illuminating the distant universe with candels · my research falls under the “cosmic dawn”...
TRANSCRIPT
Illuminating the Distant Universe
with CANDELSSteven Finkelstein
Texas A&M University ! University of Texas Austin
New Horizons at High Redshift, Cambridge, UK July 25th, 2011
CANDELS OverviewCosmic Assembly Near-infrared Deep Extragalactic Legacy Survey
CANDELS: Largest HST project ever - 902 orbits, 5 fields.Co-PIs: Sandy Faber (UCSC) and Harry Ferguson (STScI)
Primarily WFC3/IR imaging, some WFC3/UVIS imaging. Also ACS parallels.
CANDELS reference papers:Survey: Grogin et al. 2011, ApJS Submitted, astroph/1105.3753Data: Koekemoer et al. 2011, ApJS Submitted, astroph/1105.3754
Science Goals:Cosmic Dawn (z > 6): Discover robust samples of galaxies at z > 7; constrain evolution of physical properties from z ~ 4 - 8; identify high-z AGNs.Cosmic Noon (1 < z < 4): Study rest-optical morphologies; improve census of passively evolving galaxies; co-evolution of galaxies and black holes; AGN hosts.Supernovae and Dark Energy: Using multi-epoch observations in GOODS fields, provide large sample of SNe Ia at z > 1.5; Grism spectroscopy to confirm redshifts.UV Science: Measure ionizing photon escape fraction at z ~ 2.5; identify galaxies via the Lyman break at z ~ 2.
CANDELS Overview
Full survey details are given at our website: http://candels.ucolick.org
Field Bands Depth Completion
UDS JH 2 orbits Done
GOODS-S Y/JH 3.5/10 (Deep)1/2 (Wide)
Wide is complete, Deep complete 2/2012
EGS JH 2 orbits mid-2013 (50% now)
COSMOS JH 2 orbits 2/2012
GOODS-N YJH 3.5/10 (Deep)1/2 (Wide) mid-2013
Cosmic Dawn
My research falls under the “Cosmic Dawn” category.With the current CANDELS data, we can study galaxy evolution during the 4 < z < 8 epoch.
What questions do I want to ask?How do the colors of galaxies evolve?
What does this tell us about their stellar populations?How does the luminosity and stellar mass functions of galaxies evolve?
What do these tell us about reionization and the cosmic star-formation history?
previous resultsA number of papers have now published samples of galaxies at z > 7 (e.g., Bouwens et al. 2010abc; Oesch et al. 2010; Finkelstein et al. 2010; McLure et al. 2010, 2011; Yan et al. 2010; Bunker et al. 2010).It was noted that these z~7 galaxies were blue, with the faintest galaxies having ! ~ -3!
In Finkelstein et al. (2010) we used 23 galaxies at z ~ 7 selected via photo-z’s to study these colors in more detail.
Black: All
Red: H < 28.3
Blue: H > 28.3
!faint = -3.07 ± 0.51
!all = -2.44 ± 0.25
previous resultsA number of papers have now published samples of galaxies at z > 7 (e.g., Bouwens et al. 2010abc; Oesch et al. 2010; Finkelstein et al. 2010; McLure et al. 2010, 2011; Yan et al. 2010; Bunker et al. 2010).It was noted that these z~7 galaxies were blue, with the faintest galaxies having ! ~ -3!
In Finkelstein et al. (2010) we used 23 galaxies at z ~ 7 selected via photo-z’s to study these colors in more detail.
Black: All
Red: H < 28.3
Blue: H > 28.3
!faint = -3.07 ± 0.51
!all = -2.44 ± 0.25
• Conclusions:• Galaxies have evolved
significantly since z ~ 3.• Galaxies at z ~ 7 have rest-
UV colors consistent with normal star-forming galaxies, such as NGC1705 (see also Dunlop et al. 2011).
Measuring UV color
Drawbacks: Large uncertainties at high-redshift leave much to be desired when studying evolution of stellar populations.
To resolve this problem, we need deeper data and a larger sample.
We now use four fields:
HUDF: now use the full-depth HUDF data.
~0.2-0.3 mag deeper in each band compared to the data used in the first slate of papers.
ERS: Northern ~ 3rd of GOODS-S
CANDELS GOODS-S Deep: 4-orbit JH depth, full (3.5 orbit) Y-depth over 1/3 of area.
CANDELS GOODS-S Wide: Complete 2-orbit JH and 1-orbit Y depth over full area.
The much larger areal coverage of the CANDELS+ERS fields dramatically increases the numbers, especially at the bright end.
Measuring UV color
Drawbacks: Large uncertainties at high-redshift leave much to be desired when studying evolution of stellar populations.
To resolve this problem, we need deeper data and a larger sample.
We now use four fields:
HUDF: now use the full-depth HUDF data.
~0.2-0.3 mag deeper in each band compared to the data used in the first slate of papers.
ERS: Northern ~ 3rd of GOODS-S
CANDELS GOODS-S Deep: 4-orbit JH depth, full (3.5 orbit) Y-depth over 1/3 of area.
CANDELS GOODS-S Wide: Complete 2-orbit JH and 1-orbit Y depth over full area.
The much larger areal coverage of the CANDELS+ERS fields dramatically increases the numbers, especially at the bright end.
