasibility for extending relation to z=1-2.
Blain:  Would an extra factor of ten in resolution help with general picture of AGN?...
Schade:  ...maybe interferometry, since the nucleus is bright
White/Carlberg:  Probe unified models, by resolving broad-line region, etc
White:  -  Can use narrow filter on bright lines
              -  But maybe VLTI will get there before an ELT.....but an easy case to make for a science driver

Summarising and preparation for afternoon presentation.

Leibundgut:  Need list of high-priority things to do (as at start of day)...and need to cut list down
                          to highest priorities...
Sancisi:  ...and start to address what diameter is needed for each question
Renzini: Good to measure masses of globular cluster populations of galaxies as a function of
                 redshift - to be used as a tracer of the build-up of the mass in galaxies....the fact they
                 are pointlike maximises the advantage from a large telescope

Top questions:

Carlberg: Two broad categories: evolution of smooth component (e.g. evolution of cosmological parameters), and evolution of structure (inc. galaxies).
Renzini: Cepheids to z=0.8 can calibrate Type Ia SNe, which can be used to higher redshift
Leibundgut:  Is this a science driver for a billion euro project?
Carlberg:  Suggest cosmological parameter evolution may be better probed by geometry of 2D redshift-space correlation function
White:  But isn't this a 10m project? - extending Steidel et al programme
Leibundgut:  ... and from Planck, SNAP, etc
                        -  Connect growth of structure, assembly of galaxies, and structure of IGM....
                        -  But aren't all the questions identified connected to each other, at some level?
Postman:  NGST will say when galaxies form, ELT will say how - that's the soundbite
                   -  need to give detailed picture of how galaxies formed

Leibundgut:  What observations contribute to answering that question?
White:  Other top-level things, before breaking down that one (many)
                - First objects/light
White:  First light - the beginnings of galaxies
Morris:  Evolution of cosmological parameters
Renzini:  What does it add knowing parameters at z=1, etc, above and beyond knowing what it is now?
Carlberg/White:  Quintessence - probing equation of state of Universe....constraining fundamental physics

Hierarchical Summary:

Detailed picture of how galaxies form

Emergence of large-scale structure First light/objects - the beginnings of galaxies (z~10) Evolution of cosmological parameters