Paresh Prema, IoA, Cambridge, UK

1
Galaxy Formation and Evolution using
Multi-Wavelength Multi-Resolution Imaging Data in
the Virtual Observatory
SED Service Matching Models to Observed
Spectral Energy Distributions (SEDs)

• Paresh Prema
• Nicholas A. Walton
• Richard G. McMahon

2
Outline

• Scientific Context
• Technique Description and Implementation Issues
• Study Results

3
Scientific Context

• How galaxies formed and how their stellar
  populations have evolved
• - Study competing theories on formation
  hierarchical or monolithic
• Study early phases of galaxy formation and
  evolution through high-z galaxies
• Large multi-wavelength data sets now available
• (e.g. SDSS/GOODS/UKIDSS/SXDS/SWIRE/COSMOS/
• GEMS/UDF)
• Estimate star formation histories (SFH), stellar
  masses, ages, star formation rates (SFR) through
  study of stellar populations in galaxies

4
Technique Overview Spectral Energy Distribution
(SED) Service

• Data Discovery
• Create Object Catalogues Object Photometry with
  Upper Limits
• Cross-match Catalogues
• Calculate Photometric Redshift
• Sample Selection
• Population Synthesis Model Generation
• Model Fitting Best Fit and Parameter Estimation
• Outputs

5
SED Workflow

• Data Discovery
• Catalogue Creation
• Cross-Matching
• Create Photometric Redshifts
• Sample Selection
• Model generation
• Model Fitting
• Outputs

Input Parameters e.g. SFR
E.g. Spitzer, WFCAM
E.g. XMM
E.g. GOODS SDSS
SED Model GALAXEV
SED Model PEGASE
SED Model Starburst99
Other SED Model
Sextractor
Sextractor
Sextractor
GALAXEV Spectrum
PEGASE Spectrum
Other Spectrum
Starburst99 Spectrum
X-ray Object Catalogue
Optical Object Catalogue
Infrared Object Catalogue
Observational Fit to Models
Cross-Matched SED Object Catalogue
Best fit Parameterisation
Photometric Redshift Maker
Object Selection e.g. colour criteria
6
Data Discovery

• Data Availability
• - Public data sets
• Access to data sets through VO services
• - e.g. Simple Image Access Protocol (SIAP)
• Quality of Data

7
Create Object Catalogues

• Program Source Extractor (Bertin Arnouts
  1996)
• Issues
• - Resolution difference in data sets e.g.
  ISAAC 0.15 arcsec per pixel and IRAC 0.6 arcsec
  per pixel
• - How to deal with upper limits on flux
  measurements
• - Unit system for flux measurements, AB or
  Vega magnitudes

8
Cross Match Catalogues

• TOPCAT
• STILTS
• (http//www.star.bris.ac.uk/mbt/topcat/)
  
• (http//www.star.bris.ac.uk/mbt/stilts/)
• STILTS example tmatch2
• Issues
• - Suitable region for match
• - Multiple catalogue matching

9
Create Photometric Redshifts

• Various codes currently available
• - Hyperz (Bolzonella et al. 2000)
• - Bpz (Benitez 1999)
• - ImpZ (Babbedge et al. 2004)
• - ANNz (Collister Lahav 2004)
• Issues
• - Accuracy

10
Sample Selection

• User input required
• Type of input
• - Colour cut selection through colour-colour
  plots
• - Specify objects via specific RA and Dec
• or redshift

11
Population Synthesis Model Generation

• Current Models listed in AG
• - Galaxev or Bruzual and Charlot (Bruzual
  Charlot 2003)
• - Pegase (Fioc Volmerange 1997)
• - Starburst99 (Leitherer et al. 1999)
• Issues
• - Generate model on the fly or have a
  standard set of models pre-computed
• - Models vary in how they calculate
  synthetic spectra e.g. Galaxev uses a specific
  metallicity while Pegase utilises a consistent
  metallicity evolution
• - Assigning a common unit system for the
  same parameters in each model code

12
Model Fitting

• Minimisation Technique
• - Chi-squared
• Statistics Package R
• (Robert Gentleman and Ross Ihaka (R R)
  plus collaborators 1997 - http//www.r-project.org
  /)
• - Routines available for plotting confidence
  ellipses, chi-square tests plus other useful
  statistic tools
• Issues
• - Get filter information during the SED
  fitting
• - R into AstroGrid, currently not available
  - worthwhile

13
Outputs

• Output
• - Object cut-outs
• - Best fit SED plots
• - Tabulated results containing physical
  parameters such as SFH, age, stellar mass,
  metallicity and SFRs.
• Issues
• - Storage MySpace in AstroGrid
• - Format

14
Study of a sample of objects at 3 in the
GOODS-South field

• Hildebrandt et al. 2005 Sample The
  Garching-Bonn Deep Survey (GABODS)
• - WFI_at_MPG/ESO2.2m - UBVRI - 0.238 arcsec
  pixel-1
• - approx. 0.25 sq deg (900 sq arcmin)
• - Sample of 1000 3 objects selected
  through colour selection with photometric
  redshifts
• - mag limit I
• GOODS field approx. 0.046 sq deg (165 sq arcmin)
• - ISAAC_at_VLT JHK 0.15 arcsec pixel-1
• - IRAC instrument on Spitzer
  3.6,4.5,5.8,8.0 microns 0.6 arcsec pixel-1

15
Multi-wavelength Catalogue 3U-drop selection

• Original Hildebrandt et al sample 1000 U-drop
  objects
• Cross-Match
• - ISAAC data sample reduced to 73 objects
• - IRAC data sample reduced to 18 objects
• Object photometry checked for blended objects 9
  objects
• CAVEAT Sample greatly reduced due to only
  using detected objects in all bands and not using
  upper limit. This is a weakness in the data but
  will be addressed when introducing upper limits
  and position point aperture photometry

16
What is a U-dropout? Illustration
Filter transmission profiles for the WFI shifted
to the observed frame. Over plotted is a example
spectra of a Lyman break galaxy (LBG) redshifted
to z3. The Lyman break occurs at approx. 1.5
microns where the dropout technique utilises the
large break in intensity to identify these
galaxies.
Lyman break line
17
Various Data Sets covering the GOODS-South field
18
Study Results
U B V R
I J H Ks
3.6µm 4.5 µm 5.8 µm
zphot 2.52 Age 500 Myr Stellar mass
9.9e109 Msun Reduced chi-sq 1.04 SFR current
0.79 Msun yr-1
19
Study Results Contd
Confidence ellipses for estimates on stellar mass
and current SFR based on the technique used by
Eyles et al. 2005
20
Study Results Contd
The distribution of stellar masses and ages for
the 9 3 objects in the GOODS-South field
21
Summary of Results
22
Future Work

• Start to implement the steps into an AstroGrid
  Workflow
• Deal with the issues that currently reduce the
  amount of valid scientific results this technique
  would produce e.g. introducing upper limits into
  the observational data thus, increasing the
  sample size