ERS
DEEP
WIDE
HUDF
15 a
rcm
in
10 arcmin
Catalogs and Photo-z’s
CatalogsMade J+H band selected catalogs for each of the four fields.
Ran SExtractor in two-image mode on PSF-matched BViz (ACS) YJH (WFC3) images in all four fields.
Y=F105W in all but ERS, which is Y=F098M.
Photometric RedshiftsPhoto-z’s were computed using the EAZY template fitting code (Brammer et al. 2008).
Want to select independent samples at z=4,5,6,7 & 8.
Define redshift ranges with "z = 1, i.e., z=4 sample is from 3.5 < z < 4.5.
Use full P(z) information in the following way:
1) SNR in F125W & F160W ≥ 3.5.2) ∫P(z) from zsample ± 0.5 ≥ 0.25
i.e., 3.5 - 4.5 for z=4 sample
3) ∫P(z) under primary peak ≥ 0.7primary peak is dominant
Sample Selection
Want to select independent samples at z=4,5,6,7 & 8.
Define redshift ranges with "z = 1, i.e., z=4 sample is from 3.5 < z < 4.5.
Use full P(z) information in the following way:
1) SNR in F125W & F160W ≥ 3.5.2) ∫P(z) from zsample ± 0.5 ≥ 0.25
i.e., 3.5 - 4.5 for z=4 sample
3) ∫P(z) under primary peak ≥ 0.7primary peak is dominant
Sample Selection
Want to select independent samples at z=4,5,6,7 & 8.
Define redshift ranges with "z = 1, i.e., z=4 sample is from 3.5 < z < 4.5.
Use full P(z) information in the following way:
1) SNR in F125W & F160W ≥ 3.5.2) ∫P(z) from zsample ± 0.5 ≥ 0.25
i.e., 3.5 - 4.5 for z=4 sample
3) ∫P(z) under primary peak ≥ 0.7primary peak is dominant
Sample Selection
Redshift distributions
HUDF
WIDE ERS
DEEPz #
All 29114 20025 5556 2357 868 33
Measuring the UV color
The rest-frame UV color is typically parameterized by the UV spectral slope !, where f# ! #!.
This is defined spectroscopically, by fitting the spectral slope in wavelength windows defined by Calzetti et al. (1994).
Windows are designed to omit spectral emission and absorption features.
At high-z, continuum spectroscopy isn’t available, so ! is traditionally estimated from a single rest-UV color.
(e.g., Meurer et al. 1997; Hathi et al. 2008; Bouwens et al. 2010).
With this dataset, we have multiple rest-UV colors for all but the highest redshift bin.
Can we do better?
SED Fitting
We use all available colors by measuring ! during the SED-fitting process.
Find the best-fit model to a given galaxy, and measure ! from that best-fit spectrum, using the Calzetti et al.-defined windows.
Advantages:
Essentially averages over all colors.
Obtain accurate ! for the specific redshift.
Disadvantages:
Restricted to the choice of models.
We use CB07 (11?); bluest ! = -3.15.
We don’t find anything at ! < -3, so maybe its ok (for now)...
Does it make a difference?
To compare color-derived vs. SED-derived ! methods, we ran simulations.Input mock galaxies into our images, recovered them, and measured ! with both the single color, and SED fitting method.
Compared to input value of !.
No bias!
Does it make a difference?
To compare color-derived vs. SED-derived ! methods, we ran simulations.Input mock galaxies into our images, recovered them, and measured ! with both the single color, and SED fitting method.
Compared to input value of !.
No bias!
Does it make a difference?
To compare color-derived vs. SED-derived ! methods, we ran simulations.Input mock galaxies into our images, recovered them, and measured ! with both the single color, and SED fitting method.
Compared to input value of !.
No bias!
I have measured ! for the HUDF, GOODS-S Deep, GOODS-S Wide and ERS samples via the SED fitting method.
At each redshift, I’ve computed mean values of ! for the total sample, as well as two sub-samples split by the evolving value of 0.4 L* (using LF from Bouwens+07,10).
Results
SquaresRed=HUDFBlue=Wide
Green=DeepYellow=ERS
CirclesBlack: All
Red: > 0.4L*
Blue: < 0.4L*
Evolution in Beta
!faint = -2.6 ± 0.2
!all = -2.3 ± 0.1
Simple model: Assume the entire change in color is due to dust.Assumptions:
zf = 20, maximally old stellar population, Z = Z⨀, evolving with a constant SFH.
To match the observed ! of -1.82 @ z=4 and ! = -2.25 @ z=7:
AV drops from 0.4 @ z=4; " AV = 0.04 at z=7.If Z = 0.2* Z⨀:
AV drops from 0.5 @ z=4, " AV = 0.2 at z=7.
Evolution in Beta
summary
Galaxies appear to contain little dust at z=7, gaining a significant amount by z=4-5.
t(z=20 " 7) ~ 550 Myr t(z=20 " 5) ~ 1 Gyr
Consistent with a scenario where SNe seed the galaxies with a low level of dust prior to z=7, but then AGB stars really begin to contribute at z < 7.
This much larger sample size yields more precise measurements of !.Allows for robust conclusions to be made about the evolution of galaxies from z = 7 ! 4.
Results will be bolstered by folding in other CANDELS fields, as well as HUDF parallel fields.